2000 HSR&D Annual Meeting: Abstract 50-99
50. A Psychiatric Diagnoses-Based Case Mix System for Mental Healthcare Utilization Data
James N. Breckenridge, Ph.D., VA Palo Alto Healthcare System, Palo Alto, CA
Objectives: The purpose of this study was (1) to develop a psychiatric diagnoses-based case mix methodology (PDCM) for risk adjusting provider profiling comparisons of concurrent ambulatory VA mental healthcare and (2) to compare the efficacy of this methodology with the Johns Hopkins Ambulatory Care Groups (ACGs) approach.
Methods: Sample I, comprised of 41,600 patients from a tertiary care, academic affiliated VA medical center offering a broad array of mental health services was employed for the initial development of the PDCM. Results were cross-validated in a second sample of 42,484 patients from two additional academic tertiary care VA medical centers. All ICD9 diagnoses assigned to patients during any inpatient stay or outpatient encounter throughout FY98 were used for ACG assignment. A combination consensus clustering (Breckenridge, 1989, 1999) procedure and a recursive partitioning algorithm utilizing all psychiatric diagnoses was employed to develop the PDCM. For each patient, all procedures reported for any outpatient encounter during FY98 were converted from CPT4 codes to HCFA relative value units (RVUs). Both annualized Work RVUs and Total RVUs were calculated. In addition, the number of outpatient encounters, total DSS costs, and HCFA FY98 estimated total reimbursements were employed as further healthcare utilization measures. These five measures were computed for total (i.e., both medical and psychiatric) ambulatory care and for mental healthcare only. As is typical for healthcare utilization data, the distributions of all ten measures were highly skewed with heteroscedastic variances. Consequently, generalized linear regression models were employed to assess the relationship between each utilization measure and the ACG and PDCM classification variables.
Results: Roughly one third of Sample I patients carried an ICD9 psychiatric diagnosis or related V-code. Although ACGs could account for more than half the variance in measures of total ambulatory healthcare utilization for patients without psychiatric diagnoses, the goodness of fit was much poorer for patients with psychiatric problems. Thus, the amount of variance accounted for by ACGs for the psychiatric group ranged from 30 percent (for the number of encounters) to 56 percent (for Total RVUs) less than the results for the non-psychiatric patient group. The PDCM variables, however, could account for an additional 10 to 26 percent of the variance in total utilization among psychiatric patients beyond that attributable to ACG classifications. When ACG and PDCM were employed together, from roughly 40 to 59 percent of the variance in total utilization measures could be accounted for in the psychiatric subgroup. A somewhat improved fit was observed when utilization indices were restricted to mental healthcare only. Finally, cross validation in Sample II strongly supported the results observed for the development sample.
Conclusions: A significant, substantial relationship between diagnoses-based classifications and concurrent VA ambulatory mental healthcare utilization was observed, suggesting the utility of a diagnoses-based risk adjustment approach to indexing the relative illness burden of psychiatric patients.
Impact: The combined ACG/PDCM approach has been selected for use in a national study of provider profiling for VA psychologists and psychiatrists, and its efficacy as a prospective predictor of utilization is now being assessed with FY99 data.
51. Clinical Outcome, Functional Outcome, and Health Care Costs in Manic-Depressive Disorder: How Closely Related?
Mark S. Bauer, Providence VAMC, Providence, RI, G Kirk, Perry Point Cooperative Studies Coord Ctr, WO Williford, Perry Point Cooperative Studies Coord Center
Objectives: It is frequently assumed that illness severity drives functional outcome and service use in serious mental illness (SMI). However, evidenc to the contrary is beginning to develop for several types of SMI. Accordingly, for manic-depressive disorder we sought to quantify the inter-relationship of three outcome domains in manic-depressive disorder: clinical outcome (symptoms and episodes), functional outcome (social and occupational function), and health care costs.
Methods: Forty-three veterans with manic-depressive disorder were followed in an easy-access ambulatory treatment program under naturalistic conditions. Data were collected over 48 weeks using validated and reliable structured measures for clinical and functional outcome and direct treatment costs from the perspective of the payor (the VA health care system). Data were analyzed using multiple regression and related techniques.
Results: Factors that predicted clinical outcome did not predict functional outcome. Factors that predicted functional outcome did not predict clinical outcome. No factors predicted health care costs. Functional outcome was correlated only with depressive symptoms during follow-up, but much of the variance remained unexplained. Health care costs did not correlate with either clinical or functional outcome.
Conclusions: The three outcome domains studied were not highly associated, indicating that other factors than illness severity must be invoked to explain the complex phenomena of social and occupational function and health care utilization. Results are consistent with a model for schizophrenia introduced by Strauss and Carpenter in 1972 that proposed that outcome in that SMI was characterized by "linked open systems" of determinants.
Impact: Conceptually, we must revise our understanding of functional outcome and health care costs as being passively driven by illness severity. Practically, measurement of all 3 outcome domains is necessary to characterize outcome in this SMI. Therapeutically, aggressive attention to ongoing depressive symptoms is likely to be necessary to optimize functional outcome. However, the converse model may also hold: that optimizing functional status is necessary to reduce continuing depression.
HSR&D Funded: CSP #430, DEV97-015
52. Tuberculosis (TB) Prevention Measures among US Hospitals with versus without Prior Nosocomial TB Outbreaks (1995-1997)
Jennie Hounshell, VA Chicago Healthcare System--Lakeside Division, Chicago, IL, AM Sipler, VA Chicago Healthcare System--Lakeside Division, CL Bennett, VA Chicago Healthcare System--Lakeside Division
Objectives: Nosocomial TB transmission remains a concern, following multi-drug resistant TB outbreaks. Most hospitals comply with CDC guidelines for TB prevention, yet TB skin test conversion rates have increased four-fold at HIV-experienced hospitals with low-TB caseloads, while decreasing by two-thirds at HIV-experienced hospitals with high-TB caseloads (Manangan L, Chest, 1999, in press). We evaluated variations in delays in TB isolation during 1995-1997 among 1,932 HIV-infected patients presenting with TB signs/symptoms.
Methods: Rates of TB isolation and ordering AFB smears for high-risk HIV-infected individuals in the first 2 hospital days were measured at 59 HIV-experienced hospitals in Chicago, New York, Los Angeles, Miami, Seattle, Phoenix, and Tucson.
Results: All cities ordered AFB smears at a rate of >55%, with a range from 40% to 100%. County hospitals in each city, the most experienced HIV and tuberculosis providers, had the highest rates of ordering AFB smears. Private hospitals, with low caseloads of tuberculosis, generally had the lowest 2-day rates of ordering AFB smears. Low 2-day rates of TB isolation were found in Phoenix/Tucson (8%), Seattle (21%), and LA (28%), cities without prior nosocomial TB outbreaks, while cities with prior nosocomial TB outbreaks (New York, Chicago, and Miami) had rates >60%. Within cities, TB isolation rates varied much more than rates of ordering AFB smears. Only 8% of hospitals in cities with a prior TB outbreak had 2-day isolation rates that were more than 30 percentage points lower than AFB smear rate, versus 68% of hospitals in cities without a prior TB outbreak. For cities with prior nosocomial TB outbreaks, TB isolation rates were a mean of 13 percentage points lower than AFB smear rates, versus isolation rates that were 45 percentage points lower than AFB smear rates for cities without a TB outbreak (p<0.0001). Four hospitals in cities without a prior nosocomial outbreak isolated only 1, 1, 2, and 3 patients, while sending smears on 19, 23, 18, and 47 patients, respectively. One HIV-experienced hospital in Chicago had a nosocomial outbreak of multi-drug resistant TB during the study period. That hospital failed to isolate 40% of confirmed cases of HIV-infected individuals with TB, while providing care for 6 of 34 multi-drug resistant TB cases in the city. During the outbreak, skin test conversions occurred in six nurses, four physicians, and one ward secretary. One health care worker who had previously become HIV-infected following an in-hospital needle-stick, developed a multi-drug resistant tuberculosis infection.
Conclusions: We have identified clinical practices that place patients and health care workers at significant risk. As with most aspects of medicine, experience counts. Most hospitals in three cities that experienced a nosocomial outbreak of TB (New York, Chicago, and Miami) require TB isolation for patients who have sputum smears evaluated for AFB and comply with these requirements.
Impact: HIV-experienced hospitals in cities without a prior nosocomial tuberculosis outbreak should adopt this TB prevention policy and isolate all patients who have AFB smears ordered.
HSR&D Funded: HSR&D--Epidemiology
53. The Relationship between Posttraumatic Stress Disorder Severity and Health Service Utilization
Patrick S. Calhoun, PhD, Durham VAMC and Duke University Medical Center, Durham, NC, HB Bosworth, Durham VAMC and Duke University Medical Center, JC Beckham, Durham VAMC and Duke University Medical Center, TK Dudley, Durham VAMC, SC Grambow, Durham VAMC and Duke University Medical Center
Objectives: Posttraumatic stress disorder (PTSD) is one of the most common mental disorders and is associated with both increased subjective health complaints and physician diagnosed medical conditions. Despite this, there is relatively little research examining the relationship between PTSD and health care utilization. This study examined the relationship between health care utilization and PTSD severity.
Methods: Data were collected on a sample of male veterans (N=1,118) diagnosed with PTSD at a VAMC specialty PTSD clinic in the southeastern United States between 3/92 and 3/98. PTSD diagnosis was based upon structured clinical interview. Average age of the sample was 49 years. Fifty-four percent of the sample was Caucasian and 42% were African-American. PTSD severity was assessed with the Mississippi Scale for Combat-related PTSD. Objective health service use data were collected from a centralized VA database for the year prior to the patients’ PTSD evaluation. For the analysis of service utilization, a clinic classification system based upon categorization of VA clinic stops was used to categorize each stop for the purpose of either medical or psychiatric care. Logistic regression analyses were used to examine the relationship between service use and PTSD severity after controlling for demographic and service-related disability factors.
Results: The sample had a median of 15 outpatient visits during the year prior to their PTSD evaluation. The median number of physical health outpatient visits was 11 while the median number of mental health outpatient visits was 3. Twenty-two percent of patients had at least one inpatient visit during the year prior to their evaluation and the average length of stay for these veterans was 6 days. Logistic regression analyses indicated that PTSD severity was significantly associated with increased outpatient mental health use, a history of psychiatric inpatient hospitalization, and increased length of stay after controlling for race, age, education, socioeconomic status, and service-connected disability. Veterans with severe PTSD were almost two times more likely than less severe patients to have high use of outpatient mental health services. Similarly, patients with severe PTSD were almost two times more likely then less severe patients to have been hospitalized at a VA medical center in the year prior to their PTSD evaluation. Among those with an inpatient stay, patients with severe PTSD were more than two times more likely to have stays lasting longer than 6 days.
Conclusions: Veterans with PTSD are more likely to present for physical health care than mental health services. Among PTSD patients, symptom severity is associated with increased use of mental health services.
Impact: As the VA strives to improve health care for veterans, increased understanding of the potential utilization of health care among PTSD patients is critical. The current study is one of the first to examine the relationship between PTSD and health service use with objective utilization data. Results suggest that veterans subsequently diagnosed with PTSD are more likely to present at VAMC outpatient clinics for treatment of physical symptoms than for mental health services.
54. Comorbid Lifetime Alcohol Use Disorders and Depressive Disorders in a Community Population
JoAnn Kirchner, MD, HSR&D Center for Mental Healthcare and Outcomes Research, North Little Rock, AR, GM Curran, HSR&D Center for Mental Healthcare and Outcomes Research, RR Owen, HSR&D Center for Mental Healthcare and Outcomes Research, CR Thrush, HSR&D Center for Mental Healthcare and Outcomes Research, BM Booth, HSR&D Center for Mental Healthcare and Outcomes Research
Objectives: This study describes, in a community population of individuals with an alcohol use disorder (n=619), differences in sociodemographic characteristics, comorbidities, alcohol use severity, and functioning between individuals with and without a lifetime diagnosis of a depressive disorder. For the individuals with comorbid alcohol and depressive diagnoses (n=109, 17.6%), analyses compared those whose onset of depression preceded the onset of an alcohol disorder to those whose alcohol disorder was the first diagnosis.
Methods: We conducted screening telephone interviews in six southern states (Alabama, Arkansas, Georgia, Louisiana, Mississippi, and Tennessee) with a stratified random sample of community-residing individuals (age 18 or older) to identify a cohort of at-risk drinkers. At-risk drinkers were defined as individuals who either met DSM-IV criteria for past year alcohol abuse or dependence or who were at-risk for meeting diagnostic criteria in the following year. The initial bivariate analysis compared subjects with an alcohol use disorder and a comorbid depressive disorder to those without a lifetime diagnosis of a comorbid depressive disorder, and also compared the subjects with primary onset of an alcohol use disorder to those with primary onset of the depressive disorder. Multivariate analyses examined whether the relationships observed in bivariate analyses would continue to be significant while controlling for important covariates.
Results: Results showed that individuals who met criteria for a lifetime depressive disorder had lower levels of mental and physical functioning, worse illness severity (more likely to be alcohol dependent, more alcohol symptoms), more comorbid drug and psychiatric disorders (other than depression) and more chronic physical conditions. The average age of primary onset alcohol use disorders was 18.4 (SD=5.2) and of primary onset depressive disorders it was 19.2 (SD=8.2). The onset of depression in subjects with primary onset alcohol use was 28.1 (SD=8.5), whereas the onset of alcohol in subjects with primary onset depression was 23.9 (SD=9.8). Few differences were found between groups based on their order of diagnostic onset. In multivariate analysis, female gender (OR=3.3, p<0.001), alcohol dependence (OR=3.1, p<0.001) psychiatric comorbidity (OR=1.8, p<0.05), and drug comorbidity (OR=2.5, p<0.001) were significant predictors of a lifetime diagnosis of a depressive disorder. In our next series of models, lifetime diagnosis of depression was a significant predictor of current physical conditions (OR=2.7, p<0.001) and was negatively associated with current mental health functioning (the SF-36 MCS: p<0.001). In addition, participants with a lifetime diagnosis of a depressive disorder were significantly more likely to be without insurance (OR=1.8, p=0.01) controlling for age, gender, income, employment, ethnicity, marital status, and education.
Conclusions: Findings from this study both replicate and extend the findings in treatment populations to a community sample that the presence of a comorbid depressive disorder is associated with more impairment in functioning, which may lead to less access to health care.
Impact: The most compelling policy implication from these results is that the community dwelling problem drinkers who may need health insurance the most, the most severely impaired, may have less access to health care as a consequence of their associated mental health status.
55. Clinical Significance of Undiagnosed Diabetes among VA Medical Center Outpatients
David Edelman, Center for Health Services Research in Primary Care, Durham VA Medical Center, Durham, NC, LJ Edwards, Center for Health Services Research in Primary Care, Durham VA Medical Center, TK Dudley, Center for Health Services Research in Primary Care, Durham VA Medical Center, EZ Oddone, Center for Health Services Research in Primary Care, Durham VA Medical Center
Objectives: Advocacy groups recommend screening all adults over the age of 45 for unrecognized diabetes mellitus. One possible reason for screening is to improve outcomes associated with unrecognized diabetes. However, the comorbid diagnosis of diabetes changes the treatment approach for other chronic diseases (i.e. hypertension, hyperlipidemia). Therefore, another rationale for screening is to identify diabetes so that comorbid hypertension and hyperlipidemia may be treated more aggressively. Our objectives were to determine the clinical severity of unrecognized diabetes among VA medical center outpatients, and to identify the prevalence of comorbid hypertension and hyperlipidemia in patients with unrecognized diabetes.
Methods: We contacted all veterans age 45-64 who used the Durham Veterans Affairs Medical Center between 1/98 and 4/99 and asked if they had diabetes. All patients who did not report having diabetes were eligible for the screening study. We obtained comorbidity (Kaplan-Feinstein comorbidity index), quality of life (SF-36) and depression (CES-D) measures on all subjects. We screened patients by using an initial random Hemoglobin A1c (HbA1c) measurement, and then obtaining follow-up fasting plasma glucose (FPG) for all subjects with HbA1c >= 6.0%. A case of unrecognized diabetes was defined as either HbA1c >= 7.0% or FPG >= 126. We additionally obtained fasting lipid profile, serum creatinine, and quantitative urine protein measures on patients with HbA1c >= 6.0%.
Results: We enrolled 1253 patients and 56 were found to have diabetes at screening. We achieved at least partial follow-up on 238/258 patients (92%) with HbA1c >= 6.0%. The prevalence of unrecognized diabetes was 4.5% (95% CI, 3.4, 5.7). Of the 56 patients diagnosed with diabetes at screening, 23 (41%) had HbA1c levels >= 7.0%. The rates of proteinuria and renal insufficiency were similar in those patients diagnosed with diabetes compared to those subjects who screened negative for diabetes. Subjects with unrecognized diabetes had SF-36 physical functioning scores 2.8 points lower than subjects who did not have diabetes upon screening (p=0.49). CES-D depression scale scores and exercise levels were also similar among subjects with and without diabetes identified at screening. While overall comorbidity scores were similar among subjects with and without diabetes at screening, 41/56 patients with diabetes diagnosed at screening (73%) had hypertension, and 8/39 patients (21%) with a follow-up fasting lipid profile had LDL-cholesterol levels >= 130.
Conclusions: Patients identified with diabetes at screening do not have markedly different health status from those who are found not to have diabetes. Many of these patients have HbA1c levels that suggest additional therapy for their diabetes is necessary, but they rarely have complications of diabetes at the time of screening. Most patients screening positive for diabetes have hypertension and/or hyperlipidemia.
Impact: Screening for diabetes might be important for identifying patients who will benefit from more intensive treatment for the common chronic illnesses associated with diabetes. VA and other health care providers should consider issues of burden of disease and treatment of comorbid disease in deciding whether to perform systematic screening for diabetes.
HSR&D Funded: CSP 705D
56. Effect of Community-Based Outpatient Clinic Care on Patient Satisfaction and Health Status: A Randomized Trial
Steven J. Borowsky, Minneapolis Center for Chronic Disease Outcomes Research, Minneapolis, MN, DB Nelson, Minneapolis Center for Chronic Disease Outcomes Research, JL Bradley, Minneapolis Center for Chronic Disease Outcomes Research, SM Nugent, Minneapolis Center for Chronic Disease Outcomes Research, V Boulaevskaia, Minneapolis Center for Chronic Disease Outcomes Research, PR Hamann, HB Rubins, Minneapolis VAMC/VISN 13 Minneapolis Center for Chronic Disease Outcomes Research
Objectives: VHA has approved more than 230 Community-based Outpatient Clinics (CBOCs) for a national initiative to provide primary care in community-based settings. This randomized trial compares the effects of CBOCs and VA facility-based primary care on satisfaction with care and health status.
