1. Are We Improving the Quality of Nursing Home Care: The Case of Pressure Ulcers?
DR Berlowitz, MD, MPH. HSR&D Field Program CHQOER, Bedford, MA. HQ, Bezerra, GH Brandeis, JJ Anderson
Objectives: Despite recent initiatives to improve nursing home care, widespread concerns regarding nursing home quality persist and it is unknown whether care is improving. Large databases describing nursing home residents are now becoming available. We use two such databases to determine whether risk-adjusted rates of pressure ulcer development have changed over time in two large, national providers of nursing home care, the VA and a private, for profit chain.
Methods: We used the Patient Assessment File (PAF) to study VA nursing homes from 1990 through 1997, and the Minimum Data Set (MDS) to study the private nursing home chain from 1991 to 1995. Information from after 1995 was unavailable due to the switch to version 2 of the MDS. Rates of pressure ulcer development were calculated for successive six-month periods by determining the proportion of initially ulcer-free residents having a stage 2 or larger pressure ulcer on subsequent assessments. Rates were risk-adjusted for patient characteristics using previously developed models specific to each setting. The proportion of new ulcers that were deep (stages 3 or 4) were also calculated for each period.
Results: We examined risk-adjusted rates of pressure ulcer development based on 274,919 observations of VA nursing home residents, and 144,379 observations of residents from the private chain. Between 1990 and 1995, rates of pressure ulcer development declined by over 25% in both systems (p<0.05). Additionally, the proportion of new ulcers that were deep declined in the private chain from 30% to 22% (p<0.01). Beginning in 1996, though, VA risk-adjusted rates of ulcer development increased to levels similar to those in 1990. The proportion of new ulcers that were deep also increased from 40% before 1996, to 45% afterwards (p=0.01).
Conclusions: Significant improvements have occurred in the quality of nursing home care between 1990 and 1995. Further studies are required to understand why these improvements were not sustained in the VA.
Impact: These results demonstrate how large databases may be used to monitor the quality of nursing home care over time. Despite concerns over nursing home quality, our results are reassuring in that improvements have occurred. However, pressure ulcer preventive care in the VA should be reexamined.
HSR&D Funded: IIR 95-031
2. Variations in Condition-Specific Health Status Among VA General Internal Medicine Clinics
Stephan Fihn, MD, MPH, and Mary McDonell, MS. Northwest Center for Outcomes Research in Older Adults, Seattle, WA.
Objectives: Recent efforts to improve quality of care and productivity have involved measurements of patients' general health status to adjust for case-mix and assess outcomes. Measures of health status, such as the SF-36, however, may be too general to identify important differences in health and function among patients with specific medical conditions. We compared variability in general and condition-specific measures of health status among patients from a geographically diverse sample of VA primary care clinics.
Methods: As part of the Ambulatory Care Quality Improvement Project, we surveyed 38,642 active, eligible patients in 7 VA General Internal Medicine Clinics. Baseline data collected on respondents included an inventory of active medical conditions, including a depression screen (MHI5), and a demographic survey. Respondents were then mailed the SF-36 and condition-specific health status measures for any of six target conditions reported. Measures included the Seattle Angina Questionnaire (SAQ) and the Seattle Obstructive Lung Disease Questionnaire (SOLQ). Patients with a positive MHI-5 were mailed the SCL-20, a measure of depression.
Results: Sixty-six percent of patients responded to the baseline inventory. The mean age was 64 years, and 96% were male. Of the 24,287 patients who were mailed follow-up SF-36 and condition-specific surveys, 15,007 (62%) responded. Statistically and clinically significant differences among sites were observed on all SF-36 scales including Physical Function (range 38.9 to 58.6), General Health (36.1 to 54.4), Vitality (33.8 to 50.5); Bodily Pain (44.7 to 61.3). Significant but smaller variation was observed on scales of social and mental health. The relative differences among sites were consistent across scales and approximated a full standard deviation comparing the highest and lowest sites. Of 9004 patients who reported angina on the initial questionnaire, 5697 (63%) returned the SAQ. Highly significant differences among sites were also observed for scores on all SAQ scales including Anginal Frequency (69.4 to 80.5), Anginal Stability (50.6 to 64.4) and Physical Function (43.5 to 55.6). Of 5553 patients who reported COPD, 3302 (59%) returned the SOLQ. Again, highly significant differences among sites were present for all scales including Coping (59.8 to 70.7), Emotional function (51.1 to 65.5) and Physical Function (51.8 to 61.6).
Conclusions: Although potentially limited by response bias, these results suggest there are clinically important differences among primary care patients in different geographic locations with regard to both general and condition-specific health status for chronic stable angina and COPD.
Impact: Differences in health status among primary care patients in different locations must be taken into account when comparing outcomes and quality of care. These results indicate that substantially more research is required to understand and address these differences.
HSR&D Funded: SDR 96-002
3. Patient Preferences in PSA Screening: The Impact of Shared Decision-Making Videos
EG Wilkins, MD, JC Lowery, MHSA, PhD, and JB Hamill, MPH. VA Center for Practice Management and Outcomes Research, Ann Arbor, MI.
Objectives: VHA has recognized the need for a more patient-oriented approach to the delivery of healthcare, including sharing responsibility for medical decisions between providers and patients. Previous research has shown that shared decision-making (SDM) may result in a variety of benefits, including improvements in patient satisfaction and clinical outcomes. SDM programs have involved the use of educational materials, primarily videotapes, to encourage patients to actively participate in their healthcare. Although initial assessments of these approaches have been encouraging, randomized clinical trials (RCTs) are lacking for most SDM interventions. Our study used an RCT design to assess the impact of a standardized video program for patients considering Prostate Specific Antigen (PSA) screening.
Methods: Male patients eligible for PSA testing were recruited from five VA outpatient clinics and randomized to either SDM video or non-video groups. Prior to meeting with their primary care provider, those in the SDM group viewed a PSA video from the Foundation for Shared Medical Decision-Making. The control group received information (verbally and in writing) traditionally used by the participating centers for PSA test candidates. Following the provider meeting, patients in both groups were administered written questionnaires ascertaining their decisions to undergo or decline PSA testing. Consisting largely of previously validated instruments, the surveys also evaluated patients' overall satisfaction with care; knowledge about PSA testing; decision-making and information-seeking preferences; and decisional conflict. Comparisons of categorical variables between the two groups were performed using chi-square analysis, while continuous variables were analyzed using t-tests.
Results: A total of 91 patients were randomized to either SDM (n=43) or control (n=48) groups. While no significant differences between the two groups were noted in age, race, or marital status, the control group reported a significantly higher educational level than the SDM group (p=0.04). Fewer SDM patients chose to undergo testing (59.5%) compared with controls (80.4%) (p=0.099). SDM patients were more satisfied with both the information received (p=0.08) and their levels of input into the testing decision (p=0.07). Furthermore, the SDM group demonstrated significantly greater knowledge about PSA testing compared to the control patients (p=0.0001). Both groups expressed a strong desire for information; but SDM patients expressed significantly greater preferences for active participation in the PSA decision (p=0.0075). Finally, in spite of assuming larger roles in medical decision-making, SDM patients did not report significantly higher levels of anxiety or uncertainty surrounding their decisions.
Conclusions: VA patients want information to help them make decisions about their healthcare. Implementation of a video-based SDM protocol for patients considering PSA screening appears to result in significantly greater knowledge and preferences for active decision-making roles. These findings contradict reports by Ende and others (1989) who found that older, less educated patients prefer less information and participatory roles.
Impact: Our findings have important implications for designing and implementing patient-centered care in VHA. Patients want and need information to help them make decisions about their healthcare. SDM videos offer a means of conveying the results of outcomes research in a way that patients can understand and use to make decisions consistent with their preferences and values.
4. Effectiveness of VA Primary Care Firm Systems: Preliminary Findings
Elizabeth Yano, PhD, Mingming Wang, MPH, Lisa Rubenstein, MD, MSPH. Center for the Study of Healthcare Provider Behavior, Sepulveda, CA.
Objectives: As a healthcare delivery model, firm systems have been associated with improved prevention, continuity and coordination. By 1996, many VA facilities had launched firm "like" primary care delivery models, with little information about the relative effectiveness of the variations being used. We assessed the relative performance of VA firm systems in contrast with non-firms in terms of preventive and chronic disease care and patients' satisfaction with continuity and coordination.
Methods: In 1996, a national VA expert panel operationally defined a VA firm system through a modified Delphi process that examined the relative importance and contribution to the suggested effects of firms of each of 30 proposed features. Survey items based on these features were incorporated into a 50-item organizational survey, pilot tested and fielded in June 1996 among all VAMCs nationwide. A VAMC was classified as having a firm system if they had 1+ interdisciplinary primary care teams with patients indefinitely assigned to a primary care provider who was responsible for care from clinic-to-ward. We used these survey results (100% response rate) to evaluate the performance associated with VA firm systems using baseline chart-based data from the VHA External Peer Review Program and 1996 VHA National Customer Feedback Center patient satisfaction data. We compared performance associated with firms vs. non-firms and among alternate firm system specifications using the Kruskal-Wallis test.
Results: Overall, 35 (21.9%) VAMCs had firm systems. Compared to non-firms, VA firms served fewer patients (p<.05), but achieved higher proportions of patients who reported having a primary care provider (76.4% vs. 71.3%, p<.05). Preventive practices were higher in firms vs. non-firms (Prevention Index .79 vs. .74, p<.05), with specific benefits in alcohol counseling and colon-cancer screening (p<.05). The Chronic Disease Index scores were also higher among firms (.87 vs. .82, p<.05), with higher performance of sensation exams and foot pulse checks among diabetics, and exercise counseling among hypertensives and obese patients (each p<.05). On average, patients reported slightly fewer problems with continuity (.24+/-.12 vs. .28+/-.10, p=.08). VA firms were more likely to have specialty referral and notification policies (p<.01). The subset of firms with randomized patients and providers and the larger set of firms without inpatient-outpatient continuity had few performance benefits in contrast.
Conclusions: VA healthcare facilities adopting firm systems as their primary care delivery model have higher proportions of patients who report having a primary care provider, provide more preventive care, and higher quality of care for selected chronic conditions. Patients were only marginally more satisfied with their continuity of care. More research is needed to understand the key ingredients of effective delivery models to support the continued growth and development of VA primary care delivery systems, including the adoption of managed care practices, especially as they relate to interdisciplinary teams of providers.
Impact: This work will support the planning and design of VA primary care delivery systems that are associated with higher facility performance.
HSR&D Funded: MPC 97-012
5. Do Patients with Mental Disorders "Unnecessarily" Use More Medical Services? Policy Implications for Mental Health Capitation
CS Hankin, PhD, A Spiro III,PhD, and D Miller. Bedford VA Medical Center, Bedford, MA. D Mansell,MD, MPH. Assistant Professor, Birmingham, AL LE Kazis, ScD, Bedford VA Medical Center, Bedford, MA.
Objectives: It is commonly believed that patients with mental disorders "unnecessarily" use high rates of medical services; the objective of the present study was to explore this assumption. We examined patterns of medical services use among VA ambulatory care patients who screened positive for depressive, alcohol-related, or posttraumatic stress (PTSD) disorders. We hypothesized that even with adjustment for age and medical disease comorbidity, patients who screened positive for targeted mental disorders would have higher rates of medical services use that those who did not screen positive.
Methods: Baseline data were obtained from the Veterans Health Study, a longitudinal investigation of the health of 2,425 Boston-area, male VA ambulatory care patients. Screening measures were CES-D for depression, CAGE for alcohol-related disorder, and PCL-C for PTSD. Prior medical services use (number of inpatient stays in the prior 12 months, and emergency room visits, outpatient visits, or telephone contact with medical personnel in the prior 3 months) was assessed by self-report. The Disease Burden Index (DBI; Kazis et. al., 1998) was used to adjust for medical comorbidity. We compared unadjusted means between groups using t-tests. We then used general linear modeling to compare means, adjusting for age and DBI.
Results: In unadjusted analyses, patients screening positive for depression reported significantly more inpatient stays (1.86 vs. 1.53), outpatient visits (4.80 vs. 3.15), and telephone contacts (2.53 vs. 1.83); following adjustment for age and medical comorbidity, only the significant difference in number of inpatient stays (1.78 vs. 1.58) remained. Prior to adjustment, patients screening positive for alcohol-related disorder reported significantly more inpatient stays (2.16 vs. 1.57), fewer emergency room visits (1.20 vs. 2.12), outpatient visits (3.36 vs. 3.78), and telephone contacts (2.05 vs. 2.15); however, with adjustment, differences in number of inpatient stays (2.02 vs. 1.58) and outpatient visits (2.56 vs. 3.85) remained significant, but differences in number of emergency room visits and telephone contacts no longer met statistical significance. Prior to adjustment, patients screening positive for PTSD reported significantly more inpatient stays (1.82 vs. 1.59), emergency room visits (2.47 vs. 1.66), outpatient visits (4.76 vs. 3.46), and telephone contact (2.88 vs.1.88); following adjustment, differences in medical services use were no longer significant.
Conclusions: Although unadjusted analyses of medical services use revealed clear effects of the screening presence of mental disorders on most outcomes, after adjusting for age and medical comorbidity, many of these effects no longer met statistical significance.
Impact: Our findings have particular relevance for private sector managed care organizations, where restrictions on mental health benefits ("capitation") serve to reduce adverse selection of patient with mental disorders who are assumed to be poor risk because they "unnecessarily" use high rates of medical services. Our findings do not support the notion of such a discrete or direct relationship between mental disorders and medical services use. Results suggest that capitation strategies which are based upon the presumed relationship between mental disorders and medical services use should reconsider the important contributions of age, comorbid medical disorders, and specific mental disorder diagnoses.
HSR&D Funded: SDR 91-006
6. A Comparison of VA to Public Sector Mental Health Patients: the Connecticut Outcomes Study
Rani Hoff, PhD, Robert Rosenheck, MD, M Sernyak, J Steiner, S Atkins. VA Connecticut Healthcare System, West Haven, CT.
Objectives: This paper presents results from a study comparing mental health service delivery to seriously mentally ill patients at three institutions: a VA hospital and two local community mental health centers operated by the State Mental Health Agency (SMHA). The objective was to compare mental health care delivered in VA to those services delivered to a socio-demographically similar population in community mental health centers.
Methods: A random sample of 600 patients with serious mental illness were taken from the three institutions (200 at each) and interviewed about their clinical status, service use, satisfaction with services, and needs and barriers to care.
Results: VA patients were older, less likely to be non-white, had higher incomes both from entitlements and other sources, had fewer problems with housing, and were more satisfied with their housing. They also received significantly more medical care, particularly outpatient medical care, than did SMHA patients, even after controlling for age and medical conditions. There were no differences in the severity of mental health symptoms or use of inpatient mental health care across the three sites. However, VA was less community oriented in its outpatient mental health care: VA patients were less likely to have received case management and rehabilitation services, were more likely to be receiving individual therapy and medication management from a psychiatrist, and were less likely to be taking atypical antipsychotics. These differences persisted after controlling for severity of symptoms, age, race, marital status, income, and education.
Conclusions: We conclude that the integrated medical system in VA affords veterans better access to needed medical services; that the VA in Connecticut has successfully moved away from a focus on inpatient mental health care; but that VA has not made a complete transition to community oriented mental health care.
Impact: The impact of this research highlights the advantages of an integrated medical system in ensuring veterans' access to all types of health care. However, it also indicates that mental health treatment in VA has not been completely translated to a community-oriented model.
HSR&D Funded: PPR 94-002
7. Comparing Quality of Mental Health Care in Public Sector and Privately Insured Populations: First Efforts and Methodological Challenges
Douglas Leslie, PhD and Robert Roseheck, MD. VA Connecticut Healthcare System, West Haven, CT.
Objectives: Comparing quality of care between large health care systems is methodologically difficult, but is an important challenge to health care system management. This study demonstrates methods for measuring quality of mental health care and compares a sample of VA and privately insured patients using these measures.
