2000 HSR&D Annual Meeting: Abstract 100-149
100. Performance of the Mail Administered Stroke Impact Scale Compared to Conventional Measurements of Stroke Outcomes
Pamela W. Duncan PhD, PT, VAMC, Kansas City, MO and Center on Aging, Kansas University Medical Center, Kansas City, MO, DM Reker, VAMC, Kansas City, MO and Center on Aging, Kansas University Medical Center, Kansas City, KS, B LaClair, VAMC, Kansas City, MO, BB Hamilton, Center on Aging, Duke University Medical Center, Durham, NC, RD Horner, ERIC, VAMC, Durham, NC and Dept of Medicine, Duke University Medical Center, Durham, NC, TK Dudley, VAMC, Durham, NC
Objectives: Our study objectives were to 1) assess mail response rate of a new stroke specific outcome measure, the Stroke Impact Scale (SIS), 2) compare the survey responders to the non-responders, and 3) to compare conventional six month post-stroke outcomes (Functional Independence Measure (FIM) and the SF-36) with the SIS.
Methods: One hundred sixty-six patients enrolled in a VA prospective cohort study of post-stroke rehabilitation outcomes were mailed the SIS after a 6-month telephone interview. Telephone assessments included: FIM (maximum motor score = 91 and maximum cognitive =35) and the SF-36 (maximum physical score =100). The SIS was mailed within two weeks of the telephone interview followed by a reminder letter. If patients could not complete the SIS, we accepted proxy responses. The SIS is a new 64-item stroke outcome measure designed to capture a broader range of physical function as well as to assess communication, emotion, memory, and social participation. Within each of the 8 SIS domains, standardized scores range from 0 to 100. SIS responders were compared to non-responders with chi-square and t-tests.
Results: The SIS response rate was 67%. Among the responders, 51% of the questionnaires were completed by proxy respondents. Proxies were more likely to complete the survey if patients were less educated, did not live alone, had a hemorhagic stroke and were cognitively impaired. Missing responses for all SIS items were less than 5%. Prior to stroke, patients were independent in ADL. Patient demographic descriptors (mean, (sd), p) for the respective SIS responders and non-responders were: age 68.2(9.8) vs 65.7(10.2) p<.13; race (% white) 72% vs 65% p<.47; married 57% vs 51% p<.51, and prior physical function (SF-36 physical scale) 66.8(29.4) vs 62.9(27.6) p<.4. Conventional 6-month outcome indicators (mean, (sd), p) for SIS responders vs. non-responders were: FIM motor scores: 78.9(12.7) vs 70.7(21.7) p<0.01; FIM cognitive/social: 32.5(4) vs 29.9(7.1) p<.01; and SF-36 physical function: 33.9(25.7) vs 28.8(27.1) p<0.25. Functional status as measured by the 8 domains of the Stroke Impact Scale (responders) was: Strength 50.7(23.1), Mobility 52.8(23.9), ADL/IADL 64.3(25), Hand function 48.7(29.8), Combined physical 54.7(22.3), Communication 74.8(22.2), Emotion 59.9(18.9), Memory 68.4(26.1), Social Participation 52.7 (25.0). Graphical displays of the FIM motor scores and SF-36 physical scores reveals that FIM scores exhibit a ceiling effect (68% of scores between 80 and 91) and SF-36 physical scores display a floor effect (62% of scores between 0 and 30). SIS scores were normally distributed with no evidence of floor or ceiling effects.
Conclusions: Mailed administration of the SIS yielded good response rates and little missing data. Responders were less impaired functionally than non-responders. Comparison of scores across the FIM, SF-36 and the SIS reveals that the SIS captures a broader range of functional decline with residual impairments in memory, communication, and social participation. SIS scores exhibited no floor or ceiling effects.
Impact: Among patients who are less functionally impaired, the SIS captures a broader range of post-stroke sequelae than conventional outcome measures and may be administered by mail.
HSR&D Funded: ACC97-1141
101. Prevalence of and Risk Factors for 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: Despite limited evidence, advocacy groups recommend screening all adults over the age of 45 for unrecognized diabetes mellitus. Our objectives were to determine the prevalence of unrecognized diabetes among VA Medical Center outpatients, and to identify associated risk factors that may define the optimal target population for diabetes screening.
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 enrolled subjects at the time of an outpatient visit to the Medical Center. We obtained demographics, a structured medical history, height and weight for 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. Logistic regression was performed to determine independent factors associated with undiagnosed diabetes and adjusted odd ratios.
Results: We enrolled 1253 subjects. The prevalence of unrecognized diabetes was 4.5% (95% CI, 3.4, 5.7). Independent risk factors for diabetes identified by logistic regression were: self-reported hypertension (adjusted OR=2.5, p=0.004); obesity as defined by weight more than 20% above ideal body weight (adjusted OR=2.2, p= 0.02); and self-reported family history of diabetes (adjusted OR=1.7, p=0.06). Non-Caucasian race, a known risk factor for diabetes, was not associated with the new diagnosis of diabetes at screening in our sample. Considering obesity, family history, and hypertension as 3 risk factors for unrecognized diabetes, the prevalence of unrecognized diabetes varied by the number of risk factors for each patient. For patients with 0 risk factors, the prevalence of unrecognized diabetes was 1/183 (0.5%, 95% CI 0.0%-3.5%); for patients with 1 risk factor, the prevalence of unrecognized diabetes was 14/418 (3.3%, 1.9%-5.7%); for patients with 2 risk factors, 24/461 (5.2%, 3.4%-7.8%), and for patients with 3 risk factors, 17/171 (9.9%, 6.1%-15.6%).
Conclusions: Assessing risk factors prior to blood testing will improve the specificity, with little cost in sensitivity, of screening for diabetes in a medical center setting. The target population which optimizes the potential value of diabetes screening consists of patients with at least 1 of the above 3 risk factors for diabetes (obesity, hypertension, family history). Hypertension is strongly associated with unrecognized diabetes in veterans.
Impact: VA and other health care providers considering whether to perform systematic screening for diabetes should use known risk factors to identify an appropriate target population for screening.
HSR&D Funded: CSP 705D
102. A Follow-up Investigation of the Effectiveness and Efficiency of GEM Care
Joseph B. Engelhardt, PhD, VA Healthcare Network Upstate New York, Albany, New York, J Gao, VA Healthcare Network Upstate New York, R Toseland, State University New York at Albany
Objectives: This study examined the impact of Geriatric Evaluation and Management (GEM) and usual primary care (UPC) on the survival, health care utilization and health care cost of a sample of 160 frail, high use elderly outpatients, over 48 months following random assignment to GEM or UPC.
Methods: This randomized controlled study was conducted by utilizing a sample of patients recruited in a previously funded study, "The Effectiveness and Efficiency of Outpatient Geriatric Evaluation and Management". In that study, assessments at baseline, 8, 16 and 24 months were made on measures of: health and functional status, psychosocial well-being, quality of health and social care, utilization of inpatient and outpatient services, and cost of care. These results revealed that GEM was significantly more effective than UPC in increasing patient satisfaction, and improving the quality of the health and social care. GEM was not effective in reducing health care utilization or cost of care for the entire study period, but it was effective in reducing utilization and the cost of care during the sixteen to twenty-four month study period. If these utilization and cost trends were to continue, cost savings for GEM patients would begin to occur at 26 months, or about 2 months after the final data collection. This study investigated if an initial investment in GEM was recouped by subsequent lower costs, and whether there is an overall cost savings by 48 months. VAMC records and data from the Austin Automation Center were examined to collect data about survival, utilization, and cost variables. Outpatient utilization variables include: UPC or GEM clinic visits, medical clinic visits, surgical clinic visits, emergency room visits, and total clinic visits. Inpatient utilization data include: total number of admissions, hospital bed days of care, nursing home admissions, and nursing home bed days of care. Health care cost variables include: total outpatient cost, total inpatient cost, total institutional cost i.e. hospital plus nursing home costs, and total health care cost. GEM and UPC clinics continued to serve the original study patients. Survival data were available for all 160 patients enrolled in the original study. Random regression models allowed us to include data on utilization and cost from all subjects, including partial data from those who did not survive.
Results: Total cost for GEM patients was, on average, $1,154 less than for UPC patients across six, 8-month measurement periods (p=0.0001). Initially higher GEM total costs at 8 and 16-month measurement periods were recouped by 24 months. Over the first three 8-month periods GEM patients, on average, cost $ 1,451 (p=0.0001) less than the UPC patients. Lower GEM costs were generated by less inpatient utilization when compared to UPC. There was no significant difference in survival at 48 months.
Conclusions: Data suggest that an initial investment in outpatient GEM can be recouped by 24 months and significantly lower GEM costs can be sustained over the long-term (i.e. 48 months) where savings from inpatient utilization can be realized.
HSR&D Funded: IIR 95-108
103. Evaluating Screening Criteria for Adverse Outcomes in Primary Care
Ron L. Evans, University of Washington, Seattle, WA
Objectives: The purpose of the current study was to identify variables near hospital admission that could effectively discriminate patients at risk for nursing home placement, lengthy hospital stay, or readmission. The specific goals were to determine the sensitivity and specificity of differing screening strategies in predicting adverse outcomes. We also evaluated the screening criteria, including those available through hospital billing and resource usage data, to determine if a subset of generic screens might efficiently identify outcome.
Methods: Risk criteria reported in the literature were used to predict discharge destination and duration of care for 1,332 admissions. Factors that discriminated outcome included: comorbidity, mental status, living arrangement, transfer to a special care unit, prior admission within 1 year, iatrogenic trauma, and pending litigation.
Results: Sensitivity and specificity of individual screens varied widely. Prior hospitalization was the most sensitive [64%] but least specific [47%]. Transfer to a special care unit was specific [91%] but not sensitive [11%]. Transfer to another hospital was intermediate [sensitivity 32%, specificity 71%]. Combinations of screens were compared, including some using only resource usage data. The most sensitive strategy using billing data detected 44% of adverse outcomes and cost only $4 per admission [$51 per adverse event], versus $14 per admission [$94 per event] when all records were reviewed.
Conclusions: We concluded that use of cumulative risk scores can result in accurate prediction of hospital outcome, which may be useful in targeting patients for intervention.
Impact: Using screens available through billing and resource use data, although insensitive, would be the most cost effective strategy.
104. Sildenafil for Male Erectile Dysfunction: A Systematic Review and Meta-Analysis
Howard A. Fink, MD, MPH, Geriatric Research Education & Clinical Center, Section of eneral Internal Medicine, Minneapolis, MN, TJ Wilt, MD, MPH, Staff physician, Section of General Internal Medicine; Investigator, Center for Chronic Disease Outcomes Research, R MacDonald, MS, Coordinator, VA Coordinating Center of the Cochrane Review Group, IR RUTKS, B.S., Trials Search Coordinator, VA Coordinating Center of the Cochrane
Objectives: Erectile dysfunction (ED) is defined as the persistent "inability to achieve or maintain an erection sufficient for satisfactory sexual performance." (NIH. JAMA 1993). Sildenafil has generated tremendous interest regarding its utilization for treatment of ED. This systematic review of available evidence from randomized clinical trials (RCT) on the efficacy and safety of sildenafil for treatment of ED will facilitate its evidence-based use.
Methods: We performed a meta-analysis of RCTs, both published and unpublished, from January 1995 through September 1999. Trials were eligible if they: (1) were randomized; (2) included men with ED; (3) compared sildenafil with placebo or active control; (4) were at least 7 days in duration; and (5) assessed clinical outcomes related to ED. Main outcome measures were percent of successful intercourse attempts, percent of subjects achieving successful intercourse during treatment, percent of subjects reporting improvement in erectile function, responses to questions 3 & 4 of the International Index of Erectile Function (IIEF Q3 and Q4) and percent of subjects reporting adverse events.
Results: Seventeen trials involving 4110 subjects met eligibility criteria for the review. Men enrolled in the trials had a mean age of 56 years and a mean ED duration of 5 years. Trial treatment duration was 1-26 weeks. Successful sexual intercourse was achieved in 46.5% of all attempts in men randomized to sildenafil versus 19.5% for men allocated to placebo (RBI=2.94, 95%CI=2.26,3.82). Of men receiving sildenafil, 81.7% had at least one successful sexual intercourse attempt during treatment compared with 52.4% of those allocated to placebo (RBI=1.76, 95%CI=1.44,2.17). Of men randomized to sildenafil, 68.9% reported improvement in erectile function compared with 18.5% of men randomized to placebo (RBI=3.75, 95%CI=3.02,4.66). Sildenafil also was superior to placebo based on subjects’ responses to IIEF Q3 and Q4. The most common adverse events in men randomized to sildenafil were headache (18.3%), flushing (14.1%), dyspepsia (6.8%) and visual disturbances (3.9%); all occurred significantly more frequently than in men receiving placebo. Men randomized to sildenafil were significantly less likely than those allocated to placebo to drop out of trials overall (8.8% for sildenafil versus 14.6% for placebo; RRI=0.59, 95%CI=0.50,0.71) and no more likely to drop out secondary to an adverse event. In data from 12 of the trials, 5 men died on or within 30 days of study drug or withdrew from studies because of death. Four of the men had been randomized to sildenafil.
Conclusions: This meta-analysis indicates that sildenafil improves erectile function and is generally well tolerated compared with placebo. Extensive exclusion criteria within trials and incomplete data for treatment in many subgroups of men may limit generalizability of findings regarding the efficacy and safety of sildenafil treatment.
Impact: The possible association of sildenafil with mortality in men with erectile dysfunction warrants further exploration.
HSR&D Funded: Supported in part by a grant from the Technology Assessment Program of the Management Decision and Research Center, Health Services Research and Development Service of the Veterans Health Administration’s Office of Research and Development. Additional support provided by the Center for Chronic Disease Outcomes Research, Veterans Affairs Medical Center, Minneapolis.
105. A Nurse Managed Chronic Wound Telehealth Clinic
Sue Gardner, PhD, CWCN, Iowa City VA Medical Center, Iowa City, Iowa, RA Frantz, University of Iowa College of Nursing, KA Buresh, Iowa Veterans Home, JP Specht, Iowa City VA Medical Center, BJ Wakefield, Iowa City VA Medical Center, JR Flanagan, University of Iowa College of Medicine
Objectives: The purpose of this project was to pilot test a nurse managed chronic wound consultation clinic for a long-term care (LTC) population using an interactive video telehealth application. Data were collected on wound assessment accuracy, patient and nurse satisfaction, and costs.
Methods: Data were collected from patients, 7 primary nurses and one skin care nurse at the at the LTC facility, and from two consultant nurses providing telehealth wound assessment. Analyses compared on-site and telehealth assessments for 13 individual wound consultations. Satisfaction questionnaires were completed by patients and nurses following each consultation session. A cost-minimization analysis was conducted comparing in-person consultation to the telehealth consultation.
Results: Inter-rater reliability for nine different wound characteristics ranged from 54 to 100%. The costs of telehealth wound consultations were estimated to be $92.80 for each 20 minute consultation. The LTC patients found the telehealth consultation to be as good as the in-person assessment, although patients noted difficulty hearing and seeing the telehealth consultant. Nurses were equally satisfied with both the telehealth and in-person consultations and felt both consultation modes were a productive use of their time and skills. During implementation of the telehealth clinic, changes were made which may have impacted final outcomes. For example, upon finding that some long term care residents had difficulty in hearing and seeing the consultant, a larger monitor was installed and the patient was positioned closer to the telemedicine unit. Anecdotal evidence indicated that as participants learned how to use the equipment over time, their satisfaction with the telehealth clinic improved. For example, the consultant’s communication style improved and became more active and animated with larger gestures, based on feedback from the distant site.
Conclusions: Patients and nurses perceived the telehealth clinic as a convenient approach to providing access to specialized care. The advantages of implementing the telehealth chronic wound clinic included reducing travel costs and time for the consultant, and more timely follow-up for the LTC patients. Several facilitating factors and barriers to implementation of telehealth applications were identified during the course of this pilot test.
Impact: Based on this evaluation, the Chronic Wound Clinic is now an established clinic at the ICVAMC.
HSR&D Funded: DEV 97-012
106. Predictors of Pulmonary Function in Spinal Cord Injury
Eric Garshick, Medical Service, VA Boston Healthcare System, Harvard Medical School, West Roxbury, MA, EK Larkin, Research Service, VA Boston Healthcare System, CG Tun, Rehabilitation Medicine Service, VA Boston Healthcare System, Harvard Medical School, SL Lieberman, Medical Service, VA Boston Healthcare System, Harvard Medical School, R Brown, Medical Service, VA Boston Healthcare System, Harvard Medical School
Objectives: Spinal cord injury (SCI) results in reduced pulmonary function and excess morbidity and mortality from respiratory causes. The extent that neurologic level and completeness of SCI as well as factors other than SCI contribute to reduced pulmonary function is unknown. The purpose of this study was to assess predictors of pulmonary function in a cohort of subjects with SCI recruited from a VAMC SCI Service and the community.
Methods: Between 10/94 and 8/98, 295 white male subjects (>1 yr since SCI) completed a health questionnaire and had a neurologic exam and pulmonary function tests (spirometry, lung volumes by helium dilution, and maximal inspiratory pressure (MIP) as an index of inspiratory muscle strength). Analysis of variance/covariance was used to assess significant independent predictors of forced expiratory volume in 1 second (FEV1), forced vital capacity (FVC), and total lung capacity (TLC).