Methods: In 1998, two primary care CBOCs were established in Minnesota using contracts with non-VA clinics. 1452 veterans who lived within 50 miles of either CBOC site and had at least one primary care visit at any of three VA facilities during the preceding year were eligible to transfer primary care to a CBOC. In a southern Minnesota region 784 veterans were offered a consortium of 23 clinic sites; in a northern Minnesota region 668 veterans were offered a CBOC consisting of 2 clinic sites. Of 624 veterans requesting CBOC care, 66 who were >50% service-connected were assigned to their area CBOC; the remaining 558 were randomized to their area CBOC (n=334) or VA (n=224) resulting in 400 CBOC patients (200/CBOC), the maximum number that prior to the study, local VA administration had planned placing in these CBOCs. Baseline questionnaires assessed health status (SF-36V) and satisfaction with VA care [VA Customer Satisfaction Survey (CSS)]. Of 519 randomized subjects alive after 1 year, 472 (91%) completed a follow-up SF-36V. 56 of the 519 subjects were determined to have had no visits to their assigned clinic. 399 (86%) of the remaining 463 completed a satisfaction questionnaire concerning their assigned clinic. Multiple regression methods were used for SF-36V and satisfaction scale comparisons at follow-up adjusted for the following baseline variables: age, service-connected status, number of prior VA clinic encounters, respective baseline SF-36V or CSS scale score, and (for CSS scores only) SF-36V physical and mental component scores. Intent to treat analysis was used.
Results: After adjustment for baseline covariates, patients randomized to CBOC care compared to those randomized to VA care were less likely to report problems on the education/information (p< .001), visit coordination (p<.001), and overall coordination of care scales (p<.001). More pronounced CBOC-VA differences on the overall coordination scale were observed for the northern than southern CBOC (interaction p< .05). CBOC and VA patients did not differ for the access/timeliness, courtesy, emotional support, preferences, or continuity scales. CBOC patients were also marginally more likely than VA patients to rate the overall quality of their most recent visit as very good or excellent (p=.05, adjusted OR 1.6, 95%CI 1.0-2.6). CBOC and VA patients did not differ at follow-up with respect to adjusted SF-36V subscales.
Conclusions: CBOC care was associated with improved satisfaction with care in specific dimensions (patient education/information, visit-specific and overall coordination of care). CBOC patients were also more satisfied overall with their latest visit. CBOCs neither affected health status nor improved patient perceived access and timeliness of care (though few problems in this area were noted at baseline).
Impact: Although limited to two CBOCs, this study, the only randomized study of CBOC outcomes that we know of, demonstrates that CBOCs can improve some dimensions of patient satisfaction.
HSR&D Funded: CDA 98-309
57. Efficiacy of Telepsychiatry in the Treatment of Depression
Paul Ruskin, MD, VA Maryland Health Care System, Baltimore, MD
Objectives: Telepsychiatry is a novel means of providing expert psychiatric treatment to patients who live far from a source of care. The major goal of this study is to demonstrate that treatment of depressive disorders via telepsychiatry is as effective as in-person treatment. In addition, this study will show that depressed patients are as compliant and as satisfied with telepsychiatry as with in-person treatment. Another goal is to demonstrate that telepsychiatry is as cost-effective as in-person treatment.
Methods: Veterans who present with depression to any of three VA Maryland Health Care facilities, and who have a SCID-IV depression diagnosis as well as a minimum Ham-D score of 16, are eligible for participation. Eligible veterans are randomized to either in-person or remote treatment. Psychiatrists from each of the three facilities conduct both in-person and remote treatment, which involves eight sessions over a six month period. Treatment consists of psychotropic medication and psycho-education concerning the disease, medications, and side effects. Outcome measures include: Ham-D Scale, Beck Depression Inventory, Spielberger Anxiety Scale, drop out rate, and a patient satisfaction scale.
Results: Of the 46 patients who entered the study more than 6 months ago and who should have completed the study, 30 successfully completed the study while 16 discontinued. Of the 16 dropouts, 7 were in the in-person group and 9 were in the remote group. There was no statistical difference in dropout rates between the in-person and remote groups, Chi-squared =.38. As predicted, depressive symptoms on the HAM-D significantly improved over the treatment period, (F=25.9, p<.001) and this effect did not differ by treatment group. Additionally, there was no significant difference in the depression levels of the two treatment groups (F=.39, ns), and this was the case even when the initial level of depression was held constant (F=.06,ns). Similar results were found when using the BDI. As expected, the depression scores on the BDI significantly improved for the 46 patients altogether (F=12.6, p<.001) and this effect did not differ for the two treatment groups (F=1.2, ns). There were no significant differences in the depression levels of the two treatment groups (F=.06,ns) even when the initial depression levels were held constant (F=1.8, ns). Levels of anxiety over time were also examined. Anxiety levels as measured by the STAI were shown to significantly improve over time and did not differ for the two treatment groups. Finally, satisfaction levels were not statistically different for the two treatment groups.
Conclusions: The results indicate that depressed patients who see their psychiatrist via telepsychiatry show similar improvement to depressed patients who see their psychiatrist in person. Over the six month period of treatment, their depression and anxiety decreased, and they reported similar levels of satisfaction with treatment. They are no more likely to drop out of treatment than patients seen in-person.
Impact: This preliminary data is already indicating that treatment efficacy is similar for patients treated in-person and remotely. With the use of telepsychiatry, VA healthcare can be widely expanded, at great convenience to veterans and without incurring high costs.
HSR&D Funded: ACC97-341.
58. Do Automated Calls with Nurse Follow-up Improve Self-Care and Glycemic Control among VA Patients with Diabetes? Results of a Randomized, Controlled Trial
John D. Piette, PhD, Center for Health Care Evaluation, VA Palo Alto Health Care System, Palo Alto, CA, Menlo Park, CA, M Weinberger, Roudebush VA Medical Center, Indianapolis, IN, FB Kraemer, Center for Health Care Evaluation, VA Palo Alto Health Care System, Palo Alto, CA, SJ McPhee, University of California, San Francisco, CA
Objectives: Many VA patients with diabetes are at risk for inadequate self-care, poor glycemic control, and complications. Telephone care programs may improve their outcomes by fostering more effective communication with clinicians between face-to-face outpatient visits. However, such programs can be too labor-intensive to implement on a large scale. We conducted a randomized, controlled trial to evaluate the impact of automated telephone disease management (ATDM) calls with telephone nurse follow-up as a strategy for improving outcomes while using scarce nursing resources as efficiently as possible.
Methods: We enrolled patients with diabetes using hypoglycemic medications at the time of outpatient clinic visits. Participants were randomly assigned to usual care or to receive ATDM calls every two weeks. Calls included assessments of patients' self-care, glycemic control, and symptoms. Calls also included optional health education modules. Patients responded to ATDM queries using their touch-tone telephone keypad. ATDM problem reports were followed up over the telephone by a nurse who conducted more detailed assessments, provided advice and self-care education, and coordinated the intervention with patients' primary care. Outcomes measured at 12-months included glycosylated hemoglobin; serum glucose; and survey reported self-care behaviors, symptoms, and satisfaction with care. At baseline, study participants had unexpectedly good glycemic control: median HgA1c = 7.8%, interquartile range = 7.0% - 8.9%; median serum glucose = 159 mg/dl, interquartile range = 127 mg/dl - 208 mg/dl. Hence, we focused our analysis both on the overall group as well as the subgroup with HgA1c levels above the point where extra intervention is recommended (i.e., 8%).
Results: 290 patients were randomized and 272 (94%) provided outcome data at 12 months. On average, intervention patients at follow-up self-monitored their blood glucose more frequently than control patients (P = 0.03), though there were no differences between the two groups in the frequency of foot self-exams or medication adherence problems. In the sample overall, intervention and control groups had similar follow-up mean HgA1c levels (8.2% versus 8.0%, P = 0.3). However, in the subgroup of patients with baseline HgA1c > 8% (N = 123), those receiving the intervention had lower follow-up mean HgA1c levels than controls (8.5% versus 9.1%, P = 0.04) as well as marginally lower serum glucose levels (162 mg/dl versus 184 mg/dl, P = 0.09). In the overall sample, intervention patients reported fewer symptoms of poor diabetes control (3.6 versus 4.4, P = 0.03) as well as greater satisfaction with the interpersonal processes of their care (P = 0.02) and the quality of their health outcomes (0.004).
Conclusions: ATDM-supported nursing care may improve glycemic control among VA patients with diabetes who are hyperglycemic, but may not be useful for patients whose glucose levels already are well controlled. Regardless of their level of glycemic control, the intervention can positively impact symptoms and satisfaction with care.
Impact: VA is exploring creative strategies for improving the effectiveness of disease management for patients with diabetes and other chronic health problems. ATDM-supported nursing care could improve providers' ability to serve chronically-ill patients effectively while using less staff.
HSR&D Funded: IIR #95-084.2
59. Demand Characteristics of Residential Substance Abuse Treatment Programs
Christine Timko, Center for Health Care Evaluation, VAPAHCS, Menlo Park, CA, K Yu, Center for Health Care Evaluation, VAPAHCS, RH Moos, Center for Health Care Evaluation, VAPAHCS
Objectives: The purposes of this study were to (1) compare the demand characteristics of programs in which substance abuse, versus those in which psychiatric, patients were receiving treatment, and (2) examine whether receiving substance abuse care in a program with more of the features that distinguish substance abuse from psychiatric care was associated with better in-program outcomes.
Methods: A total of 1,158 patients living in 54 substance abuse and 46 psychiatric programs participated. The study used the Physical and Architectural Characteristics Inventory and the Policy and Service Characteristics Inventory to report on each program. The Resident Background Inventory and the Community Oriented Programs Environment Scale, which assesses the treatment climate, were completed by patients.
Results: Logistic random-effects regression analyses determined the physical and architectural features, policies, and services that were associated with treatment for a substance abuse rather than a psychiatric problem. When patients had a substance abuse diagnosis, their treatment program had fewer safety features (for security, monitoring, and accident-prevention), and social-recreational aids (to foster interaction and activity); in addition, program policies emphasized higher expectations for patients' functioning (physical and psychological) and more resident control (i.e., formal mechanisms for patients to influence program policies). Substance abuse patients were in programs that had fewer health and treatment services. Having determined that high-demand program characteristics distinguished substance abuse from psychiatric patients, we asked if substance abuse patients being treated in programs with these high-demand features (i.e., with few safety features, social-recreational aids, and health and treatment services, and high expectations for functioning and resident control) were better off than those in low-demand programs. Compared to substance abuse patients in low-demand programs (n = 240), those in high-demand programs (n = 201) judged the treatment climate to be more relationship oriented (more involving, supportive, and encouraging of spontaneity), more goal oriented (emphasizing patients' autonomy, and having a practical and personal problem orientation, and less anger and aggression), and more organized (having order and clarity regarding program routines). In addition, patients in high-demand programs engaged in more self-initiated activities and in more activities organized by the program.
Conclusions: Veterans in substance abuse residential care may benefit from programs with architectural features, policies, and services that have high demand characteristics, in terms of patients' functioning. This possibility should be examined in prospective studies considering alternative configurations of high- and low-demand program characteristics, and assessing patients' substance use outcomes at discharge and follow-ups.
Impact: Further development regarding the demand characteristics of substance abuse care may lead to better treatment outcomes for patients, and more efficient use of resources within the treatment system.
HSR&D Funded: IIR 95-011
60. Interventions that Increase use of Adult Immunization and Cancer Screening Services: a Meta-analysis of 218 Controlled Trials
Paul G. Shekelle, Cedars-Sinai Medical Center, Los Angeles, CA, EG Stone, Cedars-Sinai Medical Center, Morton SC, Rand, MA Maglione, Rand, BS Mittman, Greater Los Angeles VA Healthcare System-Sepulveda, LV Rubenstein, Greater Los Angeles VA Healthcare System-Sepulveda, PG Shekelle, Greater Los Angeles VA Healthcare System-West LA
Objectives: Immunizations to prevent influenza and pneumococcal pneumonia and screening for early breast, cervical, and colon cancer are key health care services for reducing morbidity and mortality. Performance rates remain below national targets. Numerous studies have evaluated interventions to improve adherance to these services, but no attempt has been made to synthesize evidence across interventions or preventive services. We used a typology based on social science theory to characterize key components and features of prevention improvement interventions and then used quantitative meta-regression of published and unpublished literature to evaluate effectiveness of intervention components and features.
Methods: MEDLINE, the Cochrane Effective Practice and Organization of Care register, prior systematic reviews, and the Medicare Health Care Quality Improvement Project database were searched for relevant studies that assessed interventions to increase any of the five preventive services that used an RCT, CCT, controlled before-after, or interrupted time series design. Two reviewers independently extracted data about study characterstics and outcomes from unmasked articles. Intervention components were classified as reminders, education, financial incentives, provider feedback, or organizational change. Intervention features were classified as the use of social influence, marketing/outreach, high visual appeal/clarity, collaboration/teamwork, design based on needs/barriers/theory, top management support, and active learning strategies. Meta-regression models were developed for immunizations and each cancer screening service to assess the effect of intervention components and features.
Results: 187 articles describing 218 studies met the inclusion criteria. Not all intervention components could be modeled for each service due to data sparseness. Adjusted odds ratios from meta-regression for effectiveness were as follows: for immunizations, organizational change=7.17, provider reminder=4.32, patient financial incentive=3.49, provider financial incentive=2.62, patient reminder=2.44, provider education=1.66, patient education=1.33, feedback=1.11; for mammography, patient financial incentive=3.57, patient reminder=2.57, organizational change=2.26, provider education=2.26, provider reminder=1.59, feedback=1.49, patient education=1.31; for pap smear, patient financial incentive=3.12, organizational change=2.65, patient reminder=1.84, provider education=1.59, patient education=1.53, provider reminder=1.40, feedback=1.12; and for fecal occult blood testing, organizational change=18.1, provider education=4.02, patient reminder=3.73, patient financial incentive=2.22, patient education=1.76, provider reminder=1.45, feedback=1.13. There was strong convergent validity for the meta-regression results from triangulation using individual studies of direct comparisons of interventions. Intervention features found to be consistently associated with success were collaboration/teamwork and design based on needs/theory.
Conclusions: Organizational change, patient financial incentives, and patient reminders were consistently more effective than education or feedback. Interventions were more effective if they were based on an assessment of needs/barriers/theory and if they incorporated collaboration/teamwork.
Impact: The study provides strong empiric support that the best strategy to increase use of preventive services is the implementation of specific organizational change that makes the identification and delivery of these services a routine part of patient care. Patient reminders and patient financial incentives are also very effective interventions. Education and feedback are consistently less effective and should not be the first choice for interventions.
61. A Pilot Study of Shared Decision Making for Prostate Cancer Patients of Low Socioeconomic Status
CL Bennett, Northwestern University Medical School, Chicago, IL, RA Schoor, Northwestern University Medical School, SP Kim, Northwestern University Medical School, SJ Knight, The Chicago VA Health Care System-Lakeside/Northwestern University Medical School, KM Collela, The Chicago VA Health Care System-Lakeside, AG Tennant, Northwestern University Medical School, CL Bennett, The Chicago VA Health Care System-Lakeside/Northwestern University Medical School
Objectives: Patients with newly diagnosed prostate cancer are often confronted with a myriad of information sources. Shared decision making with CD-ROM support may offer advantages as patients receive customized educational information. In this feasibility study of a CD-ROM technology, we sought to determine if, among veterans with prostate cancer, almost all of whom are of lower socioeconomic status, would this technology clarify or confuse patient understanding.
Methods: Newly diagnosed prostate cancer patients participated in this shared decision making project that used CD-ROM technology. Assessments included knowledge related to prostate cancer, health literacy, satisfaction, and preferred treatment.
Results: The 30 participants had a mean reading level of junior high school. All participants navigated through the CD-ROM material, but most required assistance from a research assistant. After viewing the CD-ROM, patient comprehension as measured by the prostate cancer knowledge questionnaire was poor (23.4 percent with less than 90% correct). Low literacy and educational attainment also represented major barriers for patient comprehension. Patients with a less than 9th grade literacy level had a significantly lower score on the post-intervention questionnaire than patients with a 9th grade or higher literacy level (59.9 percent versus 82.1 percent). Similarly, education had an effect on information retained from the CD-ROM tool. Patients with less than high school education scored lower (62.1 percent) when compared to high school graduates (74.1 percent) and patients who went onto advanced education (82.2 percent). Among patients with clinically localized disease (n=26), 61.4% chose treatment options that were not recommended by the clinical guidelines from the American Urological Association and were not discussed in the CD-ROM. While more than 70% were very satisfied with the shared decision aid, almost one third of the patients remained uncertain about selecting a treatment after completing the CD-ROM material.
Conclusions: This pilot study of a CD-ROM technology in the VA prostate cancer raises concern about the feasibility of shared decision making programs for patients of lower socioeconomic status, many of whom have low literacy and limited computer skills.
Impact: Although many veterans with prostate cancer express interest in becoming more educated about their medical condition, they may not benefit from shared decision making programs that have not been tailored to low literacy populations. Further studies that focus on developing tool that are appropriate for low literacy populations are needed before large randomized trials of this technology are initiated in populations of low socioeconomic status patients with prostate cancer.
63. The Effect of Preoperative Risk Factors on Hospital Departmental Costs of Coronary Artery Bypass Grafting Surgery
Chuan-Fen Liu, HSR&D, VA Puget Sound Health Care System, Seattle, WA, S Subramanian, Boston Scientific Corporation
Objectives: Coronary artery bypass grafting (CBAG) is one of the most common and the most costly surgical procedures in the United States. Thus, costs of CABG in particular are currently examined closely by insurance payers. This study examined the impact of preoperative risk factors on hospital departmental direct variable cost of CABG. Identification of high cost risk factors would help clinicians to better manage high risk patients.
