brdlgo.gif (4792 bytes)

2000 HSR&D Annual Meeting: Abstract 150-199

Paste Title Here:        

This page includes abstract 200-267. You can go directly to the abstract
you wish to view by pasting the title into this search box.

Read Other Abstracts:


150. Service Line Activity in VHA Facilities: the Development and Measurement of a Service Line Construct

Victoria A. Parker, DBA, CHQOER, MDRC, Bedford, MA, WH Wubbenhorst, MDRC, LG Pucci, MDRC, M Meterko, MDRC, ML Fonseca, MDRC, MM Byrne, NP Wray, MP Charns, HQCUS, HQCUS, MDRC

Objectives: Product lines, service lines, and patient care lines are all different names for an organizational structure that differs from traditional structures in that personnel and services are organized by a common output (e.g. mental health services) rather than by common discipline (e.g. pharmacy, psychology). This type of structure, which we call service line (SL), has been implemented in a variety of clinical areas. These implementations presuppose that SL structures will improve the quality and efficiency of care. In order to test whether SLs have any such effect, our objective was to develop a reliable method of identifying and classifying SL structures.

Methods: We conducted site visits to VHA's 22 networks to determine the clinical areas in which SLs were being employed. We then developed a survey designed to identify clinical areas in each facility where any type of SL structure had been implemented. The survey asked a series of detailed questions about the structure and management of SLs for mental health (MH) and primary care (PC), the two most prevalent SL domains within VHA facilities. Responses to these questions, particularly those regarding the supervision of employees within SLs, were coded to characterize each SL along the continuum of SL structures originally developed by Charns & Tewksbury (1993) (the CT continuum). These categories were: no SL; minimal, called a task force (TF); moderate, a team (T); and well developed, a division (D). The survey was then pilot tested and amended to improve clarity. The survey was mailed to all VHA facility/integrated system directors in late 1998.

Results: 143 of 144 facilities responded. As of December 1998,110 facilities had implemented Primary Care Service Lines (PCSLs), and 108 facilities had Mental Health Service Lines (MHSLs). There were fewer SLs in other clinical areas such as Geriatrics/extended care (47 SLs), tertiary/specialty care (29 SLs), Spinal cord injury (15 SLs) and Rehabilitation/physical medicine (10 SLs). Almost all of the PCSLs and MHSLs could be classified on the CT continuum. A small number of SLs were not classifiable because the pattern of responses for them did not fit into any of the other coding categories. We found that 22% of PCSLs and 17% of MHSLs existed prior to 1996; in fact, some dated as far back as 1988.

Conclusions: The proliferation of SLs in VHA facilities in the areas of mental health and primary care has been extensive. However, there is quite a bit of variation in how SL structures have been organized. Contrary to widespread perception that the adoption of SLs has been driven by headquarters initiatives, we found that a sizeable number existed before 1996, indicating that they were initiated at the local level.

Impact: The fact that VHA managers have adopted different types of SLs provides an opportunity to observe variation among these types. It is not known which SL structure type (TF, T, D) might lead to the best outcomes. With measures of SL structure now developed, it will be possible to analyze the effectiveness of this organizational intervention.


151. Bayesian and Hierarchical Models for Assessing Health Care Quality: What are they? Why are they needed? How are they used?

Cindy L. Christiansen, PhD, CHQOER, Bedford, MA, and Boston University School of Public Health, JF Burgess, Management Science Group

Objectives: This workshop is designed to acquaint health services researchers and decision-makers with hierarchical and Bayesian models. The objectives of the workshop are that the participants will: 1) learn some of the statistical terminology for Bayesian and hierarchical modeling. We will answers questions such as: What is a Bayesian model and how is it different from a hierarchical model? What is the 'level-of-interest'? Is a p-value the probability of extreme performance? 2) learn why these models are important for health services research. Health care assessments and decisions often are made without good understanding of statistical assumptions. When this happens, 'quality' is defined implicitly by the statistical method instead of by experts and policy-makers. Confidence intervals and p-values are difficult to understand and often are misinterpreted. We will discuss the importance of defining performance criteria and show how Bayesian and hierarchical models can improve understanding through probability statements about true performance levels. 3) understand when the models should be used to improve statistical analyses. Bayesian and hierarchical models are more complicated than standard statistical models. Is this complication necessary? Will the results change with the choice of models? We will answer these questions and others using several examples of provider assessments. 4) learn how to use the models. The output from Bayesian and hierarchical analyses is different from standard output. The statistical software needed for these models also is different. We will discuss and interpret output from several studies and will explain how to gain access to software. Several easy-to-read articles on Bayesian and hierarchical models will be distributed and web sites that offer more information will be provided.

Activities: Workshop participants will review the terminology given on the handouts. They will be asked to discuss performance criteria for several examples and they will be asked to participate in interpreting output from these examples. Handouts will include several worked examples, terminology and references.

Target Audience: The workshop is designed for health service researchers who use statistical models in their work, for decision-makers who rely on output from statistical analyses, and for applied statisticians who are unfamiliar with this area of statistics.

Audience Familiarity: Participants should be familiar with the term 'variation' and have an understanding of standard statistical methods such as t-tests and chi-square tests. They should have used or seen analyses involving confidence intervals or p-values.