Results: Mean age(+sd) was 50.7(14.6) and mean years since injury was 18.0(12.8). Twenty-eight percent were current cigarette smokers, 23% were former smokers (but smoked following injury), 17% quit at or before injury, and 32% were never smokers. The final prediction model included terms for SCI level and completeness, age, stature, weight, years since injury, cigarette smoking (current, pack-years), MIP, a history of chest injury or operation, asthma, and a recent respiratory illness. The decrement in FEV1 and FVC attributable to aging was 23 and 19 ml/yr respectively, similar to values expected in the able-bodied. There was an additional effect of years since injury, with a mean yearly decrement in FEV1 of 8 ml and FVC of 9 ml, one-third to one-half of the decrement in lung function attributable to aging. Adjusting for lifetime cigarette consumption (pack-years), current smokers had an average FEV1 203 ml (95% lower limit of 378 ml) and an average FVC 184 ml (95% lower limit of 397 ml) less than a former or never smoker. This substantial decrement in lung function attributable to active smoking was equivalent to 8%-18% of the mean FEV1 and 6%-14% of the mean FVC of subjects with tetraplegia. For FVC, the lower 95% bound attributable to tetraplegia alone was 48% to 60% predicted, with the greatest abnormality occurring in those with the highest injury level. Subjects with the greatest MIP had the greatest FEV1, FVC, and TLC, and the effect of MIP was greatest in tetraplegia.
Conclusions: Years since injury, smoking, and MIP were independent predictors of pulmonary function separate from the effects of age, neurologic level, and completeness of SCI. Pulmonary function values below the 95% limits for FVC are not likely to be due to tetraplegia alone and should prompt a search for correctable factors.
Impact: This study suggests that, in addition to the effects of aging, there is an accelerated loss of lung function in SCI following injury. The prediction model suggests that in SCI smoking cessation and inspiratory muscle training could lead to improvements in lung function. Additional longitudinal study is required to confirm these cross-sectional findings and assess whether inspiratory muscle training leads to sustained improvement in pulmonary function in SCI.
HSR&D Funded: IIR-93-155
107. Racial Variation in Cancer Patients' Health Seeking Behaviors and Acceptance of Care
Howard S. Gordon, MD, Houston VAMC, Houston, TX, DA Paterniti, Houston VAMC
Objectives: Prior studies have described variation in treatment and outcome by race for patients with cancer. Efforts to explain these variations by differences in patient characteristics have been largely unsuccessful. One of several alternative hypotheses to explain variation by race is that patients' patterns for seeking care and/or patients' acceptance of physicians' recommendation for treatment could differ by race.
Methods: We undertook a qualitative study using focus group interviews to assess the determinants of patients' decisions. We used the Health Belief Model (HBM) as the conceptual framework to build a list of guiding questions for the focus groups. Guiding questions emphasized patients' perceptions of the four principal components of the HBM (susceptibility, severity, barriers, and benefits). Focus groups were composed of patients recruited from physicians' offices and hospitals in Houston, Texas, with similar cancer diagnoses (breast, prostate, colorectal cancer), self-defined race (black, white), and gender. Eligible patients had a biopsy confirmed diagnosis of breast (female only), prostate, or colorectal cancer within one year of recruitment. All focus group interviews were audiotaped and transcribed. Two investigators (Sociologist and Physician) inductively developed categories for coding the transcripts using the conceptual components from the HBM and by noting the patterned occurrence of specific behaviors, attitudes, and values in the transcripts. Analysis was also performed using a qualitative computer program (NUD*IST) to explore relationships among conceptual components derived from the HBM and categories developed by the investigators.
Results: We conducted 8 focus groups with a total of 33 subjects. Focus groups lasted an average of 90 minutes and were structured by questions derived from the HBM. Patients discussed their experiences of health seeking and their process of decision-making for recommended treatment and cancer care. All focus group participants identified factors important to their health seeking behaviors and treatment decisions. Some of these included: feeling ill at the time of health seeking, advice from family members, information from the media, understandings of treatment effects, and knowledge about family history. Many patients felt they had "no choice" regarding treatment and care. Patients who perceived they had treatment decisions believed they had greater agency in the health seeking and treatment process (versus those who viewed themselves as having less agency). Data also indicated variation in susceptibility and barriers to care and treatment by race. Patients' perceived susceptibility included knowledge of their family history (e.g., one black man claimed that his family "didn't talk about what people died from") and perceived agency (i.e., "I can beat this thing"). Perceived barriers to care included concern about insurance and feelings of distrust regarding medical personnel.
Conclusions: Perceived severity of illness may be rooted in racial/ethnic factors. Further understanding of factors that lead to health seeking behaviors and factors that may influence perceptions about care and treatment decisions will assist health services researchers in making sense of the multi-factorial reasons for variation in treatment and outcome by race for patients with cancer.
Impact: Variation in cancer patients' willingness to seek care by race is impacted by patients' perceived severity of illness.
108. Medical Illness Predicts Psychiatric Service Utilization and Cost for Elderly Patients with Bipolar Disorder
Hillel Grossman, Providence VAMC; Brown University Department of Psychiatry and Human Behavior, Providence, RI, Heather Richards, BA, Providence VAMC, Abdul Nadeem, MD, Providence VAMC; Brown University Department of Psychiatry and Human Behavior, Mark Bauer, MD, Providence VAMC; Brown University Department of Psychiatry and Human Behavior
Objectives: In numerous studies, medical illness has been shown to be the most powerful predictor of poor outcome for patients with late-life depression. In contrast, there have been no studies of predictors of outcome for late-life Bipolar Disorder. This study aims to determine the influence of medical illness on service utilization and VA cost expenditure for older veterans with bipolar disorder.
Methods: Subjects: Veterans over 55 with a diagnosis of Bipolar Disorder on the Structured Clinical Interview for DSM-IV (SCID) were eligible for the study. 150 potential subjects were identified from hospital computer logs. Measures: Baseline measures included the diagnostic interview; assessment of demographics; review of medical history and quantification of medical illness severity [Cumulative Illness Rating Scale (CIRS)]; determination of Cerebrovascular Rick Factor Score (CVRF) derived from the American Heart Association criteria (AHA, 1990). Subjects also received the Mini-Mental Status Exam (MMSE) as measures of general cognition and The Executive Interview (EXIT), a measure of executive cognitive function. Service utilization was calculated by tabulating mental health and general medical contacts over a two-year period from the medical center computer system (CPRS), supplemented by patient interview for non-VA encounters. Costs were determined using cost distribution report (CDR) methodology to assign dollar values to services rendered over the same 2-year period. Analysis: Stepwise linear regression techniques were used to assess the independent associations of medical illness severity, CVRF, MMSE and EXIT with measures of service utilization including numbers of acute visits, scheduled medical and psychiatric visits, inpatient medical and psychiatric days and VA costs for each encounter over the past two years.
Results: Seventeen subjects meeting eligibility criteria have been studied to date. Mean age is 67.2 with a mean of 14 years of education. CIRS is the only variable to predict mental health service utilization, including the number of psychiatric acute and scheduled visits, number of inpatient days and CDR-derived expenditures for mental health. CIRS also predicts scheduled medical and surgical utilization and costs, but not emergent medical and surgical care. The latter variables are associated with age and inversely with MMSE score.
Conclusions: This pilot study found that medical illness severity correlates, understandably, with utilization of medical services but also with utilization of mental health services and cost expenditures for mental health. This suggests three, not necessarily exclusive, possibilities: - that Bipolar Disorder is a risk factor for medical illness - that medical illness is a risk factor for Bipolar Disorder - that a "third" factor (e.g. vascular disease) might exacerbate both. CVRF in this study did not predict service utilization. Age predicted only acute medical service. General cognition correlated inversely with acute medical visits as well. The EXIT, a measure of executive function, did not correlate with any service utilization parameters. Age of onset of Bipolar Disorder did not predict service utilization.
Impact: The strong association between medical illness and mental health service utilization calls for enhanced attention to medical illness in elder patients with Bipolar Disorder. This might take the form of increased screening for medical illness in mental health clinics and greater collaboration between primary care medical and mental health providers.
HSR&D Funded: DEV97-015
109. VA vs. Private Sector Mergers of Academic Medical Centers: Who's More Successful at Integration?
Marylou Guihan, PhD, MCHSPR, Hines, IL, JD Thomas, MCHSPR, S LaVela, MCHSPR, DM Hynes, MCHSPR, LM Manheim, MCHSPR and Northwestern University
Objectives: In June, 1996, VHA approved the integration of Lakeside and West Side VA Hospitals into VA Chicago Health Care System (VACHCS), as part of a larger restructuring plan. Among the 14 initial VA integrations approved/in planning at this time, this integration was unique because it involved two urban, tertiary care hospitals, with separate academic affiliations, and a common patient population. Objectives included identifying and describing key players and organizational structures and comparing VA and private sector academic medical centers (AMCs) in terms of resources, facilitators and obstacles to integrating, with special attention given to the role of graduate medical education. Specific characteristics examined included merged facility characteristics, board representation, joint recruitment, common information systems.
Methods: Descriptive analyses using qualitative data collected during in-person, semi-structured interviews were conducted with managers and service chiefs at VACHCS, and two private sector AMCs and their medical school affiliates.
Results: VACHCS’s integration was involuntary and was met with great resistance. Integration scope was similar at VACHCS but its goals were more all-encompassing. Private sector mergers resulted in relatively independently functioning hospitals, whereas VACHCS sought to achieve more comprehensive clinical integration across its two divisions. Integration in VACHCS’s administrative and ancillary services was comparable or superior. VACHCS’s integration process and stakeholder involvement was more open and public. VACHCS sought the active involvement of its academic affiliates. Leadership and communication were key factors in overcoming obstacles and in facilitating change at all three mergers. Physician leadership was lacking at VACHCS and it made fewer efforts to engage the physician community. VACHCS had a common compensation system for employees at the merging facilities. However, VA may have missed an opportunity for demonstrating a good faith effort to engage with its employee's integration-related concerns.
Conclusions: One difficulty in comparing mergers was rooted in how differently "integration" was conceptualized within VA vs. the private sector. At the private sector sites, integration was pursued in response to a clearly identified predicament, whereas most VACHCS respondents never understood what problem integration was supposed to be solving. In addition, the private sector sites identified managed care as being among their integration’s major drivers. Once integrated on paper (if not in fact), the private sector sites could declare their goals as having been met. Because VACHCS’s goal was to lower costs and increase efficiency, this required management to reduce duplication and work towards clinical integration. Differences in how integration was defined and pursued made it extremely difficult to assess which sites were furthest along. Finally, we note that these difficulties in finding comparable sites probably suggests something about the difficulties involved in achieving real clinical integration.
Impact: With the ongoing restructuring of VA, as outlined in "Vision for Change," this study provides valuable and applicable information about the structures and processes of merging two urban, tertiary care facilities. Since integrating similar VA facilities is anticipated for the future, identifying and understanding the facilitators and obstacles, and describing the lessons learned in this context will benefit the strategic planning of the next complex integration.
110. Pain Management and Control of Effects of Treatment in Hospitalized Veteran Patients with Cancer
Susan Jane Hagan, James A Haley Veterans Hospital, Tampa, Florida, SC McMillan, University of South Florida, M Tittle, University of South Florida, J Laughlin, Bay Pines Veterans Affairs Medical Center
Objectives: This study evaluated pain management outcomes in a group of veterans with cancer receiving care in the year before the Veterans Administration made its announcement of Pain as the Fifth Vital Sign. Four areas were addressed: (1) Mean pain intensity; (2) Pain interference with function; (3) Incidence of constipation in patients on opioids; and (4) Documentation of pain assessment and management.
Methods: At two veterans hospitals, all patients with cancer were asked to complete the Constipation Assessment Scale (CAS) and Brief Pain Inventory (BPI) once and a Pain Visual Analog Scale (VAS) three times during a 24 hour period. The VAS completion was separated by at least two hours between 7 AM and 9 PM. This allowed for an average pain intensity level over 24 hours. At the end of the 24 -hour period, a chart audit was completed using the Chart Audit for Pain. Frequencies, percentages, means and standard deviations were used to describe the sample. The range, mean and standard deviation was computed for the VAS, the BPI and CAS. For items on the Chart Audit for Pain, frequencies and percentages were calculated. In addition, after converting the medication and pain intensity to a 0-4 scale, the pain level was then subtracted from the severity of patient pain from 0-4 to determine adequacy of pain management.
Results: The sample included a group of 90 veterans with cancer who had been hospitalized at least 48 hours. The three measures over 24 hours were averaged to obtain a daily pain intensity score. The mean of the daily scores was 32.9 (SD+25.3) and the daily scores ranged from 0-98. Approximately one-fourth of patients (25.6%) reported average daily pain intensity at levels above the mid-point (5). On the BPI, pain at its worst averaged 5.6(SD=3.3) and least was 2.6 (SD=2.3) on a 0 to 10 scale. Average pain intensity in the past 24 hours was 4/0 (SD=2.3) Pain interference on the BPI, was 30out of 60 (SD=19.4). Greatest interference was reported to be with walking and sleep. CAS for 81 patients indicated that 80% reported at least one symptom of constipation. Of the 84 charts reviewed with the chart audit, only 36 (42.9%) had of at least one numerical pain rating recorded. Results of interventions were noted in only 22 (26.2%) of charts. VAS scores and pain medications were both converted to a 0 to 4 scale. Subtraction of the VAS from the analgesic level, indicated that 11/ 84 patients were clearly undermedicated by nurses.
Conclusions: Although the AHCPR guidelines have been published for four years, it appears that the goal of systematic assessment and management of persons with cancer has not been achieved. Continued focus on improving skills of nurses is essential to improving these outcomes.
Impact: System approaches are needed to improve pain assessment and management. Nurses need further education to overcome old misconceptions about pain and its management.
HSR&D Funded: #95-042.1
111. Patient Characteristics Associated with Provider Practices for Psychotropic Care
Cheryl S. Hankin, Center for Health Care Evaluation, VA HSR&D, Palo Alto Health Care System, Palo Alto, CA, JJ Anderson, Center for Health Quality, Outcomes, and Economic Research, Bedford, MA, L Herz, Edith Nourse Rogers Memorial Veterans Hospital, Bedford, MA, H Gao, Boston University School of Mathematics, Boston, MA, L Curran, Center for Health Quality, Outcomes, and Economic Research, Bedford, MA, D Berlowitz, Center for Health Quality, Outcomes, and Economic Research, Bedford, MA
Objectives: Although effective pharmacologic treatment of mental disorders is widely available, research conducted in the private sector indicates that elderly and lower income psychiatric patients are less likely to receive psychotropics than their younger or more affluent counterparts. Researchers conclude that providers must improve the quality of their prescribing practices. We sought to extend this body of research. Controlling for psychiatric diagnoses and medical illness severity, we examined whether specific patient demographics were associated with the likelihood of receiving psychotropics among VA psychiatric patients.
Methods: During FY98, 17,960 patients were hospitalized in VA New England facilities (VISN 1); of these, 2,443 had a primary psychiatric diagnosis at an index hospitalization. For these patients, we extracted VA pharmacy data to identify psychotropics (antidepressants, hypnotics, mood stabilizers, antipsychotics, and anxiolytics) dispensed during the two-weeks following discharge. We used data from VA centralized databases to compare classes of psychiatric diagnoses (per ICD-9 codes), medical illness severity (number of inpatient days in prior year due to a medical illness), and demographic characteristics (age, race, gender, and income) between patients who did and did not receive targeted psychotropics. In separate analysis, we constructed a regression model to examine the association between psychiatric diagnoses and likelihood of receiving psychotropics. In successive models, we added medical illness severity and demographics as predictors of receiving psychotropics.
Results: Of the 2,443 identified psychiatric patients, the majority were under age 65 (81.5%), Caucasian (86.7%), male (95.1%), and reported income under $20,000 (69.9%). Psychoses were diagnosed among 33%, depression 16%, bipolar disorder 15%, posttraumatic stress disorder (PTSD) 10%, psychoses comorbid with other psychiatric diagnoses 6%, and PTSD comorbid with other psychiatric diagnoses (excluding psychoses) 20%. There were 2,948 unique psychotropics dispensed to 59% (N=1435) of psychiatric patients during the two-week period. Among those who received psychotropics, 31% received exactly one, 39% received two, 23% received three, 6% received four, and 1% received five unique psychotropics during these two weeks. Patients who received any psychotropic were significantly younger and more likely to have a diagnosis of depression or bipolar disorder than patients who did not receive any psychotropic. Controlling for psychiatric diagnoses and medical illness severity, patients aged 65 and over were approximately half as likely to receive antidepressants (OR 0.63, 95% CI = 0.49 - 0.81), hypnotics (OR=0.54, 95% CI = 0.39 - 0.76), mood stabilizers (OR=0.58, 95% CI = 0.44 - 0.77), or antipsychotics (OR=0.42, CI = 0.32 - 0.56). NonCaucasian patients were one third less likely to receive anxiolytics (OR=0.67, CI = 0.52 - 0.82). Gender and income did not significantly contribute to the likelihood of receiving psychotropics.
Conclusions: Whereas many VA psychiatric patients received multiple psychotropics during the two weeks following discharge, elderly and nonCaucasian patients were significantly less likely to receive psychotropics. Future research is needed to replicate this study nationally and explore reasons for these apparent variations in provider prescribing practices.
Impact: Findings may be useful in profiling VA patients at risk for undertreatment. Results suggest that interventions to improve the quality of psychotropic care should target the treatment needs of elderly and nonCaucasian patients.