Methods: Using discharge abstracts, we identified individuals receiving inpatient mental health care in VA during the first six months of each fiscal year 1993 to 1995. A similar cohort of privately insured individuals was identified using MEDSTAT's MarketScan database. These individuals were tracked for six months following discharge and length of stay, readmission rates, and access to outpatient services were calculated. Means of these variables, adjusted for patient characteristics, were compared over time and between the two populations.
Results: The private sector outperformed VA with respect to most of the quality measures, although these differences were modest in magnitude and are likely explained by the fact that VA patients are generally more severely ill and far more socially disadvantaged. However, readmission rates increased considerably over time in the private sector, whereas they declined among VA patients. The outpatient measures improved in both systems, with VA improving more on the number of outpatient visits after discharge and the private sector improving more with respect to the number of days to the first outpatient visit after discharge. VA outperformed the private sector with respect to the continuity of care measure in each year. Quality measures varied by diagnosis, with VA performing better relative to the private sector in treating patients diagnosed with substance abuse and mental health disorders not elsewhere classified, but performing relatively worse in treating patients belonging to the depression diagnostic groups. Unfortunately, variables describing income, disability and homelessness were not available to adjust for major differences between patients treated in the two systems.
Conclusions: Although the private sector outperformed VA to a modest degree with respect to these quality measures (with the exception of continuity of care), VA improved markedly over time compared to the private sector, especially with respect to the inpatient quality measures, and treats a more troubled population for which adjustment was not possible.As these systems continue to adopt strategies to reduce the costs of care and as government systems are increasingly compared to their private sector counterparts, methods for comparing and evaluating the quality of care delivered become increasingly important, although methodological challenges, as illustrated here, are substantial.
Statements: This study provides a benchmark for quality of mental health care in VA as efforts are made to reduce costs. Although similarities in these quality measures between VA and the private sector are impressive (especially given the fact that VA treats a more severely ill population), this study shows that there is room for improvement.
8. Comparing Reimbursement Models for VA Mental Health Services
Douglas Leslie, PhD, Robert Rosenheck, MD, Rani Hoff, WD White. VA Connecticut Healthcare System, West Haven, CT.
Objectives: VA funding mechanisms have historically relied on previous utilization as a basis for distributing funds, and have been criticized as rewarding inefficient provision of care. In response to such criticisms, there is a desire to base these funding mechanisms entirely on patient characteristics. This study explores the impact of using various reimbursement models based on two patient classification schemes - a functional assessment measure (the GAF) and clinical diagnosis - to distribute funds for VA mental health care across VISNs.
Methods: We identify a cross-sectional sample of veterans treated in specialty mental health clinics during a two-week period in fiscal year (FY) 1991. Data from the Patient Treatment File, Outpatient Care File, and the Cost Distribution Report were combined to calculate total utilization and costs for this sample during FY 1991. We then simulate hypothetical reimbursement amounts for these patients based on five different reimbursement models: 1) reimbursement based on average total cost per capita (average cost), 2) reimbursement based on average total cost by GAF score (GAF-based), 3) reimbursement based on average total cost by diagnosis (diagnosis-based), 4) outpatient care reimbursed based on average overall outpatient costs and inpatient care reimbursed based on average inpatient costs by GAF score (GAF-based inpatient), and 5) outpatient care reimbursed based on average overall outpatient costs and inpatient care reimbursed based on average inpatient costs by diagnosis (diagnosis-based inpatient). Total simulated reimbursement by VISN for each of these models was then compared to actual resource utilization.
Results: The distribution of funds nationally across VISNs changes substantially as a result of our reimbursement models compared to actual expenditures. The percentage of total VA expenditures redistributed across VISNs as a result of our models ranges from 6.86% for the diagnosis-based reimbursement model to 8.29% for the simple average cost reimbursement model. However, using functional or diagnostic measures resulted in only small changes from simple average cost capitation. The effects on individual VISNs were substantial and consistent across reimbursement models, with some VISNs experiencing an increase in funding of over 30% and others experiencing a decrease of almost 22% in some simulations.
Conclusions: Reimbursement mechanisms based on functional measures could have a substantial impact on funding patterns compared to current regimes, although they make only minor changes compared to a simple average cost capitation rate. Ideally, one would like to base the distribution of funds on the inherent costs of treating patients in the VA system. Our use of GAF scores and diagnoses is a step in the right direction, but these measures are probably imperfect bases for classification and predicting costs because they are based on clinician judgement and are vulnerable to rater bias.
Impact: These preliminary analyses demonstrate the application of simple, readily available clinical classification schemes to reimbursing VA mental health care. Further analyses are underway to elucidate the strengths and weaknesses of this approach.
9. Impact of Primary Care Satellite Clinics on Access to General Health Care Services and Mental Health Services
Robert Rosenheck, MD. VA Connecticut Healthcare System, West Haven, CT.
Objectives: This study sought to determine whether the establishment of community-based primary care clinics (CBOCs) facilitate access to general health care and/or mental health care services, in both the general population and among people with disabling mental illness.
Methods: From the last quarter of FY 1995 through the third quarter of FY 1997, the Department of Veterans Affairs established thirty-four new community-based primary care clinics in underserved areas. Data were obtained from the 1990 Census on the number of veterans residing in each US county. Data were also obtained from VA's computerized workload data bases (the Patient Treatment File and the Outpatient Care File) on the number of veterans residing in each US County who used VA health and mental health services In FY 1995, before the clinics were established and FY 1997 and used to determine the proportion of veterans in each county who used VA health and mental health services. Analysis of Covariance was used to compare changes from FY 1995 - FY 1997 in use of VA services in counties in which new primary care clinics were located and in other US counties, adjusting for potentially confounding factors such age distribution, race and gender composition, and income distribution.
Results: Counties in which new clinics were established in late 1995-6 showed an increase in the proportion of veterans who used general VA health care services that was almost twice as large as that observed in comparison counties (2.3% vs 1.3%; F=4.94; df=1,3117; p=.03). However, introduction of these clinics was not associated with greater use of specialty VA mental health services in the general veteran population; or of either general health care services or mental health services among veterans who received VA compensation for psychiatric disorders. Interaction analysis showed that CBOCs had their greatest impact on VA service use in counties with larger proportions of low income veterans (beta = 0.006; F=2.11; df=1,3019; p=.03) and had less of an impact in counties with a high proportion of hispanics (beta=-.35; F=2.20; df=1,3019; p=.02).
Conclusions: Community-base primary care clinics successfully improve access to general health care services but do not improve access to specialty mental health care for either the general population or for people with serious mental illness.
Impact: In part as a result of this study, the Under Secretary's Special Committee on Treatment of Seriously Mentally Ill Veterans recommended that specialty mental health staff be included in planning of future CBOCs and the Undersecretary has approved the recommendation.
HSR&D Funded: PPR 94-002
10. Corticosteroid Utilization and Outcomes in HIV Associated Pneumocystis Carinii Pneumonia: Three-fold Higher Mortality among Severely Ill Patients when Corticosteroids given by CDC Guidelines
Charles Bennett, MD, PhD and TC,Ilraith. Chicago VA Hospital, Chicago, IL.
Objectives: Experience with the management of Pneumocystis carinii pneumonia (PCP) exploded with the onset of the AIDS epidemic. Anecdotal reports of improved outcome when steroids were used as adjuncts to appropriate anti-microbial therapy led to controlled trials demonstrating their efficacy.[Bozette S, NEJM] The CDC created guidelines supporting adjunctive therapy with corticosteroids. PCP management during 1995 to 1997 was evaluated for variations in guideline adherence and outcomes.
Methods: Chart reviews from 7 states, 74 hospitals (66 non-VA and 8 VA), and 1,660 empirically diagnosed or confirmed PCP cases.
Results: 735 (44.2%) met CDC guidelines for adjunctive steroids (Aa oxygen gradient > 35 mm Hg or p02 < 70 mm Hg) and 606 (82.4%) received steroids as directed by the guidelines (w/in 3 days of anti-PCP medications). Higher rates of appropriate steroid use were associated with African-American race/ethnicity (84%) versus white (81%) or Hispanics (73%) (p=0.02), younger age (p=0.002), or receiving care in Chicago or Seattle (90%) vs NY, Miami, or LA (71%) (p=0.001). Among severely ill patients with Aa oxygen gradient > 53 mm Hg, 73% were confirmed PCP cases and 82% received corticosteroids per the CDC guidelines. Among these patients, mortality rates were three-fold higher for those who DID receive steroids (18% vs 6%, p=0.02). Similar findings were noted for severely ill persons with confirmed PCP (17% vs 6%). In addition, receiving PCP prophylaxis was associated with higher mortality rates (30% vs 16%, p=.001).
Conclusions: Adherence to CDC guidelines for adjunctive corticosteroid use varied according to city and patient sociodemographics. More importantly, improved outcomes seen in randomized controlled trials were not realized in practice. Among severely ill PCP patients, mortality was three-fold higher when corticosteroids were given according to CDC guidelines. Our findings suggest that the utility of adjunctive corticosteroids in severe PCP needs to be revisited.
Impact: There was a three-fold higher mortality among severely ill patients with associated Pneumocystis carinii pneumonia when corticosteroids were given by CDC guidelines.
11. Use of Centralized VA Data Registry to Assess Quality of HIV Care
Samuel Bozzette, MD, PhD and the HIV-QUERI Executive Committee. Center for the Study of Healthcare Provider Behavior, San Diego, CA.
Objectives: To demonstrate the feasibility of extracting Quality of Care data from the comprehensive longitudinal VA Immunology (HIV) Case Registry (ICR) by assessing the use of newer "highly active" antiretroviral therapies among the HIV-infected under VHA care.
Methods: The HIV-QUERI Coordinating Center was provided with copies of the ICR. We identified unique patients and constructed ratios of the number of patients and of total outpatient visits and hospital discharges, and prescriptions for protease inhibitors or non-nucleoside reverse transcriptase inhibitors (PI-NNRTI). We compared these values with national data on the population of adults under care from the HIV Cost and Service Utilization Study (HCSUS).
Results: Approximately 11,000 unique HIV-infected patients were seen at VA facilities each quarter from the 1st quarter of 1996 (1Q96), when PI-NNRTIs were licensed, to the 4th quarter of 1997 (4Q97). Outpatient visits/100 patients fluctuated from 769 in 1Q96 to 715 in 4Q96 to 851 in 4Q97. However, discharges/100 patients declined monotonically from 30 in 1Q96 to 24 in 4Q96 to 18 in 4Q97. The patterns of use accompanied the rapid adoption of PI-NNRTI therapy in the VA as the percentage of unique patients having at least one prescription written for a PI-NNRTI reached 62% by 4Q96 and 71% by 4Q97. The former compares favorably with estimates that 52 to 61% of all HIV-infected adults under care in the US received the newer drugs in 12/96.
Conclusions: Newer antiretroviral therapies were adopted by VHA providers and patients at least as quickly as by the general population. Increasing use of these agents was accompanied by a dramatic decline in inpatient but not outpatient care.
Impact: Initial comparisons suggest that access of VHA patients to newer HIV therapies is similar to that of other American adults, and that use of these therapies is associated with a reduction in inpatient stays. The VA ICR database contains important information on care for HIV. Improving the validity and accessibility of this database is a high priority for the HIV-QUERI.
HSR&D Funded: HIV 98-000
12. The Prevalence of Hepatitis C in a Sample of Severely Mentally Ill Veterans
Andrew Muir, MD and M Butterfield,. Duke University Medical Center, Durham, NC. KG Meador, Hayden Bosworth, PhD. Durham VA Medical Center, Durham, NC. K Stechuchak, and R Frothingham.
Objectives: There is a pressing need to study Hepatitis C prevalence and risks in veterans including those with severe mental illness. The goal of this study is to establish a prevalence estimate of Hepatitis C in severely mentally ill veterans.
Methods: From March 1 1998- October 15, 1998 consecutive patients admitted to the Durham VAMC inpatient psychiatric unit with a severe mental illness (SMI) diagnosis were serotested for Hepatitis C. SMI diagnoses were defined as meeting Diagnostic and Statistical Manuel of Mental Disorders fourth edition, (DSM-IV) criteria for schizophrenia, schizoaffective disorder, bipolar disorder, or posttraumatic stress disorder (PTSD).
Results: 146 SMI veterans were serotested during the study period. Among these patients, 92.5% were male, 58.9% were African-American. The overall proportion of hepatitis C in this sample of severely mentally ill patients was 17.8% (26 of 146). There was a trend for SMI African-Americans to have significantly higher prevalence rates of Hepatitis C than for SMI Caucasians (24.42% vs. 9.62%). Among those who tested positive for Hepatitis C, 42.3% were diagnosed with schizophrenia, 3.9% with schizoaffective disorder, 7.7% with bipolar disorder, and 46.2% with PTSD.
Conclusion: SMI patients have substantially higher levels of Hepatitis C than what has been reported in community samples. There is a significant trend for SMI African-American patients to have higher Hepatitis C prevalence rates than Caucasian persons with SMI. Positive Hepatitis C was most prevalent among veterans suffering from schizophrenia or PTSD.
Implications: Hepatitis C is a potential epidemic among veterans, including those with severe mental illnesses. Veterans with severe mental illness who are infected with Hepatitis C may increasingly rely on the VA health system for care with substantial cost implications. Further studies are necessary to determine potential Hepatitis C risk factors for SMI persons who are particularly vulnerable to Hepatitis C. Racial differences in Hepatitis C among SMI veterans may exist and warrant further exploration. Increased risk behaviors (intravenous drug and alcohol use) and increased social/contextual risks (community violence poverty, and trauma) have been associated with race/ethnicity and may impact on Hepatitis C and associated risks. Research directed toward understanding Hepatitis C prevalence, transmission, risks, and progression in severely mentally ill veterans is essential.
CSP Funded: EPP 97-022 "HIV Seroprevalence and Risks in Veterans with Severe Mental Illness
13. Management of Hepatitis C at the VA Puget Sound Health Care System
Therese Dawson, BMED. VA Puget Sound Health Care System Seattle, WA.
Objectives: Hepatitis C virus (HCV) infection is emerging as a major health concern in the veteran population. Reports suggest up to 85% of those infected will develop chronic hepatitis and 20% of these will develop cirrhosis. While the prevalence in the VA patient population is estimated to be 10 times that of the general community there appears to be no uniform approach to the management of HCV. Although recent guidelines may help clarify HCV treatment several factors may impede their implementation. The aims of this study were to review the management of patients diagnosed with HCV at the VA Puget Sound Health Care System (VAPSHCS) in 1996 and identify factors influencing this management.
Methods: A retrospective, systematic analysis was made of medical, pathology and pharmacy records of all patients diagnosed with HCV in 1996 at the VAPSHCS. Patients were followed through to May, 1998. Data collected included demographics; diagnostic tests for HCV; history of drug and alcohol abuse; history of HIV, HAV and HBV (testing and status); frequency and results of liver function tests; the presence of concomitant liver disease; HCV notation in discharge summaries; hospital attendances; Gastroenterology referrals; liver biopsies and alpha interferon treatment. Descriptive statistics were performed where appropriate.
Results: One hundred and twenty seven patients were diagnosed with HCV in 1996, 125 males and 2 females (mean age 46 years). Four died during the study, 3 from liver disease. Infection with HCV was detected by ELISA in 119 (94%) and confirmed by PCR in 15 (12%), (RIBA not available). Eight patients (6%) had only PCR results recorded. A history of intravenous and alcohol abuse was recorded in 89 (70%) and 68 (53%), respectively. Thirty-nine patients were tested for HIV (2 positive), 93 for HAV IgM (all negative) and 110 for HBV (70 previous exposure, 3 chronic disease). Liver function tests were performed at HCV diagnosis in 121 patients (95%) of which 63 (52%) were abnormal. Of these, 42/121 (33%) had no subsequent LFTs. Hepatic cirrhosis was recorded in 10 (13%) and hepatocellular carcinoma in 2. Of 149 admissions among 84 patients HCV was recorded in 98 (66%) of their discharge summaries. Twenty-seven patients (21%) have been referred to Gastroenterology and 11 have undergone liver biopsy. Treatment with alpha interferon was commenced in 9 patients.