Methods: This study used micro-cost and clinical data from three hospitals participating in Health Care Financing Administration’s Medicare Heart Bypass Demonstration during the period of 1991 to 1993. The study sample included 2,828 Medicare patients who underwent CABG in DRG 106 (with catheterization) and DRG 107 (without catheterization). Hospital patient care costs were measured by direct variable costs, because these are the costs can be changed by physician practices in patient care management. Professional fees were excluded. Five major departments were included in the analysis, which were routine nursing, nursing ICU, operation room (OR) and recovery, pharmacy, and catheter lab (DRG 1060nly). The costs were adjusted to 1993 dollars using hospital input price factors. Preoperative risk factors included patient's age, gender, race, surgical priority, patient's age, gender, race, admission priority, type of coronary heart disease, risk factors, comorbid conditions, previous CABG, and ejection fraction.
Results: The impact of pre-operative risk factors differed across major departments. The costs of nursing ICU and pharmacy were affected by more preoperative risk factors than other major departments. Congestive heart failure (CHF) and chronic obstructive pulmonary disease (COPD) significantly increased the cost of nursing ICU for all three hospitals. Older age, female sex, urgent or emergent surgical priority, previous CABG, recent myocardial infraction (MI), diabetes, chronic renal failure, and preoperative insertion of intra-aortic balloon pump (IABP) significantly increased the cost of nursing ICU in two of the three hospitals. COPD and diabetes significantly increased the pharmacy cost in all three hospitals. CHF, previous CABG, recent MI, chronic renal failure, and preoperative insertion of IABP were significantly increased the cost of pharmacy cost in two of the three hospitals. Few risk factors were cost drivers for routing nursing, OR and recovery, and catheter lab. Preoperative insertion of IABP was the largest cost driver in both routine nursing and catheter lab across all three hospitals. Older age significantly increased the cost of routine nursing in two hospitals. Previous CABG was the cost driver of OR and recovery across all three hospitals.
Conclusions: The preoperative risk factors affected direct variable costs of CABG different across major departments. This study identified preoperative cost drivers by major department that can be used to identify high cost patients and to provide better patient care management for CABG patients with these high risk factors.
Impact: The results of this study can provide useful information to VA clinicians to identify high cost CABG patients and to provide better patient care management for patients with these high risk factors.
64. Cost-Effectiveness of Conservative Management of Non Q-wave Myocardial Infarction
Paul G. Barnett, VA Health Economics Resource Center, Palo Alto, CA, S Chen, VA Health Economics Resource Center, WE Boden, Syracuse VA Medical Center, NR Every, VA Puget Sound Health Care System, PW Lavori, VA Cooperative Studies Coordinating Center, MA Hlatky, Stanford University
Objectives: The VA Non-Q-wave Infarction Strategies in Hospital (VANQWISH) trial demonstrated that a conservative, ischemia-guided strategy was safe and effective for management of non-Q-wave myocardial infarction. Health services methods were used to find total cost incurred by study subjects and to analyze trial data from an economic perspective.
Methods: We limited our study to the 876 of the 920 VANQWISH participants who enrolled at VA sites so that we could track their use of VA hospitals and outpatient clinics in utilization databases. We employed a micro-cost method to estimate the cost of initial VA hospital stays, simulating costs with a function estimated with data from the Myocardial Infarction Triage and Intervention (MITI) registry. We used average cost methods to estimate the cost of subsequent care, using a Medicare based rate for acute hospital stays, and VA derived estimates for outpatient care and non-acute inpatient stays. Data also included cost-adjusted charges of non-VA hospital stays. We employed a bootstrap method to estimate the variance of the cost-effectiveness result. We tested sensitivity of results to cost methods by using an alternative method of funding the cost of acute hospital care, employing a cost function estimated with data from the VA decision support system, and a VA based rate for subsequent stays.
Results: Subjects randomized to the invasive strategy incurred an average of $41,893 in cost, significantly more than the $39,707 cost incurred by the conservative strategy group (p=.037). The initial hospital stay cost an average of $19,256 for the invasive group, significantly more than the $14,733 mean cost of the initial stay for the conservative group (p=.0001). The mean cost of follow-up care was not significantly different ($22,626 for invasive group vs. $24,974 for conservative strategy). Conservative strategy had lower all-cause mortality (relative risk ratio of .72) but this difference was not significant (p=.058). The conservative strategy was significantly more cost-effective than the invasive one (p=.044). This result held over a wide range of critical cost-effectiveness threshold values, for a different discount rate, and for the alternate cost method.
Conclusions: A conservative, ischemia-guided strategy is more cost-effective than the routine use of invasive procedures in the management of non-Q wave infarcts.
Impact: Non-Q wave infarcts account for approximately 40% of new heart attacks. We found conservative strategy saved $2,186 per heart attack treated. This cost difference applied to 40% of 10 thousand VA stays for myocardial infarction represents $9 million in annual health care cost. If results can be generalized throughout the U.S. health care system, where 750 thousand new heart attacks are treated annually, the difference represents an annual cost of $650 million.
65. Clozapine Prescription and Long-term Inpatient Service Utilization
Michael J. Sernyak, VA Connecticut Healthcare System and Department of Psychiatry, Yale School of Medicine, West Haven, CT, R Rosenheck, Northeast Program Evaluation Center, VACT, West Haven, Yale Department of Psychiatry and School of Epidemiology and Public Health, R Desai, Northeast Program Evaluation Center, VACT, West Haven, and Yale School of Epidemiology and Public Health, G Ripper, Clozapine Coordinating Center, Department of Veterans Affairs, M Stolar, Northeast Program Evaluation Center, VACT, West Haven, and Yale School of Epidemiology and Public Health
Objectives: Clozapine has been demonstrated to be superior to typical neuroleptics in treating symptoms in refractory schizophrenia with a decreased risk of extrapyramidal symptoms. However, clozapine treatment is also expensive. It has been hoped that the greater prescribing costs of clozapine would be offset by decreased costs associated with other aspects of treatment. There have been many attempts to investigate the pharmacoeconomics of clozapine. The studies with the weakest research designs (pre/post comparisons) have demonstrated the greatest savings associated with clozapine treatment, while the most rigorously designed studies, randomized clinical trials, find only modest savings. However, the conclusions reached in randomized clinical trials suffer from their limited applicability to "real world" conditions. We report on an attempt to examine the long-term treatment costs associated with clozapine using ‘propensity scoring’ and a national database from the Department of Veterans Affairs.
Methods: Data for this study were obtained from the records of all 73,700 (45,917 individuals) psychiatric hospitalizations for schizophrenia within the VA during the fiscal years from 1992 to 1995. Within this group were 1415 patients identified by the Clozapine Coordinating Center within the VA as having initiated treatment with clozapine during a hospitalization; 2830 of the other 44,502 patients were selected as controls using propensity scoring. Propensity scoring is a recently developed and well established statistical technique where a group of potentially confounding covariates in an observational study are replaced by a single score measuring the propensity of a subject to be similar to the subjects in the treatment group. This score is then used to match treatment and control groups in an attempt to minimize the effect of selection bias.
Results: The two groups were well matched on a host of demographic, clinical, and utilization variables. Clozapine-exposed patients had significantly more inpatient days than control patients at 6 months (28 v. 19, p=0.0001), 2 years (97 v. 70, p=0.0001) and even 3 years (124 v. 91, p=0.0001). They also had more outpatient appointments (26 v. 13, p=0.0001). Clozapine exposed patients were also divided into those who had received clozapine for less than a year (drop-outs) and those who were maintained on clozapine (completers). Clozapine "drop-outs" had significantly more inpatient days than clozapine "completers" and controls at 6 months (38 v. 24 v. 19, p=0.0001), 2 years (143 v. 79 v. 69, p=0.0001) and 3 years (187 v. 100 v. 91. p=0.0001).
Conclusions: The utilization of propensity scoring allowed for construction of a well-matched control group to compare with a group of veterans with schizophrenia receiving clozapine. Clozapine exposure was associated with increased utilization of treatment resources. The highest utilizers appear to be those patients receiving clozapine for less than one year.
Impact: While clozapine treatment results in real symptomatic improvement in many veterans with refractory serious mental illness, these data do not support previous claims that this pharmacotherapy results in overall treatment cost savings in actual practice.
66. Validity of Standard Gamble, Time Trade-Off, and Rating Scale Methods of Utility Assessment for Prostate Cancer Health States
K Wristers, Houston Center for Quality of Care and Utilization Studies: A VA HSR&D Center of Excellence; Department of Medicine, Baylor College of Medicine, J Souchek, Veterans Affairs Medical Center-Houston, Baylor College of Medicine, B Brody, Veterans Affairs Medical Center-Houston, Baylor College of Medicine, T Menke, Veterans Affairs Medical Center-Houston, Baylor College of Medicine, C Ashton, Veterans Affairs Medical Center-Houston, Baylor College of Medicine, N Wray, Veterans Affairs Medical Center-Houston, Baylor College of Medicine
Objectives: The value of utility elicitation has been questioned because the methods are cognitively intense. In this study we assessed the use of utility scores to guide policy or individual patient decision making in patients with advanced prostate cancer. The criterion of participant's rankings of health states was compared to utility values obtained for those states.
Methods: 120 patients diagnosed with advanced prostate cancer who accepted either orchiectomy or LHRH-analogue therapy had their utility values for eight hypothetical states measured by two of the three methods we investigated (standard gamble, time trade-off, and rating scale). To evaluate the validity of the utility values we used the rank order the patient gave to the health states as a gold standard and compared the utility values to this gold standard. Inconsistency was defined as the proportion of utility responses inconsistent with the ranks given to the health states. Differentiation was defined as the proportion of utility scores, which were different, provided the rankings were not tied. We had five pairs of health states that varied only in that one member of the pair had one or more symptoms than the other member. Three of these pairs differed by the presence or absence of bone pain, while two varied by the presence or absence of other symptoms such as hot flashes, and loss of muscle tone and body hair. We assessed the extent to which utility values were logical by calculating the proportion of cases where states with fewer symptoms were given lower utility than states with more symptoms.
Results: Inconsistency was 0.20, 0.26, and 0.34 for the Rating Scale, Standard Gamble, and Time Trade-off techniques, respectively. Differentiation was 0.71, 0.43, and 0.41 for the Rating Scale, Standard Gamble, and Time Trade-off, respectively. For logical choice test, 34.6% of the choices in the "bone pain" pairs were illogical in that the utility value for the state with bone pain was equal or higher than the state without bone pain while, 74.7% of our other group pairs were illogical.
Conclusions: In summary, our data bring to question the use of utilities as assessed by these methods to guide decision making in that the results of the utilities frequently do not discriminate as well as rankings of the health states. The utility values are also frequently illogical.
Impact: Use of utility measures for cost effectiveness analysis and other policy decisions have been widely recommended. Our data raise serious concerns about the validity of these methods.
67. A Novel Approach to the Analysis of Cost-Effectiveness Data in Psychiatric Clinical Trials: Cost-Effectiveness Acceptability Curves
Robert A. Rosenheck, VA Connecticut Healthcare System; Departments of Psychiatry and Public Health, Yale School of Medicine, West Haven, CT, W Xu, Cooperative Studies Program Coordinating Center, J Cramer, VA Connecticut Healthcare System; Department of Psychiatry, Yale School of Medicine, J Thomas, Cooperative Studies Program Coordinating Center, W Henderson, Cooperative Studies Program Coordinating Center, D Charney, VA Connecticut Healthcare System; Department of Psychiatry, Yale School of Medicine
Objectives: The statistical analysis of cost-effectiveness (CE) data from clinical trials is often complicated by low power due to relatively large standard deviations and small effect sizes. Construction of "cost-effectiveness acceptability curves" (CEACs) offers a new approach to statistical analysis of CE data that may make CE trials more informative. Although this approach was presented in the economic literature by Van Hout in 1994, it has received little use.
Methods: We used this technique to re-analyze data from a randomized trial (n=423) which compared the atypical antipsychotic medication clozapine with haloperidol in refractory schizophrenia. This study found clozapine to be more effective than haloperidol but also more expensive. Since it was associated with reduced hospital use, related savings lowered the total cost of clozapine treatment. Re-analyses using CEACs were conducted on the entire sample and on subsamples with high, middle and low levels of inpatient use and costs. The primary outcome was Quality Adjusted Life years as determined by methods described elsewhere. In generating CEACs, a bootstrap technique was used to estimate incremental cost and effectiveness of clozapine as compared to haloperidol through 3,000 replications. Next, these data were used to generate cost-effectiveness acceptability curves (CEACs) that plot the probability of various cost-effectiveness ratios, i.e. the probability that the marginal cost of achieving one unit QALY with clozapine over haloperidol will have in each of a range of non-negative values. This approach allows determination of the probability of a range of levels of cost/effectiveness, not just a single overall CE ratio.
Results: Using the standard statistical technique of O'Brien et al. (1994) clozapine had lower costs than haloperidol and greater benefits but with great uncertainty (CE ratio = -$55,556/QALY; 95%CI =-$478,688/QALY to $367,576/QALY). More precise information from CEACs shows that clozapine has a 69% probability of being cost neutral (i.e. have zero net cost), a 76% chance of costing less than $50,000/QALY (a conventional estimate of the monetized value of one QALY) and a 95% chance of costing less than $240,000/QALY. Examination of subgroups revealed that among high users clozapine has an 85% chance of cost neutrality and a 95% chance of costing less than $150,000/QALY. Among middle-level users, in contrast, clozapine has a 55% chance of cost neutrality and a 95% chance of costing less than $426,000/QALY, because there is less room for savings in patients who use fewer services. Among low users, clozapine has only an 8% change of cost neutrality, and a 95% chance of costing less than $690,000/QALY.
Conclusions: Analysis of CE studies using CEACs substantially enhances the value of resultant data because it allows one to determine the likelihood of achieving a specified improvement at a range of costs. This kind of data allows program managers to take informed risks, knowing the probabilities of various CE outcomes in using new treatments. Substantial differences in cost-effectiveness were observed across utilization subgroups.
Impact: These findings may improve the value of clozapine treatment by optimizing patient targeting.
HSR&D Funded: CSHS #17
68. Donepezil Use and Impact on Utilization and Cost among Patients with Alzheimer's Disease
William A. West, Ph.D., VA Center for Healthcare Quality, Outcomes and Economic Research, Bedford , MA, M Prashker, VA Center for Healthcare Quality, Outcomes and Economic Research, L Merriman, VA Center for Healthcare Qualitiy, Outcomes and Economic Research, D Miller, VA Center for Healthcare Quality, Outcomes and Economic Research, J Anderson, VA Center for Healthcare Quality, Outcomes and Economic Research, L Curran, VA Center for Healthcare Quality, Outcomes and Economic Research
Objectives: In 1996 the FDA approved donepezil for treatment of Alzheimer’s patients with early symptoms of memory and cognitive loss. A recent study based on clinical trial data projected that the cost of donepezil could be offset by reduced utilization over a two year period. The purpose of this study was to measure the actual use of donepezil among Veterans’ Health Administration (VHA) patients in FY98, and its impact on utilization and cost of health care services.
Methods: Patients diagnosed with Alzheimer's disease who received treatment in four VHA medical centers located across the country were identified using VHA treatment files. Costs for inpatient and outpatient care were estimated using Medicare’s reimbursement rates and data (DRG weights, Relative Value Units, and Medicare fee schedules). The cost of extended care was calculated at $236 per day (VHA national per diem). Filled prescription records were obtained from the VISTA system at each site and prescription costs were calculated using VHA Federal Supply Schedule and National Contract prices. Costs and days of care (acute inpatient plus extended care) were measured starting with first donepezil use until the end of the study period and annualized. Patients with medical contraindications, those who died during the study period, and those admitted to an extended care facility prior to the beginning of the fiscal year, were excluded from the analysis. The number of days from first recorded dementia diagnosis was used as a proxy for disease stage or severity. Comorbidity was controlled for using diagnostic data to calculate a Charlson comorbidity index. Other control variables included number of hospital stays in FY97 and dummy variables for each site. Regression analysis was used to measure the association between donepezil use and health care costs while controlling for severity, comorbidity, previous hospitalizations, and site.
Results: Donepezil was used by 12% (141) of patients with no contraindications (N=1150). Donepezil users had lower average comorbidity scores (3.27 vs. 4.09), shorter duration in days from first dementia diagnosis (306 vs. 315), and fewer prior acute care admissions (.18 vs. .40). Of the patients who used donepezil, 78% used it for at least 60 days during the study period. The regression models indicated donepezil use was associated with higher pharmaceutical costs (+$1496, p=.0001) but lower inpatient (acute plus extended care) costs (-$1386, p=.07) and days of care (-7.5, p=.01). The was no substantial difference in outpatient (+$160, p=63) or total cost (+$271, p=.75).
Conclusions: The results of this study support the hypothesis that, for donepezil users, higher pharmaceutical costs are offset by lower inpatient costs. The extent of the offset may be greater for the VHA (due to substantial discounts with drug manufacturers), than would be found in other health care systems.
Impact: Since the VHA is concerned with the total cost of treating patients, it is important to recognize that the high cost of some medications used in treating chronic disease may be offset by cost savings in other parts of the health care system. This result may have an impact on how formulary decisions are made.
69. Diabetes Costs and Control in Several VISN 4 Facilities: There is a difference
Chester B. Good, MD, MPH, VA Pittsburgh Healthcare System, Pittsburgh, PA, SL Fultz, VA Pittsburgh Healthcare System, ME Kelley, VA Pittsburgh Healthcare System, MJ Fine, VA Pittsburgh Healthcare System
Objectives: Diabetes mellitus (DM) is a prevalent condition within VA. High pharmacy costs for DM treatment must be balanced by evidence that improved glycemic control leads to beneficial health outcomes. We sought to evaluate differences in cost and glycosylated hemoglobin (HbAc) among different provider types and among different sites in VISN 4.
Methods: Two academic sites (sites A and B), and two community sites (sites C and D) were chosen for analysis, identifying all patients from each site receiving DM medications or supplies for calendar year 1997. Primary care providers (PCP) were defined if they wrote more than 50% of all DM prescriptions for an individual patient. Staff types were: staff attendings (AT), medicine house officers (HO), nurse practitioners or physician’s assistants (NP), endocrinologists (EN), or no PCP. Analysis of variance was performed to evaluate cost and glycemic control by site and providers, controlling for age, race, and marital status (all of which were found to be significant predictors of cost and HbAc). Variance component analysis was performed using site and provider within site, as independent factors to describe the amount of variance in glycemic control and cost attributable to these factors.