HSR&D Funded: CPI 99134-1
112. Perceptions of the Quality of VA Medical Services among Gulf War-era Veterans
David H. Hickam, MD, MPH, Portland VAMC and Northwest HSR&D Center of Excellence, Portland, Oregon, SK Joos, Portland VAMC and Northwest HSR&D Center of Excellence
Objectives: A major clinical initiative of the VA has been to establish a primary care delivery model. This study sought to measure whether primary care services are associated with improved patient satisfaction. The study's objective was to determine whether veterans who receive care within a VA primary care system have different perceptions of the adequacy of their health care than those who do not receive primary care services.
Methods: Study subjects included all individuals who were discharged from military service after August 1990 (Gulf War-era) and had received medical care from the Portland VAMC between October 1997 and September 1998. In January 1999, a questionnaire was mailed to 1113 veterans who met these criteria. The questionnaire included demographic items, the SF-36V health status instrument and the VA Customer Satisfaction Survey (developed in collaboration with the Picker Commonwealth Foundation). Subjects who had not returned the questionnaire after six weeks were contacted by telephone and mailed an additional questionnaire when necessary. Respondents were classified by whether they were enrolled in the facility's primary care system and had an assigned primary care provider. Comparisons of satisfaction items between patient groups were adjusted for age, gender, and health status by using analysis of covariance and logistic regression.
Results: Of 467 respondents, 15 were excluded because their age was greater than 60 years. Of the 452 remaining subjects, their median age was 38 years, 86% were male, and 84% were Caucasian. Sixty nine percent had an assigned primary care provider (PCP). Patients having a PCP had higher satisfaction with their most recent clinic visit (p=.05), but they were not more satisfied with overall VA care. Sixty two percent of patients with a PCP and 58% of patients without a PCP reported that the overall quality of their care was very good or excellent. Patients having a PCP did not consistently have more favorable perceptions of specific provider behaviors. Among patients with a PCP, 18% reported that they were not involved in decisions about their care as much as they desired, 23% reported not having complete confidence in their provider, and 23% reported not receiving as much information about their condition as they desired. These proportions did not differ significantly from those reported by patients without a PCP. However, more of the patients with a PCP (75%) than of those without a PCP (55%) reported that their providers were always familiar with their recent medical history (p < .001).
Conclusions: Among younger veterans, a VA primary care system positively affected some aspects of their perceptions of medical care but did not appear to affect overall satisfaction with VA services. For some physician communication behaviors, primary care providers do not appear to perform better than other providers.
Impact: Overall satisfaction with VA services is influenced by factors that will not necessarily be corrected by adoption of a primary care delivery model. Primary care providers may benefit from initiatives to improve communication behaviors.
113. Have Deaths Due to Medical Errors Been Overestimated?
Timothy P. Hofer, MD, MSc, VA Center for Practice Management and Outcomes Research, Ann Arbor, MI, RA Hayward, VA Center for Practice Management and Outcomes Research, Ann Arbor, MI
Objectives: Physician implicit review has been used to assess whether a hospital deaths are "preventable" and estimates from these type of reviews of the number of deaths in American hospitals that are reportedly due to medical errors is shockingly high. We sought to uncover what physicians meant when they describe a death as "preventable" in terms of the probability of a patient surviving had care been optimal.
Methods: There were 302 reviews of 68 different eligible patients who died in the hospital, at seven Department of Veterans Affairs Medical Centers. Cases with a terminal diagnosis who were admitted to the hospital predominantly for comfort care were excluded. A total of 13 physician reviewers were trained in the use of the implicit review instrument for the study. Each was a board certified internist with extensive experience in inpatient medicine. A previously reported instrument was used. The question on preventable death was rated on a 5 point scale (1=definitely yes, 2=probably yes, 3=uncertain, 4=probably not, 5=definitely not) In this study we added to the survey instrument several questions asking the probability of certain events, such as the likelihood of the prevention of death to be if care had been optimal, rated from 0% - 100%. The unit of analysis was the review although we accounted for the clustering of reviews by patient and reviewer using multilevel statistical models.
Results: 8% of the reviews rated overall care as being substandard and 22% reported care as of borderline quality or worse. In 7% of the reviews death was rated as probably or definitely preventable by optimal care and in 18% the reviews rated the preventability of death as uncertain. These ratings of quality are similar to those found in other studies. In addition, the inter-rater reliable of ratings of the preventability of death was similar to previous studies (ICC= 0.22 for two reviewers compared to 0.24 in The Harvard Malpractice study. The probability of preventing death with optimal care was estimated to be 60% (95% CI 50%, 71%), if a death was rated as probably or definitely preventable, 24% [20%, 28%] with uncertain preventability and 5%[4%, 6%] if death was rated as not preventable. However, even for the probably/definitely preventable deaths the likelihood of surviving three months with good cognitive and physical function was estimated at only 12% [3%, 20%].
Conclusions: It is necessary to qualify the alarming statistics about preventable deaths in US hospitals by pointing out that when physician reviewers rate a death as preventable it is most commonly in very sick patients whose chances of surviving to leave the hospital, even with optimal care, would be uncertain. A further implication relevant to this finding is that hospital mortality rates can not capture differences in preventable death rates to the degree that patients die before discharge even with optimal care.
114. Health-related Attitudes/Beliefs among Elderly, African-American and White Veterans with Chronic Knee and/or Hip Pain
Said A. Ibrahim, MD, MPH, Louis Stokes Dept. VAMC & CWRU, Cleveland, OH, Christopher J. Burant, MA, CWRU, Cleveland, OH, Laura A. Siminoff, PhD, CWRU, Cleveland, OH, C. Kent Kwoh, MD, Louis Stokes Dept. VAMC & CWRU, Cleveland, OH
Objectives: Numerous studies have reported racial variations in health care utilization. Observed differences in health care utilization between African-American (AA) and whites may reflect differences in health-related attitudes/beliefs. In a VA-funded study of ethnic/cultural variations in the management of osteoarthritis, we sought to examine measures of health-related attitudes/beliefs among elderly, male veterans with chronic knee and /or hip pain, comparing AA and white patients.
Methods: 228 AA (46%) and white (54%) patients (age >50) were interviewed. Patients with moderate to severe knee or hip pain for more than six months were eligible for the study. Trained research staff interviewed participants with regard to their health-related attitudes/beliefs using the following previously validated measures: Health Belief scale; Fatalism scale (modified Powe scale); Self-care scale; Krantz scale for patient desire for active participation in therapy; and components of Bill-Harvey efficacy of specific treatments for arthritis questionnaire.
Results: AA and white patients in this cohort were comparable with respect to age (67 ± 9 vs. 68 ± 9, p= 0.53), severity of arthritis [measured by Lequesne scale (mean score 11 vs. 12, p=0.09)], WOMAC scale for activity difficulty (mean score 17 vs. 18, P=0.25), Geriatric Depression scale (5 vs. 5, p=0.60), perceived disability due to pain (66% vs. 63%, p=0.50), moderate or higher impact of pain on quality of life (70% vs. 71%, p=1.00), and living alone (37% vs. 30%, p=0.26). However, AA were more likely to report lower income (41% vs. 22%,p=0.04) and employment (6% vs. 14%, P= 0.05). Compared with whites, AA patients scored higher on the Self-care scale (mean score, 5.5 vs. 6.4; P= 0.00) and exhibited a trend toward higher scores on the Health Belief scale (mean score, 6.9 vs. 7.4; P= 0.08). On the Bill-Harvey questionnaire, AAs were much more likely than whites to rate prayer as being "very helpful" in arthritis treatment (24% vs. 55%; P= 0.000). There were no significant differences in scores on the Fatalism scale (mean score, 1.98 vs. 2.04; P=0.37); and Krantz scale (mean score, 3.5 vs. 3.2; P= 0.35).
Conclusions: In this cohort of male, elderly veterans with moderate to severe chronic knee and/or hip pain, AAs were more likely than whites to score higher on the Self-care and Health Belief scales, and to believe that prayer is "very helpful" in the treatment of arthritis. Further research is needed to examine the clinical significance of these differences.
Impact: The aging of the US population, combined with the increasing representation of ethnic minorities in the US armed services, makes it likely that the VA health care system will increasingly care for older patients from diverse ethnic and cultural backgrounds suffering from chronic diseases. Understanding differences in patients’ attitudes/beliefs toward health may help explain variations in health care utilization and outcomes.
HSR&D Funded: ECV-0003
115. Moderation Management: An Alternative to AA for Patients with Alcohol Use Disorders
Elena Klaw, VA Palo Alto Health Care System, Menlo Park, CA, S Luft, VA Palo Alto Health Care System, K Humphreys, VA Palo Alto Health Care System
Objectives: The current study examines Moderation Management, a mutual-help group alternative to AA. Moderation Management (MM) believes alcohol problems can be managed as a habit disorder using social support and cognitive-behavioral principles. In contrast to abstinence based approaches, Moderation Management assumes that many problem drinkers (which MM differentiates from alcohol dependent individuals) will be able to return to moderate, non-problem drinking. This study used quantitative and qualitative techniques to 1: Explore the characteristics of individuals who telephone the MM information and referral line and participate in MM groups. 2: Contrast the principles of MM with those of AA. 3: Explore problem drinkers' reasons for choosing to participate in MM and their understandings of the ways in which MM differs from AA.
Methods: 1- Individuals telephoning the MM information and referral line (N = 467) were assessed on demographics, drinking problems and severity, co-morbid psychiatric conditions and prior treatment experiences. In addition, the survey included an open-ended question assessing hopes and motivations related to MM participation. 2-Using an expanded version of the phone survey, data was collected from current MM group participants. Within this survey, an open-ended question assessed perceptions of the differences between MM and AA. 3-Thirty life history interviews were conducted with individuals currently involved in MM. Based on a grounded theory approach (Straus & Corbin, 1990), the qualitative analysis program, NUD*IST was used to derive themes from the open-ended response materials. Exploratory qualitative findings will be presented in the context of descriptive quantitative results.
Results: Individuals who telephone the MM phone line and participate in its groups tend to have less severe drinking problems and significantly greater educational and economic resources than the alcohol dependent persons served by AA. In addition, MM callers and participants tended not to describe themselves as religious and appear to place great value on individualistic self-control. For women especially, the philosophy of MM appeared to normalize their experiences and alleviate some of the stigma commonly associated with the disease model of alcoholism espoused by AA.
Conclusions: For problem drinkers, particularly those with greater economic and educational resources, less severe drinking problems and low religiosity, MM may serve as a beneficial alternative to AA. In the current sample, participants (especially women) expressed that, in contrast to AA, MM provided a sense of confidence in their ability to manage alcohol and enabled them to prevent further drinking problems. MM might therefore be an appropriate option for VA alcohol patients who are female and/or have low levels of dependence.
Impact: As the VA Health Care system seeks to seeks to serve a broader array of veterans and expand ambulatory care and services to women, the development of appropriate screening techniques, referrals and programmatic interventions for alcohol abuse is essential. Current findings suggest that Moderation Management groups may appeal to less impaired alcohol abusers and serve as a cost-free preventive intervention for individuals beginning to experience problems with alcohol.
116. Use of Focus Groups to Identify Knowledge Gaps about Care for Spinal Cord Injured Veterans
Marcia W. Legro, PhD, Seattle, WA, F Weaver, MC Hammond, M Guihan, E Collins, E Langbein, T Pape, and S LaVela, Edward Hines, Jr. VA Hospital, Hines, IL.
Objectives: The Spinal Cord Injury Quality Enhancement Research Initiative (SCI QUERI) is part of a VA effort to improve quality of health care for specific veteran populations. A primary QUERI goal is to identify areas of study that will yield improved quality of care. Since SCI is a permanent injury requiring acute, rehabilitative, preventive, and lifelong care, there are many interventions to review for effectiveness and for their relation to the patient's health-related quality of life. The purpose of convening focus groups was to identify areas where patients and providers have questions not addressed by existing research.
Methods: Ten focus groups were convened, five with providers and five with patients at VA SCI Centers (GA, IL, NJ, WA, & WI). The resulting information was compiled and presented to the SCI QUERI Executive Committee (research and clinical experts) for consolidation into research topics.
Results: Patient groups were 90% male and with appropriate representation by age, type of injury, and living situation. Most patients received acute care at a non-VA hospital with transfer to VA care. Provider groups included dietitians, health care technicians, home care coordinators, nurses, physical therapists, physicians, physician assistants, psychologists, rehabilitation therapists, and social workers. The major topics identified by SCI veterans were: pain, sexual functioning and fertility, shoulder overuse problems, equipment needs and access, housing and transportation issues, home care and nursing home care, issues related to caregiver training and caregiver burden, and access to providers with expertise in SCI. Providers mentioned many of the same issues (e.g., pain management, overuse injuries, and sexual functioning). However, providers also indicated areas such as substance abuse, nutrition, surgery for veterans with SCI, aging with SCI, and continued problems with managing and preventing pressure ulcers and urinary tract infections.
Conclusions: Many topics are common to both SCI veterans and providers. Identified knowledge gaps will guide the formulation of new RFA/RFPs. One RFA has been disseminated on Pressure Ulcer Prevention and Management, and one is being developed on the Preoperative Risks and Outcomes of Surgery for Individuals with SCI.
Impact: The identification of knowledge gaps will promote the generation and funding of RFAs by HSR&D for research of greatest need and priority to patients with SCI and providers of SCI care. Specific education and dissemination activities will also be formulated.
117. Time Costs in Integrated Care for Depression Treatment in Primary Care
Chuan-Fen Liu, HSR&D, VA Puget Sound Health Care System, Seattle, WA, S Hedrick, HSR&D, VA Puget Sound Health Care System, E Chaney, VA Puget Sound Health Care System, N Hasenberg, HSR&D, VA Puget Sound Health Care System, B Felker, VA Puget Sound Health Care System
Objectives: The objective of this study is to describe the kinds of time costs incurred in providing treatment to VA patients with Major Depression Disorder (MDD) through integrated care in a primary care setting. VA has mandated implementation of a comprehensive guideline for care of patients with MDD. Integrating mental health providers in the primary care setting is one approach to assist in depression care. The activities in integrated care include screening, evaluation, treatment recommendation, treatment (Cognitive Behavior Therapy (CBT) and provider visits for antidepressant medication), and follow-up. A randomized clinical trial of integrated team care for MDD is being conducted at the Seattle General Internal Medicine Clinic to examine the effect of depression treatment by comparing with usual care. This study reports on time cost components for integrated team care.
Methods: All staff in integrated team treatment arm have kept track of their time using tailored time sheets. Types of staff include psychiatry, psychology, social work, health technician, and research assistants. Activities included weekly team treatment meetings, CBT group treatment, patient phone calls (screening and follow-up), patient appointments, as well as note writing and electronic chart documentation. In total, four weeks of time sheets were collected, one week for each month, from January 1999 to April 1999 for each staff member. Number of encounters for each and average time per encounter was calculated for each type of activity. Time costs per type of activity per staff member was calculated by multiplying the time used to perform the activity for the by the hourly wage rate.
Results: The average time for a team treatment meeting was 56 minutes, with an average cost of $253 and an average of 13 patients evaluated in a meeting. The average cost for a 90-minute CBT session was $97, with an average 2 patients attending. Twenty patients saw a psychologist over the four-week period, with an average 40 minutes for the visit and an average cost of $22. Follow-up phone calls were made by social workers, social work students, and other study staff with an effect on costs over time. One hundred and nineteen follow-up phone calls were made, with an average time of 24 minutes and an average cost of $10.
Conclusions: Time costs vary widely by type of provider and type of activities. Several activities are performed by more than one type of provider, and others that are performed by only one provider in this treatment team could potentially be performed by different types of providers. These data can be used in estimating the cost of integrated care for depression and in examining the staff configuration and mix needed to provide the most efficient services.
Impact: The results of this study could be used by managers in both primary care and mental health to examine the staffing mix and configuration needed to provide care to an important segment of the VA population.
HSR&D Funded: IIR#95-097
118. Comparing the Performance of Two Diagnosis-Based Case-Mix Measures in the VA
Susan A. Loveland, Bedford VA, HSR&D Center of Excellence, Bedford, MA, AK Rosen, Bedford VA, HSR&D Center of Excellence, JJ Anderson, Bedford VA, HSR&D Center of Excellence, D Berlowitz & J Rothendler, Bedford VA, HSR&D Center of Excellence, C Hankin, Palo Alto VA, Center for Health Care Evaluation, A Ash & M Moskowitz, Boston Medical Center
Objectives: Two diagnosis-based case-mix measures, Adjusted Clinical Groups (ACGs) and Diagnostic Cost Groups (DCGs), have proven valuable in profiling and predicting health care resource utilization in managed care organizations. As the VA adopts managed care practices, a detailed understanding of the disease burden of veteran users is necessary for evaluating the quality of care, predicting resource utilization, and assessing VA's effectiveness as a managed care organization. Since both ACGs and DCGs were developed on other populations, neither system may adequately discriminate among groupings of VA users. Therefore, we will compare the predictive performance of ACG and DCG scores (based on externally derived weights) and of fully reparameterized ACG and DCG models.
Methods: VA inpatient and outpatient files were used to select a 1% random sample of veterans (N=26,095) who utilized VA health services, other than dental and phone clinic stops, at some time during FY'97. We retained diagnostic information from face-to-face provider encounters only. We regressed concurrent resource utilization, measured as annualized service days (outpatient visit days + inpatient days) on four kinds of predictors: 1) an ACG score based on HMO-derived weights; 2) a DCG score based on Medicare-derived weights; 3) a full ACG model using the 32 Adjusted Diagnostic Groups (ADGs); and 4) a full DCG model using the 136 Hierarchical Condition Categories (HCCs). To compare model performance, we obtained R-squares and predicted service days for each model, grouped cases by deciles of predicted service days, and obtained mean observed service days and coefficients of variation (CVs) within these deciles. Models 3 and 4 are fully reparameterized because they estimate VA-derived weights.