Conclusions: This review provides an insight into current managementpractices of HCV in our veteran population. Although it is limited by the accuracy and completeness of records and data extraction some conclusions may be drawn. The diagnosis and management of HCV infection at the VAPSHCS lacks uniformity and direction with only 12% of patients receiving confirmatory PCRs. Even once diagnosed, 5% of patients did not have LFTs performed and despite a high incidence of abnormality, a further 33% have not had LFTs repeated. Few patients are referred to Gastroenterology and fewer receive alpha interferon treatment. That HCV has a low profile among health care providers is demonstrated by its lack of inclusion in the discharge summaries of 34% of patients.
Impact: Hepatitis C infection is over represented in the veteran population. Despite its associated morbidity and mortality HCV appears to have a low profile at the VAPSHCS. Although new treatment guidelines have been proposed it is unlikely that these will have significant impact on patient care if an active program for raising the profile of HCV and organizing and centrally implementing them is not established.
14. Hepatitis C Virus Infection and Addiction Disorders in Department of Veterans Affairs Facilities Nationwide
Gary Roselle, MD, and Marta RENDER, MD. VA Medical Center, Cincinnati, OH. SM Kralovic, LH Danko, and LA Simbart.
Objectives: An association exists between Hepatitis C Virus (HCV) infection and addiction disorders (AD), such as illicit injection drug use, inhalation of cocaine, and alcohol ingestion. Alcohol use is of special importance since data suggest an accelerated progression to cirrhosis, and perhaps, hepatocellular carcinoma. This investigation was designed to quantify the extent of comorbidities of documented AD in persons testing positive for HCV antibody (HCVAb).
Methods: The DVA has implemented a national automated electronic surveillance tool (Emerging Pathogens Initiative [EPI]) to extract information from VHA local facility computer systems. The EPI program first identifies persons with a laboratory test positive for HCVAb. Once identified, data are automatically extracted, including demographics such as age, gender, era served, race/ethnicity, and ICD-9-CM coded diagnoses for inpatients. At the Austin Automation Center data are converted to SAS format for analysis by the Infectious Diseases Program Office. Each patient was only counted once, even if the HCVAb test was positive on multiple occasions. AD for inpatients were defined by ICD-9-CM codes and included: Alcohol Only, Cocaine Only, Tobacco Only, Other Drugs Only (including barbiturates, amphetamines, narcotics, etc.), and combinations of these diagnoses. Data were collected from 9/97 - 7/98 (11 months). Statistical analysis used the chi-square test.
Results: Over 11-months, 22,676 persons were identified with a positive test for HCVAb. Mean age was 48.51 (+9.03 SD); 2,021 (8.9%) were noted as homeless; 21,811 (96.5%) were male. 15,981 (70.5%) were outpatients, 6,695 (29.5%) were inpatients; 41% were identified as White, 28.6% Black, and 6.6% Hispanic; 2.6% of inpatients with HCVAb died. 54.2% of inpatients with HCVAb had an ICD-9-CM code for an alcohol-related diagnosis. Any cocaine and any other drug ICD-9-CM diagnoses were seen in 30.5% and 35% of patients who had HCVAb, respectively. 27.9% of persons who were HCVAb positive had no documented AD. HIV infection was seen in 3.7% of inpatients with HCVAb. 81.6% of HCVAb positive persons were from the Vietnam or post-Vietnam era.
Conclusions: 54.2% of inpatients with a positive test for HCVAb had an ICD-9-CM diagnosis related to intemperate use of alcohol, a substance associated with greater severity of HCV-related liver disease. The AD seen in approximately 72% of the patients are often associated with risk behavior for transmission, such as sharing equipment for illicit injection drug use or inhalation of cocaine, and sexual risk behavior. AD in patients with HCV infection will be challenging to provision of the continuum of care; treatment regimen will require complex drug combinations with the need for close follow-up, presenting major obstacles to mitigation of disease.
Impact: The VHA has identified a large population of HCVAb positive persons, creating an inevitable patient care (>22,000 patients to date) and financial (up to 750 million dollars) burden. AD were common, including intemperate alcohol use. Intervention strategies must include VHA data driven targeted screening of risk populations (IL 10-98-013), providing treatment and education for those currently infected, and, most importantly, emphasize prevention of risk behavior that leads to transmission.
HSR&D Funded: DEV 97-032
15. Mortality after Cardiac Bypass Surgery: Prediction from Administrative versus Clinical Data
Jane Geraci, MD, MPH, Howard Gordon, MD, ML Johnson, Nancy Petersen, PhD, L Shroyer, Nelda Wray, MD, MPH. Houston Center for Quality of Care and utilization Studies, Houston, TX.
Objectives: To determine whether the administrative data contained in the VA Patient Treatment File (PTF) can adjust for patient severity and identify hospital mortality outliers identical to severity adjustment and outlier identification using clinical data from a primary data collection, the Continuous Improvement in Cardiac Surgery Program (CICSP).
Methods: We developed logistic regression models to predict operative mortality following cardiac bypass surgery (BYPASS) in 15,288 veterans who underwent bypass from 10/93-3/96. Two models were developed: one from PTF data and one from the CICSP data. Observed-to-expected (O/E) ratios for operative mortality in each of 43 hospitals performing BYPASS were calculated, and hospitals were identified as outliers if the 90% confidence interval for their O/E ratios did not include 1.0. Hospital outlier status was compared for the O/E ratios calculated from each of the 2 logistic models. Because the PTF model and CICSP models did not identify the same hospitals as outliers, clinical variables from the CICSP data set were added, in stepwise fashion (entry criterion pú0.1), to the PTF model to determine whether it could be "enriched" to equal the CICSP model in outlier assignment. Both the PTF and the CICSP models were developed on 2/3 of the study population and validated on the remaining 1/3.
Results: The PTF and CICSP models had comparable predictive power (c statistics=0.767 and 0.760, respectively). The PTF model identified 3 of 6 CICSP high outliers and 1 of 4 CICSP low outliers, plus 2 unique high and 3 unique low outliers. Two additional CICSP variables added to the PTF model resulted in identification of 4 of 6 high and 2 of 4 low outlier hospitals; addition of other CICSP variables increased the PTF modelÆs predictive power (final c statistic=0.811) but did not further improve outlier identification.
Conclusions: PTF data have predictive power for BYPASS operative mortality. Addition of a few clinical variables improved the PTF modelÆs hospital outlier assignment, but differences between the PTF and the CICSP model in this performance remained.
Impact: Administrative data have an ability to predict operative death following BYPASS that is comparable in predictive power to that of clinical data. The comparable predictive power yet unique hospital outlier identification suggests that the PTF model and CICSP model adjust for different things. The addition of a small number of clinical variables not currently available in the PTF improved its ability to detect hospital outliers, as measured against the CICSP data.
HSR&D Funded: 95-005
16. The Houston Approach to Optimal Risk-Adjustment Using Administrative Data
Howard Gordon, MD, Jane Geraci, MD, MPH, ML Johnson, Nancy Petersen, PhD, L Shroyer, and Nelda Wray. Houston Center for Quality of Care and Utilization Studies, Houston, TX.
Objectives: To develop a comprehensive risk adjustment that takes full advantage of the data available in an administrative database.
Methods: We analyzed 15,288 veterans who underwent cardiac bypass surgery (BYPASS) in the Department of Veterans Affairs (VA) between 10/93-3/96, using data from the Patient Treatment File (PTF), the administrative data base on inpatient hospitalizations maintained by the VA. Data included death within 30 days of BYPASS, age, date of surgery (weekend versus other), ancillary procedure performance before or on the day of surgery (coronary angioplasty or intraortic balloon pump (IABP)), and the ICD-9-CM principal and up to 9 comorbid diagnosis codes. Diagnoses identified as possible complications of BYPASS were not used for risk-adjustment. The frequency of post-operative death for principal and comorbid diagnoses was ascertained in a developmental set of FY92-93 BYPASS cases. Principal diagnoses that occurred in 30 or more cases were coded individually. Less frequent principal diagnoses were grouped together into a "rare principal" variable. Comorbid diagnoses were ranked in five groups by risk of death, for codes occurring in at least 30 cases. Less frequent comorbid diagnoses were grouped together into a "rare" group. Patients without any comorbid diagnosis codes were placed into another group. Stepwise logistic regression was used to choose important independent predictors of death within 30 day of BYPASS. In addition, we explored whether a count of comorbid conditions had independent predictive power, and whether a count within each of the comorbidity groups was predictive. The model was developed on 2/3 of the study sample and validated on the remaining 1/3. Data are presented from the full study sample.
Results: Eight independent predictors of postoperative death were identified: age, weekend surgery, angioplasty on the day of surgery, IABP before or on the day of surgery, and 4 variables representing principal or comorbid diagnoses. Two principal diagnosis codes were protective against operative death: angina pectoris (413) and "other forms of chronic ischemic heart disease" (414). A count of the comorbidities within the group having the lowest mortality rate in the developmental set also was protective. Having any comorbidity from the rare group increased the risk of postoperative death. The model c statistic=0.767, Hosmer-Lemeshow goodness-of-fit statistic=4.4 (p=0.82).
Conclusions: Administrative data have moderate ability to predict operative death following BYPASS. Even in a homogenous group of patients undergoing the same surgical procedure the principal medical diagnosis influenced the risk of death, and should be handled separately from comorbid conditions. Some comorbid conditions were more influential than others in predicting death, illustrating that a unit-weighting of them does not accurately reflect their risk.
Impact: Administrative data have moderate ability to predict operative death following BYPASS. Even in a homogenous group of patients undergoing the same surgical procedure the principal medical diagnosis influenced the risk of death, and should be handled separately from comorbid conditions. Some comorbid conditions were more influential than others in predicting death, illustrating that a unit-weighting of them does not accurately reflect their risk.
HSR&D Funded: 95-005
17. A Markov Model of Severity of Illness States and Hospital Mortality
ML Johnson, W Chan, DH Kuykendall, and Carol Ashton, MD, MPH. Houston Center for Quality of Care and Utilization Studies, Houston, TX.
Objectives: To model the course of hospitalization from admission to discharge or death using a Markov model of illness severity states derived from laboratory data.
Methods: Laboratory data was collected using the Health Summary component of VISTA for 476 consecutive patients admitted with non-psychiatric principal diagnoses to the general medicine section of the Houston VAMC. Laboratory test results were scored to create APACHE-L severity of illness measures for each day of the hospital stay that laboratory tests were conducted. Patients were classified into Low, Medium, or High severity states based on the APACHE-L scores. A Markov chain is proposed to model the transition of patients between severity states over the course of the hospital stay, until reaching states of discharge or hospital death, defined as death in the hospital or within 30 days of discharge. A total of 12 possible transitions can occur from day to day: from Low to Medium (LM), Low to High (LH), Low to Discharge (Ldisch), or Low to Dead (Ldead); and similarly four each from Medium and High. A split sample technique was employed to validate predictions of transition rates between severity states, and from severity states to discharge or death.
Results: In the development sample, 44.5% of patients were admitted in severity state Low (106/238), 46.2% in Medium (110/238) and 9.2% in High (22/238). The overall hospital mortality rate was 9.2% (22/238), with a 3.8% mortality rate occurring in patients admitted in Low state (4/106), 9.1% in Medium (10/110) and 36.4% in High (8/22). Median waiting times were 4 days in a Low state, 2 days in Medium and 1 day in High. Example transition rates were: Ldisch 8.7%, meaning patients in a Low state had a 8.7% probability of discharge the next day; Hdead 6.25%, meaning patients who ever reach a High severity state faced a 6.25% chance of dying before the next day. Rates were applied to the validation sample and accurate predictions were found, including: 149.55 expected discharges from Low state with 149 observed; 9.6 expected Hdead transitions, 6 observed; and 5.8 Ldead transitions, 6 observed. Stationary transition probabilities obtained from the limiting probability distribution were 6.1% deaths from Low state, 9.0% Medium and 21.6% High for an entire course of hospitalization.
Conclusions: A Markov model can be constructed from laboratory data to accurately predict transitions between severity of illness states and from severity states to discharge or death. Waiting times indicate relatively quick departure from High and Medium severity states, but relatively long stays in a Low severity state before being discharged, with risk of becoming more ill or even dying.
Impact: Further research is needed to explore methods to use results of Markov modeling as quality of care indicators. In addition to hospital mortality rates, performance could be monitored on rates of improvement or worsening during the course of hospitalization. To improve quality of patient care, a computerized system could track severity states of patients daily to provide a synthesis of laboratory data and risk of mortality to physicians.
18. Hospital Profiles of Mortality from Cardiac Bypass Surgery: Effect of Definition of Outcome based on Administrative versus Clinical Data
ML Johnson, JM Geraci, NP Wray, HS Gordon, MJ Petersen, AL Shroyer. Houston Center for Quality of Care and Utilization Studies, Houston, TX.
Objectives: To determine whether using a definition of post-operative mortality based on administrative data from the Patient Treatment File (PTF) affects identification of hospital mortality outliers derived from a model of post-operative mortality based on clinical data from a primary data collection the Continuous Improvement in Cardiac Surgery Program (CICSP).
Methods: A sample of 15,288 patients from the CICSP data collection who underwent cardiac bypass surgery from 10/93 to 3/96 were studied. The CICSP defined post-operative mortality as any death occurring within 30 days of the operation, or any post-surgical death attributable to surgical complication by medical peer review. Post-operative death was defined using administrative data as any death occurring within 30 days of surgery. Agreement of the outcome variables defined by administrative data versus clinical data was assessed. A logistic regression model was constructed to predict the dependent variable, operative mortality, as defined by CICSP, using the CICSP clinical data as independent variables in a backward stepwise selection (Model 1). Observed-to-expected (O/E) ratios for operative mortality in each of 43 hospitals performing bypass were calculated, and hospitals were identified as outliers if the 90% confidence interval for their O/E ratios did not include 1.0. A second logistic model was constructed using the same independent variables from Model 1, but substituting the dependent variable with post-operative mortality as defined by the PTF to determine the unique effect of the definition of outcome on hospital profiles (Model 2).= Hospital outlier status was compared for the O/E ratios calculated from each of the 2 logistic models.
Results: CICSP classified 557 patients as post-operative deaths, and the PTF classified 498 as post-operative deaths. Sensitivity of the PTF definition to the CICSP was 89.4% and specificity was 100%. Model 1 identified 6 hospitals as high outliers and 4 hospitals as low outliers. Model 2 identified 5 high outliers, 3 of which were original outliers, and 3 low outliers, 2 of which were outliers originally.
Conclusions: Agreement of classifications of post-operative death was very high between administrative data and clinical data. However, using the definition of death from administrative data resulted in 10.6% fewer patients classified as a post-operative mortality, due to the inability of the administrative data to capture post-30 day deaths attributed to the surgery by medical review. This discrepancy in the definition of death further resulted in different hospitals being identified as outliers than originally derived from clinical data alone.
Impact: Further research to examine the sensitivity of hospital profiles to changes in outcome definition is needed to determine best use of both administrative and clinical data. The PTF is a highly accurate source of identifying post-operative death from cardiac bypass surgery overall; however, the administrative database does not contain the clinical information to identify which post-30 day deaths are due to surgical complications. Hospital profiles developed from clinical or administrative models may differ because of differences in the definition of outcome even when the independent variables are identical.
HSR&D Funded: IIR 95-005
19. Accuracy of Computer Identified Diagnoses in a VA General Medicine Clinic
Herbert Szeto, MD, MPH and Mary Goldstein, MD. VA Palo Alto, Palo Alto, CA.
Objectives: Many existing and proposed critiques of medical care rely on DHCP diagnoses. Whereas the accuracy of VA inpatient diagnoses in VISTA/DHCP versus charts have been well studied, little is known about the accuracy of outpatient diagnoses. This study sought to examine the accuracy of diagnoses for hypertension-relevant disorders recorded in DHCP when compared to the outpatient medical record.