Results: 4458 patients were identified (site A=1957, site B=1136, site C=611, site D=754) with adequate information (demographic data, cared for on site) for cost analysis; 3612 (81%) of these patients had laboratory data for analysis of HbAc. For total pharmacy cost (DM medications and supplies) there was a significant site effect (p<.0005), indicating site D costs ($26.86 per patient per month) were > site A ($15.85), which was > sites B ($9.43) and C ($8.44). Likewise, there was a significant staff type effect (p<.0005), indicating that EN ($25.68) had greater costs than all others (HO= $16.56, no PCP= $14.98, NP= $13.39, AT= $12.63). Use of diabetic supplies varied among sites, from 73.5% of patients (site A), 71.6% (site D), 37.1% (site B), to 26.3% (site C). For total cost, 7.3% of variance was attributable to site, 3.5% to provider, and 89.2% to factors associated with individual patient characteristics. For supply costs, there was greater variance attributable to site (38.3%) and provider (7.1%), with 54.6% attributable to patient characteristics. For glycemic control, there was a significant site effect (p< .0005) for site A with greater HbAc (8.27) than the other sites (site B= 7.41, site C= 7.40, site D= 7.22). There were no significant differences in HbAc for staff type (EN= 7.34, AT= 7.54, no PCP= 7.67, HO= 7.98, NP= 8.07). For overall HbAc, 5.6% of variance was attributable to site, 1.8% to provider variance, and 92.7% to individual patient characteristics.
Conclusions: Two sites (B and C) with lowest costs also had better glycemic control than site A. Although patient factors accounted for most of the difference in medication costs across sites, there was substantial variation in the use of diabetic supplies.
Impact: Since supplies account for a substantial part of overall cost for DM care, decreasing use of glucose test strips may result in savings without adversely effecting glycemic control.
70. Risk of Hospitalization and Mortality In-Patients with Chronic Obstructive Lung Disease: The Impact of Quality of Life
Vincent S. Fan, MD, VA Puget Sound Health Care System, HSR&D, Seattle, WA, J. Randall Curtis, MD MPH, Division of Pulmonary Medicine, Univ. of Washington, Shin-Ping Tu, MD MPH, Division of Internal Medicine, Univ. of Washington, Mary B. McDonell, MS, VA Puget Sound Health Care System, Stephan D. Fihn, MD MPH, VA Puget Sound Health Care System
Objectives: The majority of costs of caring for patients with chronic obstructive pulmonary disease (COPD) are spent on hospitalizations. If disease-specific quality of life (QOL) instruments are accurate predictors of an increased risk of hospitalization or death in these patients, these instruments could be used to target patients for preventive measures.
Methods: We conducted a prospective cohort study of outpatients in general internal medicine clinics at seven Veterans Affairs hospitals who were enrolled in the Ambulatory Care Quality Improvement Project (ACQUIP). Of 31,175 outpatients who completed a health checklist questionnaire, 3418 reported having COPD, and were followed for 12 months. Baseline QOL was measured with the Seattle Obstructive Lung Disease Questionnaire (SOLDQ), which is comprised of four scales: emotional function, physical function, coping skills, and satisfaction. Outcomes studied were any hospitalization, hospitalization for a COPD exacerbation, and all-cause mortality. We performed logistic regression to calculate the odds ratios (OR) and confidence intervals (CI) for the outcomes comparing patients with scores in the lowest quartile to the highest quartile.
Results: During the follow-up period 615 (18%) patients were hospitalized, 145 (4.2%) for COPD exacerbations, and 170 (5%) patients died. In a multivariate model including comorbid diseases, chronic steroid use and prior hospitalization, all three scales remained statistically significant risk factors for the outcomes. The risk of any hospitalization for patients with emotional, physical, and coping skills scores in the lowest quartile compared to the highest was 1.7 (95% CI, 1.2-2.3), 2.2 (95% CI, 1.6-2.9), and 1.8 (95% CI, 1.4-2.3). When only COPD-related hospitalizations were analyzed, the risk of hospitalization for the emotional, physical, and coping scales scores was 2.4 (95% CI, 1.3-4.6), 4.8 (95% CI, 5.5-9.1), and 2.8 (95% CI, 1.6-4.9). Similarly, patients with scores in the lowest quartile of the emotional, physical and coping skills scales also had an increased risk of death [OR 4.0 (95% CI, 2.1-7.7), 5.9 (95% CI, 2.9-12.2) and 3.5 (95% CI, 2.0-6.0)].
Conclusions: In conclusion, lower quality of life is a significant risk factor for hospitalization and all-cause mortality.
Impact: Lower quality of life is a significant risk factor for both hospitalization and mortality in patients with COPD. Disease-specific quality of life instruments may provide an opportunity to identify patients who would benefit from preventive measures.
HSR&D Funded: SDR96-002
71. Functional Status in Veterans after Discharge for Acute Myocardial Infarction versus Unstable Angina Pectoris
Pete Pitaya Fong, VA Puget Sound Healthcare System, Seattle Center of Excellence, Seattle, WA, LA Parsons, VA Puget Sound Healthcare System, Seattle Center of Excellence, AE Sales, VA Puget Sound Healthcare System, Seattle Center of Excellence, JA Spertus, Mid-America Heart Institute, DM Caldwell, VA Puget Sound Healthcare System, Seattle Center of Excellence, NR Every, VA Puget Sound Healthcare System, Seattle Center of Excellence
Objectives: Unstable angina pectoris (UAP) is a common and costly condition. Although functional status in patients diagnosed with acute myocardial infarction (AMI) has been studied and shown to be similar in some comparisons with age matched controls, little is known about functional status in patients with UAP. Therefore, the objective of this study is to compare the functional status outcomes in patients after hospital discharge for AMI versus discharge for UAP and to relate these findings to treatment strategies.
Methods: This study used data from the VA ACCESS to Cardiology project. VA ACCESS is a cross-sectional observation study of all patients discharged from March 1998 to February 1999 with ICD-9-CM codes of 410 (AMI) or 411 (UAP) from VA medical centers in six Veterans Integrated Service Networks in the western and central U.S. Functional status data using the Seattle Angina Questionnaire and SF-36 surveys were then collected seven months post-index hospital discharge to determine cardiac-specific and generic health status. Analyses were performed on the subscales of the SAQ and SF-36 comparing scores in patients diagnosed with AMI versus UAP and subsets of these groups.
Results: Of 2968 patients discharged after AMI or UAP, 2612 patients were alive at seven months to be surveyed. We have collected data on 1805 participants with 984 having been diagnosed with AMI and 821 diagnosed with UAP. UAP patients scored significantly worse in all of the SAQ subscales when compared with AMI patients. The largest differences are observed in the anginal frequency score [mean difference 10.2 points, P<0.0001 (10 point difference considered clinically significant)] and the anginal stability score (mean difference 7.4 points, P<0.0001). There was a statistically significant impairment in the physical function, bodily pain, general health, vitality, and social function scores of the SF-36 in UAP patients. Subset analysis of incident AMI and UAP cases showed impairment in all of the UAP subscale scores in both the SAQ and SF-36 except for the disease perception / quality of life score (mean difference 4.8, P=0.092). Notably, in the subset of patients who were revascularized with percutaneous transluminal coronary angioplasty (PTCA) or coronary artery by-pass grafting surgery (CABG), UAP patients scored significantly lower only in the anginal frequency domain of the SAQ (mean difference 6.4, P=0.034) and in several of the SF-36 subscales. Fewer UAP patients were revascularized by PTCA (22.0% AMI v. 10.8% UAP, P<.0001). There was a substantial difference in mortality at seven months (19.3% AMI v. 6.3% UAP, P<0.0001).
Conclusions: Patients discharged with a diagnosis of UAP have a poorer functional status than patients discharged with a diagnosis of AMI. These findings may be related to a difference in revascularization frequencies and mortality between the two groups.Impact: The moderate impairment in functional status observed between UAP patients versus AMI patients may be explained by lower rates of revascularization performed in UAP patients overall. These results may argue for a more aggressive strategy in the management of patients discharged with a diagnosis of unstable angina pectoris to help improve quality of life.
HSR&D Funded: 97-079
72. Quality of Life and 5-Year Mortality among Veterans with Coronary Artery Disease: Results from the Ambulatory Care Quality Improvement Project (ACQUIP) Pilot Study
Mary B. McDonell, VA Puget Sound Health Care System, Seattle, WA, JA Spertus, University of Missouri, Kansas City, SD Fihn, VA Puget Sound Health Care System and University of Washington
Objectives: Measures of patient reported quality of life (QOL) are potentially useful as predictors of outcomes. Few studies, however, have examined the relationship between disease-specific health status and long-term mortality. This study examined the relationship between QOL and all-cause 5- year mortality among VA outpatients with coronary artery disease (CAD).
Methods: Three VA General Internal Medicine Clinics (Puget Sound Health Care System - Seattle, Boston, and White River Junction, VT) participated in the study as pilot sites for the Ambulatory Care Quality Improvement Project (ACQUIP). This prospective cohort study examined 1679 patients with CAD who responded to QOL surveys mailed between November 1993 and April 1994. Quality of life was measured using 4 scales of the Seattle Angina Questionnaire (SAQ): physical function, disease perception, angina frequency and angina stability. All scales were scored 0 (worst) to 100 (best). Co-morbid conditions and demographic information were obtained from a health history completed by patients at entry into ACQUIP. Death information was obtained from local VISTA systems and by linking to the VA Beneficiary Identification and Record Locator System (BIRLS). We included mortality from all causes. Data were examined using survival analysis and Cox Proportional Hazards regression.
Results: The cohort was 98% male and 92% Caucasian with a mean age of 67. Overall, 469 (28%) of the 1679 died during the five year follow-up period. The cumulative proportion surviving at five years was 0.69. Patients were stratified into 2 groups for all scales: those scoring =< 50 and those scoring > 50. The cumulative proportion surviving at 5 years was 0.63 for patients scoring =< 50 on the physical scale compared to 0.79 for those scoring > 50 (p=.000). For the frequency scale, the cumulative proportion surviving was 0.63 for patients scoring =< 50 and .70 for those scoring > 50 (p=.005). These differences remained significant when stratified by age and co-morbid conditions. Differences in survival were not significant for the disease perception and stability scales. In a multivariate model including physical function, age, and co-morbid conditions, the relative risk of death from all causes was 2.0 (95% CI=1.62-2.46) for patients scoring =< 50 compared to those scoring > 50 (p=.000). In a model including angina frequency, age, and co-morbid conditions, the relative risk of death was 1.4 (95% CI=1.13-1.76) for those scoring =< 50 (p=.003).
Conclusions: Lower quality of life, as measured by the SAQ physical function and frequency scales, appears to be a significant predictor all-cause mortality in veterans with coronary artery disease. QOL as measured by the disease perception and symptom stability scales did not significantly relate to risk of death.
Impact: If disease specific QOL instruments such as the SAQ can predict mortality, they can potentially be used to identify patients at risk for adverse outcomes. If identified, such patients could benefit from targeted interventions and preventive measures.
HSR&D Funded: SDR 96-002 and CSHS 91-007
73. Determinants of Alcohol Advice or Treatment among At-Risk Drinkers in the Outpatient Setting. Results from the ACQUIP Study
Marcia. L. Burman, MD, Ambulatory Care Fellow. Northwest Center of Excellence. VA Puget Sound Health Care System, Seattle, WA, MB McDonell, MS, Northwest Center of Excellence. VA Puget Sound Health Care System, SD Fihn, MD, MPH, Director. Northwest Center of Excellence. VA Puget Sound Health Care System, KA Bradley, MD, MPH, Northwest Center of Excellence. VA Puget Sound Health Care System
Objectives: Randomized controlled trials have demonstrated that brief counseling interventions by primary care providers improve health outcomes in at-risk drinkers. However, the majority of at-risk drinkers never receive such advice and little is known about factors that lead primary care providers to counsel patients about their drinking. We examine rates of alcohol-related advice and treatment reported by male at-risk drinkers who receive primary care in the VA, and identify factors associated with at-risk drinkers’ reports of not receiving any alcohol-related advice or treatment in the past year.
Methods: We surveyed patients followed at seven geographically diverse VA General Internal Medicine Clinics participating in the VA Ambulatory Care Quality Improvement Project (ACQUIP). At-risk drinkers, identified using a validated augmented CAGE questionnaire, were sent the ACQUIP Drinking Practices Questionnaire (DPQ) which includes the Alcohol Use Disorders Identification Test (AUDIT) and three questions about alcohol-related advice or treatment. Other data collected included demographic characteristics, socioeconomic status, patient reports of health problems, questions about health-related quality of life (SF-36), and patient satisfaction. Site investigators and coordinators indicated whether each clinic had instituted a standard alcohol-screening program.
Results: Of 6,788 eligible at-risk drinkers, 3,891 returned the DPQ (57% response). DPQ respondents were primarily male (98%) and Caucasian (62%), with a mean age of 61. The mean AUDIT score was 7.2 with 819 (21%) reporting at least one symptom of dependence in the past year. Among DPQ respondents, 700 (18%) indicated that in the past year they had been advised by their primary care provider to drink less, 661 (17%) had been advised to quit, and 156 (4%) had received alcohol treatment. Excluding those treated in the past year, a total of 784 (21%) received advice to change their drinking. Advice was more common among patients who were younger, reported less education, were unmarried, had lower income, reported liver disease, or smoked. Among at-risk drinkers who had not received alcohol treatment in the past year, multivariate logistic regression revealed that the following groups were significantly more likely to report not receiving any alcohol-related advice in the past year: patients who drank <14 drinks/week (OR 1.99), those without symptoms of alcohol dependence (OR 3.33), those who did not self-report a drinking problem (OR 4.22), those who did not smoke (OR 1.32), and those who reported their health as very good or excellent (OR 1.62). Implementation of clinic-based screening with the CAGE questionnaire was not significantly.
Conclusions: Primary care providers are providing alcohol counseling to less than 25% of at-risk drinkers. Our findings also suggest that primary care providers are more likely to counsel patients with more severe problem drinking. Clinic-based screening alone was not associated with increased counseling of at-risk drinkers.
Impact: Primary care providers are providing alcohol counseling to less than 25% of at-risk drinkers. Our findings also suggest that primary care providers are more likely to counsel patients with more severe problem drinking. Clinic-based screening alone was not associated with increased counseling of at-risk drinkers.
HSR&D Funded: SDR 96-002
74. The Impact of Chronic Liver Disease on Health-related Quality of Life (HRQOL)
Ian M. Gralnek, MD, MSHS, Greater Los Angeles VA Healthcare System, Division of Gastroenterology, Los Angeles, CA, RD Hays, UCLA Division of General Internal Medicine and Health Services Research, A Kilbourne, VA Pittsburgh Healthcare System, HR Rosen, Portland VA Medical Center, EB Keeffe, Stanford University Medical Center, P Martin, UCLA Division of Digestive Diseases and UCLA-Dumont Transplant Center
Objectives: Chronic liver disease (CLD) is a significant health problem within VA. Moreover, projections of the potential future burden of CLD within VA are concerning and we will be caring for an increasing number of such effected veterans. Presently, physician rating of CLD severity is almost exclusively based upon a derived score utilizing traditional physiologic endpoints, Child-Pugh Score (CPS). In contrast, patient self-reported HRQOL data provides information on functional status and well-being that transcends CPS and may allow for further understanding of disease impact. However, there are few data evaluating the HRQOL of persons with CLD and no data comparing CLD with other chronic diseases. The purpose of this multicenter study is to compare the HRQOL of patients with CLD to previously reported HRQOL data for the U.S. general population and selected chronic diseases.
Methods: We compared the HRQOL of n=221 consecutive ambulatory adults referred for primary orthotopic liver transplantation (OLT) evaluation with SF-36 data for the U.S. general population (n=2474), GERD (n=516), diabetes mellitus (DM, n=541), and dialysis dependent, end-stage renal disease (ESRD, n=165) pts. The SF-36 is a reliable and valid self-report measure of HRQOL which consists of 8 scales evaluating physical functioning (PF), role limitations-physical (RP), bodily pain (BP), general health (GH), emotional well-being (EWB), role limitations-emotional (RE), energy/fatigue (EF), and social functioning (SF). SF-36 scores range from 0-100 with higher scores indicating better HRQOL.
Results: CLD cohort demographics: 64.1% male; median age=51 years; 68.9% white; Child-Pugh Class A=29% (mild disease), B=52% (moderate disease), C=19% (severe disease). CLD patients had significantly lower mean SF-36 scores on each of the 8 scales (worse HRQOL) compared with age and gender adjusted U.S. general population data, p < 0.001. Similarly, as compared with both moderate to severe symptomatic GERD patients and DM patients, we found that the CLD patients reported significantly worse HRQOL on all scales of the SF-36, p < 0.001. Compared to dialysis-dependent ESRD patients, CLD patients scored significantly worse on 6 of the 8 SF-36 scales, except physical functioning and role limitations due to physical health.
Conclusions: Patients with CLD referred for OLT evaluation have significantly worse HRQOL compared with the U.S. general population and selected chronic diseases. This is evident across both physical and emotional scales, particularly energy/fatigue, overall perceived general health, and role limitations due to physical health. These HRQOL data offer insights into the significant debilitating effects and impact of chronic liver disease on patient functional status and sense of well-being. Furthermore, these data provide patient-centered information not captured utilizing CPS alone and suggest that additional criteria apart from CPS should be included in assessing the impact of CLD on patients' lives.
Impact: As more and more veterans with chronic liver disease are being cared for within VA: (1) Patient centered data such as HRQOL can allow for a more comprehensive evaluation of severity of symptoms and disease impact beyond what traditional physiologic endpoints (Child-Pugh score/class) presently offer. (2) HRQOL data can help improve the evaluation, management and resource utilization in this burgeoning patient group.
75. Prevalence of Diabetic Nephropathy in the Veterans Administration Health Care System
Bessie Ann Young, HSR&D, VA Puget Sound Health Care, Seattle, WA, JA Pugh, Department of Medicine, University of Texas Health Science Center at San Antonio, San Antonio, TX, EJ Boyko, HSR&D, VA Puget Sound Health Care, Seattle, WA, GE Reiber, HSR&D, VA Puget Sound Health Care, Seattle, WA
Objectives: Diabetic nephropathy (DN) is the leading cause of endstage renal disease in the US, and contributes significantly to the cost, morbidity and mortality associated with diabetes. We describe the frequency and sequelae of diabetic renal disease in the Department of Veteran Affairs.