Results: R-squares for ACG models increased from 0.146 to 0.215 with reparameterization, and from 0.171 to 0.296 for DCGs. Reparameterized models had better overall CVs of service days compared to original models (e.g., in the reparameterized DCG model, CVs for deciles ranged from 0.9-2.4; in the original DCG model, they ranged from 1.7-3.3). Further, reparameterized models showed better discrimination across deciles than original models. Although DCG models had higher R-squares and lower CVs of service days per decile than ACGs (e.g., CVs per decile for DCGs were 2-35% lower than for ACGs), ACG models had better discrimination in the lower deciles.
Conclusions: DCGs generally performed better than ACGs based on R-squares and CVs, suggesting that DCGs produce more reasonably homogeneous groupings of VA patients with respect to resource utilization. Although reparameterization of both the DCG and ACG models increased model performance, the R-squares obtained were lower than those found in HMO populations. However, these results are preliminary because of the small sample. We will refit, reparameterize, and validate these models on approximately 1.5 million veterans.
Impact: Since weights derived from other populations explain only a small percentage of the variation in VA utilization, ACG and DCG resource-predicting scores derived on other populations should be recalibrated for use in the VA.
HSR&D Funded: MPC-97-009
119. Impact of Dropout Rates on Effect Size in Randomized Controlled Trials
Michael F. Luther, South Texas Veterans Health Care System, Audie L. Murphy Memorial Veterans Hospital Division, San Antonio, Texas, JE Cornell, GRECC/STVHCS-ALMD, Medicine/UTHSC-SA, CD Mulrow, South Texas Veterans Health Care System, Audie L. Murphy Memorial Veterans Hospital Division
Objectives: Our objective is to describe the effects of dropout rates on summary statistics in meta-analyses. Although dropout rates are routinely cited as threats to the internal validity of randomized controlled trials (RCTs), their actual influence on estimations of treatment effects have not be documented empirically.
Methods: The database for these analyses included 44 RCTs that compared a newer antidepressant with placebo. These trials were identified and described in a recent evidence report of treatments for depression. Briefly, English and non-English literature was searched using the Cochrane Collaboration Depression, Anxiety, and Neurosis Group's registry. All RCTs of six weeks or greater duration that evaluated one of 32 specific newer antidepressant or herbal treatments were included. The measure of treatment effectiveness was based on significant clinical improvement in depressive symptoms: i.e., a 50% reduction in HAMD scores. Meta-analysis regression was used to examine the influence of total dropout rate and dropout rates due to adverse effects or lack of efficacy on estimates of treatment efficacy. Odds ratios based on both intention-to-treat data and data from participants completing studies (completors) were analyzed.
Results: Thirty-six of the studies were in populations of patients with major depression, 3 were primarily in dysthymics and 5 were conducted in mixed populations. Diagnosis and a measure of severity of depression at the beginning of the study were included in the meta-regression. Important relationships between dropout rates and both intention-to-treat and completors effect sizes were verified. Initially the models included a single variable to obtain an estimate of the influence of the difference between the treatment and placebo dropout rates on effect size. When the analyses indicated these estimates were significant, the models were modified to allow examination of the treatment and placebo dropout rates separately. Examination of the modified intention-to-treat odds ratios revealed that as the treatment dropout rate increased the effect size decreased. This relationship was seen with both the total dropouts and dropouts due to lack of efficacy. The opposite was observed with the placebo dropout rate. As it increased the odds ratio also increased. These associations reversed when dropouts due to side effects were analyzed. The effect size increases along with the treatment dropout rate but decreases as the placebo dropout rate increases. Analyses of the odds ratios based on completors were considerably different. The effect size increases as the treatment dropout rate increases and decreases as the placebo dropout rate increases when examining both the total dropouts and dropouts due to side effects. No relationship was seen between dropouts due to lack of efficacy and the odds ratio when examining completors.
Conclusions: Dropout rates have significant and varying effects on the summary estimates of RCTs.
Impact: This is one of the first empiric demonstrations to document the affect dropout rates have on summary estimates of trial results. It confirms that in intention-to-treat analyses, increased
dropout rates in treatment groups lead to underestimation of effect size while increased dropout
rates in control groups leads to overestimation of effect size. In completor analyses, the opposite effects are seen.
HSR&D Funded: HFP 98-002
120. Characteristics and Perceptions of Spouse Caregivers of Dementia Patients
Melitta K. Maddox, RN, CNS, MSN, CCDOR & GRECC, Minneapolis VAMC, Minneapolis, MN, MA Kuskowski, GRECC, Minneapolis VAMC, MJ Bauer, Research, Minneapolis VAMC
Objectives: The mission of the Minneapolis GRECC includes the development of models of care for community-dwelling persons with dementing disorders. The GRECC clinical model emphasizes patient function and quality of life while recognizing the central role of the family caregiver. The adverse effects of caregiving derive from multiple sources, but the principal source of burden is the caregiver’s inability to manage and control the persistent, troubling behaviors.
Methods: As part of a larger study, we interviewed the spouse caregivers of dementia patients entering the GRECC memory loss clinic. Identifying the characteristics, knowledge, and perceptions of the caregiver provides an assessment-based way to address caregivers’ dementia management problems and produce positive changes for patients and families. The caregiver interviews included data regarding depression, burden, and attitudes toward and preparedness for caregiving. The Revised Memory and Behavior Problems Checklist (Teri et al., 1992) allowed the caregivers to report patient behaviors and the extent to which the behaviors affected them.
Results: The 61 veterans with mild to moderate dementia whose spouse caregivers consented to be interviewed for this study had the following characteristics: mean age of 76 years (SD = 6.9), mean years of education 12.3 years (SD=3.2). The mean MMSE (Folstein, 1975) score was 20.5 (SD=6.1). The mean GDS score for 44 patients who were given the Geriatric Depression Scale (Yesavage, 1986) was 3.3 (SD=2.6). Median income for the couples was reported between $20,000 and $29,000. The mean length of time in their current residence was 24.25 years (SD=17.5). Forty-six (76.7%) remained in their own homes, five (8.3%) currently owned condominiums, and six (10%) lived in apartments. The mean age of the spouse caregivers was 72.3 years (SD=7.3) and their mean years of education was 12.7 years (SD=2.1). Only eleven spouses worked outside the home. Most caregivers listed at least one chronic health condition. The most frequently reported were hypertension (26.7%), arthritis (25.0%) and heart disease (23.3%). Twenty of the 36 patient behaviors were significantly correlated with caregivers' reports of distress Spearman rho, all p<0.05. As expected, caregiver depression measured by the CES-D correlated with the Burden Interview (p=.000). The preparedness scores correlated negatively with the burden scores (p=.002) and with depression (p=.003).
Conclusions: Although there was little correlation between the mental status of the patients and other patient or caregiver data, certain behaviors were stressful and seemed to contribute to the burden and depression of caregivers. Further analyses illustrated the relationship between the frequency and intensity of each of the behaviors reported. Caregivers who felt competent to handle the patient changes and prepared for the caregiving role exhibited less depression and less burden.
Impact: The caregivers’ stress did not necessarily increase with the progression of the disease. Assessing the qualities and concerns of the caregiver as well as the patient’s stage of disease can lead to interventions tailored to the specifics of the situation.
HSR&D Funded: NRI S95-143-2
121. Application of HEDIS Measures within a Veterans Affairs Medical Center
Nancy M. Mencke, Altanta Veterans Affairs Medical Center, Decatur, Georgia, LG Alley, Altanta Veterans Affairs Medical Center, J Etchason, Kerr L. White Institute for Health Services Research
Objectives: As the Veterans Health Administration (VHA) attempts to (1) benchmark its performance against the private sector and (2) assess the quality of care provided by community-based outpatient centers, the ability to generate standardized performance data has become increasingly important. The state of the art in performance measurement is the Health Plan Employer Data and Information Set (HEDIS), a set of validated quality indicators designed to evaluate and compare health plan performance. The objective of this research was to pilot test selected HEDIS performance measures within the Atlanta Veterans Affairs Medical Center (AVAMC). The three measures assessed include: (1) beta blocker treatment after a heart attack, (2) ambulatory follow-up after hospitalization for mental illness, and (3) cervical cancer screening.
Methods: The study design consisted of a retrospective review of administrative data. One measure also had a medical record review component, performed for purposes of data validation. Eligibility criteria differed by HEDIS measure. The eligible populations included: (1) patients 35 and older who were hospitalized and discharged alive with a primary diagnosis of acute myocardial infarction (MI) in calendar year 1996; (2) patients who were hospitalized for treatment of selected mental health disorders in fiscal year 1997; and (3) female patients ages 18-64 who were enrolled in VA primary care clinics during the fiscal year 1995-1997 period. Data were collected in accordance with HEDIS 3.0 1998 specifications for administrative data, comprised of the following steps: (1) identification of the eligible population; (2) search of transaction data systems to identify numerator events for all members in the denominator; (3) search of transaction data for members with contraindications to the service/procedure being measured; (4) exclusion of members with identified contraindications; and (5) calculation of the rate.
Results: The rate at which beta blockers were prescribed following discharge for acute MI was 85.5 percent based on administrative data only; using the hybrid methodology, the rate of beta blocker therapy was somewhat higher (88.1 percent). The rate of 30-day, ambulatory follow-up for selected mental health diagnoses was 61.3 percent. The cervical cancer screening rates were as follows for the various primary care (PC) teams: Purple Team PC, 29 percent; Green Team PC, 35 percent; Red Team PC, 40 percent; and Women's Wellness/Gynecology, 91 percent.
Conclusions: With few or no modifications to the specifications, the three HEDIS measures were successfully adapted to the AVAMC setting. Comparative data indicate that, in some areas, AVAMC performance compares favorably to or exceeds that of the private sector. In addition, opportunities for quality improvement were identified.
Impact: HEDIS has applicability within the VHA, both as a means of quantifying performance and as a tool for improving the quality of care. However, hindrances to HEDIS implementation within the VHA (e.g., organizational structure, information systems issues) must be adequately addressed, if VHA is to take advantage of the full measurement capabilities of HEDIS. The VHA stands only to gain from the adoption of standardized performance measurement systems and the enhancements in information systems standardization would require.
122. Do Veterans Want to Participate in Medical Decisions?
Melissa R. Partin, Minneapolis VA Medical Center, Minneapolis, Minnesota, N Dillon, Minneapolis VA Medical Center, D Nelson, Minneapolis VA Medical Center, S Nugent, Minneapolis VA Medical Center, P Brugge-Wiger, Minneapolis VA Medical Center, H Rubins, Minneapolis VA Medical Center
Objectives: The objectives of this study were to assess veterans’ preferences regarding participation in medical decisions, and to determine whether these preferences vary by patient demographic and other characteristics.
Methods: All patients checking in for primary care appointments at the Minneapolis VA Medical Center during a six-day study recruitment period in June of 1999 were asked to fill out a brief written survey. The survey included questions on education, race/ethnicity, and care arrangements, and asked patients "which of the following best describes how you usually like to make medical treatment decisions? (a) I like to decide, by myself or with my family, which treatment I will get, (b) I like to share responsibility for deciding which treatment I get with my doctor, (c) I like to share responsibility for deciding which treatment I get with my doctor and family, (d) I like to leave all decisions about my treatment to my doctor". Responses to this survey were linked to another VISN wide survey on functional status. Information on patient age and marital status was extracted from outpatient records. Chi-squared tests and analysis of variance procedures were used to test for significant associations between patient characteristics and their decision making preferences
Results: A total of 489 surveys were completed. The vast majority of patients completing the surveys (81%) indicated they wanted to participate in treatment decisions. Approximately 18% indicated they preferred to leave all treatment decisions to their doctors. Decision making preferences varied significantly by patient age (p =.01), marital status (p = .001), and care arrangements (p = .001) but not by race/ethnicity, education or functional status. Patients who indicated they preferred to share treatment decisions with their family +/or doctor were significantly younger than patients who indicated they preferred to make decisions by themselves or with their family. Married patients and patients reporting they had someone helping them with their day to day health care decisions were more likely than other patients to indicate they prefer to share treatment decisions with their doctors and families. Divorced and never married patients, and patients indicating they did not have someone helping them with their day to day health care decisions were more likely to indicate they prefer to share treatment decisions with their doctor only.
Conclusions: Most veterans want to participate in medical decisions. Married veterans and veterans who have assistance with day to day health care decisions want to include family in decisions about their treatment.
Impact: The increasingly endorsed shared decision making approach to physician-patient interaction assumes patients want to participate in decisions about their care. The above findings suggest this assumption holds for the majority of veterans visiting a VISN 13 primary care clinic, but that preferences regarding whether to share decisions with family and/or doctors vary by patient characteristics. Patient satisfaction will likely be maximized if patient decision making preferences are used to tailor physician-patient interactions.
123. The Accuracy of Coding Chronic Heart Failure in the VA Outpatient Clinic File
Nancy J. Petersen, PhD, VA HSR&D Center of Excellence, Houston, TX, J Souchek, VA HSR&D Center of Excellence, JM Geraci, VA Medical Center, D Thalberg, Baylor College of Medicine, W Foster, VA HSR&D Center of Excellence, CM Ashton, VA HSR&D Center of Excellence
Objectives: In FY97, the Department of Veterans Affairs (VA) began collecting diagnostic information on all patients treated in its outpatient setting. However, little is known about the accuracy of the diagnosis codes in this administrative database. It is important to assess whether the OPC can be used by health services researchers for common purposes such as identifying patients with a specific disease or risk-adjusting outcome measures used for evaluating the quality of VA healthcare. The purpose of this study was to examine the accuracy of the coding of the diagnosis of chronic heart failure (CHF) in the VA Outpatient Clinic File (OPC). In particular, we wished to see if patients identified in the OPC as having CHF would be identified as having heart failure by physicians' review of the patients' medical records.
Methods: We identified veterans with any of 14 ICD-9 codes for heart failure in the FY98 OPC file (these were considered as CHF patients from OPC) and veterans with no CHF diagnosis (these were considered as not CHF patients from OPC). We randomly selected 250 patients from 11 VA medical centers and requested their medical record information from the Veterans Integrated System Technology Architecture (VISTA) computer system. The patients were selected from the OPC so that approximately half were considered as a CHF patient and half would not be considered to be a CHF patient. An abstraction form containing criteria for judging if a patient had CHF was developed by 2 physicians and a physician assistant (PA). The form included information on patient demographics, medications, comorbidities, medical history, symptoms, signs, social history, hospitalization diagnoses, and cardiology procedures and findings. The PA abstracted the data from VISTA, and the 2 physicians classified each patient into 1 of 4 categories - evidence of CHF, no evidence of CHF, equivocal, or not enough information to classify. The PA and physicians were blinded to whether the patients had been identified in the OPC as having a diagnosis of CHF or not.
Results: The 2 physicians both classified 248 patients into the same category. They disagreed among themselves on 2 cases. Their classification of the 248 patients agreed with the classification in OPC in 179 cases and disagreed in 9 cases. The MDs considered 17 cases as equivocal, and VISTA lacked enough information in 43 cases. Sensitivity was 97.3% and specificity was 92.4%.
Conclusions: Overall, coding in the VA OPC was adequate in identifying patients with CHF. We found both high sensitivity and high specificity. Further research is needed to determine if VA facilities differ in the accuracy of their coding. Also, other diagnoses including comorbidities should be studied to determine their accuracy.
Impact: When evaluating quality of care among providers, it is essential to adjust for differences in the patients' case-mix and severity of illness. Unless the data are accurate, risk adjustment will be poor and unfair comparisons of providers may result. Although coding of CHF seemed adequate, VA should institute a continuing mechanism of training in coding diagnoses and of regular review of diagnostic accuracy.
124. Comparison of Risk Adjustment Methods for Capitated Payments in VHA
Laura A. Petersen, MD, MPH, Houston Center for Quality of Care and Utilization Studies, Houston, TX, M Byrne, Houston Center for Quality of Care and Utilization Studies, Houston, TX, K Pietz, Houston Center for Quality of Care and Utilization Studies, Houston, TX, M Thompson, Houston Center for Quality of Care and Utilization Studies, Houston, TX, JA Tuchschmidt, Portland VAMC, Portland OR, CM Ashton, Houston Center for Quality of Care and Utilization Studies, Houston, TX
Objectives: The goal was to determine the ability of several candidate clinical risk adjusters to explain variation in VA health care costs.
Methods: We identified users of the VA health care system in fiscal year (FY) 1995-97 who survived through the end of FY 1997. Diagnostic data from both inpatient and outpatient encounters from the VA Patient Treatment File and Outpatient Clinic File were used to generate Adjusted Clinical Groups (ACGs) by the Austin Automation Center (AAC) and Diagnostic Cost Groups (DCGs) by the VA Management Sciences Group. Resource Utilization Groups (RUGs), and Functional Independence Measure-Function Related Groups (FIM-FRG) scores were obtained from the AAC. Patient-level cost data were obtained from the Allocation Resource Center for use in concurrent models (FY 1997 costs) and prospective models (FY 1998 costs). Linear regression analysis generated R-squared values for each candidate risk adjuster or combination. We compared these to the existing three-category capitation method (Veterans Equitable Resource Allocation [VERA]), age/gender models, and prior year cost models.