Methods: We undertook a cross-sectional chart review to assess the accuracy of common medical diagnoses that may affect the choice of anti-hypertensive therapy. Eight clinical conditions, asthma/COPD , atrial fibrillation, BPH, congestive heart failure, coronary artery disease, diabetes mellitus, gastroesphogeal reflux disease (GERD), hyperlipidemia, and hypertension were assessed in the general medical clinic at VA Palo Alto. Charts of patients scheduled to one of the half-day clinics of 16 residents, 5 nurse practitioners (NPs) and 5 attending physicians (MDs) were reviewed for the week of May 18-22, 1998. A half-day clinic schedule was chosen by convenience for MDs and NPs who had more than one general medicine clinic that week. Clinic charts of 137 of the 148 scheduled established patients were available for review. A diagnosis was considered to be present if it was listed either in the past medical history or as part of the problem list for any past note in the clinic record. A list of diagnoses for each patient was then obtained through DHCP and compared to diagnoses listed in the chart.
Results: The prevalence of the conditions varied from 5.1% for CHF to 51.1% for hypertension. Using the outpatient record as the gold standard, the sensitivity/specificity of DHCP for the conditions are as follows: asthma/COPD (.43/.96), atrial fibrillation (.89/.97), BPH (.3/.95), congestive heart failure (.57/.97), coronary artery disease (.57/.98), diabetes mellitus (.77/.95), GERD (.58/.97), hyperlipidemia (.73/.88), and hypertension (.74/.79). The specificity of DHCP was generally greater than .95, but was as low as .79 for hypertension. The positive predictive value for a diagnosis listed in DHCP varied from .50 for BPH to .89 for CAD. One of the limitations of this study is that patients were chosen through scheduled appointments with specific providers. As is typical of university affiliated VA clinics, our clinics are staffed by resident and staff physicians. Since resident physicians have clinic scheduled only once a week, the sample was disproportionately shifted toward patients cared for by residents. Moreover, the patient population may be biased toward patients with a greater number of medical problems, who seek medical attention more frequently. Coding of diagnoses for these patients may be less complete since providers are less likely to list secondary diagnoses related to any one visit.
Conclusions: This pilot study suggests that significant discrepancies may exist between clinic records and the DHCP system. Diagnoses were often omitted from DHCP, and some diagnoses present in DHCP were not present in the clinic record.
Impact: As we move into an era where electronic clinical databases may be used to assess quality, risk adjust, or as a basis for automated treatment recommendations, careful attention must be paid to the accuracy of such data.
20. Prostate Cancer Quality of Life and Outcomes Research among Patients with Low Socioeconomic Status: An Overview of the VA Cancer of the Prostate Outcomes Study (VA CaPOS)
Simon Kim, MPH. Chicago, VA Hospital. SJ Knight, E Moran, CN Robertson, and JE Smith. Charles Bennett, MD, PhD. Chicago VA Hospital, Chicago, IL.
Objectives: Outcomes assessment for prostate cancer are important, because of debates over the benefits and costs of alternate treatments and outcomes. Because of a lack of evidence of survival benefits with specific therapies, quality of life (QOL) evaluations have taken on increased importance. QOL is rarely assessed among racial/ethnic minorities and men of lower socioeconomic status, who make up a disproportionately large part of the prostate cancer burden. We have initiated the first multi-center QOL outcomes study of lower socioeconomic status men, the VA Cancer of the Prostate Outcomes Study (VA CaPOS).
Methods: VA CaPOS QOL information is collected from prostate cancer patients, spouses, and physicians at six VA medical centers. Because of low rates of literacy, interviewers assess QOL, involvement in care, and the relative importance of likely outcomes following alternative treatments. Spouses provide proxy ratings of patient QOL. Physicians provide information on patients' performance status and the patients' perceived preferences for alternate outcomes. Medical records and electronic databases are reviewed for sociodemographic characteristics and relevant clinical characteristics.
Results: Currently, 601 men with prostate cancer are included in the VA CaPOS, over half of whom are African American. The mean time since diagnosis was 1.4 month for newly diagnosed patients and over 4 years for the rest. QOL responses were most favorable for newly diagnosed, intermediate for stable metastatic disease, and poorest for progressive metastatic disease patients, most of whom had been followed for several years. Spouse emotional well-being assessments were significantly worse than those of individual patients. While patients were not able to provide reliable estimates of their own preferences for future QOL states, they were able to respond reliably to questions phrased as a comparison of the preferences of two hypothetical patients. While African American prostate cancer patients were more likely to have advanced stage disease at the time of diagnosis, after adjustment for differences in health literacy, race was no longer a significant predictor of having advanced prostate cancer.
Conclusions: The VA CaPOS provides useful information on health status, QOL, and low literacy for VA prostate cancer patients. Our results indicate that valid and reliable assessments in low literacy populations are feasible, but that long-term evaluations are needed to detect clinically meaningful information on QOL as the disease progresses. Alternative sources of QOL information, such as spouses, provided results that had poor concordance for emotional and social functioning, but were generally valid for other dimensions of health. The reliability of patient ratings of future QOL states was increased when questions were based on two hypothetical friends rather than consideration by the patients themselves of two potential, but different, future health states. Observational database efforts are potential sources of important information for lower socioeconomic status patients who are faced with difficult therapeutic decisions, limited financial resources, and concerns over both quantity and quality of life outcomes with alternative therapies.
Impact: VA CaPOS provides useful information on health status, QOL, and low literacy for VA prostate cancer patients. It also provides useful information about spouse proxy ratings of patient QOL.
21. Prevalence of Lower Urinary Tract Symptoms and Associated Resource Utilization among VA Primary Care Patients
Elizabeth Yano, PhD, MSPH. VA Greater Los Angeles Healthcare System, Sepulveda, CA. AB Lanto. Timothy Wilt, MD, MPH. VA Medical Center, Minneapolis, MN. Lisa Rubenstein, MD MSPH. VA Greater Los Angeles Healthcare System, Sepulveda, CA.
Objectives: In 1997, nearly 750,000 male veterans had outpatient diagnoses of benign and malignant prostate disease, yet little is known about their presenting symptoms in primary care or associated resource utilization, information that would be useful in the design of programs to systematically screen and treat them. We assessed the prevalence of lower urinary tract symptoms consistent with the presentation of prostate disease among VA primary care patients and assessed their VA healthcare utilization.
Methods: These data were drawn from a longitudinal cohort of randomly sampled male veterans in no acute distress who were visiting the Primary Ambulatory Care and Education (PACE) Center from March-June 1993. Using the Survey of Health and Medical Care designed for the evaluation of PACE, a primary care firm system, we conducted a telephone survey of 1,849 veterans with 3+ primary care visits regarding their health status and symptoms. Lower urinary symptoms (e.g., nocturia) were queried using items developed by a national panel of academic urologists. Cutpoints were selected to identify the most clinically meaningful symptom burden. Survey data were linked with VA administrative data to evaluate general medicine and subspecialty outpatient visits (particularly urology), and admission rates during the following year. We used logistic regression to evaluate the predictors of symptom burden among primary care patients.
Results: For specific symptoms, 24.4% of men reported sometimes-to-frequently having urinary urgency (11.2% frequently), 12.1% reported sometimes-to-often having problems urinating in the previous month (4.0% often), and 24.1% reported urinating 1+ times/hour (4.4% >2 times/hour). Nearly half of those surveyed (47.5%) reported that they urinated 2+ times/night (22.2% with 3+ trips), while over half (56.2%)reported moderate-to-slow streams (11.2% slow and strained). Taking the most severe category for each symptom, we found that 32.9% had 1+ clinically meaningful lower urinary tract symptoms (21.0% with 1, 6.7% with two symptoms). As clinically expected, hypertensives were 53% more likely to report 1+ symptoms (1.53, 95%CI 1.18-1.98), diabetics were 75% more likely (1.75, 95%CI 1.32-2.31), and men reporting shortness-of-breath at rest (1.80, 95%CI 1.28-2.52) or on exertion (2.03, 95%CI 1.49-2.77) were more likely to report 1+ lower urinary tract symptoms (p<.0001). Adjusting for these comorbid conditions, men with 1+ symptoms had significantly more general medicine (p<.01) and urology visits (p<.00001) in the following year compared to men without urinary symptoms. The more symptoms reported, the higher the general medicine (p<.005), subspecialty (p<.0001) and urology (p<.0001) visit rates, as well as medical-and-surgical admission rates (p<.0001).
Conclusions: Male primary care patients have a significant lower urinary tract symptom burden consistent with underlying prostate disease that is associated with significantly higher utilization of VA healthcare resources. Clinical guidelines and pathways that address the optimal role of primary care practitioners and urologists in a coordinated system of care are needed.
Impact: VHA has identified prostate disease as a high impact target condition for QUERI. These analyses are helpful in quantifying the needs of veterans and in assessing how to manage patients in primary and specialty care.
22. Relative Effectiveness of Androgen Suppressive Therapies for Advanced Prostate Cancer
Timothy Wilt, MD, MPH. VA Medical Center, Minneapolis, MN. N Aronson, J Seidenfeld, D Samson, V Hasselblad, and P Albertsen. Charles Bennett, MD, Chicago, IL. PhD Alan Garber, MD. PhD. Center for Health Care Evaluation, Palo Alto, CA.
Objectives: To determine the relative effectiveness of alternative strategies for androgen suppression as treatment of advanced prostate cancer.
Methods: We conducted a systematic review of the evidence from randomized controlled trials Options included monotherapy with surgical or medical orchiectomy (diethlystibesterol, [DES], luteinizing hormone-releasing hormone [LHRH] agonist, or antiandrogens) or combination therapy with additional suppression of adrenal androgens (Combined Androgen Blockade) [CAB]. Three issues were addressed: (1) the relative effectiveness of monotherapy (orchiectomy, DES, LHRH agonists, and antiandrogens); (2) the effectiveness of CAB compared to monotherapy; (3) the effectiveness of immediate androgen suppression for PSA rise compared to androgen suppression deferred until clinical progression. Outcomes included overall, cancer-specific, and progression-free survival; time to treatment failure; adverse effects and quality of life.
Results: There are wide differences in costs of androgen suppression: annual costs were DES = $109; LHRH agonists = $5,000; antiandrogens = $3500; CAB = $8500; surgical orchiectomy per operation = $3000. Survival after treatment with an LHRH agonist is equivalent to survival after orchiectomy or DES (Hazard ratio [HR] = 1.11; 95% CI= 0.89,1.39). The available LHRH agonists are equally effective regarding survival and no LHRH agonist is superior to others when considering adverse effects. Survival may be lower with use of a nonsteroidal antiandrogen (hazard ratio relative to orchiectomy = 1.22; 95% CI = 0.97, 1.54). There is no statistically significant difference in survival at 2 years between patients treated with CAB or monotherapy (HR = 0.97; 95% CI=0.87,1.09). Meta-analysis of the limited data available show a statistically significant difference in survival at 5 years that favors CAB but this is of questionable clinical significance (HR = 0.91; 95% CI = 0.85,0.99). Data suggest that monotherapy has fewer adverse events and results in improved quality life. More men randomized to CAB (10%) withdrew from therapy due to adverse events than men randomized to monotherapy (4%). There is no evidence from randomized trials to compare early androgen suppression initiated upon PSA rise to androgen suppression deferred until clinical progression. There was a significant difference in 5-year overall survival in favor of androgen suppression initiated at the time of radiation compared to radiation alone followed by androgen suppression initiated at clinical progression (HR = 0.63; 95% CI = 0.48, 0.83; NNT at 5 years = 12).
Conclusions: Androgen suppression with orchiectomy or DES provides equivalent survival at lower cost than LHRH agonist, antiandrogens or CAB. Adverse events and quality of life favored monotherapy.
Impact: Androgen suppression for advanced prostate cancer with orchiectomy or DES provides equivalent survival and quality of life compared to LHRH agonists, antiandrogens or CAB. Treatment with orchiectomy or DES would result in considerable cost savings. Despite widespread practice of initiating androgen suppression for rising PSA levels in the absence of clinical signs or symptoms there is no evidence from randomized trials comparing this strategy to androgen suppression deferred until clinical evidence of disease progression. Patients who undergo immediate treatment will have a longer duration of therapy in which they experience the adverse effects and increased costs of androgen suppression.
23. Men's Beliefs about the Benefits of and Professional Recommendations for PSA Screening
Judith Zemencuk MA, and Rodney Hayward MD. Center for Practice Management and Outcomes Research, Ann Arbor, MI. SJ Katz
Objectives: Despite the many uncertainties surrounding screening for prostate cancer, few published studies have examined men's beliefs about the pros and cons of prostate-specific antigen (PSA) testing. This study assesses and describes the beliefs men hold with regard to the benefits of, controversies about, and professional recommendations for screening for prostate cancer by PSA testing.
Methods: Using a self-administered questionnaire developed to measure men's knowledge and understanding regarding PSA testing, we collected data from a sample of men seeking care at a VA medical center general medicine outpatient clinic or at one of two primary care sites of its university affiliate. Of 390 consecutive men, 45 years or older, at the VA medical center site, 270 (69%) consented to participate and returned completed questionnaires. At its university affiliate sites, 119 of the 162 (73%) eligible patients completed questionnaires.
Results: Over 42% of all men reported that they had had a PSA test, while 16% were unsure. Nearly 77%, however, indicated a desire for PSA testing within the next year or two, including 75% of men 70 years and older (for whom the test is not generally recommended even by screening advocates). In general, most men believed regular PSA testing to be highly beneficial. For example, over 80% of all men thought that medical studies have shown that regular PSA testing can reduce a man's chance of dying from prostate cancer by 50% or more. Most men were unaware of the uncertainties surrounding the use of PSA as a screening test. For example, although nearly half of men reported reading or hearing reports about PSA testing, nearly two-thirds were unsure as to whether there was a controversy among medical professionals concerning when and how often a man should get regular PSA testing. Beliefs regarding recommendations for PSA screening for men their age were inconsistent. For example, while 62% of all men thought that the medical profession in general recommends that men their age get regular PSA testing, 46% were unsure as to whether their own physician endorsed a similar recommendation. There were no significant differences in responses at the VA vs. non-VA clinics.
Conclusions: Most men were uncertain as to whether screening for prostate cancer by PSA testing is controversial, while a majority believed that PSA testing is recommended by the medical profession in general. A substantial proportion of men were uncertain as to whether their own physician recommended PSA testing. Still, a large majority believed PSA testing to be substantially beneficial and wanted PSA screening.
Impact: Uncertainties concerning the benefits of screening for prostate cancer using prostate-specific antigen (PSA) have led many to suggest that men be involved in the decision to screen. Indeed, the VA has mandated documentation of PSA counseling. Our study's findings indicate that men may not have sufficient information to make an informed decision regarding screening for prostate cancer by PSA testing and highlight the need to improve men's knowledge and understanding of this screening test.
24. Preventing Urinary Tract Infection using Urinary Catheters Coated with Silver Alloy: A Cost-Effectiveness Analysis
Sanjay Saint, MD, MPH. Center for Practice Management and Outcomes Research, Ann Arbor, MI. DL Veenstra, SD Sullivan, C Chenoweth, and AM Fendrick.
Objectives: Up to 25% of hospitalized patients have a urinary catheter placed at some time during their stay. The overall incidence of nosocomial urinary tract infection (UTI) among patients with an indwelling catheter is about 5% per day. UTIs account for up to 40% of nosocomial infections, with bacteremia complicating nosocomial UTI in about 4% of cases. A recent meta-analysis indicated that silver alloy urinary catheters were efficacious in preventing UTI compared with standard urinary catheters; however, providers must decide whether the efficacy of such catheters is worth the extra cost of approximately $5 per catheter. We performed an analysis to determine the incremental cost and the incremental incidence of catheter-related UTI, bacteremia, and death associated with the use of silver alloy urinary catheters versus standard urinary catheters.
Methods: The decision analytic model, performed from the healthcare payer=92s perspective, was based on patients admitted to acute care hospitals on general medical, surgical and urological services requiring indwelling urethral catheterization between three and six days. Three interventions were compared: 1) universal silver alloy catheter use; 2) silver alloy catheter use only in those at high risk for catheter-related UTI, defined as patients not on systemic antimicrobials; and 3) standard (non-coated) urinary catheter use in all patients. Outcome estimates were derived from several quantitative syntheses of published reports. Cost estimates were based on University of Michigan Health System costs and literature review. Sensitivity analyses were performed and included scenarios in which estimates were set to favor either the universal use of silver catheters or the universal use of standard catheters.