Methods: Veterans with a diagnosis of diabetes and diabetic nephropathy, endstage renal disease (ESRD) and other kidney diseases were identified from 1997 to 1998 using the national VA databases. Outpatient Care files (OPC) and patient Treatment Files (PTF) were utilized to identify veterans with diabetes and renal disease based on ICD-9 and CPT codes for diabetes, diabetic nephropathy, ESRD and other kidney diseases. Veteran characteristics, disease prevalence and one year mortality data were analyzed.
Results: A total of 415,910 persons were identified with diabetes. Of those 44,671 (10.7%) were found to have diabetes and a concomitant diagnosis of renal disease. The majority of veterans were male (98.4%), elderly (66.6 +/- 10.5 yr.) and Caucasian (C) (59.7%). Diabetic nephropathy (DN) was the renal diagnosis in 56.6% of veterans with diabetes and renal disease; 39.5% of those veterans had a diagnosis of ESRD. An age-standardized prevalence of 72.6/ 1000 was found for all veterans with a diagnosis of diabetes and any renal disorder. More than half (58.5/1000) were veterans with DN. African American (AA) veterans were 1.5 times more likely to have DN, and 1.7 times more likely to have a diagnosis of ESRD compared to Caucasian veterans. Male veterans had 1.8-fold higher rate of DN and 2.1-fold higher rate of DN and ESRD when compared to women veterans. The overall age-adjusted 1-year mortality was high (10.7%), but not significantly different between Caucasians, AA or women veterans. When compared to diabetic veterans without a diagnosis of renal disease (DM), diabetic veterans with renal disease (DMR) were significantly older, and had more comorbid conditions such as cardiovascular disease (59.1% vs. 38.6%, p<0.0001), stroke (12.6% vs. 6.8%, p<.0001), hypertension (81.8% vs. 65.6%, p<.0001) and cancer (18.3% vs. 10.7%, p<.0001, DMR vs.DM respectively). Hospitalizations over 1 year were significantly higher (43.5% vs. 15.4%, p<0001, DMR vs. DM respectively), while survival at one year was significantly lower in those veteran with a diagnosis of renal disease compared to diabetic veterans without a diagnosis of renal disease (89.3% vs. 97.1%, p<0.0001, DMR vs. DM respectively).
Conclusions: Diabetic nephropathy is a major problem in the veterans with diabetes. There appears to be significant elevation of comorbid diagnoses associated with having any renal disease, as well as an associated lower one-year survival. Inference regarding causation from these differences are limited by the nature of the administrative databases, but are significant and deserve further evaluation and validation.
Impact: Diabetic nephropathy (DN)is a major problem in the veterans with diabetes. African American (AA) veterans were 1.5 times more likely to have DN, and 1.7 times more likely to have a diagnosis of ESRD compared to Caucasian veterans. There appears to be significant elevation of comorbid diagnoses associated with having any renal disease, as well as an associated lower one year survival.
76. Primary Care Providers and Their Patients Who Drink Alcohol: Knowledge, Attitudes and Practice
Joseph Conigliaro, VA Pittsburgh Healthcare System/University of Pittsburgh, Pittsburgh, PA, M Kelley, University of Pittsburgh, N Brucker, University of Pittsburgh, N Day, University of Pittsburgh, W Kapoor, University of Pittsburgh
Objectives: Brief interventions (BI) can persuade hazardous drinkers to reduce intake to low risk levels and to use less healthcare resources. Use of BIs, however, is not part of routine practice. The aim of this study was to examine the knowledge, attitudes and practice of primary care providers with respect to patient’s alcohol problems.
Methods: National cross sectional survey of primary care providers using a commercial medical marketing service. Providers included physicians (internists or family practitioners), nurse practitioners and physician assistants who described themselves as providing adult primary care. Main outcome measures were knowledge; attitudes regarding diagnosis and treatment; self-reported use of screening tools and interventions for alcohol problems.
Results: Sixty three percent of 2902 providers (1830) returned the survey. Only 16% of the respondents received training in BIs. Eighty nine percent strongly agreed or agreed that advising patients about safe levels of consumption can decrease alcohol-related morbidity. While 96% strongly disagreed or disagreed that advising patients about safe levels of consumption was not their responsibility only 77% felt physician counseling was effective in changing behavior. Only 36% of providers screened 81-100 % of their practice and 20% screened 0-20%. When providers did screen 89% asked about use and 81% asked quantity and frequency questions most or all of the time. Only 11% and 1% of providers used the CAGE or AUDIT most or all of the time and 53% and 90% used them none of the time. Providers answered that a safe and hazardous level of consumption was 6.8 drinks/wk and 14.2 drinks/wk for a healthy 45 year-old man and 5.8 drinks/wk and 12.4 drinks/wk for a healthy, non-pregnant 35-year-old woman. When respondents were asked what recommendation they would offer hazardous drinkers, 44% recommended treatment by an addictions specialist, 22% recommended abstinence and 32% recommended moderation. On multivariate analysis, attitudes were the strongest predictors of screening followed by knowledge and training. Attitudes and training also predicted an intervention.
Conclusions: Few providers have received training in BIs. Most agreed that addressing alcohol problems is important but less felt that it would change behavior. Few providers screened a majority of patients and some screened very few. They rarely used the CAGE or AUDIT. Providers correctly knew what safe and hazardous levels of consumption were but recommended treatment more in line with alcohol abuse or dependence. Specific training and knowledge predicted increased screening practice and intervention. However, attitudes also play a role in screening and intervention practices.
Impact: Interventions aimed at improving provider attitudes as well as increasing competence through training may predict an increase in primary care screening and intervention of alcohol problems.
77. Comparing VA to the Group Health Population on the Chronic Disease Score Categories
Kevin L. Sloan, M.D., VA Puget Sound Heath Care System, Seattle, WA, AE Sales, VA Puget Sound Heath Care System, PA Fishman, Center for Health Studies, Group Health Cooperative of Puget Sound, JD Malkin, Covance Health Economics and Outcomes Services Inc
Objectives: The Chronic Disease Score (CDS) is a pharmacy based risk adjuster developed by Group Health Cooperative of Puget Sound (GHC) that has been used in several organizations to predict costs and utilization. Compared to other available instruments, the CDS appears to have similar explanatory power but does not depend on diagnostic data and is available in a non-proprietary, open-source form that can be tailored to the specific needs of organizations. The authors sought to translate the CDS to the VHA Northwest Network (VISN 20) formularies then, as a validity check, to directly compare the VISN 20 patient population with that served by GHC.
Methods: We compared the CY 1997 pharmacy records for 226,675 GHC enrollees with those for 179,109 individuals seen in VISN 20 during CY 1996-8 (using the three-year administrative definition of a VHA user to more closely mimic the GHC definition of "enrollment"). We mapped each drug in the GHC CDS to its equivalent in the VA formulary and constructed the 29 CDS categories. We then compared absolute and relative proportions of individuals in each category as well as the age/sex categories frequently used for demographic risk adjustment. We considered absolute differences of 5% or proportional differences of a factor of 2 between GHC and VA patients in a category to be significant.
Results: Although only 45% of the 3-year VA patients were classified by the CDS compared to 56% of GHC patients, 38% of VA patients classified received >=4 CDS categories compared to only 21% of GHC patients. Additionally, if one examines only those 126,134 individuals seen at the VA during CY 1997, fully 64% received at least one CDS category. Examining absolute proportions, compared to GHC there were 8.7% more VA Gastric Acid Disorder category patients but 5.0% fewer VA Pain and 5.3% fewer VA Pain/Inflammation category patients. Examining relative proportions, the VA population had 2-3.3 times as many Bipolar Disorder, Epilepsy, Gastric Acid Disorder and Liver Failure patients, while having only 0.4 times the number of individuals with Thyroid Disorder. Excepting only Parkinson's Disease (which lacked sufficient numbers), classification rates of all other categories were similar across the two systems.
Conclusions: This direct comparison between two populations supports previous findings that the VA population is overall sicker than many other populations. However, comparing between the two groups that fall into one of the CDS categories, most numbers are similar, and the distribution among categories is not strikingly different. This comparison demonstrates that VA users can be categorized into CDS categories in a manner similar to the private-sector GHC population.
Impact: The CDS offers a promising methodology for risk adjustment to the VHA system. The high proportion of same-year VA users who can be classified into at least one category of the CDS, even without tailoring the CDS to the VA, suggests that this method might be much more useful in the VA than it has been in the private sector.
78. Medicare Enrollment Status among VA Users by VISN
Steven M. Wright, PhD, VA Boston Healthcare System, West Roxbury, MA, RP Lamkin, VA Boston Healthcare System, LA Petersen, Houston Center for Quality of Care and Utilization Studies
Objectives: There is wide variation in the proportion of VA users receiving VA healthcare services in the 22 Veteran Integrated Service Networks (VISN). As this population grows older, more VA users will be eligible for Medicare services. The extent to which Medicare enrollment rates vary by VISN has not been examined. The aims of this study were to determine the proportion of VA users who were enrolled in the Medicare program, assess the type of Medicare enrollment, and to identify factors which may influence enrollment rates across the 22 VISNs.
Methods: We identified all unique users of VA inpatient or outpatient care between FY95 and FY97 who were alive and age 65 or older in FY97. We linked this cohort with the Denominator file from the Health Care Financing Administration (HCFA) merging by SSNs, gender, and date of birth. We determined the proportion of VA users enrolled in Medicare and identified enrollment type (fee-for-service [FFS] vs Health Maintenance Organization [HMO]). We calculated the percent (rate) of VA users in each VISN who were enrolled in Medicare. Rates were further stratified by patient characteristics.
Results: There were 1.5 million VA users who were alive and age 65 or older in FY97. Of this cohort, 1.3 million (84%) were enrolled in Medicare in 1997. Of those enrolled in Medicare, 88% were enrolled in FFS versus 12% in a Medicare HMO. Most VA users were initially entitled to Medicare benefits because of age versus disability (81% vs 19%). Medicare enrollment was significantly lower in female vs male VA users (59% vs 86%, p<0.001) and significantly lower among non-service connected vs service connected VA users (85% vs 88%, p<0.001). There was a significant difference in the rate of Medicare enrollment by VISN (low VISN = 78%; high VISN = 88%, P<0.001). There was significant variation across VISNs in the percentage of VA users who were Medicare FFS (low VISN = 49%; high VISN = 86%) versus enrolled in Medicare HMOs (low VISN = 1%; high VISN = 33%).
Conclusions: Among VA users age 65 and older, a high proportion were enrolled in the Medicare program in 1997. We found significant variation in the rate of Medicare enrollment among elderly veterans by VISN. VISNs also varied in the percent of VA users enrolled in Medicare FFS compared to Medicare HMOs.
Impact: Use of non-VA services has implications for the effective and efficient management of resources within each VISN, and impacts the budgets established for each VISN. Regional differences in Medicare enrollment rates and the type of Medicare enrollment (FFS vs HMO) may reflect regional differences in patterns of health care utilization among veterans. More research is needed to examine variation in use of non-VA care at the network level.
79. Telemedicine Clinics: Successes and Failures
Bonnie J. Wakefield, PhD, Iowa City VAMC, Iowa City, Iowa, JR Flanagan, University of Iowa College of Medicine, JP Specht, Iowa City VAMC
Objectives: The goal of this project is implementation and evaluation of specific telehealth applications, including triage of patients in the Iowa Veteran’s Home (IVH) by primary care staff in the Urgent Care Clinic (UCC) of the Iowa City VA Medical Center (ICVAMC); and follow up evaluation of IVH patients in subspecialty clinics.
Methods: For the UCC portion of the study, the initial plan was to use 2 months on and 2 months off as the mechanism for randomization for UCC visits. Consultation procedures were developed and pilot tested in May 1998. The Specialty Clinic study was initiated in November 1998 at both the ICVAMC and the Des Moines VAMC (DMVAMC). Using established procedures, specialty clinic appointments are scheduled, except that the patient remains at a distant site. For both studies, data are collected on the outcome of the consultation (admission, follow-up clinic appointment, change in treatment), providers present, and patient and provider satisfaction.
Results: For the UCC study, no patients were subsequently referred for telemedicine urgent care consults. Even though patients were quite favorable to the technology, IVH physicians have not used telemedicine for UCC referrals. For the specialty clinics through October 1999, 75 individual consultations have been completed. Most of these are follow up visits (97%, n=73) in urology, neurology, cardiology, and general surgery. The most frequent outcome is a change in treatment plan and the patient remains at IVH (n=32, 43%) or no change in treatment (n=21, 28%). Physicians ratings are: 76% good to excellent for usefulness in developing a diagnosis; 85% good to excellent for usefulness in developing a treatment plan; 76% good to excellent for quality of transmission; and 84% good to excellent satisfaction with the consult format. Overall, 72% of patients were satisfied with the consult format, while 8% were neutral and only 6% were somewhat dissatisfied (14% did not respond to this question). 100% of patients believed the specialist understood their problem and 92% felt it was easier to get medical care via telemedicine.
Conclusions: Telemedicine applications have been successful in specialty clinics, yet remain unused for UCC. As follow up to the UCC study, we plan to interview physicians at the three sites (IVH, ICVAMC, and DMVAMC) to gain insight into the perceptions of telemedicine by physicians. Several facilitators, barriers, and logistical issues to use of telemedicine were identified during the course of these studies.
Impact: The promise of telemedicine is considerable. Telehealth has the potential to improve access to specialized care for veterans, enhance training of health professions students, provide patient and staff education, improve patient satisfaction and clinical outcomes, and relieve family/caregiver burden. As evidenced by the results of these studies, lessons can be learned from both successful and failed attempts to implement telemedicine applications. While there is great potential, it needs to be clear how and under what conditions telemedicine contributes to improved patient care.
HSR&D Funded: DEV 97-012
80. Incorporating Missing Data Information in Modeling Facility Case-Mix Adjusted Decline Rates in Functional Status
Bei-Hung Chang, ScD, CHQOER, Bedford VA Medical Center, Bedford , MA, AK Rosen, PhD, CHQOER, Bedford VA Medical Center, C Rakovski, MS, CHQOER, Bedford VA Medical Center, J Wu, MS, CHQOER, Bedford VA Medical Center
Objectives: Decline in functional status is used as a quality indicator in long-term care facilities in both VA and non-VA settings. To assess decline within a specific time period, it is necessary to have functional status measurements at two time points. However, many residents in long-term facilities lack a second assessment due to death, transfer to other facilities, or return to the community. Facility comparison of decline is, therefore, restricted to only those residents who have two assessments. We developed a method to incorporate this missing data information in modeling facility case-mix adjusted decline rates.
Methods: 121 VA long-term care facilities were included in the analysis. Decline in functional status was defined as a worsening of 2 points in the activities of daily living (ADLs) summed score between two time points of 4/1/96 to 10/1/96. Facility observed decline rates are the proportion of patients who decline. Facility expected decline rates were estimated by a logistic regression model that included residents’ medical conditions, diagnoses, and demographic characteristics. The case-mix adjusted decline rate was calculated as the difference between the observed and expected rate. Missing information on those residents without a second assessment was summarized by 4 variables that indicated the percentage of residents who died, transferred to community, another facility, or another hospital. These 4 variables were then used to model variation in facility case-mix adjusted decline rates. A Hierarchical Bayes Linear Mixed Effects Model was used. This method models the "true" case-mix adjusted rate which does not include random error.
Results: Based on the Bayesian method the mean of the true case-mix adjusted decline rate was estimated to be -.003 (3 fewer patients per 1,000 declined compared to what would be expected giving the patient characteristics in the facility) with a range of -.068 to .187. The reason for missing the second assessment is associated with the true case-mix adjusted decline rate. Specifically, facilities with more patients transferred to communities had significantly lower case-mixed adjusted decline rates (beta=-.001, probability of >0 is .023).
Conclusions: We have shown that the reason for the lack of a second assessment in functional status is associated with the facility case-mix adjusted decline rates. Since it is reasonable to assume that facilities that provide good care are likely to result in having more residents transferred to the community, our finding that facilities with more patients transferred to communities had lower case-mix adjusted decline rates therefore supports the use of functional status decline as a quality indicator.
Impact: To avoid biased results it is important to take into account missing data information when using functional status decline rates as a quality indicator of long-term care. Our method of modeling facility case-mix adjusted rates by proxy variables indicating type of missing data verifies the use of decline rate as a quality indicator.
HSR&D Funded: IIR 96-065
81. Severity of Acute Inpatients and Length of Stay in VA
Yujing Shen, Bedford VA Medical Center, Bedford, MA, A Hendricks, Bedford VA Medical Center
Objectives: To assess the severity of acute inpatients in VISNs and examine the relationship of severity to length of stay (LOS) across VISNs using 3M APR-DRG software for all VA acute inpatients for the first time.
Methods: Using 3M APR-DRG software which generates severity levels within each APR-DRG according to multiple comorbidities, age, procedures, principal diagnosis and discharge status, we assigned VA inpatients in FY 1998 to one of 355 base APR-DRGs and one of four severity levels: 1 minor, 2 moderate, 3 major or 4 extreme. This assignment required generating acute inpatient episodes based on bed section information within each inpatient stay. Within each stay in the VA Patient Treatment File, we defined all sequences of acute bed section stays as one acute inpatient episode. We identified 59 major APR-DRGs each of which accounted for at least 0.5% of the overall acute inpatient episodes or the overall length of stay (total 65% of workload). For each major APR-DRG, we compared severity level in each VISN with the overall mean based on Z-value. We also compared the observed average LOS in each VISN with the expected LOS with/without case-mix adjustment based on Z-value. For each APR-DRG, we case-mix adjusted LOS by regression on patient severity, age, sex and number of days in other bed sections during a stay using untrimmed and trimmed data from which we deleted outliers (log of LOS greater than 3 times SD).