Results: The study cohort contained 3,442,311 veterans. The mean age was 58. 95.2% were male. All patients were assigned an ACG or DCG. Only 4.7% had RUGs and 0.1% had an FRG. In linear regression analyses, age + gender models had an R-squared of 0.006 for FY 97 and 0.004 for FY 98 costs. VERA models had an R-squared of 0.174 for FY 97 and 0.078 for FY 98 costs. A model using FY 97 cost had an R-squared of 0.238 for FY 98 costs. A model with only RUGs had an R-squared of 0.110 for FY 97 and 0.073 for FY 98 costs. ACG models had an R-squared of 0.226 for FY 97 and 0.100 for FY 98 costs. Models with ACGs + RUGs had an R-squared of 0.317 for FY 97 costs and 0.142 for FY 98 costs. DCG models had an R-squared of 0.314 for FY 97 costs and 0.120 for FY 98 costs. DCG + RUGs models had an R-squared of 0.427 for FY 97 costs and 0.206 for FY 98 costs. The addition of FIMS did not improve the R-squared value in any models.
Conclusions: Models utilizing clinical risk adjusters had higher R-squared values than the existing VERA method. DCG + RUGs had the highest R-squared value for future year costs.
Impact: We plan future work to explore model performance across different patient diagnostic categories (e.g. mental health patients) and with incorporation of multi-year diagnostic information.
HSR&D Funded: RCD 95-306
125. Conceptualizing Spiritual Well-Being: A Focus Group Study of Older Adults with Chronic Illness
Kenneth E. Rosenfeld, MD, VA Greater Los Angeles Healthcare System, Los Angeles, CA, M Kagawa-Singer, UCLA School of Public Health and Asian American Studies, RE Patterson, VA Greater Los Angeles Healthcare System
Objectives: Improving the health-related quality of life (HRQOL) of individuals with chronic illness has become a major goal of contemporary health care, creating the need for measurement tools that accurately assess individuals’ HRQOL. Current HRQOL assessment tools have focused nearly exclusively on its physical, emotional, and social components, despite research evidence that spiritual well-being is important in the setting of chronic or terminal illness. As the first phase of a project to develop a spiritual well-being instrument this study sought to construct a conceptual model for spiritual well-being, from the perspective of older adults with chronic illness.
Methods: Eight focus groups of older individuals with chronic illness were conducted. A diversity of perspectives was sought through stratification by one or more of the following variables: age (65-74, greater than 74); race (Caucasian, African-American, Spanish-speaking Latino); chronic illness category (heart disease, lung disease, arthritis, visual impairment). Participants were asked about the meaning of spiritual health; the characteristics of spiritually healthy people; what they did to maintain spiritual health; and how aging and chronic illness impacted on spiritual health. Focus groups were audiotaped and transcribed verbatim (and translated for Spanish-speaking groups), then coded for thematic structure by two independent coders trained in content analysis. "Trustworthiness" (the qualitative equivalent of validity and reliability) of the analysis was sought through checks on interrater reliability, respondent feedback to the findings, and concurrence by a panel of spirituality experts.
Results: A total of 65 older adults participated in one of 8 focus groups. Participants expressed a wide range of perspectives on spiritual health. Many participants equated spirituality with formalized religious commitment, although a variety of non-religious elements were also identified. Analysis of the focus group transcripts yielded themes in five major areas, with substantial overlap between them: 1) spiritual orientation (e.g. religious or spiritual practices, connection to God); 2) existential state (e.g. meaning and purpose in life, attitude towards death); 3) interpersonal relationships (e.g. loving and being loved, helping others); 4) outlook (e.g. "glass half full") and 5) affect (e.g. inner peace, joy).
Conclusions: Spiritual health was an important dimension of older, chronically ill individuals' HRQOL. Contrary to historical attempts to define spirituality in terms of its religious elements, our analysis suggested that a definition that includes its psycho-developmental elements may better represent patients' own models of spiritual health. Several themes identified in this study (e.g. affect, interpersonal relationships) correspond to previously identified HRQOL dimensions (emotional well-being, social well-being), although other identified areas (spiritual orientation, existential, positive outlook) are not currently represented in HRQOL assessment tools. Expanding the conceptualization of HRQOL to include spiritual well-being may be appropriate in studies of HRQOL in aging and chronic illness.
Impact: If spiritual well-being represents a unique dimension of HRQOL, then current instruments may not fully capture the impact of chronic illness and its treatment on patients' HRQOL. The conceptual model developed in this study will be used to develop an instrument to measure spiritual well-being, and to test its unique contribution to HRQOL.
126. National Hepatitis C Surveillance Day in the Department of Veterans Affairs (DVA)
Gary Roselle, MD, VAMC Cincinnati, University of Cincinnati Div of Infectious Diseases, Cincinnati, OH, SM Kralovic, VAMC Cincinnati, University of Cincinnati Div of Infectious Diseases, LH Danko, VAMC Cincinnati, University of Cincinnati Div of Infectious Diseases, LA Simbartl, VAMC Cincinnati, University of Cincinnati Div of Infectious Diseases, TL Wright, VAMC San Francisco, University of California at San Francisco, KW Kizer, VA Headquarters, Washington, DC
Objectives: Using the Emerging Pathogens Initiative (EPI) computer-based surveillance system, the Veterans Health Administration (VHA) of the DVA determined a case rate for hepatitis C virus antibody (HCVAb) in the patient population served nationwide on a single day (March 17, 1999).
Methods: Inpatients or outpatients seen at any VHA medical care facility who had any blood testing ordered that day were offered HCVAb testing, with education and handout information about hepatitis C disease.
Results: Nationwide 139,054 patients had contact with the VHA that day (130,951 outpatients, 8,103 inpatients). 138 of the 142 (97.2%) reporting sites had HCVAb data available for analysis. 26,102 tests for HCVAb were performed that day and a positive HCVAb test occurred in 1,724 persons (6.6%). This represents an underestimate of the true case rate as persons most at risk for HCV (e.g., injection drug users) are unlikely to have frequent laboratory testing done and would not be captured by this testing schema. 75.9% of the positive tests occurred in outpatients. 1,472 (85%) of those testing positive for HCVAb on March 17 were new positive persons identified. For those with a positive HCVAb test on March 17, 1999, the mean age was 53.8 yrs; 6.3% had a homeless designation. Era of military service revealed 58.7% were Vietnam era. Of note, in FY 1996, Vietnam era veterans made up only 30.2% of the VHA-served population. Race/ethnicity were reported as 46.0% White-non-Hispanic, 29.4% Black-non-Hispanic, 4.9% Hispanic, 0.2% American Indian/Alaskan Native, and 19.6% unknown.
Conclusions: The EPI was able to rapidly provide a case rate for HCVAb in a VHA population including basic demographic information. Despite the predicted underestimate of the true prevalence, this information will serve as the basis for projections for patient care and administrative purposes.
Impact: Estimating costs of care and approaches to maximize effectiveness of care requires understanding the scope of the problem and patient characteristics. EPI is a surveillance system designed to meet the data requirements for planning VHA programs in emerging pathogens.
HSR&D Funded: DEV97-032
127. HEDIS 3.0 Cardiac Procedure Rates in the VA: Variation Among VISNs and Effect of Accounting for Non-VA Care on Overall Measures
James A. Rothendler MD, Bedford VA Medical Center, Bedford, MA, B Chang, Boston University School of Public Health, C Comstock, Bedford VA Medical Center, D Jones, Bedford VA Medical Center, A Hendricks, Bedford VA Medical Center
Objectives: HEDIS 3.0 "quality" indicators include rates of cardiac catheterization (CATH), coronary angioplasty (PTCA) and coronary surgery (CABG). We sought to determine rates of these procedures in the VA and whether there were significant differences among VISNs. We also investigated the relative utilization of CABG vs. PTCA, a selection with important clinical and economic implications. To assess the impact of non-VA care on the aggregate rates of PTCAs and CABGs, we evaluated the contribution of these procedures performed under Medicare in two VISNs.
Methods: For FY 1998, we determined VISN-level procedure rates per 1000 veterans using VA services in FY 1996+1997. In the following analyses, we excluded one VISN that did not perform PTCAs or CABGs as well as veterans with care in more than one VISN (8.3%). For two VISNs, we also analyzed Medicare data for calendar year 1996 for veterans with VA utilization during FY 1996.
Results: There was wide variation among VISNs (n=21) in rates of CATH (5.8-17.5, mean 11.5, sd 3.2, p<.001), CABG (1.2-3.9, mean 2.5, sd 0.89, p<.001) and PTCA (0.8-4.2, mean 2.4, sd 0.97, p<.001 ). With the exception of several VISNs, there were relatively minor changes in VISN-level rates when adjusted for the prevalence of diagnoses for acute myocardial infarct and ischemic heart disease among veterans using VA services in FY 1996+1997. There was also a wide range in the ratios of CABG to CATH (0.1 to 0.38) and PTCA to CATH (0.12 to 0.32). Although there was moderate correlation between rates of CABG and PTCA among VISNs (R=0.52, p<.02), there was a large variation in the CABG/PTCA ratio (0.41-2.3, mean 1.2, sd 0.48). There was moderate negative correlation between CABG/PTCA ratios and rates of CATH (R=-0.37, p=0.10), suggesting that VISNs with higher rates of CATH had lower CABG/PTCA ratios. For the two VISNs in which Medicare data was analyzed: Using only VA data, CABG rates were 3.4 and 2.5, PTCA rates were 2.3 and 1.1, and CABG/PTCA ratios were 1.5 and 2.3. When VA and Medicare procedures were combined CABG rates increased to 8.0 and 7.4, PTCA rates increased to 6.1 and 4.0, and CABG/PTCA ratios decreased to 1.3 and 1.9.
Conclusions: a) HEDIS-defined cardiac procedure rates and CABG/PTCA ratios vary widely among VISNs. b) Higher rates of CATH tend to be associated with lower CABG/PTCA ratios. c) For two VISNs studied, there were substantially more CABGs and PTCAs done under Medicare compared to VA. d) Because of the large numbers of cardiac procedures done on veteran "enrollees" outside the VA, differences among VISNs in procedure rates and CABG/PTCA ratios are difficult to interpret without taking into account such "non-VA" procedures.
Impact: Given that much care of VA enrollees is performed outside the VA, measures of cardiac procedures contained in VA databases are invalid measures of the aggregate care that these veterans receive. Accurate comparisons of these overall rates will require the use of databases containing both VA and non-VA care.
HSR&D Funded: MPC 97-002
128. Cholesterol Screening for Primary Prevention: An Effect of Computerized Reminders?
Anne E. Sales, PhD, VA Puget Sound Heath Care System, Seattle, WA, KL Sloan, VA Puget Sound Heath Care System, NR Every, VA Puget Sound Health Care System
Objectives: National guidelines recommend cholesterol screening every five years for men ages 35 to 65 and women ages 45 to 65. The VA has developed and distributed to each medical center computerized health factor reminders to be used in conjunction with ambulatory care encounters, which can be activated at the medical centers' discretion. This paper describes results of a natural experiment in VISN 20, the Northwest Network, of having these reminders activated at five of eight facilities. The results described here cover all primary care patients in the target age ranges in the eight facilities, measured twelve to eighteen months after reminders were turned on.
Methods: As the result of a VISN-wide decision-making process on guideline implementation, five of eight facilities in VISN 20 decided to activate the primary prevention cholesterol screening reminders in the Health Factors package distributed through VistA at each medical center. Three of the eight decided, for a variety of reasons, not to activate the reminders at that time. VISN-wide, population based data were extracted on all currently active primary care and cardiology patients in the target age range (35-65 for men, 45-65 for women) using the CHIPS data warehouse, a relational database developed in VISN 20 that contains all available data in most clinically relevant tables from VistA systems in the eight facilities in VISN 20. Data were extracted between 12 and 18 months after the Health Factors reminders were activated.
Results: 29,870 patients met inclusion criteria during the year prior to data extraction. Screening rates in the target populations in the five facilities that activated the reminders ranged from 92-98% screened while the rates for the three facilities not activating the reminders in the same time frame ranged from a low of 82% to a high of 87%. Treating the populations as samples for the purpose of statistical testing, the differences in mean screening rates between the group of facilities using the reminders is significant, despite the fact that numbers of affected patients are roughly equal in the two groups (14,648 in the group with reminders, 15,222 without).
Conclusions: Although there were some other differences in process between the facilities using the reminders and those without reminders, the major difference was the use of reminders. The high rates of screening overall attest to a high level of attention to this important primary preventive measure. Even when overall screening rates are high, use of computerized reminders appears to have an additional positive effect in this natural and uncontrolled experiment.
Impact: The health factors reminders used in these facilities are available throughout VHA. Activating them is not a difficult process. The evidence of these data suggests that they should be used routinely throughout the VHA system.
HSR&D Funded: IHS 99-233-1
129. General Life Adjustment in a National Sample of Vietnam Veterans
Dawn M. Salgado, MA, National Center for Posttraumatic Stress Disorder, Brookline, MA, MK Suvak, National Center for Posttraumatic Stress Disorder, VW Savarese, National Center for Posttraumatic Stress Disorder, DW King, National Center for Posttraumatic Stress Disorder, DM Salgado, National Center for Posttraumatic Stress Disorder
Objectives: A number of researchers (e.g., Holahan, Moos, & Schaeffer, 1996; Pavot & Diener, 1993; Ryff, 1995; Seeman, 1989) have emphasized the importance of understanding positive health and well-being as a complement to the study of vulnerabilities and illnesses. Despite this recommendation, studies of positive life adjustment among Vietnam veterans are scarce. The purpose of this project (supported by VA HSR&D and ERIC) is to gain a better understanding of satisfaction, psychosocial adaptation, and attainment among Vietnam veterans and to document premilitary, military, and postmilitary factors that appear to contribute to adjustment within this population.
Methods: This research project is best classified as observational and cross-sectional/prevalence. Data from the National Vietnam Veterans Readjustment Study (NVVRS, Kulka et al., 1990), the only nationally representative sample of Vietnam veterans available today, are being utilized in the current project. Classical test theory principles (Nunnally, 1978) were used to develop indicators of key variables from the collection of survey responses. A variety of multivariate data analyses are currently being conducted in a series of investigations to identify factors related to positive adjustment within this sample. The final phase of the research will consist of testing comprehensive models that will include all identified variables that influence positive adjustment among Vietnam veterans.
Results: Findings from one of our studies suggest that Vietnam veterans experience levels of positive life adjustment (e.g., marital, parenting, and general life satisfaction and occupational and educational attainment) similar to those of veterans who were not deployed to Vietnam and to matched civilian controls. Another set of analyses examining the relationship between veterans’ war-zone coping styles and general life adjustment found an emotion-focused coping style to be associated with poorer adjustment, while a problem-focused coping style tended to be more effective for veterans who experienced moderate levels of combat. Also, endorsing benefits attributable to military service seems to relate positively to general life adjustment, while endorsing setbacks attributable to military service appears to relate negatively to general life adjustment.
Conclusions: Findings to date suggest that Vietnam veterans, on the whole, have adjusted reasonably well since returning to civilian life. Individual differences in the way veterans coped with war-zone stress and in the way they attribute meaning to their military service, as assessed by perceived benefits and setbacks, appear to influence levels of postmilitary general life adjustment.
Impact: According to the World Health Organization (1964), health is defined as "the state of complete physical, mental, and social well-being and not merely the absence of disease and infirmity" (Seeman, 1989, p. 1100). This statement has strong implications concerning the development and implementation of treatments designed to address the mental health and psychosocial adjustment of today’s military veterans. An important tool in designing comprehensive treatment programs that go beyond alleviating maladaptive symptoms to promote "complete physical, mental, and social well-being" is a thorough understanding of positive life adjustment. By documenting premilitary, military, and postmilitary factors that influence general life adjustment among Vietnam veterans, this projects aims to provide valuable information to health care policy makers and providers.
HSR&D Funded: JP05-Massachusetts Veterans Eric (MAVERIC)
130. Racial Differences In Use of Lumbar Spine Radiographs: Results From the Veterans Health Study
Alfredo J Selim, Center for Health Quality, Outcomes, and Economic Research, VA Medical Center, Bedford, Massachusetts, G Fincke, Center for Health Quality, Outcomes, and Economic Research, VA Medical Center, Bedford, Massachusetts, XS Ren, Center for Health Quality, Outcomes, and Economic Research, VA Medical Center, Bedford, Massachusetts, RA Deyo, Center for Health Quality, Outcomes, and Economic Research, VA Medical Center, Bedford, Massachusetts, K Skinner, Center for Health Quality, Outcomes, and Economic Research, VA Medical Center, Bedford, Massachusetts, L Kazis, Center for Health Quality, Outcomes, and Economic Research, VA Medical Center, Bedford, Massachusetts
Objectives: Relatively little is known about the effects of race on the use of lumbar spine radiographs. In a prior study, we reported that non-white patients received more lumbar spine films than did white patients. It was not obvious whether the difference reflected the clinical characteristics of patients. In this study, we used comprehensive case-mix adjustment to answer this question.
Methods: We analyzed longitudinal data from the Veterans Health Study, an observational study of patients receiving VA ambulatory care. Five hundred and seventy four patients with low back pain (LBP) were identified through patient report of ever having had LBP, LBP that began more than three months ago, and a health care visit for LBP in the past year. Four hundred and one of these patients were available for follow-up at 12 months after the initial study visit. Participants were mailed the Medical Outcome Study Short Form Health Survey (SF-36) and had scheduled interviews that included the completion of a low back questionnaire, a comorbidity index, and a straight leg raising (SLR) test. We defined two groups based on race, whites and non-whites. We compared these two groups in terms of x-ray use, sociodemographics, comorbid conditions, low back pain intensity, radiating leg pain, straight leg raising, HRQoL, and low back disability.