Results: In the base-case analysis, the strategy of universal silver catheter use resulted in an expected cost savings of almost $6 per catheter compared with universal standard catheter use and a 45% relative decrease in the incidence of symptomatic UTI, bacteremia, and death. The strategy of targeted silver catheter use resulted in cost savings of $3.75 per catheter used compared to universal standard catheter use and an expected 20% relative decrease in symptomatic UTI, bacteremia, and death. Sensitivity analyses across the reasonable range of outcome and cost estimates did not dramatically alter these findings.
Conclusions: The results of our analysis indicate that strategies using either universal silver catheters or targeted silver catheters in patients requiring urinary catheterization between three and six days are cost-saving compared to standard catheters and should reduce the incidence of symptomatic UTI, bacteremia, and death. The appropriate use of silver catheters in patients requiring catheterization for less than three days or more than six days remains unclear.
Impact: The universal use of silver alloy-coated urinary catheters in patients requiring catheterization between three and six days should be strongly considered. The base-case analysis suggests that for approximately every 1000 silver-alloy catheters used, an average of $6500 will be saved overall, and 9.7 cases of symptomatic UTI, 1.7 cases of bacteremia and .2 deaths will be avoided.
25. HSR&D Development Center Workshop
Douglas D. Bradham, DrPH. HSR&D Developmental Center, Capital VISN (5), Baltimore, MD.
Purpose: To provide an open discussion (from the trenches) of key issues facing these Centers as they mature into HSR&D infrastructure for their VISNs.
Objective A: To encourage interchange among the current (10) HSR&D Developmental Centers by discussing their approaches to several key issues including:
a) Attracting a critical mass of clinician investigators.
b) Training and mentoring these investigators toward successful LOIs and IIRs.
c) Hiring talented staff in critical areas of:i) VA Database extraction and programming. ii) Biostatistics and methodologists. iii) Health economists. iv) Others?
d) Acquiring and holding on to space!
e) Maintaining linkage with R&D and VISN-level operational concerns.
f) Gaining VISN-level financial support.
Objective B: To improve the next round of HSR&D Developmental Center applications by discussing the current Centers' approaches to issues including:i) Critical items in the proposals. ii) Critical items in the organizational phase. iii) Critical items in the implementation phase. iv) Other critical items for success.
Approach: Each topic will be examined by the 10 directors along characteristics of:
a) What works and why?
b) What does not work?
c) Who is helpful and why?
Activities: Each Developmental Center Director will be asked to participate. Modifications in the issues listed above will be generated. Prior to the Annual Meeting, each Director will be asked to lead a posting of ideas for group contributions for a particular topic. The posting of ideas will be recorded and a summary report will be assembled. That report will be made available for the attendees and for HSR&D Headquarters staff for use in encouraging new applicant's success.
Audience: Current and future of applicants for developmental center funding.
Audience Familiarity: Individuals who are considering application for Developmental Center funding should attend.
26. Internet-Based Education for Epidemiology
Michel A. Ibrahim, MD, University of North Carolina at Chapel Hill, Chapel Hill, NC .
Abstract: This workshop is intended to demonstrate the utility of Internet-based courses in basic and continuing education of epidemiology. As an illustration, one module for an outbreak investigation will showcase the technology and demonstrate the interactive and stimulating nature of this mode of learning. A module is a self-contained unit of study that requires advanced readings and consists of an audio tutorial and a review of a case study with a set of questions and answers. An entire course consists of several related modules. You need not to bring anything with you to benefit from this demonstration. However, if you want to have a hands-on experience by actual participation you should bring a laptop with Windows95, Intel Pentium, floppy drive, 16Mb of RAM, 16-bit sound card with headphones, Netscape Navigator version 3.x or 4.x, RealAudio Player and CD-Rom drive. You may download Netscape from http://home.netscape.com/ and RealAudio Player from http://www.real.com - both these programs are free and are needed for the demonstration. No Internet link is necessary during the demonstration. In order to be able to use the technology effectively to complete an actual module or an entire course, the "student" must have the skills and understanding of Internet jargon, Internet navigation, discussion forums, listservs, and other related topics.
Target: The target audience includes The Veterans Administration clinicians, researchers and administrators. The educational offerings would vary in terms of the nature of topics covered as well as the degree of complexity of content from the introductory to the advanced, depending on the particular audience.
27. Using National VA Data in health Services and Epidemiologic Research
Reiber, Gayle, PhD, MPH, and Jennifer MAYFIELD, MD, MPH. VA Puget Sound Health Care System, Seattle, WA. Charles Maynard. Michael Chapko, PhD, VA Puget Sound Health Care System, Seattle, WA.
Objective: To describe and illustrate basic considerations and strategies for use of national VA data in health services and epidemiologic research.
Activities: In this workshop, participants will have an opportunity to follow the steps involved in procuring data from national VA databases. Two clinical examples will provide participants with an overview of the process and specific methods used to obtain the data. The first example is ô what was the prevalence of esophageal cancer in the VA from 1993-1997 and what was the survival experience of these patients? Ö The second example is ô how many lower limb bypass and amputation procedures were performed in VA hospitals between 1992 and 1996? Participants will receive information on administrative and human subjects clearance, database considerations, downloading strategies, sample programs for data extraction, and file linkage and data quality considerations.
Target: Health services researchers with no prior experience using national VA databases.
28. Continuity of Care as a Determinant of Patient Satisfaction. Results from the ACQUIP Study
Marcia Burman, MD, Mary McDonell, MS, and Stephan Fihn, MD, MPH. VA Puget Sound Health Care System, Seattle, WA.
Objectives: Patient satisfaction has been positively associated with improved patient compliance and improved clinical outcomes and negatively associated with malpractice claims. Factors that have been found to be related to patient satisfaction include sociodemographic factors, payment source, both patient and physician rated health status and utilization measures. We examined relationship between continuity of care and patient satisfaction.
Methods: We surveyed 38,642 General Internal Medicine Clinic (GIMC) patients followed at 7 VAs. Data gathered included demographics, active medical problems and satisfaction with care using the Seattle Outpatient Satisfaction Questionnaire (SOSQ) which measures satisfaction with humanistic and organization aspects of care. Patients who reported angina, COPD, diabetes or hypertension also received condition-specific questionnaires (e.g. the Seattle Angina Questionnaire and the Seattle Obstructive Lung Disease Questionnaire) which include measures of condition-specific satisfaction. All scales were scored from 0 (worst) to 100 (best). Patients were also asked to rate their continuity of care.
Results: 14,865 patients responded with 11,711 reporting one or more of the following disease conditions: angina (n=5544), COPD (n=3189), diabetes (n=3249) or hypertension (n=8059). The percent of patients reporting seeing the same provider "always", "most of the time", "sometimes", and "rarely or never" were 38, 39, 14 and 6 respectively. There was no difference in the distribution of continuity scores between the disease conditions. Mean scores on both the SOSQ humanistic scale and condition-specific scales were strongly related to perceived continuity of care with satisfaction scores ranging from 86.7 to 61.4 among veterans who reported "always" seeing the same provider. Satisfaction scores among veterans who reported "rarely or never" seeing the same provider ranged from 69.5 to 41.9 (p<.001). These findings persisted after adjusting for age, education, race, income, VA facility, and length of time receiving care at the VA and SF-36 scores.
Conclusions: Continuity with the same provider is highly related to Patients' general and condition-specific satisfaction.
Impact: Since higher patient satisfaction has been associated with improved outcomes in chronic disease and improved medical compliance efforts to improve patient satisfaction might be reasonable interventions for improving outcomes. Further studies are needed to evaluate the contribution of various components of continuity of care (for example convenience, access, ease of negotiating the system, etc) to patient satisfaction.
HSR&D Funded: SDR 96-002
30. A Disease-Targeted Measure of Health-Related Quality of Life (HRQOL) for Patients with Chronic Liver Disease the LDQOL 1.0.
Ian Gralnek,MD, MSHS. West Los Angeles VA Medical Center, Los Angeles, CA. RD Hays, HR Rosen, EB Keeffee, DM Jensen, and P Martin.
Objectives: The development and validation of a patient-centered HRQOL outcomes measure is timely and needed for individuals with chronic liver disease. Disease-targeted measures can capture small, yet clinically meaningful changes in patients' health status due to an intervention or disease progression that a generic instrument may fail to detect. Therefore, the objective of this study is to evaluate the psychometric properties (reliability and validity) of a newly developed disease-targeted HRQOL instrument (the LDQOL 1.0) for individuals with chronic liver disease.
Methods: Disease-targeted items in the LDQOL 1.0 were developed from focus groups of patients with chronic liver disease awaiting liver transplantation, expert hepatology panel input, and an extensive review of the literature. Cognitive interviews were conducted to detect potential problems with instrument design or wording of items. The HRQOL instrument was then constructed consisting of 36 generic items (SF-36) supplemented with 77 disease-targeted items. A multicenter, cross-sectional field test was conducted.
Results: 221 consecutive ambulatory adult patients being evaluated for liver transplantation participated in this field test (64.1% male; median age = 51 years (range 23-78 years); 68.9% white, 6.8% Asian/Pacific Islander, 3.9% frican-American, 1.9% Native American, 18.5% other or multiracial). The LDQOL 1.0 is a self-report measure that includes 21 multi-item scales (number of items): physical functioning (10), role limitations-physical (4), pain (2), liver disease-related symptoms (19), emotional well-being (7), role imitations-emotional (3), energy (4), cognitive function (6), memory (4), concentration 3), hopelessness (7), loneliness (6), stigma of liver disease (8), social function (2), quality of social interaction (5), sexual function (3), sleep 6), general health perceptions (7), health distress (4), effects of liver disease 9), and impact of liver disease (4). Internal consistency reliabilities (Cronbach's alpha) ranged from 0.67 to 0.95 (median=0.86); 20/21 reliability estimates were excellent, alpha >= 0.70. All 21 scales were significantly (p<0.05) associated with self-reported severity of symptoms and 4 scales with duration of liver disease (better HRQOL related to less severity and shorter duration of liver disease). Role limitations-physical was most strongly related to severity of symptoms (p<0.01); sleep was most strongly associated with duration of disease (p<0.05). Worse physical functioning (p<0.01), worse sexual functioning among males (p<0.01), role limitations-physical (p<0.01), more liver disease-related symptoms (p<0.05), and greater negative effects of liver disease (p<0.05) were all significantly associated with higher Child-Pugh class.
Conclusions: This multicenter study demonstrates the high degree of reliability and construct validity of the LDQOL 1.0 for individuals with chronic liver disease. This HRQOL outcomes instrument is able to measure significant impairment of daily functioning not detected by more traditional clinician-rated methods (e.g., Child-Pugh classification). The LDQOL 1.0 is now ready for implementation into prospective, longitudinal studies.
Impact: 1. The LDQOL 1.0 will allow for a better understanding of HRQOL in patients with chronic liver disease. 2. The LDQOL 1.0 fits well with the VHA's initiative to build a system of data collection that will integrate both generic and disease-targeted functional status instruments into a routine process of HRQOL data collection. 3. HRQOL data collection in chronic liver disease will allow for case mix comparisons, evaluation of changes in patient functional status over time and its potential relationship to processes of care, and generation of patient summary information for the clinician in their care of patients with chronic liver disease. 4. The LDQOL 1.0 may provide important information on resource utilization within the VA health care system such as response to therapeutic interventions (e.g., therapies for chronic viral hepatitis and liver transplantation).
31. Measuring the Job Satisfaction of Providers in VA Primary Care: the Seattle Provider Satisfaction Questionnaire (SPSQ)
Mary McDonell, MS. VA Puget Sound Health Care System, Seattle, WA. J Marshall. Stephan Fihn, MD, MPH. VA Puget Sound Health Care System, Seattle, WA. KE Kilpatrick, and DS Lessler.
Objectives: Determining the satisfaction of health care providers is critical given continual changes in organization and expectations that often dramatically affect providers and influence their perceptions about their ability to deliver high quality care. We developed and validated a survey to measure multiple dimensions of job satisfaction among VA primary care providers.
Methods: Open-ended interviews designed to elicit provider perceptions of organizational characteristics that contributed to work satisfaction were conducted with 20 providers from a large group practice and VAPSHCS. Interviews and published questionnaires were used to generate a prototype 78-item questionnaire. As part of the Ambulatory Care Quality Improvement Project (ACQUIP) pilot study, the survey was tested among 236 general internists at 8 facilities (VA and non-VA) and condensed to 50 items. The 50-item questionnaire has now been administered to 220 providers at 7 ACQUIP sites and 792 providers at 56 VA facilities participating in the PRIME program. We performed exploratory (EFA) and confirmatory factor analyses (CFA) to analyze dimensionality of the SPSQ. Convergent and discriminate validity of the SPSQ were assessed using CFA as an approximation of the multitrait-multimethod approach. ANOVA and linear regression were used to examine subscale and global satisfaction scores and to adjust for demographic characteristics. Scales were scored from 0 (least satisfied) to 100 (most satisfied).
Results: The sample of 1012 providers was comprised of physicians (69%), nurse practitioners (NPs) (22%) and physician assistants (PAs) (9%). 49% were female. Ages ranged from 20-65 years; 40% were between 36 and 45. 77% were Caucasian, 12% Asian, and 4% African American. 44% had been at their jobs for >5 years. Six dimensions (subscales) emerged from factor analysis and all demonstrated high internal consistency: clinical staff communication (Cronbach's alpha=.86); relationships with specialists (.89); hassle-free work environment (.89); patient characteristics (.77); philosophy of practice (.84); and colleagues (.84). The CFA also demonstrated both convergent and discriminant validity (p<.001). Mean scores for all scales were low: staff communication (38.6); specialists (41.5); assle-free (37.7); patient characteristics (49.3); philosophy (66.4); and colleagues (58.4). Subscale and global satisfaction scores did not differ significantly according to gender, age, ethnicity, time on the job or provider type (resident, staff MD, PA, NP). Comparing different VA facilities, there were significant differences for all scales that were most dramatic (i.e., >25 points) for philosophy and colleagues. When the six domains were entered into a regression model predicting global satisfaction, the philosophy domain (belief in organization's philosophy and ability to practice according to personal standards) accounted for 57% of the variance (p < .001). Clinic organization and clinical staff communication were also significantly related to overall satisfaction (p< .001).
Conclusions: The SPSQ is a reliable, valid measure of multiple domains of provider satisfaction. The extent to which a provider's practice standards and philosophy are aligned with the organizations is a key determinant of overall satisfaction.
Impact: The SPSQ is potentially useful for monitoring outcomes and evaluating changes in clinic organization.
HSR&D Funded: SDR 96-002
32. Patient Characteristics and Patterns of use of Lumbar Spine Radiographs: Results from the Veterans Health Study
Alfredo Selim,, MD, MPH. Boston VA Medical Center, Boston, MA. G Fincke. X Ren, PhD. Edith Nourse Rogers Memorial Veterans Hospital, Bedford, MA. A Lee, and C Skinner. Lewis Kazis, ScD. Edith Nourse Rogers Memorial Veterans Hospital, Bedford, MA.
Objectives: Understanding patient factors that affect patterns of use of lumbar spine films is essential for clinicians, hospital administrators, and managers in order to adopt the best and most effective practice of x-ray utilization. Very little is known about patient factors that affect patterns of x-ray use such as no use and repeat lumbar spine films. In this study, we examined a number of patient characteristics in groups of low back pain patients as defined by their pattern of x-ray use after 12 months of prospective follow-up.
Methods: We analyzed longitudinal data from the Veterans Health Study, an observational study of patients receiving ambulatory care. Five hundred and seventy four patients with LBP were identified through self-report of ever having had 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 12 months after the initial study visit. Participants were mailed a health-related quality of life questionnaire (HRQoL), and had an interview that included a LBP questionnaire and a straight leg raising test. We defined four groups based on the patterns of use of lumbar spine films (new, repeat, no repeat, and no use of lumbar spine films. We compared these groups in terms of sociodemographics, comorbid conditions, low back pain intensity, radiating leg pain, straight leg raising, HRQoL, and low back disability. We examined predictors of repeat lumbar spine radiographs during the 12 months of prospective follow-up and no use of lumbar spine films.