Results: The mean severity levels in APR-DRGs ranged from 1.2 (drug/alcohol abuse) to 3.2 (respiratory system diagnosis w ventilator support 96+hours). On average, severity levels in VISNs were not significantly different from the means for 77% of the APR-DRGs. In the VISNs, up to 29% (average 11%) of the 59 APR-DRGs had significantly (p<=0.05) higher severity levels than the overall means, and up to 22% (average 12%) had lower severity levels. For all the regressions, severity was highly correlated with LOS, but R-squares ranged from 0.003 to 0.36 (mean 0.07, median 0.05). Moving from moderate to major severity, LOS increases varied across APR-DRGs with a maximum of 17 days (tracheostomy except for face/mouth/neck diagnoses), a minimum of 1 day (disorder of personality/impulse control) and an average of 5 days. LOS varied considerably across VISNs. Up to 68% (average 14%) of the 59 APR-DRGs had significantly (p<=0.05) higher LOS than the expected LOS with case-mix adjustment, and up to 83% (average 27%) had lower LOS. Except for 8% of all 1298 cases (22 VISNs*59 APR-DRGs), comparisons based on expected LOS with/without adjustment had identical results. Trimming the data did not change the results.
Conclusions: Inpatient severity varies across VISNs. Severity is a good predictor for LOS at patient level but does little to explain the difference in LOS across VISNs.
Impact: To our knowledge, this is the first time that anyone has applied APR-DRGs severity measurement to all VA acute inpatients. This provides VA researchers with another valuable method to identify utilization patterns, evaluate resource utilization and outcomes.
HSR&D Funded: VERA Assessment (MRR 99-008)
82. The Nature and Prevalence of Patient Expectations in a VA General Medicine Clinic
Bob Mitchell Peck, Durham VA Medical Center, Durham, NC, PA Ubel, Philadelphia VA Medical Center, DL Roter, Johns Hopkins University School of Public Health, SD Goold, University of Michigan Medical Center, DA Asch, Philadelphia VA Medical Center, AL Shearin, SC Grambow, CM Koropchak, JA Tulsky, Durham VA Medical Center
Objectives: Patients' expectations may drive medical decision-making. Clinicians are more likely to prescribe new medications, order tests, and make referrals when patients request these services. Fulfilled expectations are associated with higher satisfaction and adherence to therapy. Nevertheless, relatively little is known about patients' specific expectations for care. We conducted this study to describe the nature and prevalence of expectation patients present to their clinicians and how fulfillment of these expectations affects satisfaction.
Methods: Participants were recruited from a random sample of patients visiting a VA general medicine outpatient clinic. Interviewers administered a face-to-face, structured questionnaire designed to assess patients' expectations before the visit. Afterwards, patients identified the services they received and rated their satisfaction with the visit using the ABIM Patient Satisfaction Questionnaire (PSQ) and Ware Visit Specific Questionnaire (VSQ).
Results: We approached 401 patients, of whom 253 (63%) completed both the pre- and post-visit interviews. The remaining 148 patients either refused (16%), did not complete the interview (21%), or were unable to understand the questions (<1%). The median age of the study population was 65; 68% were white and the majority had completed a high school education. Thirty-one percent of patients reported one expectation and 25% had two or more expectations. The expectations were extremely varied, falling into 40 different categories of tests, referrals, and new medications. Furthermore, patients' expectations for clinic visits were often vague, with 40% of responses not falling within any specific category. Fifty percent of patients who expressed an expectation did not receive the service they desired. Nevertheless, patient satisfaction was very high. The median score on the VSQ was 1.5 and on the PSQ was 1.0 (1=excellent, 5=poor). Satisfaction, however, was not related to whether expectations were met or unmet. Patients with one or more unmet expectations did not differ significantly from patients with all expectations met on the VSQ and PSQ (p=.308 and p=.502, respectively). Generally, there were no differences between patients who wanted a particular service and received it and those who wanted the same item but did not receive it. The one exception, however, was new medications. Patients who expected a new medication but did not receive it reported lower levels of satisfaction (p=.002).
Conclusions: Patients present to their VA primary care providers with multiple and varied expectations for tests, referrals, and new medications, and many of these expectations are not met. However, patient satisfaction remains high and is largely unrelated to fulfillment of expectations. Further research should examine attributes of the clinical encounter that result in high satisfaction despite unmet expectations.
HSR&D Funded: MPC97-011-1
83. Insurance Coverage and Access to Care for Non-Elderly Veterans
Chuan-Fen Liu, VA Puget Sound Health Care System, Seattle, WA, ML Maciejewski, VA Puget Sound Health Care System, A Sales, VA Puget Sound Health Care System
Objectives: The objective of this study is to compare the insurance coverage and access to care for non-elderly veterans and non-elderly people in the general population. This study also compares the insurance coverage and access to care for veterans who are uninsured, publicly insured, and privately insured. In addition, we explore the factors associated with uninsurance in the non-elderly veteran population.
Methods: Data from the 1996 Medical Expenditures Panel Survey (MEPS) was used to identify veterans under age 65 and their insurance coverage, access to care, demographic characteristics, and employment status. There were 1,496 veterans under age 65 who were surveyed in MEPS, representing 17.9 million of non-elderly veterans. The MEPS survey was not able to differentiate VA users and non-users, so results are generalizable to the entire population of non-elderly veterans.
Results: Our results indicate that a vast majority of non-elderly veterans had private insurance in the first half of 1996 (81%). The observed uninsured rate in MEPS was 14.7% for non-elderly veterans as compared to 20.6% for the general non-elderly population. The uninsured non-elderly veterans were more likely to be of Hispanic origin, unmarried, less educated, between ages 16-44, residing in non-MSA and unemployed as compared to privately insured veterans. In addition, 80% of veteran reported having a usual source of care as compared to 77% for the general non-elderly population. Uninsured, non-elderly veterans were less likely to have a usual source of care than those with private or public coverage.
Conclusions: The VA system provides a special safety net for non-elderly veterans, which is not available to most of Americans. Most veterans have alternative mechanisms to obtain health care and do not have to rely on the VA as a safety net. By reaching out to publicly and privately insured veterans, the VA would be able to increase the number of veteran users and increase the percentage of total revenue obtained from non-appropriated sources, two goals in the VA’s 1998-2003 Strategic Plan.
Impact: The results from this study can be used by VA administrators to target specific groups of veterans as new users of VA services, and at the same time further the VA’s mission as a health care safety net.
84. Variation among VA Hospitals in Risk-Adjusted Pulmonary Complication Rates following Major Noncardiac Surgery
Ahsan Mohammed Arozullah, MD, MPH, University of Illinois at Chicago, VA Chicago, Westside Division, Chicago, Illinois, W Henderson, Cooperative Studies Program Coordinating Center, Hines VA Hospital, S Khuri, Veterans Affairs Boston Healthcare System, West Roxbury, MA , J Daley, Partners Healthcare System and Veterans Affairs Boston Healthcare System, West Roxbury, MA
Objectives: 1. To determine if variation exists in postoperative pneumonia (POP) and respiratory failure (RF) rates among VA hospitals. 2. To determine the effect of adjustment for patient-specific risk factors on the hospital rankings for postoperative pneumonia and respiratory failure rates.
Methods: Cases were selected from those enrolled in the National Surgical Quality Improvement Program. POP was defined as positive sputum culture with antibiotic treatment or an infiltrate on chest x-ray diagnosed as pneumonia or pneumonitis following surgery. RF was defined as mechanical ventilation greater than 48 hours postoperatively and/or reintubation and mechanical ventilation subsequent to postoperative extubation. Logistic regression models using patient-specific risk factors to predict POP and RF were developed using Phase I cases (10/91-12/93, 44 VA hospitals) and validated using Phase II cases (1/94-8/95, 123 VA hospitals). The models were used to estimate the probability of POP and RF for each case with the sum within each hospital designated as the ‘expected’ number of complications. Observed to expected (O:E) ratios and 95% confidence intervals (CI) were computed for each hospital. Hospitals were designated as ‘high’ outliers if the CI was > 1, or ‘low’ outliers if the CI was < 1.
Results: Phase I POP rates ranged from 0.8% to 8.5%. After risk adjustment, 7 hospitals (16%) were high outliers, 12 hospitals (27%) were low outliers, and 9 hospitals (20%) changed rank by > 5. Phase I RF rates ranged from 1.0% to 6.0%. After risk adjustment, 5 hospitals (11%) were high outliers, 7 hospitals (16%) were low outliers, and 21 hospitals (48%) changed rank by > 5. No high outlier hospital for RF was a high outlier for POP. Four of the 7 low outlier hospitals for RF were also low outliers for POP. Phase II POP rates ranged from 0% to 10.5%. After risk adjustment, 26 hospitals (21%) were high outliers, 21 hospitals (17%) were low outliers, and 38 hospitals (31%) changed rank by > 10. Phase II RF rates ranged from 0% to 6.0%. After risk adjustment, 15 hospitals (12%) were high outliers, 23 hospitals (19%) were low outliers, and 44 hospitals (36%) changed rank by > 10. Nine of the 15 high outlier hospital for RF were high outliers for POP. Eleven of the 21 low outliers for POP were low outliers for RF.
Conclusions: Significant variation exists in risk-adjusted rates of POP and RF among VA hospitals. There is only moderate correlation between the POP and RF rate within each hospital. Twenty to forty-eight percent of hospitals had a significant change in ranking following adjustment for patient-specific risk factors.
Impact: Hospital POP and RF rates are potential outcome-based measures that could be used in addition to mortality rates to evaluate surgical quality of care. By focusing future studies on the impact of hospital structure and process of care measures on POP and RF rates in high and low outlier hospitals, we may gain insight into how to improve perioperative care across the entire VA healthcare system.
85. Impact of Hypertension Guideline Implementation on Patients’ Blood Pressure
Mary K. Goldstein, MD, VA Palo Alto and Stanford University, Palo Alto, CA, RW Coleman, VA Palo Alto, P Munroe, VA Palo Alto, P Gholami, VA Palo Alto, P Lavori, VA Palo Alto and Stanford University, BB Hoffman, VA Palo Alto and Stanford University
Objectives: Clinical practice guidelines are widely promoted as a means to improve quality of care, but little is known about their impact on patients’ health status. VISN 21 mandated performance standards for drug therapy of hypertension. This abstract focuses on how implementation of these standards affected blood pressure of previously treated hypertensives at VA Palo Alto.
Methods: VISN 21 performance standards for choice of drug therapy (drug regimen for HTN patients to include: thiazides or beta blockers for those without diabetes (DM) or heart failure (HF); beta blocker for those with coronary artery disease where not contraindicated; ACE inhibitor or angiotensin receptor blocker for those with DM or HF) were implemented as follows at VA Palo Alto: all primary care clinicians were sent announcements of the performance standards, a copy of the VA and JNC hypertension guidelines, and a report on the proportion of hypertensive patients at the facility in the previous year prescribed antihypertensive drugs that were in concordance with the VISN standards. To focus attention on hypertension, clinicians were given a listing of each patient’s antihypertensive drugs at each of their visits. In addition, as part of a randomized trial of supplemental hypertension advisories focusing on drug choice, half the clinicians were advised at each patient’s visit whether or not the patient’s antihypertensive drugs were concordant with VISN standards.
Results: Guideline concordance improved substantially, e.g., from 48% at baseline to 61% after implementation in the category of hypertensives without DM or HF. As reported previously, the improvement was greater in the group receiving the patient-specific advice at each clinic visit (Med Dec Mkg, 10/99). 1538 patients had BP measurements listed in their computer records for initial and at least one follow-up visit. Mean systolic and diastolic BP decreased from 147(sd 20) and 78(sd 13) to 143 and 74, respectively (paired 2-tailed t-test for each: p<.001). Systolic BP decreased for the ‘no change in guideline concordance’ (n=1358), and ‘discordant to concordant’ (n=127) groups, but not in the ‘concordant to discordant’ (n=53) group, with mean decreases of 4.5, 7.6, and 0.3 respectively (N.S.). Mean decreases in diastolic BP were 3.4, 6.1, and 0.2, respectively; the decrease in the ‘discordant to concordant’ was significantly greater than in the ‘concordant to discordant’ (Post-Hoc test with Bonferroni correction, p<.05)
Conclusions: A hospital-wide hypertension guideline implementation was associated with improved BP control. Diastolic BP decreased more in patients whose treatment was altered in concordance with guidelines than in those whose treatment was altered away from guidelines. The guideline drugs have different effectiveness in preventing cardiovascular morbidity but do not have different efficacy in lowering BP, so perhaps guideline-concordance may be a surrogate for good overall patient management encouraged by the guideline implementation.
Impact: The overall decrease in BP encourages us to believe that guideline implementation may actually lower BP, and justifies clinical trials of methods of guideline implementation. The greater improvement in BP in patients whose drug regimen was changed from guideline discordant to concordant suggests that guideline-concordance of drug regimen may be useful as a surrogate measure of quality of care.
HSR&D Funded: CPG 97-006
86. Examining the Effectiveness of Academic Detailing in the Implementation of VA Hypertension (HTN) Clinical Practice Guidelines
Bradley J. Brazill, VA Northern California Health Care System, Martinez, CA, JL Meier, VA Northern California Health Care System, JR Lopez, VA Northern California Health Care System, S Lee, VA Northern California Health Care System, MK Goldstein, VA Palo Alto Health Care System, D Siegel, VA Northern California Health Care System
Objectives: VA clinical practice guidelines for HTN recommend the use of beta-blockers and thiazide diuretics as first-line therapy for HTN, depending on co-morbidities, and do not recommend calcium antagonists as preferred first-line therapy for any hypertensive. A VISN research grant was awarded to study the effect of academic detailing (AD) in the implementation and promotion of the HTN clinical practice guideline in the five facilities throughout VISN-21. We examined the impact of this AD on prescribing patterns for antihypertensive drugs.
Methods: Four hypertensive populations were targeted: use of beta-blockers and thiazide diuretics as first-line therapy for hypertensives excluding those with congestive heart failure (CHF) and diabetes mellitus (DM); beta-blockers for hypertensives with coronary artery disease (CAD), excluding patients with chronic obstructive pulmonary disease (COPD), asthma, DM, or CHF; angiotensin converting enzyme inhibitors (ACEI) or angiotensin II antagonists (ARB) for hypertensives with CHF or DM. AD interventions at each facility included lectures, poster presentations, distribution of educational materials, profiling of provider prescribing patterns, and individualized meetings with providers to promote the guideline. For analysis, researchers used ICD-9 codes from encounter forms to identify patients with HTN, DM, CHF, CAD, COPD and asthma between 4/01/97 and 9/30/98 for the baseline period, and between 4/01/98 and 9/30/99 for the post-implementation period. Prescribing patterns for HTN with and without co-morbid conditions was analyzed by relating ICD-9 codes to prescriptions for beta-blockers, alpha-1 blockers, calcium antagonists, thiazide diuretics (alone or in combination with a potassium sparing diuretic), loop diuretics, ACEI, and ARB filled over a six month period (4/01/98 to 9/30/98 for baseline and 4/01/99 to 9/30/99 for post-implementation). Pharmaceutical cost differences between the baseline and post-implementation periods were calculated as the difference between actual drug acquisition cost in the post-implementation period and what would have been spent had the changes in prescribing not occurred.
Results: Of 16,865 hypertensives (excluding those with CHF and/or DM) at baseline, 9146 (54.3%) were receiving a beta-blocker and/or thiazide diuretic; 2717 of 3880 (70.0%) hypertensives with CHF and 5643 of 8172 (69.1%) hypertensives with DM received an ACEI or an ARB. Of the 3164 hypertensives with CAD (excluding patients with COPD, asthma, DM, or CHF), 1663 (52.6%) were receiving a beta-blocker. Post-implementation, 11,890 of 19,544 (60.8%) hypertensives (excluding those with CHF and/or DM) received a beta-blocker and/or thiazide diuretic, a 12.0% increase (p<0.001); 2765 of 3717 (74.4%) hypertensives with CHF and 6699 of 9197 (72.8%) hypertensives with DM received an ACEI or an ARB, a 6.2% (p<0.001) and 5.5% (p<0.001) increase, respectively. Of 3890 hypertensives with CAD (excluding patients with COPD, asthma, DM, or CHF), 2255 (58.0%) received a beta blocker, a 10.3% (p<0.001) increase. Over the 1 year period, $177,000 in pharmaceutical costs were saved.
Conclusions: The AD intervention led to a change in the antihypertensive prescribing patterns of providers, reflecting a closer adherence to VA and JNC VI recommendations. There was a substantial cost savings associated with this change in prescribing.
Impact: AD is an effective method for improving adherance with treatment guidelines, and could be used in many therapeutic areas.
87. Comorbidity, Severity, and Outcome of Patients with Acute Myocardial Infarction in VHA Compared with Medicare
Laura A. Petersen, MD, MPH, Houston Center for Quality of Care and Utilization Studies, Houston, TX, SLT Normand, Harvard Medical School, Boston, MA, M Volya, Harvard Medical School, Boston, MA, C Brown, Center for the Study of Practice Patterns in AMI, Boston, MA, BJ McNeil, Harvard Medical School, Boston, MA
Objectives: Many clinicians believe that patients cared for in VHA suffer greater comorbidity and severity of illness than patients cared for under Medicare financing. To compare comorbidity, severity, and outcome of acute myocardial infarction (AMI) in VHA and non-VHA health care settings, we used clinical data collected by chart review from two groups of patients.
Methods: The first group was a stratified national random sample of male veterans >= age 65 discharged with the primary diagnosis of AMI (ICD-9-CM-410) in 1994-95 from non-psychiatric VAMCs with a length of stay (LOS) 2-180 days. The second group was a random sample of males age>= 65 discharged with AMI from the states of CA, FL, MA, NY, OH, PA, and TX in 1994-95 with a LOS 3-180 days cared for under Medicare financing. Data on comorbid conditions and severity of AMI were collected on both samples using the Cooperative Cardiovascular Project data collection instrument.
Results: The sample consisted of 2,486 veterans from 81 VA Medical Centers and 32,141 males from 1,530 non-VHA hospitals. The mean age of the VHA patients was 73.4 years (+ 5.7) and of the Medicare patients 75.6 years (+ 7.0) (p<0.001). There were more blacks in the VHA group than the Medicare group (13.6% vs. 3.8%; p<0.001). LOS for the index admission was longer in VHA (13.5 days [+ 11.0] vs. 9.6 days [+ 7.4]; p<0.001). Veterans with AMI were more likely than Medicare patients to have a recorded history of hypertension (64.3% vs. 57.2%), chronic obstructive pulmonary disease (COPD) (30.9% vs. 23.9%), stroke (20.4% vs. 14.3%), prior MI (36.2% vs. 32.8%), dementia (7.2% vs. 5.1%), or diabetes (27.0% vs. 24.3%) (all p<0.001). Fewer VHA than Medicare patients had a recorded history of cancer (2.5% vs. 3.3%) or of prior coronary angioplasty (6.1% vs. 7.6%) (both p <0.05). On presentation with AMI, veterans were more likely to have ST elevation (46.0% vs. 39.8%) (p<0.001). Mean peak CK-MB was 12.5 (+ 17.5) in the VHA sample vs 14.4 (+ 33.2) in the Medicare sample (p<0.001). There was no significant difference between the two groups in those initially presenting with a recorded history of cardiac arrest, shock, pulmonary edema, or tachycardia. During the hospital stay, veterans were more likely to develop shock (10.8% vs. 8.5%), recurrent angina (37.2% vs. 26.7%) or reinfarction (20.9% vs. 3.4%) (all p<0.001). There was no significant difference in age-adjusted risk of 30-day mortality in VHA vs. Medicare (Cochran-Mantel-Haenszel relative risk [95% Confidence Interval] 0.9 [0.9-1.0]).