Results: Non-white patients had lumbar spine films more often than white patients (48% compared with 27%, p= 0.02). Non-white patients had higher pain intensity scores (63±21 vs. 48±21, p<0.01) and were more likely to have radiating leg pain (76% compared with 57%, p=0.01) than white patients. Non-white patients had worse physical functioning (p=0.01), general health perception (p=0.05), social functioning (p=0.02), and role limitations due to emotional problems (p<0.01). At each low back pain intensity level, non-white patients received more lumbar spine films than did white patients (low pain intensity, 57% compared with 22%; high pain intensity 45% compared with 34%, p<0.01). Among patients with positive SLR test, non-white patients also had lumbar spine films more often than white patients (60% compared with 45%, p<0.01). However, after simultaneous adjustment for multiple clinical characteristics, race was no longer found to be an independent predictor of lumbar spine radiograph use. A positive SLR test remained associated with higher x-ray use, whereas better mental health status was associated with lower x-ray use.
Conclusions: There was greater use of lumbar spine radiographs by non-white patients compared to white patients. This remained true when patients were subcategorized by severity of low back pain or positive straight leg raising test. However, race had no influence when multiple clinical characteristics of the patients were controlled for simultaneously.
Impact: This study demonstrated the importance of careful and comprehensive case-mix adjustment when assessing apparent differences in use of medical resources.
HSR&D Funded: SDR No. 91-006.S
131. Contacting and Enrolling VA Physicians in Research: Role of Gender and Specialty Status
Stephanie Jean Smith, Center for Health Quality, Outcomes and Economic Research and Boston University School of Public Health, Bedford, MA, S Frayne, Center for Health Quality, Outcomes and Economic Research and BU School of Medicine, K Skinner, Boston University School of Public Health, K Freund, Boston University School of Medicine, M Moskowitz, Boston University School of Medicine
Objectives: While a growing number of studies include physicians as subjects, the methodologic issues in recruiting this population have not been well documented. We assessed variability by physician gender and specialty status and report the outcome of our efforts to contact and enroll physicians in our ongoing VA Variability Study (VAV).
Methods: We developed a protocol for the VAV (HSR&D GEN 97-0231) to meet the following study goals: (1) successfully reach physicians by telephone (e.g., gain the cooperation of clinic clerks, attentiveness to time of recruitment call); and (2) enlist physicians’ participation in a one hour face to face interview (e.g., informing the physician that he/she was randomly selected, offering a textbook in acknowledgement of the contribution of time). Our analytic plan for the results presented here was to: (1) determine the average number of calls necessary for contacting male/female, generalist/specialist physicians; (2) calculate the percent of male/female, generalist/specialist physicians agreeing to participate, and their reasons for refusal.
Results: This was a labor-intensive process that required the effort of a full-time Master’s level research assistant for six months. On average, fewer calls were necessary to reach women physicians (9.5 calls) than male physicians (15.5 calls), with male specialists requiring the most calls (17.0 calls). Timing of calls (e.g., before 8:00a.m.) influenced the likelihood of success. Of 83 eligible physicians, 80% (n=66) agreed to participate. Women physicians were more likely than men to agree to participate (89% vs. 74%). Women specialists had higher participation rates than women generalists (92% vs. 85%), while male generalists more often agreed than male specialists (82% vs. 64%). The most common reasons for refusal were time constraints (82 %); off-site transfer (12%); and lack of interest (6%).
Conclusions: Women generalists required the least number of calls and women specialists had the highest participation rate. These data can help inform future researchers seeking to contact and recruit male/female, generalist/specialist physicians as research subjects. Our experience demonstrates that the effort necessary to contact physicians and the success of the recruitment process varies by gender and specialty status of the physician participants.
Impact: Since many studies seek to enroll physicians as research subjects, researchers should be aware of the degree to which gender and specialty status can influence their accessibility by telephone and their willingness to participate. Limiting the number of telephone attempts may result in unintended under-representation of male specialists. Researchers seeking to enroll physicians as subjects need to be attuned to the potentially disproportionate non-participation rates of male specialists.
HSR&D Funded: GEN 97-0231
132. Prevalence of PTSD Among VA Ambulatory Care Patients
Avron Spiro III, MAVERIC, Boston VAMC, Boston, MA, CS Hankin, CHQOER, Bedford VAMC, LM Leonard, CHQOER, Bedford VAMC, I Stylianou, CHQOER, Bedford VAMC
Objectives: Posttraumatic stress disorder (PTSD) is a mental disorder that occurs in response to traumatic events such as combat. In addition to psychiatric sequelae, PTSD is associated with decrements in health status and increased medical services use. Thus, many patients suffering from PTSD are likely to seek medical care. Although several screening measures have been developed to assess PTSD, most have been examined among patients seeking or receiving mental heath treatment. We studied the utility of a screening measure to detect PTSD among patients in ambulatory care.
Methods: The Veterans Life Experiences Study (VLES) is examining the effects of mental disorders on health among male VA ambulatory care patients. We randomly selected 858 patients from the Veterans Health Study (VHS), a longitudinal study of 2,425 male Boston-area ambulatory care patients. We administered structured clinical interviews for PTSD (Brief Trauma Interview; Clinician-Administered PTSD Scale) and a self-report screening questionnaire (PTSD Checklist [PCL]) to 469 of these patients (mean age 65). We used Receiver Operating Characteristic (ROC) analysis to determine the optimally efficient cutpoint for the PTSD screening questionnaire.
Results: Ninety-four percent of patients reported lifetime exposure to one or more of the following traumatic events: combat (reported by 64%), life-threatening illness (46%), transportation accident (43%), witnessing a traumatic event (41%), natural disaster (33%), criminal victimization (22%), childhood physical abuse (13%), sexual assault (9%), or other traumatic event (8%). Based on the clinical interview, the prevalence of current PTSD was 10% in this sample of VA ambulatory care patients. Based on the ROC analysis, we identified the score of 42 on the PCL (range 17 to 85) as providing the optimally efficient cutpoint (sensitivity=.89, specificity=.90, kappa=.59). Based on this cutpoint, the screening prevalence of PTSD among the entire VHS cohort was 21%.
Conclusions: Traumatic exposure is nearly universal among VA ambulatory care patients. The PTSD screening prevalence of 21% among our sample of VA patients is approximately 20 times that found among community-residing men, and 50% higher than the prevalence among Vietnam theater veterans. Our findings suggest that PTSD is a common disorder among VA ambulatory care patients.
Impact: It is estimated that as many as 1 million VA patients who were exposed to combat may have PTSD. Patients with PTSD have been designated a "Special Care" population whose chronic and complex health care needs require extensive resources. Because the VA emphasizes quality of care in ambulatory services and allocates resources based upon the special needs of patients (e.g., those with PTSD), it is increasingly important to accurately identify patients with PTSD, to make the best use of these resources. Supported by VA HSR&D IIR 96-030.1 & SDR 91-006.S
HSR&D Funded: IIR 96-030.1
133. Can Integrating Psychiatric Clinical Nurse Specialists into VA Primary Care Clinics Improve the 3-month Outcomes of Patients with Depression? A Randomized Controlled Trial
Ralph W. Swindle Jr. PhD, Eli Lilly & Co; Roudebush VAMC, Indianapolis IN, M Weinberger, Roudebush VAMC and Indiana University School of Medicine, A Helmy, Roudebush VAMC and Indiana University School of Medicine, J Rao, Durham VAMC and Duke University School of Medicine, L Adams, Roudebush VAMC, A Zhou and G Eckert, Roudebush VAMC and Indiana University School of Medicine
Objectives: Depression is a highly prevalent, yet treatable, psychiatric disorder in VAMC and non-VAMC primary care settings. Unfortunately, despite efficacious pharmacologic and psychotherapeutic interventions, it is undertreated. This study examines whether integrating clinical nurse specialists (CNS) into primary care can improve the outcomes of veterans with depression.
Methods: We randomized two VAMC General Medicine Clinics (GMC). After training primary care physicians on PRIME-MD diagnostic criteria and pharmacologic and psychotherapeutic treatment options for depression, the two GMCs were randomized to the intervention (INT) or control (CON) group. GMC patients with regularly-scheduled appointments and who screened positive for depression by PRIME-MD criteria were potentially eligible. We excluded patients currently receiving mental health treatment, with histories of psychoses, at risk of suicide, or with life expectancies of less than six months. Patients providing informed consent completed the Beck Depression Inventory (BDI) and the RAND Patient Satisfaction Scale. In the CON GMC, PRIME-MD results were provided to physicians prior to seeing the patient. In the INT GMC, these results were provided to both the CNS and primary care physician. The CNS’s role was to: make treatment recommendations for depression to the physician, offer to help implement the recommendations, and monitor patients’ response between GMC visits. CNSs were not to change regimens without physician permission. CNSs were expected to follow patients for a minimum of two visits following enrollment. Blinded research assistants administered the BDI and RAND instruments via telephone interview 3 months after enrollment. Our primary analysis involves change scores (3 month minus baseline) for these two outcomes.
Results: 271 patients with depression were enrolled; to date, we have 3-month data for 240 (89%) patients. Change scores were not different for BDI (CON=0.05;INT=-0.07, p=0.68). Change in the 5 RAND sub-scales also showed no difference (p-values range from 0.17 to 0.52).
Conclusions: Integrating a CNS within a primary care setting was not effective in improving depressive symptoms or satisfaction of veterans with depression. Further analyses will examine 12-month patient outcomes (including quality of care), as well as process variables that may influence the impact of the intervention (e.g., the extent to which the intervention was delivered patient compliance with recommended treatment).
Impact: Integrative care for depression may fall well short of improving patient outcomes. Effective strategies may need to incorporate more direct strategies to implement clinical guidelines for depression. At a minimum, these preliminary study findings suggest great caution should be used when deciding about implementing integrated psychiatric care for depression within VA primary care settings.
HSR&D Funded: MSC 97-0101
134. A Case Finding Tool to Identify General Internal Medicine Patients with Congestive Heart Failure using VISTA Pharmacy Data
Edmunds M. Udris, VA Puget Sound Health Care System, Northwest HSR&D Center of Excellence, Seattle, WA, MB McDonell, VA Puget Sound Health Care System, Northwest HSR&D Center of Excellence, DH Au, VA Puget Sound Health Care System, Northwest HSR&D Center of Excellence; University of Washington, LC Chen, University of Rochester Medical Center, WM Tierney, Roudebush VA Medical Center; Indiana University School of Medicine, SD Fihn, VA Puget Sound Health Care System, Northwest HSR&D Center of Excellence; University of Washington
Objectives: An often-overlooked problem with the implementation of computerized clinical practice guidelines is correctly identifying patients to whom the guideline should be applied. The purpose of this study was to develop and validate a computerized tool to identify patients with congestive heart failure (CHF) who would be candidates for management according to a computerized version of a clinical practice guideline.
Methods: We conducted a retrospective cohort study in the General Internal Medicine Clinic at the VA Puget Sound Health Care System, Seattle Division. Subjects having at least one echocardiogram (Echo) or radionuclide ventriculogram between 1996 and 1998 were included. CHF was defined by left ventricular systolic dysfunction (LVSD) on either cardiac study. Subjects were randomly divided into two samples; one for prediction model development and the second for model validation. To determine the utility of clinical data in discriminating between patients with and without CHF, we examined demographic variables, pharmacy data, and inpatient and outpatient primary or secondary ICD-9 codes (428, 398.91) that were available from VISTA. The medications of interest included ACE inhibitors, AII antagonists, beta-blockers, calcium channel blockers, digoxin, diuretics, and nitrates.Discriminant analysis was used to build predictive models to identify patients with LVSD using pharmacy and demographic data. Accuracy of both the prediction model and diagnosis in VISTA were then compared to the presence of LVSD as determined by cardiac imaging.
Results: We identified 1314 subjects who had at least one cardiac imaging study. 536 (40.8%) subjects met study criteria for LVSD. The mean age was 68, 84% were Caucasian, and 97% males. Subjects with LVSD were slightly older than subjects without LVSD (69 versus 67, p=0.004), but were similar with regard to gender, race and marital status. Subjects with LVSD were more likely to have prescriptions for ACE inhibitors, aspirin, digoxin, loop diuretics, hydralazine and nitrates. (Carvedilol was not on formulary during the study period.) ACE inhibitors, aspirin, digoxin, loop diuretics, hydralazine, nitrates and age were included in the final predictive model. Use of ACE inhibitors, loop diuretics and digoxin exerted the greatest predictive power. The predictive model was 70% accurate (72% sensitivity, 69% specificity), while ICD-9 codes were 72% accurate (74% sensitivity, 71% specificity).
Conclusions: A model using pharmacy data identified most subjects with CHF, and performed equally well compared to ICD-9 codes in VISTA. Although the model did not demonstrate improvement over ICD-9 codes, it was able to predict LVSD using a relatively small number of the potential covariates. Considering that limited covariates were used for this study, the possibility for model refinement is considerable.
Impact: Before computerized guidelines can be effectively implemented, accurate methods of identifying patients must be developed. Although heart failure is the most common diagnosis in hospital subjects age 65 years and older, the clinical recognition and treatment of heart failure remains a difficult problem. Sub-optimal patient care can occur when CHF is unrecognized or misclassified. Whether misclassification occurs as a result of data errors in the clinical information system or clinician judgement, lack of recognition may adversely impact patient care and case-mix adjustment.
HSR&D Funded: CPG-970011
135. Health and Functioning among Four Eras of Veterans: Examining the Impact of Social and Economic Determinants
Valentine M. Villa, PhD, Veteran Identity Program, VA Greater Los Angeles Health Care System, Los Angeles, CA, N Harada, Veteran Identity Program, VA Greater Los Angeles Health Care System, J Damron-Rodriquez, Veteran Identity Program, VA Greater Los Angeles Health Care System, D Washington, Veteran Identity Program, VA Greater Los Angeles Health Care System
Objectives: The Purpose of this analysis is to document the level of health and functioning among WWII, Korea, Vietnam, and Persian Gulf era United States veterans participating in the National Survey of Veterans in 1993.
Methods: The data for this analysis come from the National Survey of Veterans (NSV) that was conducted in 1993 by the Department of Veteran Affairs. The survey obtained information on the economic, health status, and service use of 11,620 U.S. veterans. Utilizing descriptive and multivariate statistical techniques we examine the health status of WWII, Korea, Vietnam, and Persian Gulf veterans across three domains of health: self rated health, level of functioning i.e., activities of daily living (ADL), instrumental activities of daily living (IADL), and prevalance of chronic and acute disease. Through the use of ordinary least square (OLS) regression we examine the impact that war era, social, and economic determinants have on the health status indicators.
Results: The results reveal that Vietnam era veterans report lower self rated health and more difficulty with functioning across all ADLs and IADLs than the other eras of veterans. The Vietnam era veterans' disadvantage relative to WWII veterans persists even when controlling for age, income, education, and race/ethnicity. The WWII era veterans report higher levels of chronic diseases including high blood pressure and heart disease than other eras of veterans. Additionally, African American and Latino veterans of all four eras fare worse across all health measures when compared to White/Anglo veterans.
Conclusions: While some of the health disadvantages experienced by Vietnam veterans can be explained in part by social and economic indicators, they remain disadvantaged specifically in the area of functioning even when these indicators are held constant. Further analysis should examine the impact that other determinants including disease prevalence and health habits have on health status among this population.
Impact: Examination of the level of health of U.S. veterans as well as factors that are associated with poor health are important for designing health promotion and disease prevention programs that mesh with the needs of the veteran population, and therefore lead to improved health outcomes.
HSR&D Funded: ECV97-028
136. Diagnostic and Patient-Centered Approaches in Assessing Case-Mix: How Are They Correlated?
Margaret C. Wang, Center for Health Quality, Outcomes, and Economic Research, Bedford, MA, AK Rosen, Center for Health Quality, Outcomes, and Economic Research, S Loveland, Center for Health Quality, Outcomes, and Economic Research, LE Kazis, Center for Health Quality, Outcomes, and Economic Research, W West, Center for Health Quality, Outcomes, and Economic Research, D Berlowitz, Center for Health Quality, Outcomes, and Economic Research
Objectives: Approaches to case-mix adjustment have mostly focused on diagnosis-based models such as the Diagnostic Cost Groups (DCGs). Recently, the VA has shown increasing effort to capture patient self-perceived health status, resulting in the use of patient-centered methods such as the Veterans SF-36 (Short Form Survey). However, the relationship between diagnosis-based and patient-centered approaches to describing case-mix has not been explored. We sought to determine the level of correlation between these two different measures of case-mix.
Methods: Veterans SF-36 data were collected from 31,432 veterans nationwide based on a national probability sample of ambulatory visits. The SF-36V was used to evaluate the physical component (PCS) and mental component (MCS) of patient self-reported functional status (lower scores reflect poorer health). A DCG relative risk score (i.e., a measure of patient's health status relative to a given population as benchmark) was calculated for each respondent, using the DCG/HCC model which captures inpatient and outpatient diagnoses from one year (Jan. 97' to Jan. 98'). The relative risk scores were standardized to the Medicare population where the benchmark mean was equal to one. Higher relative risk scores reflect higher predicted cost for the current year.
Results: The VA sample was 95.3% male, with a mean age of 64 years; the Medicare standard population was 41.3% male with a mean age of 72. Therefore, the VA population was on average much younger, with a larger percentage of males than the Medicare population. The average DCG relative risk score of the VA sample was 0.912, slightly lower than the Medicare population. The VA sample had an average PCS score of 32.4 and MCS score of 42, both below the U.S. national average of 50. Correlation analyses showed a highly significant relationship between PCS scores and DCG scores (R=-0.206, P<0.0001) and a smaller association between MCS scores and DCG scores (R=-0.125, P<0.0001).