Results: Patients with new lumbar spine films showed worse physical impairment and psychological distress than the other groups. Compared to other groups, patients with no lumbar spine films reported minor physical impairment. Patients with repeat lumbar spine films showed evidence of psychological distress which can not be explained by the small differences in physical health between these patients and those who had no have repeat films. Using regression models, straight leg raising, radiating leg pain, and mental health were found to be independent predictors of repeat use of lumbar spine radiographs (c-statistic of 0.73). We found that low back pain intensity, vitality, and mental health were independent predictors of no use of lumbar spine films (c-statistic of 0.72).
Conclusions: This study found that the use of lumbar spine film is strongly influenced by patient characteristics. Given the physical and emotional toll that low back pain takes on these patients, our data illustrated that physicians rely primarily on the degree of physical impairment to request initial roentgenographic evaluations. Physicians appear to give the benefit of the doubt to patient in psychological distress in considering repeat lumbar spine films. However for many of these patients, education is needed instead of lumbar spinal films.
Impact: Careful consideration of both physical and psychological characteristics of patients with LBP is essential when evaluating and comparing use of lumbar spine films.
HSR&D Funded: SDR 91-006
34. Measuring Patient Expectations: Does the Measurement Instrument Affect the Number of Expectations Elicited and Patient Satisfaction?
Bob Peck, MS. Durham VA Medical Center, Durham, NC. DA Asch, MD, MBA. Leonard Davis Institute of Health Economics, Philadelphia, PA. SD Goold, and D Roter. Peter Ubel, MD. Philadelphia VA Medical Center. James Tulsky, MD. Durham VA Medical Center, Durham, NC.
Objectives: Fulfillment of patient expectations can be used as an indicator of quality of care, and may influence health care utilization and costs. Previous research has shown that the number of expectations fulfilled is positively associated with medical visit satisfaction. However, research also suggests that how expectations are measured may create expectations for care. We conducted this study to examine the relationship between the type of instrument used to measure patients' expectations, the method of eliciting expectations, and patient satisfaction. In particular, we sought to determine whether instruments that elicit different numbers of expectations create bias by yielding different patient satisfaction.
Methods: We randomly assigned patients in a single VA outpatient clinic to one of three methods of expectation assessment. One group received a questionnaire that asked only about 3 specific expectations for care: tests, referrals, and new medications. A second group received a questionnaire that asked about the same three expectations; however, they were nested within a longer list of expectations. The third group received no pre-visit measure of expectations. All patients were interviewed after the clinic visit to assess whether expectations were met and to assess satisfaction with the visit using the ABIM and Ware Visit Specific Questionnaire (VSQ).
Results: 409 consecutive patients in the outpatient clinics were approached to participate. 291 (71%) patients completed both the pre- and post-visit interviews. The remaining 118 patients either refused (20%), did not complete the post-visit interview (8%), or failed the enrollment criteria (1%). The mean patient age was 60, 99% were male, 69% were white and, on average, the patients had completed 12 years of education. The treatment groups did not differ significantly by any demographic variables. The type of instrument used to measure expectations for care was significantly related to the number of expectations reported. The group receiving the specific questionnaire reported an average of 0.74 expectations for tests, referrals, and new medications. The group receiving the nested questionnaire reported an average of 1.65 expectations for the same items (p<0.001). Despite the differences in expectations elicited, reported satisfaction did not differ among the groups. The mean ABIM satisfaction scores for the three groups (specific, nested, control) were 13.86, 13.83, 14.42 respectively (p=.660). The mean Ware VSQ scores for the three groups were 16.09, 16.57, 16.42 (p=.897).
Conclusions: The type of instrument used to measure patient expectations affects key outcome variables. A questionnaire with more expectations to choose from generates more expectations than a questionnaire with fewer expectations. Different measurement instruments, however, do not affect patient satisfaction.
Impact: Investigators studying patient expectations must contend with an "instrument" effect when measuring the number of expectations. However, they need not worry that the instrument will affect patient satisfaction with the medical encounter.
HSR&D Funded: MCV 97-011
35. Barriers to Walking for Exercise in Elderly Primary Care Patients
Patricia Dubbert, PhD. GU "Sonny" Montgomery VA Medical Center, Jackson, MS. DE Bilbrew, KM Cooper, and ER Meydrech.
Objectives: Walking is a good exercise recommendation for many older patients, but not all are able or willing to adopt a regular walking habit. Interventions to increase physical activity in older patients at risk for increased disability require understanding of potential barriers to regular activity. The purpose of this study was to identify barriers to unsupervised walking for exercise as viewed from two perspectives: the health care provider and the patient.
Methods: Participants were patients being screened for the Seniors Telephone Exercise Primary Care Study (STEPS), which is designed to test the effectiveness of nurse telephone counseling to increase walking for exercise. To determine the relative frequency of medical exclusion barriers identified by health care providers, medical records of 658 veterans aged 60-80 and enrolled in VA primary care clinics were selected randomly on the day of clinic visits and reviewed by a nurse researcher. Conditions which would exclude the patient from counseling for an unsupervised walking program ere recorded. Subsequently, patients who passed the medical record prescreen and attended a research clinic visit completed an instrument listing 18 potential barriers to exercise. Patients indicated the degree to which they agreed or disagreed with each potential barrier. Responses were ranked within patients and analysis of variance used to identify the most highly ranked barriers.
Results: 217 (33%) of the 658 medical records reviewed by the research nurse over a period of several months revealed one or more medical exclusions for nurse counseling for unsupervised walking. The most frequent were: severe cardiovascular disease (41%), severe orthopedic problem/joint disease (21%), and severe pulmonary disease (15%).. Ratings by patients of the potential barriers revealed 8 were most endorsed. Ranked in order these were: 1) no good place to exercise; 2) exercise makes me tired; 3) too much exercise could be bad for my heart; 4) exercise makes me hurt more; 5) exercise is hard work for me; 6) family members don't encourage exercise; 7) too much exercise could make me sick; 8) persons close to me don't encourage exercise. Rankings of barriers were not associated with age, ethnicity, or urban vs. rural home location.
Conclusions: The STEPS program was designed to provide nurse counseling for physical activity for patients with chronic medical conditions, but we found that one-third of elderly primary care patients had to be excluded from consideration due to health barriers to unsupervised exercise or inability to walk. Among eligible patients who attended a screening visit, health concerns were highly ranked as perceived barriers to exercise
Implications: Although the majority of elderly primary care patients could be eligible for nurse counseling for home-based exercise, a substantial minority (1/3 in our sample) could not be referred for a walking program. Future prevention efforts need to include consideration of appropriate physical activity designed to maintain functional status and the psychosocial benefits of activity for those who cannot engage in a walking program. Effective methods of addressing patientsÆ perceived barriers, including concerns about health risks of exercise, are needed to optimize prevention counseling for those patients who can walk for exercise.
HSR&D Funded: NRI 95-022
36. Validity of Physical Activity Recall in Urban and Rural Elderly Men
Patricia Dubbert, PhD. GU "Sonny" Montgomery VA Medical Center, Jackson, MS. KE Kerr, DE Bilbrew, KA Kirchner, EF Meydrech, B Shaw.
Objectives: To evaluate the impact of low levels of physical activity (PA) on health and the effects of interventions to increase PA, investigators must often rely on self-report measures such as the 7 day physical activity recall (PAR). The reliability and validity of this measure has not been established in older and rural populations. The purpose of the study was to evaluate the temporal stability and the validity of PAR in elderly urban and rural dwelling elderly men.
Methods: 142 men, average age 68.7 (S.D. 4.6) years, 27% minority, 54% rural, 46% < HS education, who were being screened for the Seniors Telephone Exercise Primary Care Study (STEPS) completed the PAR; a subset of 97 of these subsequently completed a submaximal treadmill test. 138 completed a second PAR 2 to 4 weeks after the first. PARs were administered and scored by trained interviewers using standardized protocols. Energy expenditure (EE) was estimated for several intensity levels of PA. Temporal stability was estimated with Pearson correlations of EE at different intensities of PA across the 2 occasions. Validity coefficients were the Pearson correlations of estimated EE with cardiorespiratory fitness as measured by the submaximal treadmill performance (METs).
Results: Most participants reported little or no hard intensity PA. Temporal stability was 0.45 (P<.01) for total weekly EE; 0.44 (P<.01) for moderate PA; and 0.65 (P<.01) for light-moderate PA; comparing favorably with results of previous studies with young, healthy, active
samples. In urban participants, total EE (0.33, P<.03), hard PA (0.36, P=.02), and moderate PA (0.31, P=.04) were each significantly correlated with treadmill performance with validity coefficients similar to those observed in previous research with younger samples. In rural participants, however, only estimated EE for light-moderate PA was significantly correlated with treadmill performance (0.29, P=.03). Total EE estimates and submaximal treadmill performance were similar for both groups. Validity coefficients were improved by coding specifically for both light-moderate and moderate PA.
Conclusions: Stability of moderate and light-moderate PA was similar in urban and rural dwelling elderly men and stability of moderate and light-moderate PA recall in both groups was similar to results from previous research. In our sample, the performance of the PAR against treadmill fitness as a validation standard was better in urban men. It may be more difficult to accurately estimate duration and intensity (and EE) of the variety of PA described by rural men.
Impact: Measurement of the effectiveness of preventive counseling to promote healthful PA requires measures which are reliable and valid in the diverse populations served. Failure to accurately assess the health behaviors of elderly and rural dwelling populations will make it
difficult to evaluate effectiveness of interventions to decrease health risks. Future research should examine differences in types of PA which could account for the findings as well as evaluate the accuracy of EE estimates by using motion sensor and other approaches for measuring PA which do not depend.
37. Multidisciplinary Rehabilitation Versus Medical Care use of Meta-Analysis to Determine the Clinical Effectiveness of Physical Medicine
Ron Evans, MSW. VA Puget Sound Health Care System, Seattle, WA.
Objectives: Research studies in physical medicine have not demonstrated the effectiveness of inpatient rehabilitation services, primarily due to differences in methodological approaches which have led to inconsistent findings. To address this problem, research studies comparing the clinical effects of rehabilitation with medical care were evaluated for 3 uniformly available outcomes: survival, functional ability, and discharge location. Published trials were obtained from citations in Medline and Nursing and Allied Health Abstracts covering the period from 1974 to 1995 using the descriptors: rehabilitation, physical medicine, activities of daily living, and functional ability.
Methods: We used meta-analysis of completed clinical trials to test the hypotheses that rehabilitation results in better health care outcomes. To be included studies had to prospectively evaluate a multidisciplinary program of physical rehabilitation by comparing outcomes with a control group and report results in a manner that allowed quantitative analysis of uniformly available variables: e.g., mortality, function, and residence. Eleven studies met this rather stringent criteria.
Results: Results indicated that rehabilitation services are significantly associated with better rates of short term survival (86% versus 81%, Chi-square=7.30, p < 0.01) and improved function during hospital stay (Fisher Combined Test, Chi-square=67.12, p < 0.01), but significance was not observed at one year. Also, rehabilitation patients returned home and remained there more frequently than controls (Odds ratio=2.08, Chi-square=40.15, p< 0.001).
Conclusions: We concluded that patients who participate in rehabilitation: 1] function better at hospital discharge, 2] have a better chance of short term survival, and 3] return home more frequently than non-participants. Contrarily, long term survival and function were unaffected by the relatively short term treatment [3-5 weeks].
Impact: The sustaining benefit of returning home may justify provision of inpatient rehabilitation. However, the lack of other long term benefits suggests that services may need to be continued at home or in subacute care settings to optimize their effectiveness. Future clinical research might include social and behavioral outcomes so that results of rehabilitation research can be applied to clinical settings and can more accurately reflect psychosocial objectives.
HSR&D Funded: IIR 94-125
38. Racial Variation in Congestive Heart Failure Mortality among the Elderly
Said Ibrahim, MD. University Hospitals of Cleveland and Cleveland VA Medical Center, Cleveland, OH. EF Cook C. Kent Kwoh, MD and Gary Rosenthal, MD. Cleveland VA Medical Center. DW Baker.
Objectives: Prior studies have suggested a significant difference between blacks and whites in in-hospital mortality from congestive heart failure (CHF). Whether this difference is sustained in the long-term and what clinical factors account for this difference remain unclear. The objective of this study was to compare risk-adjusted, post-hospitalization short-term and long-term mortality from congestive heart failure in elderly black and white patients. This is a 2-year follow-up of a cohort for whom an in-hospital mortality was reported in a published study in 1996.
Methods: The study sample consists of 12,991 patients with the principal discharge diagnosis of congestive heart failure (CHF) following an admission to one of 30 hospitals in Northeast Ohio between 1991 and 1993. All patients were Medicare enrollees 65 years of age or older.
Data Collection: This data comes from the Cleveland Health Quality Choice Program (CHQC). The reliability of data collection for CHQC has been confirmed in prior studies. The database includes information on demographics, comorbid conditions, laboratory test results and detailed clinical information abstracted from patients' hospital records. Analysis: Crude and adjusted 30-day and 2-year survival curves were compared for black and white patients using Kaplan-Meier method. Multivariate analysis was performed using Cox Proportional Hazard models to determine adjusted hazard ratios for the relationship between race and 30-day, 90-day and 2-year mortality. Important clinical variables such as age, gender, comorbid medical conditions, ejection fraction, albumin level, creatinine level and serum sodium level were included in the models.
Results: Black and white patients in this cohort were comparable with respect to age, and gender. Blacks were more likely to have hypertension ( blacks 47%, whites 31% ), and less likely to have Ischemic Heart Disease (blacks 30%, whites 58%). Other comorbidities were comparable. In the one third of the cohort who had ejection fraction recorded, blacks were just as likely to have left ventricular systolic dysfunction compared with whites. During the first 30 days, blacks showed better survival compared to whites (adjusted hazard ratio 0.7 95% CI 0.60-0.83). However, over time the survival advantage for blacks diminished, although it remained significant. During the first 90 days the adjusted hazard ratio for blacks was 0.80 (95% CI 0.60-0.90), and during the first 2 years it was 0.90 (95% CI 0.83-0.97).
Conclusions: This study is one of the first to compare short term and long-term CHF mortality between blacks and whites. Although in the short-term elderly blacks have lower mortality, in the long-term their mortality approaches that of whites. However, because the majority of patients did not have left ventricular ejection fraction measured, further studies are necessary to understand whether racial differences in survival are explained by differences in left ventricular function.
Impact: This study adds to our understanding of the racial variations in outcomes of chronic medical conditions. As the minority patient population in the VA health system increases, understanding racial variations in outcomes will help us devise approaches that will reduce black and white disparities in health outcomes.
39. African American Veterans with Coronary Artery Disease Report Worse Health and Less Satisfaction with Care than Whites: Results from the ACQUIP Survey
Catarina Kiefe, Mary McDonell, MS, Donald Martin, PhD, and Stephan Fihn, MD, MPH. VA Puget Sound Health Care System, Seattle, WA.
Objectives: Racial differences in management of coronary artery disease (CAD) are well known but poorly understood. Several studies document less intense management of African Americans (AA) compared to Whites, even within the Department of Veteran Affairs (DVA). We examined AA vs. White differences in self-reported health status and satisfaction among veterans with CAD.
Methods: Cross-sectional analysis of surveys mailed to 38,642 General Internal Medicine Clinic patients at 7 VAs. Survey information included socioeconomic factors and a checklist of medical conditions, including CAD. Patients reporting CAD received follow-up surveys including: the Seattle Angina Questionnaire (SAQ); the Seattle Outpatient Satisfaction Questionnaire (SOSQ) which measures satisfaction with humanistic and organizational aspects of care; and the Medical Outcomes Study SF36. All scales were scored from 0 (worst) to 100 (best).