Conclusions: Elderly veterans with AMI are more likely to suffer from comorbid conditions such as COPD, diabetes, hypertension, stroke, and prior MI than males cared for under Medicare financing. Veterans are also significantly more likely to develop a number of complications such as shock or reinfarction during the hospital stay, possibly accounting for the longer mean LOS in the VHA sample.
Impact: Despite these clinical differences, the age-adjusted risk of short-term mortality was the same for both groups, suggesting similar quality of care.
HSR&D Funded: IIR 94-054; RCD 95-306
88. Development and Testing of a Clinically Detailed System to Assess Quality of Care in Hypertension
Steven Asch MD MPH, VA Greater Los Angeles Healthcare System, LA, CA, Eve Kerr MD MPH, VA Ann Arbor, Beth McGlynn PhD, RAND
Objectives: Guidelines for care of hypertensive patients have proliferated recently, yet quality assessment remains difficult in the absence of well-defined measurement systems. As part of a larger effort, we developed indicators for measuring process quality in hypertension. We tested this system on hypertensive women in one site of a large group model health plan.
Methods: We selected indicators based on the explicit a priori criteria of clinical importance, outcome link, evidence strength, operational ease, provider accountability, reliability and adaptability. A 1996 expert panel rated the indicators using the modified Delphi method. 1 screening, 11 diagnostic, 5 treatment, and 2 follow-up indicators (including subparts) survived this process. Patients were eligible if they had a visit in 1996-97, were continuously enrolled for at least 13 months, and were under 65 and female. Trained nurses used a laptop-based tool to abstract data from medical and administrative records for a two-year period. We sampled 467 out of 15004 eligibles, oversampling patients with hypertensive and other chronic disease administrative codes; all these patients were eligible for the screening indicator. 234 women with average blood pressure >=140/90 or a documented diagnosis of hypertension were evaluated for the remaining indicators.
Results: We analyzed 19 indicators and calculated weighted adherence proportions for each indicator. Ninety-seven percent of hypertensive patients were eligible for at least 1 non-screening indicator. Between 0% to 93% of eligible patients received the indicated care. Eight-four percent of all patients presenting for care had at least an annual blood pressure measurement. Among hypertensives, only 7% had an adequate history as defined by documentation of at least 2 of the following elements: family or personal history of premature CAD, VA, diabetes or hyperlipidemia. Only 30% had annual urinalysis, 29% annual potassium, and 86% had annual cholesterol. Just over 3% of prevalent and 33% of new mild to moderate hypertensives had chart evidence of being offered lifestyle modification before pharmacotherapy. While 93% had an annual visit, only 37% of hypertensives noted to have persistent elevations >160/90 had changes in therapy or lifestyle recommended. We also calculated overall adherence proportions (number of indicators passed/number of indicators eligible) for each person. The average adherence proportion was lower in patients with poor control (average BP>140/90) than those with good control (54% vs. 73% p<.001).
Conclusions: Quality of hypertensive care falls far short of indicators based on national guidelines. Poor performance in essential care processes is associated with poor blood pressure control. Future research should examine patient and system level predictors of adherence and extend measurement to older women, men and other delivery systems.
Impact: We have developed and tested an evidence-based system for measuring process quality in hypertension. Because hypertension is widely prevalent in the VA, quality assessment based on this system may prove useful in improving hypertensive care.
HSR&D Funded: CDA
89. Health Status and 2-year Outcomes among Veterans with Coronary Artery Disease: Results from the Ambulatory Care Quality Improvement Project (ACQUIP)
John A. Spertus, University of Missouri - Kansas City, Kansas City, Missouri, MB McDonell, VA Puget Sound Health Care System, DH Au, VA Puget Sound Health Care System and University of Washington, VS Fan, VA Puget Sound Health Care System and University of Washington, SD Fihn, VA Puget Sound Health Care System and University of Washington
Objectives: Little is known about the predictive validity of condition-specific measures of health status in relation to subsequent morbidity and mortality. Using data from the ongoing Ambulatory Care Quality Improvement Project (ACQUIP), we evaluated the performance of the Seattle Angina Questionnaire (SAQ), a brief, self-administered measure of condition-specific health status, in predicting subsequent death from all causes and ischemic cardiac events among VA outpatients with coronary artery disease. Methods: Surveys were mailed to General Internal Medicine Clinic patients from 7 VA medical centers. Baseline SAQ scores for 5,391 patients with self-reported coronary artery disease were analyzed. Patients were grouped according to their SAQ Physical Limitation and Angina Frequency scores (range 0-100, higher scores = better function). Scores were collapsed into 4 intervals: 0-24, 25-49, 50-74, and 75-100. All-cause mortality and admissions for acute coronary syndrome (ACS = acute MI or unstable angina) were accessed from VISTA and the VA Beneficiary Identification and Record Locator System. Rates of death and admission for ACS were determined for a 2-year period following administration of the SAQ. Logistic regression was used to estimate risk ratios with the subjects in the top interval (scores of 75-100) serving as the reference group.
Results: Overall, 7.7% of patients died, and 3.0% had an admission for myocardial ischemia. For the physical limitation scale, from lowest (0-25) to highest (75-100) score interval, the all-cause death rates were 13.3%, 9.0%, 5.2%, and 3.5%; the admission rates were 4.4%, 3.9%, 2.4% and 1.7%. The SAQ angina frequency scale produced similar results; the all-cause death rates were 12.4%, 10.6%, 6.6%, and 7.3%, and the admission rates were 5.5%, 5.0%, 3.7% and 2.2%. For the physical limitation scale, the unadjusted relative risk of death was 3.78 (p=.000, 95% CI=2.59-5.51) for patients scoring 0-24, 2.68 (p=.000, 95% CI=1.90-3.78) for patients scoring 25-49 and 1.48 (p=.045, 95% CI=1.01-2.18) for those scoring 50-74. For the angina frequency scale, the unadjusted relative risk of death was 1.88 (p= .0025, 95% CI=1.25-2.84) for patients scoring 0-24, 1.51 (p=.0047, 95% CI=1.13-2.01) for patients scoring 25-49 and 0.90 (p=.4026, 95% CI=0.71-1.15) for those scoring 50-74. For the physical limitation scale the unadjusted relative risk of admission was 2.86 (p=.0002, 95% CI=1.64-5.00) for patients scoring 0-24, 2.37 (p=.0007, 95% CI=1.44-3.90) for patients scoring 25-49 and 1.42 (p=.2155, 95% CI=.82-2.47) for those scoring 50-74. For the angina frequency scale, the unadjusted relative risk of admission was 2.31 (p= .0111, 95% CI=1.21-4.42) for patients scoring 0-24, 2.36 (p=.0001, 95% CI=1.54-3.62) for patients scoring 25-49 and 1.74 (p=.0000, 95% CI=1.22-2.47) for those scoring 50-74.
Conclusions: A cross-sectional survey of disease-specific health status was associated with subsequent survival and admissions for ACS. The use of patient-centered health status offers a powerful mechanism for identifying high-risk patients.
Impact: If high-risk patients with coronary artery disease can be identified through the use of health status instruments, then treatment efforts can potentially be directed more intensively to those patients. This could impact the quality of care delivered and ultimately patient outcomes.
HSR&D Funded: SDR96-002
90. Secondary Prevention of Ischemic Heart Disease – Guideline Concordance
Kevin L. Sloan, MD, VA Puget Sound Heath Care System, Seattle, WA, AE Sales, VA Puget Sound Heath Care System, NR Every, VA Puget Sound Heath Care System, GV Martin, VA Puget Sound Heath Care System, H Sun, VA Puget Sound Heath Care System, S Pineros, VA Puget Sound Heath Care System
Objectives: Ischemic heart disease (IHD) is a significant public health problem for which secondary prevention is effective in reducing morbidity and mortality. Although practice guidelines exist, it is unclear to what degree have they already been incorporated into routine clinical practice and at what point(s) in the process of care one should intervene to improve guideline compliance. The QUERI IHD project examined concordance with the IHD secondary prevention guidelines in eight VA medical centers in the Northwest Network (VISN 20).
Methods: Utilizing data contained on the CHIPS relational database server (which mirrors the VistA systems at the eight VA medical centers in the VISN 20), we targeted all currently active primary care or cardiology patients with a face-valid diagnosis of IHD. The IHD secondary prevention guidelines were broken down into three steps: (1) LDL obtained within the last 18 months, (2) if LDL was elevated, some lipid lower agent (LLA) was prescribed and (3) if LDL continued to be elevated after the initial LLA prescription, the dose was titrated upward and/or a second agent was started.
Results: 12,632 current primary care/cardiology patients with IHD were found. 97.8% were male, 84.9% were white, the mean age was 68.8 (SD 10.4), 58.2% were married and 23.7% divorced. 59.6% were either service connected or had a NSC pension. Within this group, fully 40.9% had no LDLs drawn within the past 18 months. Even excluding one site where LDLs were not an orderable lab item until 10/1/98, 35.9% of those remaining had no LDLs drawn within the past 18 months. Of those with current LDLs, 40.4% were <=100 (within guideline), 35.0% were between 100 and 130 (borderline) and 24.5% were >130. Strict application of the IHD clinical guidelines would rate all 59.6% of the individuals with LDLs >100 as guideline discordant. LLA prescription fared no better. Even in the LDLs>130 group, 34.6% were on no LLAs while only 18.3% were on more than starting doses of any LLA. Although sites varied somewhat, the striking pattern of little apparent relationship between the measured LDLs and aggressiveness of LLA therapy was consistent across the VISN.
Conclusions: In clinical guideline driven treatment of individuals with known ischemic heart disease, there is a great deal of room for improvement in all three phases of treatment including (1) measuring of LDLs, (2) starting of LLAs and (3) following-up persistently elevated LDLs with more aggressive pharmacotherapy. Most striking, however, was the apparent disconnection between the measured LDL value and the clinical response.
Impact: These findings provide a baseline benchmark against which improvements in the process of care may be measured. Additionally, they highlight an apparent disconnect between laboratory results and clinical response, perhaps suggesting the need for clinical reminders that are more sensitive to magnitude of findings (e.g., LDL value) rather than simply presence or absence of a test or procedure.
HSR&D Funded: IHS 99-233-1
91. Variations in the Management and Outcome of Patients with AIDS-Related Pneumocystis Carinii Pneumonia (PCP) in New York versus Other U.S. Cities: 1995-1997
Charles L. Bennett, Northwestern University, Chicago, IL, TJ Sloan, Northwestern University, JS Chmiel, Northwestern University, C Chan, Northwestern University, AM Sipler, Northwestern University, MB Goetz, VA Greater Los Angeles Healthcare System, CL Bennett, VA Chicago Health Care Systems - Lakeside Division
Objectives: Despite dramatic declines in morbidity and mortality rates for AIDS-related complications over the last few years, PCP remains the most common and costly complication. We evaluated if geographic variations in the management and outcome of AIDS-related PCP in the 1980s (Cohn et al., J Acquir Immune Defic Syndr Hum Retrovirol 1996;13:408-15) were still present in the 1990s.
Methods: Data from 1,985 medical records of patients hospitalized for AIDS-related PCP between 1995-1997 were abstracted at 78 hospitals in New York, Miami, Chicago, Los Angeles, Seattle, Phoenix, Tucson, Nashville, and Memphis.
Results: Among cities, New York had the lowest percentage of hospitalized PCP patients with private health insurance (9.7% versus range of 14.9% to 39.4% for the other cities, p<0.001), the highest number of injection drug users (37.4% versus range of 6.8% to 16.1% for the other cities, p<0.001), and the lowest rates of prior usage of PCP prophylaxis (41.2% versus range of 44.8% to 60.6% for the other cities, p<0.001). Diagnostic bronchoscopy rates in New York were lowest (11.6% versus range of 13.4% to 41.8% for the other cities, p<0.001), but isolation for tuberculosis was second highest (64.3% versus range of 7.5% to 65.3% for the other cities, p<0.001). Overall, in-hospital mortality did not vary across cities (range of 7.5% to 14.9%, p=0.664), while length of stay in New York was longest (15.2 versus range of 6.6 days to 9.8 days for the other cities, p<0.001) and discharge against medical advice rates were highest among injection drug users (24.5% versus range of 0.0% to 20.0% for the other cities, p=0.01).
Conclusions: New York hospitals continue to have a disproportionately large share of the AIDS-inpatient burden, comprising individuals with poor access to care. Among HIV-related PCP patients, New York hospitals had high rates of tuberculosis isolation, low rates of PCP confirmation, high discharge against medical advice rates, and long lengths of stay.
Impact: Opportunities to initiate inpatient drug treatment programs in New York are potentially missed, especially among the large numbers of injection drug users who leave against medical advice.
HSR&D Funded: HSR&D Epidemiology
92. Time Spent by Interns with Hospitalized Patients: Association with Physician Characteristics, Patient Length of Stay, and Patient Satisfaction
Marvin J. Bittner, VA Medical Center, Omaha, NE, EC Rich, VA Medical Center, Omaha, RL Recker, VA Medical Center, Omaha, PD Turner, Creighton University, RL Rospond, Drake University, MW Lubeley, VA Medical Center, Omaha
Objectives: Previous studies found that internal medicine interns spend only about 13% of their time with patients. We sought to learn whether interns who spend more time with their patients are more efficient in using resources and whether they achieve higher patient satisfaction ratings.
Methods: We studied patients of Creighton medical interns assigned to non-intensive care unit inpatient floors at the Omaha VA Medical Center July 6, 1998 through June 30, 1999. We used length of stay (LOS) as a measure of resource utilization. We studied satisfaction with the American Board of Internal Medicine’s Patient Satisfaction Questionnaire (PSQ). This ten-item instrument assesses physicians’ interpersonal skills. Each item has five responses, from 1 (excellent) to 5 (poor). We analyzed responses from patients who completed the questionnaire in a face-to-face interview the morning after admission; however, for patients admitted after 5 pm, the interview occurred the second day after admission. For a week while awake and in the hospital, each intern carried a pager that went off, at random, an average of 3.2 times an hour. Whenever the pager went off, the intern made note of the person with whom he or she was in contact. We also obtained DRG weights and information on these intern factors: census size, number of admissions during the 24-hour admitting period, postgraduate training prior to this internship, faculty ratings of humanistic qualities, and in-training examination scores.
Results: Of the 25 interns, 64% were men; 32% were in the categorical program, 16% in medicine-pediatrics, and 52% in a preliminary year. Four reported previous postgraduate training. Preliminary interns spent less time with patients (5.3% vs. 16.0%, p = 0.009). The fraction of time spent with patients was higher with more experience before starting the internship (r = 0.67, p < 0.001) and better examination scores (r = 0.42, p = 0.036). Among the 90 patients with PSQ data whose interns carried the workload pager the day they were admitted, the mean PSQ was 1.80 (s.d. 0.67). Their mean LOS was 7.63 days (s.d. 10.0). Interns caring for them had a mean of 2.5 admissions (s.d. 1.5) and a census of 5.1 (s.d. 2.1). The mean proportion of contacts with patients was 11.6%. LOS was related to DRG weight (Pearson r = 0.22, p = 0.37) but there was no significant correlation with admissions, census, or proportion of contacts with patients. Neither was there a significant correlation of PSQ with those three variables.
Conclusions: As in previous studies, internal medicine interns devoted less than 15% of their time to patient contact. Preliminary year interns spent less time with patients. There was no correlation of time spent with patients to patient satisfaction or length of stay. Nor were these outcomes related to intern census or admissions.
Impact: The fraction of their time that internal medicine interns spend with patients is small. However, simply increasing that proportion—without training in resource utilization or communication—may be unlikely to reduce length of stay or improve patient satisfaction.
93. Men’s Perceptions of the Outcomes of Early Prostate Cancer: Identifying the Domains of Quality of Life.
Barbara G. Bokhour, Center for Health Quality, Outcomes and Economic Research/Boston University School of Public Health, Bedford, MA, JA Clark, Center for Health Quality, Outcomes and Economic Research/ Boston University School of Public Health, TS Inui, Harvard Medical School, JA Talcott, Harvard Medical School
Objectives: Recent prospective studies, using generic measures, suggest that early prostate cancer and the physical complications of treatment, e.g., urinary, bowel, and sexual dysfunction, have no substantial impact on overall quality of life. However, these generic measures of HRQoL (e.g., SF-36) may fail to capture significant qualities of men's experiences. We sought to derive sensitive measures of prostate cancer-related quality of life from men's accounts of their experiences following diagnosis.
Methods: We conducted 7 focus groups, organized by age (50-69 vs. 70-79) and race (white vs. black), with men treated for early prostate cancer 12-36 months previously in 2 settings (private HMO and Veterans Affairs). Treatments received included radical prostatectomy, external beam radiation, brachytherapy and expectant management ("watchful waiting"). Discussions, led by a medical sociologist and a medical oncologist, were semi-structured to address potential areas of concern following diagnosis and treatment. Transcribed narratives of diagnosis, and the sexual, urinary and other problems men encountered were analyzed qualitatively. Codes summarizing substantive issues raised by participants were developed through consensus amongst four behavioral and medical researchers.
Results: We identified three major domains of quality of life: sexuality, urinary control, and uncertainty. Sexuality encompassed (1) anxiety about physical intimacy and sexual performance (2) uncomfortable awareness of the loss of sexual interest, (3) radically changed grounds for relationships with women, and (4) diminished sense of masculinity. Urinary control included (1) discomfort and embarrassment with leakage, (2) preoccupation with urination and the maintenance of control in public, and (3) feelings of frustration with activity limitations. Uncertainty referred to chronic concerns regarding the status of their cancer (e.g., Was it under control?) and the quality of the treatment they had undergone (e.g., Did it work? Did I make the right choice?).