Conclusions: We found a significant but small correlation between the patient-centered SF-36V and diagnosis-based DCGs. Our results suggest that diagnosis-based and patient-centered risk adjustment approaches may reflect different dimensions of disease burden in the VA population, particularly for patients with mental health problems, since the association between MCS scores and DCG scores was smaller.
Impact: The VA has recently invested efforts to employ diagnosis-based and patient-centered risk adjustment approaches. Further research is required to understand how well each independently predicts resource utilization and patient outcome.
HSR&D Funded: MPC-97-009
137. Brief Screening for Anxiety, Depression and Substance Abuse in Spinal Cord Injury
Kenneth R. Weingardt, PhD, VA Palo Alto Health Care System, Menlo Park, CA, J Hsu,VA Palo Alto Health Care System, M Dunn, VA Palo Alto Health Care System
Objectives: 1) To evaluate whether systematic, brief screening for anxiety, depression and substance abuse disorders in a VA SCI inpatient population can help identify patients with unmet treatment needs. 2) To ascertain the prevalence of, and covariation among these disorders in this population. 3) To explore the interrelationships among demographic variables, hospital utilization variables, coping strategies, and screening measures.
Methods: Participants were 117 veterans who were admitted as inpatients to the SCI Service at the VAPAHCS from 10/98 through 5/99. Screening interviews were completed at bedside within one week of admission. The brief (10 minute) interview included the Alcohol Use Disorders Identification Test (AUDIT), and screening items for depression, anxiety, panic, PTSD, and substance abuse. Patients were also asked to describe the ways that they typically cope with life stress, and were asked to rate their perceived health relative to ambulatory and disabled populations.
Results: The mean age of participants was 57.4 (SD=14.2), with a mean length of injury of 19.8 years (SD=14.7). 47% of the sample was Paraplegic, and 53% Quadraplegic. 16.2% of patients screened positive for depression, 8.5% screened positive for anxiety, 24% screened positive for Panic, and 21% screened positive for PTSD, 6.2% of patients screened positive for alcohol problems. During follow-up interviews, many positive screens (particularly those for panic and PTSD) were found to be "false positives" i.e. the individual screened positive, but did not actually meet criteria for that disorder. Individuals who screened positive for depression and anxiety were more likely to endorse "escape or avoidance" and "self-controlling" ways of coping with stress.
Conclusions: Total number of admissions in the previous year was identified as the only significant predictor of screening positive on any psychological or substance abuse measure. Consequently, the authors recommend that psychosocial staff spend more time with those patients who are admitted more than once per year as they are much more likely to suffer from anxiety or depression. The relatively low rates of alcohol and other substance abuse problems are likely attributable to the demographic characteristics of the sample: being older and having been injured for a long period of time.
Impact: While brief screening has demonstrated great utility in primary care settings, the present results suggest that systematic brief screening for psychological and substance abuse disorders in an older SCI inpatient population is probably unwarranted. The present results suggest that SCI psychosocial resources might best be concentrated on those patients who are admitted multiple times per year.
138. What is Important to Patients Deciding Whether to Undergo Coronary Revascularization
Jeff Whittle, MD, MPH, VA Pittsburgh Healthcare System, Pittsburgh, PA, J Conigliaro, VA Pittsburgh Healthcare System, CB Good, VA Pittsburgh Healthcare System, KC Goldberg, Durham VA HSRD Center of Excellence, P Sankar, University of Pennsylvania, M Skanderson, VA Pittsburgh Healthcare System
Objectives: Studies of patient decision making regarding coronary revascularization are handicapped by lack of understanding of the factors patients find most important. We assessed what factors most influenced such decisions overall and in subgroups defined by age, race and site of care.
Methods: We prospectively enrolled consenting patients who were offered or received percutaneous angioplasty (PTCA), coronary bypass grafting (CABG) or minimally invasive coronary bypass (midCAB)) at a VAMC or its university affiliate (U). We attempted to survey patients prior to the procedure when possible, but interviewed them post procedure if necessary. Patients were asked to name the most important consideration in their decision whether to undergo the procedure in an open-ended fashion, then asked to rank (1 to 10) the importance of specific factors, including expected survival benefit, symptom relief and the physician’s recommendation, among others. We dichotomized the responses to these latter based on whether they were >=8.
Results: This report includes the first 348 participants (PTCA 194, CABG 144. midCAB 10), all of whom accepted the procedure. The most common factors cited in open ended responses were improved survival (39.1%) or symptoms (31.9%), trust in the doctor or hospital (19.5%), and the results of diagnostic tests (12.1%). There was no difference in the frequency of these responses between blacks and whites, between persons older than or younger than 60 years old, and between persons treated at VA or U, except that patients treated at the U were more likely to cite test results (14.8 vs 6.3%, p = 0.02) or trust (22.8 vs 12.6%, p = 0.03). Trust was more likely to be an important factor for persons who completed the survey after the procedure was completed (25.4 vs 8.3%, p < 0.001). Patients endorsed more specific factors as important (rating >=8) than they had named in open-ended questions. Most patients endorsed the importance of the doctor’s recommendation (90.4%), improved survival (85.3%), and improved symptoms (78.6%) or functional status (80.9%). However, as many as 40% of patients endorsed other factors. Patients >59 years of age were more likely to endorse the importance of the experience of others (37.7 vs 22.0%, p = 0.005), but were otherwise similar to those under 60. Although there were only 21 blacks in this sample, they were less likely than whites to endorse doctor’s recommendation as important (76.2 vs 91.3%, p = 0.04). VA patients were more likely to endorse several factors, including medications and procedural risk but less likely to endorse doctor’s recommendation and family input (all p < 0.05).
Conclusions: Physician recommendations, and improved survival and symptoms are similarly important in all groups studied. Patients, who differed in age, race and treatment site were more alike than different in the reasons they considered important, though VA patients had several areas of difference.
Impact: The factors that should be discussed with patients are similar in groups that differ in their likelihood of undergoing procedures. Work to enhance doctor patient communications is likely to be broadly applicable, though the VA population may require special attention.
HSR&D Funded: ECV 97 - 026
139. The Relationship Between Post-stroke Depression and Quality of Life in Veterans with Stroke
Linda S. Williams, Roudebush VAMC, Indianapolis, IN, G Redmon, Roudebush VAMC, M Weinberger, Roudebush VAMC
Objectives: Post-stroke depression (PSD) occurs in at least 30% of patients within the first 6 months after ischemic stroke and has been linked to worse functional disability and increased mortality. The relationship between PSD and health-related quality of life (HRQL) has not been explored. The purpose of this study was to investigate the influence of PSD on specific domains of post-stroke HRQL.
Methods: Ischemic stroke survivors with no significant language or cognitive impairments were recruited from one VAMC as part of an ongoing study of version 2.0 of the SS-QOL, a newly-developed and validated stroke-specific HRQL scale. The SS-QOL has 12 individual domains; overall scores and individual domain scores range from 1.0 (worst) to 5.0 (best). The following measures were interviewer-administered at 1, 3 and 6 months post-stroke: SS-QOL, SF-36, CES-D (depression scale), NIH Stroke Scale (stroke impairments) and the Barthel Index (ADLs). PSD was defined as CESD score >= 16. Demographic and depression history variables, as well as overall SS-QOL score and domain scores, were compared using Chi-square or t-tests between patients with and without PSD. Instrument scores were also compared between patients rating HRQL the same as before the stroke vs. those with worse HRQL.
Results: Characteristics of the first 36 subjects were: 32 (89%) male, 28 (78%) white, 20 (56%) with an identified proxy. Most patients had mild to moderate stroke, with mean 1-month NIH Stroke Scale 2.4 (0-42, lower better) and Barthel 94 (0-100, higher better). Despite the relatively mild stroke severity, HRQL was significantly affected with mean (sd) 1-month SS-QOL 3.7 (.68), SF-36 physical function score 50 (25), and SF-36 role physical score 22 (37) . At 1 month, 13 (36%) screened positive for PSD. Baseline variables significantly associated with PSD were family history of depression (46% vs. 0, p= .001) and African-American ethnicity (49% vs. 10%, p= .02). SS-QOL score was significantly lower in those with PSD, and there was a trend toward worse ADL function in PSD patients (Barthel 89 vs. 96, p = .14). Overall SS-QOL score and all SS-QOL domains except vision and language were significantly lower in patients with PSD, with energy (1.9 vs. 3.0, p = .009) and mood (3.0 vs. 4.0, p = .001) domains most strongly related to PSD. CES-D and SS-QOL scores, as well as mobility, work, energy, mood and social roles domains, were significantly worse in patients who rated 1-month HRQL worse than pre-stroke. No SF-36 domain scores were significantly lower in PSD patients or in those with worse HRQL.
Conclusions: PSD is common in veterans after stroke and has pervasive influence on most aspects of HRQL. Patients with a family history of depression and African Americans may be more likely to develop PSD. The SS-QOL is more sensitive than the SF-36 to changes in mood and HRQL post-stroke.
Impact: Effective depression treatments exist but have not been well studied in PSD. A RCT of PSD treatment is needed since PSD strongly influences HRQL, functional recovery and mortality post-stroke.
HSR&D Funded: CD 98307-2
140. Diagnosing and Treating Chronic A Bacterial Prostatitis: A Systematic Review of the Literature
Timothy J. Wilt, MD, MPH, Veterans Affairs/VISN 13 Center for Chronic Diseases Outcomes Research, Minneapolis, MN, R MacDonald, MS, Veterans Affairs/VISN 13 Center for Chronic Diseases Outcomes Research, Minneapolis, MN, M McNaughton Collins, MD, MPH, Massachusetts General Hospital, Boston, MA
Objectives: The optimal management of chronic abacterial prostatitis is not known. Medical literature was systematically reviewed to answer the following two questions about patients with abacterial prostatitis:1) Are there accurate, reliable tests to diagnose this condition? 2) Are there effective therapies to treat it?
Methods: Studies were identified through a search of MEDLINE (1966-99), the Cochrane Library, bibliographies of identified articles and reviews, and contact with an expert. Diagnostic test articles were included if they were randomized or controlled trials. There were no language restrictions. Two investigators for each selected article on diagnosis and treatment independently extracted key data on study design features, subject characteristics, diagnostic test and treatment characteristics, and outcomes of the studies.
Results: 19 diagnostic test articles and 14 treatment trials met the inclusion criteria. The disparity between studies did not permit quantitative analysis or pooling of the findings across trials. The diagnostic articles evaluated: infection, inflammation, immunology, and biochemistry; psychological factors; and ultrasonography. There were a total of 1384 men (mean age range 33-67). The treatment trials involved: medications used to treat benign prostatic hyperplasia; anti-inflammatory medications; antibiotics; thermotherapy; and miscellaneous medications. There were a total of 570 enrollees (mean age 38-45). None of the trials were done in the United States.
Conclusions: There is no gold standard diagnostic test for chronic abacterial prostatitis and the available diagnostic studies are of low methodologic quality. The treatment trials are few, weak methodologically, and involve small sample sizes. The routine use of antibiotics and alpha blockers for chronic abacterial prostatitis is not supported by the existing evidence.
Impact: The results of this systematic review identify the gaps in knowledge regarding diagnosis and treatment of chronic abacterial prostatitis. Closing these gaps will improve health care delivery for men with prostate diseases.
HSR&D Funded: MDR 96-001
141. Information-Seeking Behaviors and Knowledge of Surgical Treatment Choices in a Population-Based Sample of Breast Cancer Patients
Judith K. Zemencuk, VA Center for Practice Management and Research Outcomes, Ann Arbor, MI, PM Lantz, The University of Michigan School of Public Health, Ann Arbor, MI, RA Hayward, VA Center for Practice Management and Research Outcomes, Ann Arbor, MI, SJ Katz, The University of Michigan Department of Internal Medicine, Ann Arbor, MI
Objectives: Women constitute the fastest growing segment of VA eligible veterans, increasing demand for breast cancer care at VAMCs. Most women with breast cancer will undergo surgery and be asked to choose between surgical treatments with different risks and benefits. We describe information-seeking behaviors and knowledge of the risks and benefits of surgical treatment options for early-stage breast cancer among women who report being given a choice between surgical treatments.
Methods: A sample of women newly diagnosed with ductal carcinoma in situ or non-metastatic invasive breast cancer was selected from the Metropolitan Detroit Cancer Surveillance System (a SEER tumor registry). Of 242 eligible cases, 183 (76%) completed a self-administered questionnaire that collected information on factors related to surgical treatment choice. Of these, 122 (67%) reported being given a choice between surgical treatments.
Results: Of women who actively chose between surgical treatments, 49% (60)had lumpectomies and 51% (62) had mastectomies. Nearly two-thirds (64%) requested opinions from more than one physician. For example, in addition to their own surgeon, 42% saw another surgeon, while 40% sought advice from an oncologist. Most women took an active role in gathering information regarding surgical treatments. For example, 80% reported reading magazines and articles about breast cancer and its treatments and 62% reported reading scientific articles about surgical treatments. The vast majority also reported asking a lot of questions concerning treatment options. For example, 86% reported asking about the risks involved with each surgical treatment alternative and 90% wanted to know the possible outcomes of each. Despite their reported efforts at collecting information, women's knowledge about differences in the risks and benefits of surgical treatments was low. For example, 25% believed the risk of recurrence to be the same for lumpectomy with radiation and for mastectomy, while 45% did not know if there is a difference in recurrence rates between the two treatments. Almost half of women (45%) did not know if survival rates differed between the two treatments. Nonetheless, the vast majority felt they had enough information to make a decision between treatments. For example, 94% felt that they were adequately informed about the issues important to their decision and 89% agreed that their surgeon had given them a balanced description of all of their surgical options.
Conclusions: Most women reported taking an active role in the decision-making process. Many sought additional opinions, asked questions about the risks of each treatment, and read articles about breast cancer treatments. Despite their efforts, knowledge about differences between surgical treatments was low.
Impact: With an aging population, more women veterans will face a breast cancer diagnosis. Our study's findings indicate that women may not have sufficient information to make an informed decision regarding surgical treatment options for breast cancer. While not all women may wish to take an active role in their treatment decision, assuring that those who do want to take an active role have the information necessary for making a choice that best meets their medical and personal needs is critical to the provision of quality care.
142. Older Adults with Osteoarthritis: Pain and Disability
Kathryn J. Burks, Harry S Truman Memorial Veterans Hospital, Columbia, MO, V.S. Conn, University of Missouri-Columbia
Objectives: Older adults with osteoarthritis often encounter disease-related pain and disability that have profound effects on quality of life and independence. The purpose of this study was to examine the ability of active pain coping, passive pain coping, self-care activities, depression, extent of osteoarthritis involvement, and overall self-rated health to predict older adults' pain and disability.
Methods: Previously developed and tested instruments were used to measure active and passive pain coping (Vanderbilt Pain Management Inventory), pain intensity (McGill Pain Questionnaire), depression (Profile of Mood States), and disability (Arthritis Impact Measurement Scales). A researcher developed measure assessed the extent of arthritis (number of joints), self-care behavior (actions weighted by frequency of performance), and self-rated health. Interviewers administered these instruments to 93 men and women aged 60 years and over.
Results: The mean age of participants was 73 and they reported an average of three chronic illnesses. Results of the regression analysis demonstrated that depression, passive coping, and health were significant predictors of pain (R-squared=0.49) and disability (R-squared=0.45). Further analysis demonstrated that passive coping was positively correlated to pain (r=0.58) and disability (r=0.51), while active coping was negatively correlated with pain (r= -0.21) and disability (r= -0.28).
Conclusions: When passive coping strategies are used, the control of pain is often relinquished to others and pain is allowed to interfere with other life areas. In contrast, active coping strategies suggest greater control by the older adult and only a limited impact of pain on other areas of the person's life.
Impact: The findings of this study suggest that coping strategies should be further examined as a potential domain for intervention research to reduce pain and disability in this vulnerable population. Due to the fact that self-management strategies are often considered to be the foundation of treatment for people with osteoarthritis, interventions desgined to enhance active coping strategies, decrease depression, and increase self-rated health status may have the overall effect of decreasing pain and disability as well.
143. Use of Depot Antipsychotic Medication in Seriously Ill Schizophrenic Patients
Laurel Copeland, MPH, VA Center for Practice Management and Outcomes Research, Ann Arbor, MI, M Valenstein, SMITREC, VA Center for Practice Management and Outcomes Research, R Owen, Center for Mental Healthcare and Outcomes Research, F Blow, SMITREC, VA Center for Practice Management and Outcomes Research, S Visnic, SMITREC, VA Center for Practice Management and Outcomes Research
Objectives: Anti-psychotic medication is an efficacious treatment for schizophrenia, with the majority of patients showing symptomatic improvement with these medications. However, medication non-compliance is common. The Schizophrenia Patient Outcome Research Team (PORT) recommends that clinicians strongly consider depot neuroleptics for patients who have difficulty complying with oral medication. In this study, we examined use of depot neuroleptics among a seriously mentally ill cohort of veterans in 14 VAMCs.
Methods: A cohort of 1303 veterans with schizophrenia were enrolled in a study of enhanced programming or standard care between January 9, 1991 and December 19, 1995. These veterans had had either 150 or more documented days of hospitalization or 5 or more inpatient admissions in the year prior to enrollment. At the time of enrollment, clinicians completed an instrument that included patient medications and an assessment of patient compliance within the last year and adverse consequences due to non-compliance. Patients completed an assisted survey that included a report of medication side effects.
Results: Only 17.8% of this severely ill schizophrenic cohort were receiving depot antipsychotic medication at enrollment. However, 40% of enrolled patients were judged to have had compliance problems in the last year, resulting in adverse consequences, including 39% percent of patients who were still on oral medications at enrollment. Cross-sectionally, patients on depot medications complained of more medication side effects and had BPRS scores that were similar to patients on oral medications.