Results: 66% of patients responded to the baseline surveys; 4,204 (16.4%) classified themselves as AA, and 15,391 (63.5%) as White. AAs compared to Whites were younger (58.9 vs. 62.8 years, p < .001), less likely to use non -VA health care (27.9% vs. 37.2%, p < .001) or to be married (44.8% vs. 59.2%, p < .001); more likely to have income below $10,000 (39.9% vs. 24.8%, p < .001), to smoke (33.6% vs. 25.6%, p < .001), and to screen positive for drinking problems (39.2% vs. 30.4%, p < .001). AAs were also more likely to report diabetes (25.8% vs. 20.2%, p < .001) and hypertension (65.4% vs. 52.0%, p < .001), but less likely to report CAD (26.6% vs. 37.9%, p < .001). CAD was reported by 9,287 patients (36%); 1,118 AAs, 5,836 Whites and 4,642 other. The differences between Whites and AAs reported above were also present in CAD patients. Among CAD patients, 61.3% returned the SAQ, SOSQ, and SF36. AAs scored lower than Whites on the Exertional Capacity (48.6 vs. 52.4, p=.003) , Disease Perception (58.8 vs. 64.1, p<.001) and Satisfaction (77.1 vs. 84.4, p<.001) domains of the SAQ (p < .01), but did not differ on Angina Frequency and Stability. AAs were also less satisfied than Whites with humanistic (65.6 vs. 74.3, p < .001) and organizational (58.7 vs. 64.2, p < .001) aspects of care. AAs scored higher than Whites on the Vitality scale of SF36 (40.9 vs. 35.2, p < .001) and lower on Emotional Function (50 vs. 54.7, p < .001); AAs tended to score lower on the other SF36 scales but differences were not significant at p<0.01. Differences persisted after adjustment for age, income, education, and marital status.
Conclusions: Compared to Whites, AAs in this DVA sample had poorer socioeconomic circumstances. Among those with CAD, there were no racial differences in overall health or severity of angina symptoms. Yet, even after adjustment for socioeconomic factors, AAs had worse scores on exertional capacity, disease perception, and satisfaction, both general and related to their CAD.
Impact: Racial differences in health status and satisfaction parallel differences in treatment observed in other studies. Longitudinal analyses should explore potential causal relationships between these different instances of racial differences.
HSR&D Funded: SDR 96-002
40. Racial Variation in Mortality Following Elective Surgery for Vascular Disease
Tracie Collins, MD, MPH, Howard Gordon, MD. Houston VA Medical Center, Houston, TX. M Johnson. Jennifer Daley, MD. Brockton, MA. W Henderson, and SF Khuri.
Objectives: Racial variation in mortality for common medical interventions such as elective surgery has not been well investigated. The purpose of this study is to assess differences in 30-day postoperative mortality between whites and blacks after elective surgery for vascular disease.
Methods: Data from the National VA Surgical Quality Improvement Program were used. A prospective, multicenter observational study of risk adjusted surgical outcomes after major surgery. Race was defined as black, not of Hispanic origin or white, not of Hispanic origin as identified by the physician, physician assistant, or clinical nurse specialist. Cases involving other races were excluded. The following elective surgeries were evaluated: abdominal aortic aneurysm repair (AAA; n=1401;94.2% white and 5.8% black), and lower extremity revascularization surgery (LEVS;n=2773;81.7% white and 18.3% black), and lower extremity amputations (AMP;n=2336;70.7% white and 29.3% black). The outcome of interest was all cause mortality within 30 days of the index surgery. For each surgery, the independent relationship of race and mortality was examined with multivariable logistic regression models. Models were adjusted for co-existing illnesses and markers of baseline health that were univariately significantly associated with 30-day mortality and/or clinically relevant.
Results: The overall 30-day mortality was 4.0% for AAA, 2.1% for LEVS, and 7.0% for AMP. The 30-day unadjusted mortality rate in black compared to white patients was higher (9.5 vs. 3.5%;p=.013) for AAA, and was similar for other vascular disease related surgeries (1.7 vs. 2.2,p=.517, LEVS;8.0 vs. 6.6%,p=.253,AMP). The multivariable odds ratio for mortality in blacks compared with white patients undergoing AAA was 2.6 (95% CI 1.11-6.05) after adjusting for age, hypertension, dyspnea, creatinine >= 1.2, history of cerebrovascular accident, diabetes, and ASA-class.
Conclusions: For black patients who undergo elective AAA, there is a higher rate of both unadjusted and adjusted mortality compared to white patients. Following LEVS and AMP, there were no statistically significant increased risks of mortality among blacks. The increased risk of mortality following AAA could not be accounted for by co-morbidities. Further research is needed to address issues of access to care, unmeasured risk factors specific to African Americans, and/or process of care, which could account for the increased 30-day mortality following elective AAA.
Impact: Understanding the factors related to increased 30-day mortality for a minority population following an elective medical intervention can help to improve health care within the Veterans Administration.
HSR&D Funded: NSQIP
41. Antimicrobial Prophylaxis for Urinary Tract Infection in Persons with Spinal Cord Dysfunction
Sally Morton, PhD. Rand Statistics Group, Santa Monica, CA. Paul Shekelle, MD. West Los Angeles VA Medical Center, Los Angeles, CA. JL Adams. Charles Bennett, MD, PhD. Chicago VA Medical Center, Chicago, IL. B Dobkin and BG Vickrey.
Objectives: Recurrent urinary tract infections (UTIs) occur frequently in persons with neurogenic bladders due to spinal cord dysfunction, resulting in increased morbidity and cost. To assess the potential benefits and possible harms of antimicrobial prophylaxis to prevent recurrent UTI, we conducted a meta-analysis of controlled trials.
Methods: MEDLINE and EMBASE were searched from 1966 and 1974 respectively to January 1998, and CINAHL was searched from 1982 to July 1998 using a broad and inclusive strategy with no language restriction. Additional articles were identified by experts and via a hand-search of reference lists. All controlled trials of prophylaxis of UTI that included bacteriuria or UTI as an outcome were included; two reviewers were used. Nineteen trials were identified. Two reviewers abstracted data. Studies were graded according to the quality criteria of Jadad and Schulz. Effect sizes of differences in weekly infection rates (treatment minus control) were pooled using a random effects model. Specific subgroup analyses (asymptomatic versus symptomatic infections; acute versus non-acute patients; high quality, defined as a Jadad score greater than or equal to three, versus low quality) were identified a priori. Sensitivity analyses were also conducted.
Results: Antimicrobial prophylaxis is associated with a reduction in asymptomatic bacteriuria among patients in the acute phase (less than 90 days) following spinal cord injury. The difference in weekly infection rates was equal to -0.27 (95% confidence interval of -0.40 to -0.15; p < 0.05). This means that one patient would require 3.7 weeks of treatment on average to prevent one infection. For non-acute patients the reduction in asymptomatic infections was -0.06 (95% confidence interval of -0.12 to 0.00; p = 0.06). However, antimicrobial prophylaxis did not significantly decrease symptomatic infections among any patients. For acute patients, the difference was -0.03 (95% confidence interval of -0.08 to 0.02; p > 0.05), while for non-acute patients, the difference was -0.04 (95% confidence interval of -0.11 to 0.04; p > 0.05). The study quality subgroup analyses showed no differences from the overall pooled results. In addition, antimicrobial prophylaxis resulted in a two-fold increase in the proportion of antimicrobial-resistant bacteria.
Conclusions: The regular use of antimicrobial prophylaxis for reducing symptomatic infections in most patients with neurogenic bladder due spinal cord dysfunction cannot be supported.
Impact: Routine antimicrobial prophylaxis should not be given to patients with neurogenic bladder due to spinal cord dysfunction. Future research to identify subgroups of patients who might benefit from antimicrobial prophylaxis is needed. Such research should focus on randomized controlled trials that employ rigorous methods; have adequate statistical power; define the study population and outcomes carefully; measure cost and quality-of-life outcomes; and measure potential harms, particularly the development of antimicrobial resistant organisms.
42. Predictors of Outcomes in Patients with Type 2 Diabetes Mellitus
Virginia Baldwin, PhD, RPL. Portland VA Medical Center, Portland, OR. SK Joos and DH Hicham
Objectives: Quality improvement programs for the management of diabetes mellitus (DM) often emphasize the importance of regularly assessing hemoglobin A1c (HgA1c) and maintaining good glycemic control, but there is only limited evidence that HgA1c accurately predicts outcomes for patients with type 2 DM. The purpose of this study was to investigate the relationships among HgA1c, severity of disease, and outcome measures for a group of patients with type 2 DM.
Methods: In 1994-96, 212 veterans with confirmed type 2 DM were enrolled in the study and followed for an average of 3.2 years. Patients were recruited during regularly scheduled visits to a general medicine clinic at the Portland VAMC. Upon study entry, subjects completed the Diabetes Severity Index (DSI) Questionnaire, the SF-36, the GHAA Consumer Satisfaction Survey, and a series of demographic questions. In September 1998, information on death from all causes, medication use, and HgA1c lab values was downloaded from VISTA. A DSI score, composed of five subscales (Visual, Renal, Vascular, Neurological, and Metabolic), SF-36 scores, and satisfaction scores were calculated for each patient. Based on mean HgA1c values, patients were classified as well-controlled (WC) (HgA1c<=7.0), moderately controlled (MC) (HgA1c>7.0 and <9.0), or poorly controlled (PC) (HgA1c>=9.0).
Results: The patients' mean age was 64.4 years, and 95% were male. Forty-two (19.8%) died during the follow-up period. Sixteen percent of patients were classified as WC, 59% as MC, and 25% as PC. WC patients had significantly shorter duration of DM than did MC or PC patients (5.7 vs. 10.2 and 10.7 years, respectively) and tended to be treated with oral diabetic agents (40%) or diet alone (50%). PC patients more often were treated with oral medications (64%) and/or insulin (68%). Mean HgA1c values were not significantly associated with death, any of the eight SF-36 subscales, any of the seven GHAA patient satisfaction measures, or any of the DSI disease severity subscales with the exception of the metabolic subscale, which consists primarily of glycemic control variables. Several DSI measures of disease severity were significantly associated with patient functional status (SF-36) and satisfaction. The overall DSI score and the neurological and vascular subscales were significantly and negatively associated with all eight SF-36 subscales (Spearman's rho [r] ranged -0.20 to -0.40) and nearly all of the seven patient satisfaction subscales (r -0.15 to -0.22) , and positively associated with death (p<0.05). However, the metabolic subscale was not associated with any satisfaction or functional status measures, or death.
Conclusions: HgA1c measures were poorly associated with severity of disease and patient outcomes, such as functional status, satisfaction with care, and mortality. Other measures of diabetes severity were much more strongly associated with outcomes for these patients with type 2 DM.
Impact: In these older type 2 diabetic patients with relatively low functional status, glycemic control as measured by HgA1c appears to be unrelated to patient outcomes. However, a multidimensional measure of severity of diabetes does predict outcomes. Quality improvement efforts for type 2 DM should focus more attention on clinical factors other than glycemic control.
43. Variations in Institutional Review Board (IRB) Approvals for Allowing Third Party Medical Record Reviews of HIV-infected Individuals without Patient Consent. The Multicity PCP Quality of Care Study.
Alison Sipler, BA. Chicago VA Medical Center, Chicago, IL
Objectives: Evaluation of medical care depends on record reviews and Institutional Review Board (IRB) approval. There is increasing concern about medical record confidentiality, especially among HIV-infected individuals. We evaluated variations in IRB decisions for access to medical records of HIV-infected individuals as part of a multi-city study of quality of HIV care.
Methods: The research setting was a random sample of VA and non-VA hospitals in New York, Miami, Chicago, Los Angeles, Raleigh-Durham, Seattle, and Arizona in 1990 (n=109) and 1997 (n=84). IRB decisions were evaluated in response to a single protocol requesting permission for third party reviewers to abstract medical record information for patients with HIV-related Pneumocystis carinii pneumonia (PCP) without written consent.
Results: In New York, IRB approval rates increased from 74% in 1990 to 100% in 1997. In Chicago, Los Angeles, and Miami, IRB approval rates in 1990 were high (80%, 89%, and 92%), but decreased by 18% to 47% in 1997. In Seattle, Raleigh-Durham, and Arizona and in all 14 VA hospitals in the study, IRB acceptance rates were 100% in both time periods. Over half of the IRBs expedited approval. Of 67 IRBs that received requests for collaboration in both periods, 70% approved the project twice, 24% approved the study in 1990 and rejected the study in 1997, and 6% rejected the study in 1990 and approved the study in 1997.
Conclusions: Despite assurances of confidentiality, IRB decisions varied markedly for permission for the same study protocol. While legislation that addresses confidentiality of medical record information is currently being proposed, proposals that require written consent before medical information could be disclosed to third party researchers would virtually end the ability to conduct large scale health services and epidemiologic research studies.
Impact: Concerns over medical privacy for persons with HIV-infection are widespread. Inappropriate disclosure of confidential medical information by medical researchers or providers, similar to that which has already occurred with electronic data bases, can affect the employability, insurability, and social situation of large numbers of individuals. Despite legislative progress in the area for HIV-infected individuals and in IRB practices in the first two decades of the AIDS epidemic, more uniformity is needed to assure patient, physician, and hospital confidentialty and to enable large scale health services and epidemiologic research studies to be continued.
44. Tailoring Smoking Cessation Treatments Based on Nicotine Dependence Level and Depression Status of Patients
Arthur Garvey, PhD. Harvard School of Dental Medicine, Bedford, MA. Pantel Vokonas, MD. Boston VA Medical Center, Boston, MA.
Objectives: Cigarette smoking is highly prevalent in the VA patient population. In an attempt to improve the delivery of smoking cessation interventions for Veteran patients, we conducted a study with a 1-year follow-up period which involved delivery of different degrees of nicotine replacement for low- and high-dependence smokers, and for smokers who were depressed or nondepressed.
Methods: The Fagerstrom Tolerance Questionnaire was used to classify subjects by level of dependence, and the Center for Epidemiological Studies Depression Scale (CES-D) was used to assess depression status. Subjects were randomly assigned to placebo, 2 mg, or 4 mg nicotine gum treatment. We collected both objective and self-reported indices of withdrawal, and we also collected saliva samples which were assessed for cotinine content by radioimmunoassay procedures so that we could determine the percent of nicotine replacement achieved by each subject. Survival analyses, logistic regression analysis, and repeated measures analysis of variance were used to assess significance.
Results: Major differences in outcomes were found for patients with different characteristics. Highly- dependent smokers had significantly better success rates at 1-year post-cessation with the 4 mg gum, while low-dependence smokers were benefited equally by either the 2 mg or 4 mg gum. Either dose resulted in significantly better outcomes than did placebo gum. Depressed smokers given placebo gum had exceptionally high relapse rates (94%), but depressed smokers given nicotine gum (either 2 mg or 4 mg) had significantly better outcomes (30% abstinent at 3-months post-cessation). A higher proportion of nicotine replacement was also significantly related to better cessation rates, and significantly reduced objective signs of withdrawal such as heart rate decline and weight gain.
Conclusions: Our findings strongly suggest that smoking cessation outcomes will be significantly improved if treatment is tailored, or individualized based on specific patient characteristics. Highly- dependent smokers need greater nicotine replacement, while those low on dependence can succeed using a lower dose of nicotine. Depressed smokers, a major sub-population among VA patients, must receive greater nicotine replacement if they are to quit smoking successfully.
Impact: Results from our study clearly demonstrate that the "one size fits all" model of smoking cessation treatment simply will not succeed in helping smokers quit successfully. Smoking cessation interventions need to be tailored to smokers based on at least two critical dimensions: (a) their nicotine dependence level, and (b) their depression status. Highly dependent patients must receive a higher dose of nicotine replacement, and depressed smokers clearly need nicotine replacement therapy if they are to be helped to stop smoking.
45. The Effectiveness of Bupropion vs. Bupropion plus Nicotine Patch in a Smoking Cessation Clinic.
Scott Sherman, MD, MPH. VA Greater Los Angeles Healthcare System, Sepulveda, CA. S Takayesu, P West, and A Chapman.
Objectives: Several research trials have examined the efficacy of bupropion. While efficacy studies such as these are helpful, they unfortunately often do not translate into effectiveness in actual clinical practice. We examined the effectiveness of bupropion vs. bupropion plus nicotine patch in the routine practice of a smoking cessation clinic.