Conclusions: Our findings provide definition to the effects of treatment, explicate what has been previously glossed as urinary or sexual "bother," and suggest that the outcomes of early prostate cancer include significant impacts on quality of life that have gone unnoticed in previous studies. Moreover, men disclosed irremediable uncertainties about their health and their own actions that began when they were presented with the diagnosis and asked to choose a treatment.
Impact: Incorporating the domains we have identified may significantly improve the assessment of the quality of life of patients with early prostate cancer. Our findings also highlight issues that physicians may wish to address in helping patients to make informed treatment decisions and cope with the outcomes of their choices. Hypothesized scales assessing these qualities of life are being tested in a survey of 500 early prostate cancer patients and 500 age- and race- matched controls.
HSR&D Funded: ECV-97081-1
94. Diabetes Nurse Specialists and a Computerized Patient Tracking System Improve Guideline Adherence
Mark Bradley, MD, VA Roseburg Healthcare System, Roseburg, OR, DE Watson, VA Roseburg Healthcare System
Objectives: To reduce morbidity and mortality associated with diabetes by implementing evidence-based standards of care. The ADA has published standards of practice recommendations since 1987 but these have not been successfully incorporated into clinical practice. Diabetes specialists perform better but most patients see generalists. Implementing behavioral change through traditional routes, such as CME, is marginally successful. We evaluated the effectiveness of diabetes nurse specialists and a computerized patient tracking system in a large primary care setting.
Methods: The VA Roseburg Healthcare System has over 1500 patients with diabetes enrolled in primary care. In 1995 the medical staff adopted the ADA’s Clinical Practice Guidelines. We created a glycemic management algorithm, foot examination template and Diabetes flow sheet. The primary care clinics identified a nurse to champion diabetes management. We added no additional nursing staff. Nurses assessed glucose control and recommended medication changes based on treatment algorithms. They performed foot examinations, ordered necessary labs (e.g. A1c, microalbumin, creatinine, and cholesterol) and consultations (e.g. podiatry and optometry). We searched the VARHS pharmacy database to identify patients receiving diabetes-related medications or glucose testing supplies. We then searched for hospital admissions, HbA1c and urine protein results, and diabetes education and ophthalmology/optometry appointments. We assessed guideline adherence with a scoring system. Each of the six measures was scored on a scale from 0 (good adherence with guideline) to 10 (poor adherence). The diabetes nurses received computer search reports quarterly. They contacted patients with total scores >30 (High-Risk) to initiate necessary care. They also saw patients referred for diabetes management.
Results: We initiated the above changes in 1st quarter of fiscal year 1995 (FY95Q1). The FY95Q1 vs. FY99Q4 data are: HbA1c within last year (70 vs. 88%), HbA1c < 10% (56 vs. 92%), diabetes education within 4 years (30 vs. 67%), and eye examination within last year (21 vs. 59%). Each parameter has shown progressive improvement. One of the 4 primary care teams dedicated significantly less nursing resources to diabetes management. This team also showed the least improvement. No relationship between adherence to the guidelines and rate of hospitalization could be identified.
Conclusions: We believe that our diabetes nurse specialists and the computerized tracking system are responsible for this improvement. Since this was not a randomized-controlled trial other factors may have contributed to our results. With the computerized tracking system, nurses focused on those patients in most need. Based on our success, we are adding blood pressure, foot examination, lipids, and ASA use to our tracking system. We have also added real time guideline reminders to our electronic medical record system.
Impact: VARHS patients with diabetes now have better glucose control and are more likely to receive preventive screening. The limited improvement by one of the primary care teams supports the importance of active nurse case management. Our results are consistent with other trials of nurse case management. However, we have applied this to a larger population with the aid of a computerized tracking system.
95. Binge Drinking Among Female VA Patients
Katharine A. Bradley, MD, MPH, HSR&D Center of Excellence, Seattle, Washington, KR Bush, HSR&D Center of Excellence, TM Davis, HSR&D Center of Excellence, DJ Dobie, HSR&D Center of Excellence, CM Rutter, HSR&D Center of Excellence, DR Kivlahan, HSR&D Center of Excellence
Objectives: Two studies, one of college students and one of past-problem drinkers, suggest that women who drink 4 or more drinks on an occasion risk developing alcohol-related problems. No study has evaluated whether the same relationship exists in female patient populations. The purpose of this report is to evaluate the relationship between drinking 4 or more drinks on an occasion, often referred to as binge drinking, and health in female VA patients.
Methods: In April 1998, we surveyed all female veterans who had received care at VA Puget Sound Health Care System between October 1996 and January 1998. The mailed survey included questions about medical, gynecologic, and preventive health history, and current psychiatric symptoms and health-related quality of life (SF-36). Two questions asked about typical quantity and frequency of drinking, and another question asked about the frequency of drinking 4 or more drinks on an occasion in the past year (binge drinking). A validated screening questionnaire for problem drinking (TWEAK questionnaire) was also included. We used age-adjusted logistic regression to evaluate the odds of alcohol-related problems, other high-risk behaviors, and health among women who reported binge drinking monthly or more often.
Results: Of 1,935 surveyed women, 1,259 (65%) returned completed questionnaires; 1,177 women who responded to all three alcohol consumption questions are included in these analyses. The majority of respondents (72.9%) are White, with a mean age of 46. Forty-eight percent are married or widowed, 32.5% are divorced or separated, and 83.6% have attended college. Among 759 (64.5%) women who report any drinking in the past year, only 46 women (3.9%) report typically drinking over 10 drinks a week, but 292 (24.8%) report binge drinking; 215 (18.3%) report binge drinking less than monthly, and 77 (6.5%) report binge drinking monthly or more often. Compared to women who never binge, women who report binge drinking monthly or more often are at increased risk for: smoking (OR 3.3; 95% CI 2.0-5.4), using drugs more than intended (OR 19.6; 95% CI 8.3-46.3); having more than one sexual partner in the past year (OR 4.0 95% CI 2.1-7.7), and reporting > 3 injuries in the past 3 months (OR 2.2; 95% CI 1.3-3.9). Report of binge drinking at least monthly is also associated with increased reports of: close friends or relatives being worried about patient’s drinking (OR 33.1; 95% CI: 16.0-68.7); arguments after drinking (OR 12.6; 95% CI: 6.5-24.4); morning drinking (OR 30.6; 95% CI 9.7-96.2); and blackouts after drinking (OR 5.0; 95% CI 3.0-8.2).
Impact: These data support findings from several other non-patient populations that women who drink 4 or more drinks on an occasion at least monthly are at increased risk for alcohol-related problems and other health risks. This threshold for female binge drinking can inform alcohol screening and preventive counseling among female veterans.
HSR&D Funded: GEN-97-002-1
96. The Cost of Healthcare for Homeless Veterans at the Atlanta VA Medical Center
Kathleen A. S. Cannella, Atlanta VAMC and Emory University, Decatur, GA, FM Cordi, Atlanta VAMC and Emory University, JS Noblin, Atlanta VAMC, JL Stephenson, Atlanta VAMC, SM Aguayo, Atlanta VAMC & Emory and Morehouse University
Objectives: Of the estimated 250,000 homeless men, women, and children living on the streets of our nation, one-third of the adult homeless population and 40% of homeless males are military veterans. In the Atlanta Metro Area alone, an estimated 6,000-8,000 veterans are homeless on any given night. It is also estimated that homeless veterans in Atlanta may cost the public at least $78,000,000 a year – that’s about $11,000 per year, per veteran! Mental illness and substance abuse have been reported as major causes of homelessness among veterans. Our study began with three questions: 1. What is the resource consumption at the Atlanta VA for homeless veterans (HV)? 2. How do these costs compare to overall cost estimates at the Atlanta VA? 3. What diagnoses were frequently associated with admissions of homeless veterans?
Methods: Using Decision Support Service (DSS) software from Transition Systems, Inc., we obtained data on outpatient and inpatient encounters for the 12-month period prior to each veteran’s initial homeless program visit. Data for these initial visits were coded in DSS beginning with fiscal year (FY) 1998. We knew these veterans were homeless at the time of their initial visit and were sheltered after this visit. We assumed that they also might have been homeless for at least some period of time during the preceding year, so for each veteran we obtained data for the 365-day period prior to their initial visit. For this study we compared the data reported on these homeless veterans with local Atlanta VA data for the period 10/97 through 6/99 (FY1998 and three-fourths FY1999).
Results: For this 21-month time period, the Atlanta VA spent $6,066,003 on 62,226 outpatient encounters and 354 admissions for homeless veterans. These outpatient encounters accounted for 3.9 percent of Atlanta VAMC total outpatient encounters and 2.5 percent of the total outpatient costs. The admissions accounted for 3.2 percent of Atlanta VAMC total admissions and 1.9 percent of the total inpatient costs. Their average length of stay was 8.6 days. At our VA, 60 percent of homeless veterans’ admissions involved mental illness (such as psychoses and neuroses) or substance abuse (alcohol and drugs), as did the majority of total admissions to the medical center overall in both 1997 and 1998. The average cost inpatient and outpatient care per HV was $5115, whereas the average cost per individual veteran overall was $4767. Acknowledged limitations of our data include inappropriate coding, variable and unknown duration of homelessness immediately preceding each veteran’s initial visit, and the variability in time periods for homeless veterans’ data.
Conclusions: Per veteran, homelessness was associated with a higher than average resource consumption at the Atlanta VAMC. Mental illness and substance abuse were the most frequent admission diagnoses for our homeless veterans. This parallels overall Atlanta VAMC admissions and reports in the literature.
Impact: Future studies will examine these costs in more detail, as well as the effectiveness of interventions designed to increase the likelihood of successful rehabilitation for homeless veterans.
97. Screening for Problem Drinking among Female VA Patients Using the TWEAK
Tania M. Davis, PhD, VA Puget Sound Health Care System, Seattle, WA, KR Bush, VA Puget Sound Health Care System, DR Kivlahan, VA Puget Sound Health Care System, KA Bradley, VA Puget Sound Health Care System
Objectives: Male VA patients have high rates of at-risk and problem drinking, but little research has focused on the drinking habits and related problems of female VA patients. The purpose of this study is to estimate prevalence rates among female VA patients of problem drinking and related patterns of psychiatric and medical comorbidity.
Methods: Women were eligible for participation if they received health care from any VA Puget Sound Health Care System facility between October 1, 1996 and January 1, 1998, had accurate addresses, lived in the region, and were healthy enough to complete the survey. The 16-page mailed survey included items regarding personal and family health history, preventive health screening, health behaviors, health-related quality of life (SF-36), and patient satisfaction. The Patient Problem Questionnaire (a recent version of the PRIME-MD mental health screening questionnaire), and the Post Traumatic Stress Disorder Checklist were also included in the survey. Alcohol-related questions included three screening questions regarding heavy drinking from the Alcohol Use Disorders Identification Test and a validated screening questionnaire for drinking-related problems in women (TWEAK).
Results: Of 1,935 eligible female VA patients, 1,259 returned surveys (65% response rate). Respondents were primarily Caucasian (73%), middle-aged (mean = 47, range 22-94), and college-educated (83%). Among the 1,259 respondents, 781 (62%) reported drinking any alcohol in the previous year. Of these 781 drinkers, 51 (7%) reported drinking over 10 drinks a week on average, and 304 (39%) reported drinking four or more drinks on any occasion in the past year. Of all respondents, 268 (21%) screened positive for problem drinking (TWEAK score > 2). Most of these women (203; 76%) were under 50 years old. Of the 226 women who screened positive on the TWEAK and reported drinking alcohol in the past year, only 51 (23%) reported prior involvement in alcohol treatment or Alcoholics Anonymous. Many TWEAK positive women also screened positive for other mental health problems, including anxiety (28%), PTSD (28%), previous drug problems (28%), and depression (38%). Frequently reported health issues among TWEAK positive women included high blood pressure in the past year (21%), pain in three or more body regions (26%), more than three injuries in the past 3 months (28%), and inadequate food or money to buy food in the past year (36%).
Conclusions: Over 20% of female veteran respondents screened positive for problem drinking and many of them reported mental health and medical issues. Younger women are at highest risk, with 76% of women under 50 who report drinking alcohol in the past year screening positive for problem drinking and less than a quarter of them having received alcohol treatment or attended AA.
Impact: The prevalence of problem drinking in female veterans is substantial, and women should be screened routinely for alcohol problems. Many of those women who screen positive for alcohol problems report other mental health and medical problems. Low rates of reported alcohol treatment warrant further study of barriers to existing treatment services and need for alternative treatment approaches.
HSR&D Funded: GEN 97-022
98. Evaluating Competing Algorithims for Identifying VA Diabetics Using Computerized Medical Information System Data
Jennifer Davis, VA Center for Practice Management and Outcomes Research, Ann Arbor, MI, SL Krein, VA Center for Practice Management and Outcomes Research, RA Hayward, VA Center for Practice Management and Outcomes Research
Objectives: Accurately and efficiently identifying patients with diabetes within a health system is critical to monitoring health care processes and outcomes and ultimately improving performance. The purpose of this project was to evaluate the accuracy of different approaches for identifying patients with diabetes in VA using existing computerized data.
Methods: FY98 Pharmacy and laboratory data were obtained, in collaberation with the National Center for Cost Containment (NCCC) from computerized files at seven medical centers comprising one VISN. Diagnosis data were obtained from national VA databases. The algorithims compared inclusion criteria of: 1) patients who had one or more prescriptions for diabetes medications or home glucose monitoring supplies; 2) patients who had a hemoglobin A1c (HbA1c) result > 6.7; 3) patients who had an inpatient or outpatient diabetes diagnostic code (ICD9 codes 250 - 250.99, 357.2x, 362.0x, 366.41); and, 4) combinations of the above. Chart reviews were done for a sample of patients at one medical center to verify diagnosis, and assess the accuracy of prescription and lab data.
Results: The prescription, HbA1c, and diagnostic criteria identified 19,104, 12,208, and 22,066 diabetics respectively. Adding HbA1c, diagnostic or both criteria to the prescription criteria increased the number of diabetics identified by 5%, 22%, and 24%, respectively, over prescriptions alone. Preliminary chart review (n=251) results indicate that the prescription only criteria has a very small false positive rate (less than 2% have a questionable diagnosis of diabetes). However, the prescription criteria alone appears to miss some mild diabetics. Using diagnostic codes to detect diabetics picked up many patients who have glucose intolerance, without true diabetes, especially if only one outpatient diagnosis was used. Using diagnosis only also identified patients who are newly diagnosed or new to the VA system. Although those identified by the HbA1c criteria alone were virtually always true diabetics (>98%), adding the HbA1c criteria did not contribute many additional diabetics to the dataset over using one or more prescriptions or 2 or more diagnostic codes.
Conclusions: The absolute number of patients identified as having diabetes varied up to 24% based upon the identification strategy used. However, the best criteria may depend on the purpose for identifying the cohort. For example, to identify patients at risk for cardiovascular problems, using diagnostic codes may suffice, since mild diabetics and patients with glucose intolerance are also at risk for these problems. However, if you are interested in microvascular complications such as blindness and amputation rates, using prescriptions and HbA1c criteria might be preferred, since this method appears to identify only true diabetics. Based on this study, the best compromise criteria is to use a combination of one or more prescriptions or 1 or more diagnoses, and at least one outpatient visit during the past year.
Impact: The approach used to identify diabetics should be matched to the purpose for identifying the cohort. After validation of these results at additional facilities, this information should help researchers and administrators produce more complete and valid estimates of performance and outcomes measures for patients with diabetes.
HSR&D Funded: DIB 98-001
99. Effects of Nurse Counseling for Primary Care Patients: Adherence to a Home-Based Walking Program
Patricia M. Dubbert, PhD, VA Medical Center, Jackson, MS and Univ MS Sch Med, Jackson, MS, Deborah Bilbrew, VA Medical Center, Jackson, MS, Karen Kerr, VA Medical Center, Jackson, MS, KM Cooper, VA Medical Center, Jackson, MS, K Kirchner, VA Medical Center Jackson MS and Univ MS Sch Med, EF Meydrech, Univ MS Sch Med
Objectives: Surveys suggest elders prefer home-based exercise and advice from health care providers to assist them in increasing physical activity. This study examined adherence to a prescribed home-based walking program among patients receiving different levels of nurse counseling.
Methods: 60-80 year old patients with chronic but stable health conditions who could increase walking were referred by VA primary care providers. After patients received theory-based individualized counseling from the research nurse with a goal of walking at least 20 minutes 3 times a week, they were randomized to receive brief personal phone calls from the nurse (PPC), personal phone calls alternating with automated calls (APC), or no further contact with the nurse (NC). Patients kept weekly walking diaries which were mailed to the data collector. Over the next 6 months, days per week walking were counted from the diaries with missing data interpreted as not walking. 3 months after randomization, significant others identified by patients at baseline were contacted to validate walking. Data from Tritrac activity monitors worn during the week prior to 6 mo followup were also examined to validate walking episodes.
Results: The 212 patients randomized were: mean age 68.5 (SD 4.7) years of age, 29% minority, 80% married, 56% rural, and most had >= 2 medical conditions including arthritis, obesity, hypertension, diabetes, heart disease, and lung disease. The percent of patients walking >= 20 minutes at least 1 day per week declined gradually from 75% for PPC and APC and 60% for NC at baseline to 58% for APC, 45% for PPC, and 36% for NC at week 25. The percent walking >=3 days per week was stable in all groups across time, with 40-50% in APC and 20-30% in PPC and NC. The percent walking >= 5 days a week was also stable with 25-30% in APC and 7-15% in PPC and NC. Significant others validated walking for 83% in PPC, 91% in APC, and 71% in NC groups. Tritrac activity monitor samples clearly validated a sample of 60% of concurrent diary episodes.
Conclusions: Most patients in all 3 intervention groups initiated home-based walking at least 1 day per week following nurse counseling and > 30% began walking at least 3 days a week. Patients receiving both personal and automated calls from the nurse walked more frequently than those in the other groups.
Impact: Regular moderate activity such as walking can decrease risk of functional impairment and improve quality of life in older adults. Nurse counseling, including automated delivery of motivational messages, appears to be an effective method of initiating a walking program in elderly patients.
HSR&D Funded: NRI 95-022
|| | | | ||