Conclusions: Depot neuroleptic medication may be underused in SMI veterans. Many patients with compliance problems remain on oral antipsychotics even in this high-utilization group. However, patient perceptions of increased side effects on depot medication may prove a barrier to their use. Cross-sectionally, levels of psychotic symptoms are similar for patients on depot or oral antipsychotic medication. Longitudinal studies are needed to examine the relationship between depot use and changes in symptoms in these severely ill patients.
Impact: Depot neuroleptic medication may be underused in SMI populations, including veteran populations with high rates of hospitalization. Quality improvement programs may wish to encourage providers to consider depot medications and develop systems that facilitate their use. However, QI programs and providers will also need to pay close attention to patient complaints of side effects.
144. Determining Screening Classification of PSA Tests and Digital Rectal Exams: Developing an Algorithm to Differentiate Screening and Diagnostic Tests
Preet K. Dhillon, University of Washington, Dept. of Epidemiology, School of Public Health and Community Medicine, Seattle, Washington, GE Reiber, Seattle VA, Associate Professor Health Services, Epidemiology, University of Washington, School of Public Health and Community Medicine, NS Weiss, Professor of Epidemiology, University of Washington, School of Public Health and Community Medicine
Objectives: The objectives of this study were to develop, as part of a case-control study of the efficacy of screening for prostate cancer, an algorithm for determining whether prostate-specific antigen (PSA) tests and digital rectal exams (DRE) in study subjects were or were not intended as screening tests.
Methods: Veteran and nonveteran male members of the Kaiser Permanente population in the Northern California region were eligible for this study. Study subjects include men aged 50-84 years who died of prostate cancer between 1996-1997, and men who were alive and free of prostate cancer on the diagnosis date of the fatal case. Veteran status is obtained by linking personal identifiers to a database of servicemen and VA services recipients. Information on the subject's PSA screening history is based on chart reviews of outpatient medical records, lab reports and available computerized data. It is important to reliably distinguish whether a test was done for screening or to evaluate a symptom for diagnosis, since failing to do so will result in a spuriously high frequency of tests attributed to screening. Our algorithm includes selection of a reference date that marks when the suspicion of cancer (or manifestation of cancer) and investigation leading to diagnosis first began. The correct classification of each test prior to and including this reference date depends on multiple factors such as, reason for visit, presence of prostatitis/BPH, previous test results, patient complaints/symptoms, the nature of clinical signs and symptoms over time, and clinical suspicion. This algorithm also considers an exhaustive list of voiding, pain, sexual dysfunction, gastrointestinal/rectal and other symptoms.
Results: In older men, defining screening tests is difficult since other more common conditions such as BPH, can produce the same symptomatology. We evaluated the likelihood that a particular symptom was a manifestation of prostate cancer given the high prevalence of BPH and other conditions (classification of symptoms based on likelihood of being due to cancer, BPH or other). The classification scheme includes an exhaustive list of voiding, pain, sexual dysfunction, gastrointestinal/rectal and other symptoms. Our algorithm divides PSA tests & DRE's into one of four categories: routine screening, probable screening, unlikely screening and not screening. The algorithm is based on reason for visit (ie-routine, referral based on suspicious sign/symptom); previous PSA/DRE results; whether prostate-related symptoms were present and if so, the nature of these symptoms (new, stable or worsening); DRE results; and clinical suspicion (flow charts not included).
Conclusions: In this study, we have developed an algorithm to determine the screening status of PSA tests and DRE's administered during a 10-year reference period. Applying this algorithm to all study subjects while blinded to case-control status, will minimize misclassification of exposure (screening tests).
Impact: Results on PSA screening efficacy are vital for making future recommendations, affecting the delivery of care in both HMO and VA settings alike. By employing the methods described above, we will hone in on a more valid measure of screening status in both asymptomatic and symptomatic men. This algorithm may be widely applicable to other areas of screening efficacy research as well.
HSR&D Funded: (received partial supplemental funding recently, no study number as of yet)
145. The Impact of Depressive Symptoms on Health Status in Patients with Chronic Pulmonary Disease
Bradford Felker, Dept of Veterans Affairs Puget Sound Health Care System, Seattle, WA, W Katon, University of Washington Department of Psychiatry and Behavioral Sciences, SC Hedrick, Department of Veterans Affairs NW, HSR&D Center of Excellence, Seattle, WA, J Rasmussen, West Los Angeles Healthcare Center Division, General Internal Medicine, MB McDonell, Department of Veterans Affairs NW, HSR&D Center of Excellence, Seattle, WA, SD Fihn, Department of Veterans Affairs NW HSR&D Center of Excellence, Seattle, WA
Objectives: This study evaluated the impact of depressive symptoms on self-reported measures of general and condition-specific health related quality of life (HRQoL) in patients with chronic pulmonary disease.
Methods: This is a cross-sectional analysis of HRQoL measures completed by patients enrolled in the Department of Veteran Affairs Ambulatory Care Quality Improvement Project (ACQUIP). At the time of this study, ACQUIP involved 25,420 patients who were followed in six different VA General Internal Medicine Clinics. From this population, 3281 patients reported a history of chronic pulmonary disease and returned the Seattle Obstructive Lung Disease Questionnaire (SOLDQ). Of these 3281 patients, 733 had a score of 1.75 or greater on the Hopkins Symptom Checklist indicating significant depressive symptoms.
Results: Patients with depressive symptoms were younger, less educated, current smokers, and had more co-morbid medical conditions than those patients without significant depressive symptoms. After controlling for these demographic differences, patients with depressive symptoms had clinically and statistically worse general and pulmonary health as reflected by lower scores on all sub-scales of both the Medical Outcomes Study Short Form 36 and the SOLDQ.
Conclusions: Patients with chronic pulmonary disease and depressive symptoms reported significantly more impaired functioning and worse health status when compared to those patients without depressive symptoms.
Impact: Because there are highly effective treatments for depression, selective screening of patients with chronic pulmonary disease would identify a group that could potentially benefit from treatment interventions.
HSR&D Funded: SDR-96-002
146. A Comparison of Methods for Summarizing Studies of Diagnostic Test Performance
Michael K. Gould, MD, MSc, VA Palo Alto Health Care System, Palo Alto, CA, C Maclean, VA Palo Alto Health Care System, DK Owens, VA Palo Alto Health Care System
Objectives: To compare methods for constructing summary receiver operating characteristic (ROC) curves to describe the performance of two diagnostic tests for mediastinal lymph node staging in patients with non-small cell lung cancer.
Methods: We performed a meta-analysis of 16 studies that compared two imaging technologies, computed tomography (CT) and positron emission tomography (PET), for the diagnosis of mediastinal lymph node metastasis in patients with non-small cell lung cancer. We constructed 2x2 tables for CT and PET in each study. We calculated the true positive rate (TPR) and false positive rate (FPR) for CT and PET in each study. We then determined the log odds TPR/FPR for both imaging tests in each study. We estimated the common log odds ratio for CT and for PET by using standard least squares regression, by the Mantel-Haenszel method, and by calculating the median and the mean. We then transformed each summary log odds ratio to define the maximum joint sensitivity and specificity (MJS&S) for CT and PET, the point on the summary ROC curves where sensitivity and specificity are equal.
Results: For CT, the mean and median log odds TPR/FPR were 1.83 and 1.72, respectively. The common log odds ratio was 1.89 by least squares regression, and 1.83 by the Mantel-Haenszel method. The corresponding MJS&S ranged from 0.70 (median) to 0.72 (regression). For PET, the mean and median log odds TPR/FPR were 4.87 and 5.20, respectively. The common log odds ratio was 4.90 by least squares regression, and 4.48 by the Mantel-Haenszel method. The corresponding MJS&S ranged from 0.90 (Mantel) to 0.93 (median). For PET, the Mantel method underestimated the log odds ratio as determined by least squares regression, despite the fact that PET sensitivity and/or specificity was perfect in six studies, requiring us to add 0.5 to each cell in the 2x2 table for each of these studies before performing regression. This continuity correction biases the results of the regression method to produce lower values for the log odds ratio.
Conclusions: For highly sensitive and/or specific diagnostic tests, such as PET in this example, different methods for constructing summary ROC curves produce different results. The Mantel-Haenszel method is thought to be useful when studies demonstrate perfect sensitivity and/or specificity, because this method does not require the use of a continuity correction that will bias the results downwards. In this example, the Mantel method seriously underestimated the common log odds ratio for PET. We suspect that the estimate obtained by least squares regression also underestimates the true common log odds ratio, due to the use of the continuity correction. For tests with intermediate sensitivity and specificity, such as CT, the choice of method does not appear to be as important.
Impact: Better methods are needed for summarizing the results of studies of highly sensitive or specific diagnostic tests.
HSR&D Funded: RCD-99023-1
147. Barriers and Facilitators to HAART Therapies
Jeanne K. Kemppainen, PhD, RN, VA Palo Alto Health Care System, Palo Alto, California
Objectives: The objective of this study is to identify barriers and facilitators to the newer highly active antiretroviral therapies (HAART) as perceived by culturally diverse patient groups and their health care providers. Specifically, this research examined unique factors associated with HAART adherence in culturally diverse VA populations related to gender, age, ethnicity, and IV drug use. While much information is known about rates of adherence to HIV/AIDS medication regimens, little is known about the full range of patient behaviors or circumstances that influence HAART adherence in these diverse populations.
Methods: The critical incident technique developed by John Flanagan (1954) was used as a methodology for this qualitative study. Through a series of brief, focused interviews, patients with HIV/AIDS were asked to recall specific incidents that affected their adherence to HIV medications. The sample consisted of 50 outpatients receiving care through a HIV/AIDS clinic in northern California. Participants included male and female patients over age 50, patients of racial and ethnic diversity, IV drug users and their physician and nurse providers. Data obtained through the critical incident interviews was analyzed through an inductive classification process and listings of adherence barriers and facilitators were generated. Data was compared across the diverse patient groups.
Results: Barrers affecting medication adherence included forgetting, lack of knowledge about HIV medications, fatigue, using street drugs, medication side effects, life stress, and depression. Strategies that patients used to facilitate medicine adherence included the use of pagers, timers, alarm clocks, medisets, reminders from friends and family, and adjusting medication schedules to fit daily routines. Relationships with health care providers play a key role in facilitating adherence.
Conclusions: Study findings provide important information on patient perceptions about medication adherence as well as data for designing interventions aimed at enhancing adherence to antiretrovirol therapies.
Impact: This study was prompted by the current major social and financial impact of this issue on the VA health care system, individual patients, and their families. In addition, the potential threat to public health and safety created through the development of a treatment resistant strain of the HIV virus made this study even more imperative. Recognition of the varied responses across diverse population groups provides an essential foundation for research aimed at developing and testing multifaceted adherence interventions.
148. Adherence of HIV+ Patients to Highly Active Antiretroviral Therapy (HAART)
Martha Jane Mohler, So. AZ VA HCS and University of Arizona, Tucson, AZ, NA Ampel, So. AZ VA HCS and University of Arizona, Tucson, AZ, C Wendel, So. AZ VA HCS and University of Arizona, Tucson, AZ, K Kroesen, So. AZ VA HCS and University of Arizona, Tucson, AZ, AL Gifford, VA San Diego HCS and University of California San Diego, San Diego, CA, J Bormann, VA San Diego HCS and University of California San Diego, San Diego, CA
Objectives: Describe patient factors associated with HIV viral load, self-reported adherence (SR-ADH), and Medication Event Monitoring System adherence (MEMS-ADH) to HAART (two protease inhibitors or three or more antiretroviral medications) in HIV-infected veterans receiving care at the So. AZ VAHCS HIV Clinic. Aims include 1) develop a clinical screening tool used to identify those at risk of non-adherence, and 2) develop interventions to maximize HAART adherence.
Methods: This cross-sectional study of HIV-positive veterans on HAART regimens examined regimen, patient-provider relationship, and patient-related predictors related to: HIV viral load, weekly SR-ADH, and monitored adherence of one antiretroviral medication within the HAART regimen by MEMS-ADH, using logistic regression odds ratios (OR). 67 subjects completed self-report and 60 MEMS cap data.
Results: Subjects were male and 78% Anglo. 54% were gay (without intravenous drug use), 21% intravenous drug users, and 10% both. 70% completed >=12 years of education. Antiretroviral history averaged five years, and 43% had an AIDS diagnosis. 35/67 (52%) had an HIV <=50 copies (undetectable). 41/67 (62%) had 100% SR-ADH over the previous seven days, and 41/60 (68%) had MEMS-ADH >=90%. SR-ADH and MEMS-ADH were correlated (r=.56 p<.0001). SR-ADH >=90% was associated with an undetectable viral load (OR 3.3, CI 1.2-9.2), and there was a trend toward an association between MEMS-ADH as a continuous variable and log viral load (p= 0.074). MEMS-ADH >=90% was not associated with an undetectable viral load (OR 1.2, CI .39-3.5). Disruption of pill-taking by activities of daily living was associated with a six-fold increased risk of viral load >50 copies (OR 5.9, CI 1.4-24.8). Belief in one's ability to take medication as ordered (OR 32, CI 4.4-234) was highly associated with SR-ADH >=90%. Employment was associated with much lower odds of SR-ADH >=90% (OR 0.12, CI .02-.67). Non-white subjects had lower odds of MEMS-ADH >=90% (OR 0.18, CI .04-.93). A 10-point increase in HIV-MOS energy/fatigue score (OR 1.46, CI 1.04-2.1), and belief in one's ability to take medication as ordered (OR 2.5, CI 1.2-5.0) were associated with greater odds of MEMS-ADH>=90%.
Conclusions: SR-ADH and HIV viral load were associated. Fewer than 70% of patients adhered at >=90% based upon SR-ADH and MEMS-ADH. Poor self-efficacy, and disruption of pill-taking by activities of daily living, may affect non-adherence. Assessing patient self-efficacy (e.g., belief in ability to take medications as ordered) may help to determine HAART ADH behavior.
Impact: VA guidelines recommend HAART for HIV-positive veterans with CD4 counts <500 cells/mm3. Concomitant HAART adherence support programs are necessary to optimize this policy. This study has enhanced our knowledge of remediable factors related to HAART ADH. The next phase of this project includes qualitative exploration of associated factors in both high-ADH and low-ADH individuals. Findings will assist in the design and implementation of VA-based HAART ADH programs. These interventions should result in optimized adherence, improved clinical and functional status, quality of life, and cost-effectiveness. In addition, study results may also contribute to understanding of adherence in other complex regimens such as diabetes, and congestive heart failure.
HSR&D Funded: IIR 98-011-1
149. The Measurement of Brain Functioning in Low Level Neurological States
Theresa Louise-Bender Pape, Edward Hines Jr. VAH and Northwestern University, Chicago, IL
Objectives: Severe brain injury results in a state of impaired consciousness, which consists of three sub-syndromes: comatose, vegetative (VS) and minimally conscious states (MCS). Under-standing progression through these sub-syndromes into consciousness is uncertain because there are no valid measures of brain functioning and because there is little evidence to indicate which factors are important for recovery. The absence of reliable and valid measurements of brain functioning limits our ability to predict the course and rate of neurological recovery. This study evaluated the reliability and validity of the Disorders of Consciousness Scale (DOCS); a clinical tool for measuring brain functioning.
Methods: Severely brain injured participants aged 18 years and older (n = 25) were recruited from the Minneapolis VAMC and the Rehabilitation Institute of Chicago. Serial DOCS evaluations were conducted yielding 43 response sets. A Glasgow Coma Scale (GCS) score and an in-dependent classification of Coma, VS or MCS were obtained. Each participant was observed for indications of consciousness. Scaling analysis, a methodological approach to analyzing responses to test items, was used to examine the quality of the DOCS measurements. Traditional analyses were used to evaluate the predictive potential of the DOCS.
Results: The DOCS employs a three-point rating scale to describe cerebral functioning (i.e., 0 = No Response (NR); 1 = Sub-cortical Response (SCR); 2 = Cortical Response). Results indicate that the raters used the scale as was intended, to reflect progressively improving levels of cerebral functioning. Inter-rater reliability of the DOCS was good for discriminating between NR and SCR (k = .94) and good for discriminating between sub-cortical and cortically mediated responses (k = .78). The separation index (separation = 1.39) tells us that the cerebral functioning test items reliably fall within 2 groups. This value is analogous to a KR-20 = .66. Concurrent validity was evaluated by computing correlation coefficients between the DOCS and GCS measures; a moderate association was observed (r = .48; n = 20 persons). Evidence of construct validity includes the fit of each item relative to the other items and should be < 1.3 to reflect coherence between item ratings and the overall state of impaired conscious-ness and >= .7 to reflect independence. 17 test items had fit statistics >=1.3 and < .7 suggesting redundancy or non-relevance. A single DOCS measure and the recovery of consciousness are related (R = .2; p< .10; OR = 3.9; 90% CI = 1.03, 15.3; n = 20 persons). A patient was 3 times more likely to have recovered con-sciousness if they received a high DOCS rating.
Conclusions: The DOCS & the GCS measure distinct but related constructs of brain functioning. The cerebral functioning test items reliably discriminate between VS and MCS. The DOCS shows promise for predicting the recovery of consciousness.
Impact: The indices comprising the DOCS measure brain functioning. Providers can use the DOCS to measure changes in cerebral functioning. Further refinement of the DOCS should allow providers to develop more accurate prognoses and to determine the level and intensity of interventions.
HSR&D Funded: LIP 42-063
|| | | | ||