Methods: The Sepulveda VA Smoking Cessation Clinic was established in 1992 and has consistently received approximately 60 referrals/month. Patients attend seven visits over two months and receive individual counseling in a group setting from both a health educator and a clinical pharmacist. Compliance is assessed by carbon monoxide meter at each visit, and patients with repeated high readings are dropped from the clinic. Since April, 1998, we have randomly assigned all new patients to either bupropion or bupropion plus nicotine patch ("combined therapy"). Patients were assigned to only nicotine patches if they had a history of seizures or refused bupropion. We assessed how patients were doing at each visit and whether or not they were tolerating their therapy. We also determined the percent of patients who successfully completed the two-month program. No extra resources or funding were used, so that we were comparing these two regimens in routine practice. We calculated approximate direct costs of the medication and counseling for each regimen.
Results: During the first four months, 231 patients were referred, of whom 140 (61%) attended at least one session. 62 patients were assigned to bupropion, 54 to combined therapy, and 22 to nicotine patches. Two patients chose not to use any pharmacologic therapy. Side effects were noted in 21 people taking bupropion (34%), 14 people taking combined therapy (26%), and 3 people taking nicotine patches (14%). 18% of patients started on bupropion were switched to another regimen (10 to nicotine patches, 1 to combined therapy), compared to 15% of patients started on combined therapy (6 to nicotine patches, 2 to bupropion), and 0% of patients started on nicotine patches. The rate of successfully completing the two-month program (by initial treatment assignment) was as follows: bupropion 13/62 (21%), combined therapy 21/54 (39%), and nicotine patch 5/22 (23%). The cost of medication and counseling for one complete two-month course was $190 for bupropion and $310 for combined therapy. Factoring in the unsuccessful patients, the cost/successful completer was approximately $447 for bupropion and $513 for combined therapy.
Conclusions: We conclude that in routine practice, patients taking combined therapy tolerate it as well (when measured by side effects or treatment switching) as patients taking bupropion. The success rate appears to be higher in patients initially assigned to combined therapy, making the cost/successful completer just slightly higher for combined therapy.
Impact: These data are helpful in determining what regimen to use within a smoking cessation clinic, especially since there are so few trials of combination therapy. We are continuing to accrue approximately 40 patients/month, which will allow us to more precisely compare these two regimens. We do not yet have data on long-term abstinence to determine whether the relapse rate differs by initial treatment assignment.
46. Reliability and Accuracy of Dermatologists' Digital Image Consultations
John Whited MD, MHS. Center for Health Services Research in Primary Care, Durham, NC. DL Simel, RP Hall, ME Foy, and KM Stechuchak. Ronnie Horner PhD, Center for Health Services Research in Primary Care, Durham, NC
Objectives: Teledermatology, using digital imaging technology, is a potentially important alternative means of dermatologic health care delivery. While teledermatology consultations are technically feasible, they have not been critically evaluated against traditional clinic-based consultations comparing diagnostic and management outcomes. Our study asked the question, how reliable and accurate are dermatologists' diagnoses and treatment plans generated from digital image consultations when compared to clinic-based consultations?
Methods: Patients with skin lesions being referred for a diagnostic question to the dermatology consult service from ambulatory care clinics at the Durham VAMC were included in the study. Each patient underwent a standardized history assessment and digital imaging of the referred condition. Two dermatologists independently evaluated the patients in a traditional clinic setting. Three different dermatologists independently reviewed the standardized histories and digital images. Examiners were asked to provide a single most likely diagnosis and differential diagnoses and suggest a management plan (medications, diagnostic testing, and therapeutic interventions). We compared diagnostic reliability, management plan reliability, and diagnostic accuracy between clinic-based examiners and digital image examiners.
Results: One-hundred sixty-eight lesions found on 129 patients were included in the study. A comparable level of diagnostic reliability was found among the different examiners. The simple proportion agreement between the clinic-based examiner pair for their single most likely diagnosis was 0.54 (95% CI 0.46-0.61), and was 0.92 (95% CI 0.88-0.96) when both the single most likely and differential diagnoses were considered. Only one pairing that compared the reliability of clinic-based examiners' diagnoses with digital image examiners' diagnoses (in the category that included single most likely and differential diagnoses) resulted in a lower level of agreement (0.79, 95% CI 0.72-0.85) than that found between the two clinic-based examiners. Additionally, the diagnoses offered by different digital image examiners were equally reliable with each other and did not differ from the diagnostic reliability of the clinic-based examiner pair. Recommendations for medical therapy were also found to be reliable. Therapeutic recommendations were less reliable than diagnoses and medical therapy recommendations, and diagnostic testing decisions were unreliable. Accuracy of the single most likely diagnosis did not differ among the five examiners (p = 0.21).
Conclusions: Compared to traditional clinic-based consultations, dermatologic consultations that use digital imaging technology provide reliable diagnostic outcomes. Management recommendations were not as reliable which may reflect, in view of diagnostic reliability, the different practice settings of the dermatologist examiners. Our study also suggests that diagnostic accuracy does not differ by consult modality.
Impact: This study may have a considerable impact on the future of teledermatology and telemedicine health care delivery and research in the VA. With the knowledge that digital imaging results in reliable and accurate diagnostic outcomes in dermatology, the clinical utility of digital imaging as a consult modality can be justified. This essential information allows for further investigations into the health services implications of functioning telemedicine consult systems.
HSR&D Funded: 95-045
47. Use of VA Services by Medicare HMO Enrollees
Steven Wright PhD. VA Medical Center, West Roxbury, MA. ES Fisher, R Lamkin, C Cagan, D Hults.
Objectives: It is well established that HMO plans receive capitated payments intended for the comprehensive care of Medicare enrollees at the same time those enrollees receive care from other federal providers such as the VA. Regional studies in Florida and California have found that a high proportion of Medicare-eligible VA patients were enrolled in Medicare HMOs at the same time they received VA care. The objective of this study was to determine time trends and patterns of health care utilization by all veteran patients of the VA according to their enrollment status in Medicare HMOs.
Methods: HCFA Medicare enrollment files were linked with a denominator file Of all veterans who used the VA for inpatient or outpatient services between 1992 and 1996. We identified VA patients 65 and older who were eligible for Medicare during this period. Enrollment in Medicare risk-contract HMOs was then determined using the monthly enrollment indicator on the Medicare enrollment files. The proportion of Medicare-eligible patients who were enrolled in Medicare HMOs were calculated for each year. We used VA PTF and OPC administrative files to determine VA resource use and the national CDR (Cost Distribution Report) to calculate costs for patients who enrolled in HMOs throughout FY1996.
Results: The proportion of VA patients who were in Medicare HMOs increased from 3.5% in 1992 to 10.2% in 1996. In FY 96, there were 56,776 VA patients over 65 who were enrolled in Medicare HMOs throughout the year and an additional 32,415 who enrolled at some point during the year. In several states, the proportion of Medicare-eligible VA patients enrolled in Medicare HMOs exceeds 20%; in other states the rates of enrollment doubled between 1994 and 1996. In FY 96 rates of outpatient visits were similar between HMO enrollees compared to HMO non-enrollees (10.8 vs 10.6 per capita). Hospitalization rates in the VA were lower for HMO enrollees compared to non-enrollees (150 vs 310 per 1,000). For VA patients covered by HMOs throughout FY 96, the total costs to the VA for their care was approximately $177 million.
Conclusions: There was rapid national growth in Medicare HMO enrollment by VA patients between 1992 and 1996. There was also marked geographic variation In patterns of HMO enrollment by states. VA patients enrolled in HMOs receive a substantial amount of VA care at significant cost to the federal government.
Impact: Rates of HMO enrollment in the general Medicare Population have continued to increase suggesting that the rates for VA patients have continued to increase. These patients will likely also continue to use VA care under the current federal system. These trends have implications for the coordination, continuity, and cost of care to VA patients. Further research will be required to determine why veterans enrolled in Medicare HMOs continue to seek care in the VA.
48. Validating Ascertainment of Mortality Using Multiple Administrative Data Sources
Steven Wright, PhD. VA Medical Center, West Roxbury, MA. R Lamkin, LA Petersen
Objectives: The VA conducts research on cardiovascular disease using large administrative databases. Ascertainment of veteran deaths is accomplished using the Beneficiary Identification and Record Locator Sub-system(BIRLS), the national VA index of all deaths. In-hospital deaths are also coded in the PTF, the national VA inpatient utilization database. The objective is to examine the validity of BIRLS for Health Services Research.
Methods: We created a cohort of 43,514 veterans age 65 hospitalized with AMI in the VA or under Medicare financing from 1992-95. Site of initial hospitalization classified veterans as VA index or Medicare index. We compared 30 day and 1 year mortality rates using BIRLS and/or PTF files against mortality rates obtained from combining BIRLS, PTF and HCFA Medicare Denominator Files (all source, gold standard). We examined sensitivity and specificity of the BIRLS and PTF stratified by index hospitalization and patient characteristics.
Results: Sensitivity of BIRLS for VA index cases was 93% for 30 day Mortality and 92% for 1 year mortality. Sensitivity of BIRLS for Medicare index cases was 79% for 30 day mortality and 81% for 1 year mortality. Sensitivity in the VA index group improved to 98% for 30 day mortality and 95% for 1 year mortality when both BIRLS and PTF were used. Specificity of BIRLS was 100% for both groups. A greater percentage of Medicare index cases coded as dead from all sources but alive in BIRLS were from the western region of the country (22 vs 16%). There were no differences in the sensitivity of BIRLS by age or race.
Conclusions: Sensitivity of BIRLS data in ascertaining death is not equivalent among all VA patients. Sensitivity of BIRLS was lower for patients hospitalized with AMI in Medicare-financed hospitals compared to patients hospitalized with AMI in VA hospitals.
Impact: BIRLS death status is obtained directly from funeral home directors. Differences in the reporting of deaths for our index groups raises concern about the validity of BIRLS for different patient subsets. We encourage researchers to use multiple sources for ascertaining mortality in veterans.
49. Veterans' Preferences for Transfer of Primary Care to Community Based Outpatient Clinics (CBOC): Association with Health Status, Satisfaction with VA Care, and Prior Use of VA Care
Steven Borowsky MD, MPH. Center for Chronic Disease Outcomes Research, Minneapolis, MN. DB Nelson, SM Nugent, PR Hamann, CJ Stolee, and JL Bradley. Hanna Rubins, MD, MPH. Center for Chronic Disease Outcomes Research, Minneapolis, MN.
Objectives: CBOCs are a national VHA initiative intended to enhance access to care. This study examines whether veterans' preferences to transfer primary care to a CBOC are associated with health status, satisfaction with VA care, and prior utilization of VA care.
Methods: We surveyed 1469 veterans in March 1998. Included with the first of two survey mailings was an application to transfer primary care to a contract CBOC. Veterans located in southern Minnesota (n=784) were offered a consortium of 23 clinics; veterans in northern Minnesota (n=685) were offered a CBOC consisting of 2 clinics. The sample included all veterans who lived within 50 miles of either CBOC area and had >= 1 primary care visit at any of three Network 13 facilities during the preceding year. The dependent variable was whether or not veterans requested transfer to a CBOC. We used bivariate analyses and multivariate logistic regression to evaluate associations between preference for CBOC and health status (SF36-V), VA Customer Satisfaction Survey (CSS) scales, inpatient and outpatient VA utilization during the preceding 12 months (data from Austin Automation Center), demographics, and service connected status. Variable selection for the multivariate analysis used a model score criteria.
Results: 1246 (85%) responded to the survey. Of these, 53% requested transfer of primary care to a CBOC. On bivariate analyses, compared to veterans not requesting CBOC, those who chose CBOC had fewer prior VA outpatient encounters (19.9 vs. 26.2, p<.0001), and had higher CSS scores (lower satisfaction with VA care) on 7 of 8 scales (p values ranging from <.0001 to < .05). CBOC preference was greater in the north than south (59.5% vs. 47.4%, p< .001), and a possible association was seen between CBOC preference and SC status (55.4 % NSC, 52.7% 0-50% SC, 45.2% 50-100% SC, p=.08). Seven of 8 SF36-V scales did not differ between veterans who did and did not request CBOC. Multivariate analysis confirmed that preference for CBOC varied by CBOC area (south vs. north OR=0.61 95%CI 0.47-0.79) and SC status (50-100% SC vs NSC OR=0.57 95%CI 0.38-0.85). Number of prior VA outpatient encounters was negatively associated with preference for CBOC (OR 0.98 95%CI 0.97-0.99). Three satisfaction scales were retained in the final model, each demonstrating an association between lower satisfaction and preference for CBOC (for 0.1 increase on 0-1 scale, higher=less satisfied: "access" OR 1.11 95%CI 1.01-1.23, "coordination" OR 1.11 95%CI 1.01-1.22, "preferences" OR 1.19, 95%CI 1.02-1.38). Three SF36V scales were retained in the model but did not indicate a strong association between health status and CBOC preference.
Conclusions: Preference for contract CBOCs is associated with less prior VA use, lower satisfaction with VA, and lower SC status but is not strongly related to self-reported health status.
Impact: Contract CBOCs may attract patients who use less VA care and who are less satisfied with VA care. CBOCs are an alternative that may improve access and satisfaction for this subpopulation. Comparisons of CBOC vs. VA-based care may not be confounded by large differences in patient health status.
HSR&D Funded: CDA 98-309
50. Effect of Facility-Level Characteristics on Primary Care Enrollment and Bed-days of Care
Margaret Byrne, PhD, Carol Ashton, MD, PhD, and Laura Hechtel, PhD. Houston VA Medical Center, Houston, TX.
Objectives: Performance standards (e.g. reducing bed days of care [BDOC], increasing enrollment in primary care [PCE]) have been established for all VA network and hospital facility directors. To the degree that these parameters are determined by facility characteristics, their use to estimate management performance may be biased. Therefore, we undertook this study to assess the extent to which PCE and acute BDOC are determined by station-level characteristics such as facility type (free-standing outpatient clinics [FSCs], hospitals), facility size (number of beds and users), focus of care (GM&S, psychiatric, long term care), or hospital complexity (acute, intermediate, tertiary).
Methods: The study was conducted on FY 1997 data. Patients' care and outcomes are assumed to be the responsibility of the predominant outpatient care provider. Therefore, using the Outpatient Clinic database, all patients were assigned to a unique station (5-digit identifier code) where they received the plurality of visits to primary care clinics or the plurality of total visits for those veterans without any primary care stop. Veterans were classified as "enrolled in PC" if they had any PC clinic stop. Acute BDOC rates were determined from the Patient Treatment File as the total acute VA bed days for veterans assigned to that station, regardless of which VA hospital was used.
Results: Over all stations providing ambulatory care (n=391), 70% of veterans were enrolled in primary care; 72.8% enrollment for hospitals (n=158), 68% for FSCs (n=231), 87% for independent outpatient clinics (n=3), and 99% for mobile clinics (n=4). Larger FSCs (more than 100 assigned users), however, had a 86.2% PCE rate. This was significantly greater than for hospitals (p<0.0001). Hospital complexity was associated with PCE; both acute (p<0.01) and intermediate (p<0.001) hospitals had higher PCE than tertiary hospitals. Hospitals which focus on psychiatric care had higher PCE than did general hospitals (p<0.02). The acute BDOC rate (per 1000 assigned users) over all VHA stations was 2,143. Overall, patients of FSCs had significantly fewer BDOC than hospitals (p<0.003), although small FSCs had higher BDOC (3,070) than did larger FSCs (1,267). Patients of psychiatric hospitals had higher BDOC (3,563) than other hospitals (2,352) (p<0.001). Regression analyses, controlling for characteristics found significant in bivariate analyses, showed that station size was predictive of BDOC. Patients of stations with more assigned users (p<0.003) and hospitals with more total beds (p<0.01) had higher BDOC rates.
Conclusions: We have shown that station characteristics (type, complexity, focus of care, and size) are important predictors of PCE and acute BDOC rates. This finding suggests stations may have structural characteristics that provide an advantage or disadvantage in meeting standards, regardless of management effectiveness. This brings into question the practice of using these indicators, unadjusted for hospital characteristics, to assess performance.
Impact: Currently, unadjusted PCE and BDOC rates are being used to judge facility managers' proficiency. Our study suggests that this may lead to invalid conclusions due to important effects of facility level characteristics. Future standards should take into account the effect of these facility characteristics before judgements are made.
HSR&D Funded: LFP 98-001