152. Data Mining using VISTA for Decision Making and Quality Improvement

Philip R. Foulis, MD MPH, James A. Haley Veterans's Hospital, Tampa, FL

Objectives: At the completion of this workshop, the participant will be able to:

1. Define patients within VISTA with specific diagnoses

2. Create a customized Clinical Health Summary

3. Extract the information from the Clinical Health Summary into a relational database

4. Perform statistical analysis and generate reports from the patient data

Activities: VISTA contains extensive information regarding patients including: pharmacy, laboratory, radiology, vital signs, outpatient encounters, and inpatient discharges. Although detailed information describing a patient can be obtained from VISTA, the usual retrieval methods are labor intensive and involve requesting individual patient records. More automated methods of obtaining information generally involve Fileman routines, require a detailed knowledge of the structure of VISTA, and utilize extensive computer resources. Several modules within VISTA permit searching for patient diagnoses, pathology diagnoses, results of diagnostic tests or patients taking specific drug classes. Unfortunately, it is difficult within VISTA to develop complex interrelationships between elements of patient information without extensive programming experience. For example, although information is contained regarding diagnoses and treatment, the task of finding individuals with a particular diagnosis receiving a specific treatment would be difficult. Likewise, it is complicated and time consuming to select patients with certain results and relate this outcome to utilization of resources.

Using existing VISTA functions, this workshop will demonstrate the ability to identify target patient populations, create Clinical Health Summaries, move the information into a personal computer, create a relational database from the information, and analyze the patient information. An example set of data will be created for patients with a diagnosis of diabetes. Information of interest will be collected using the ad hoc Clinical Health Summary function in VISTA. After requesting patients, the information will be moved to a personal computer and loaded into a relational database. Several analyses of the information will be demonstrated including; generation of report interrelating patient data, summary information describing patients, statistical analysis of patient outcomes.

Target Audience: Persons benefiting from this course would be those responsible for:

1. Defining patient populations with specific illnesses within a facility or facilities

2. Investigating patient interventions and outcomes

3. Researching resource utilization of defined populations

4. Utilizing retrospective patient information to verify practice guidelines

5. Gathering knowledge for the development of evidence-based decision making.

Audience Familiarity: The participants should be familiar with clinical portions of VISTA, particularly the components of the Clinical Health Summary. In addition, a working knowledge of Microsoft Office (TM) will be beneficial. Finally, an elementary understanding of database design will be helpful.

153. Longitudinal Data Analysis Techniques

Kimberly Wristers, PhD, Houston Center for Quality of Care and Utilization Studies: A VA HSR&D Center of Excellence, Department of Medicine at Baylor College of Medicine, Houston, Texas, Nelda P. Wray, Chief, Section of Health Services Research, Department of Medicine, Baylor College of Medicine

Objectives: Health changes constantly. Studying patient changes over time is of great importance to physicians and researchers. One way researchers study this change is by collecting outcome data over time. Yet, measuring change using longitudinal data is not easy. The purpose of this workshop is to show how individual growth curve modeling provides clinical researchers with powerful techniques for investigating research questions about systematic changes over time in individual patients. This lecture and demonstration will compare three techniques for analyzing longitudinal outcome data. We will first demonstrate the limitations of the traditional technique of comparing treatment group means at each occasion. We will then show how a traditional repeated measures approach corrects for some of these problems. Finally, we will demonstrate advantages of individual growth curve modeling over the traditional methods. Actual data from a longitudinal VA Medical Research study entitled, "Arthroscopic Treatment of Osteoarthritis of the Knee" will be used to illustrate the advantages and limitations of the three techniques. We will discuss the conceptual advantages for framing change as a continuous process and statistical advantages for modeling change at the individual level.

Activities: The Longitudinal Data Analysis Techniques workshop will compare 3 techniques for analyzing changes in health outcomes - 1) comparison of treatment group means at each occasion, 2) repeated measures analysis, and 3) individual growth curve modeling. We will provide a descriptive overview of each technique and then apply the techniques to two examples from a VA Medical Research Study. We will show how the first two techniques focus on group means over time and how the individual growth curve modeling technique focuses on individual scores over time. We will discuss the types of questions that are best investigated with these techniques, and demonstrate the flexibility of individual growth curve modeling and how it offers researchers the opportunity to investigate questions about individual change and correlates of change.

Target Audience: This workshop is designed for clinical researchers who conduct research related to patient-level changes. The target audience includes researchers who are interested in investigating questions such as, " Do all patients have the same pattern of change?" and " Are variations in change parameters systematically related to known characteristics of patients?"

Audience Familiarity: The presenters of the Longitudinal Data Analysis Techniques workshop will concentrate more at a conceptual level than a statistical level. Audience members who are familiar with significance testing and basic regression techniques will have the necessary information for the workshop.

154. Estimating VA Health Care Costs

Gary Nugent, Cincinnati VAMC, Cincinnati, Ohio, TH Wagner, VA HSR&D Health Economics Resource Center, E Franchi, Milwaukee VAMC, LB Nugent, Dayton VAMC, ML Render, Cincinnati VAMC, C Phibbs/AM Hendricks/P Barnett, VA HSR&D Health Economics Resource Center/Bedford HSRD/Health Economics Resource Center, VA Palo Alto Health Care System

Objectives: To provide researchers with tools for estimating the cost of VA health care using pseudo-bills and average costing.

Activities: This workshop is a joint presentation by researchers from the "Evaluating VA Costs" project and economists from the VA Health Economics Research Center (HERC). HERC will provide an overview of services that the Center provides to VA researchers. Staff economists will describe a database they are developing to provide the cost of every VA health care encounter. These costs are estimated with an average cost method, combining VA utilization data, cost data from the Cost Distribution Report, and non-VA data on the relative cost of different services. They will describe their econometric studies of Medicare discharge data to determine the effect of diagnosis and length of stay on the cost of acute short-term hospital stays. The presentation will describe the analysis and assumptions needed to create the cost estimates, situations in which the estimates might be appropriately used, and when the researcher should turn to direct measurement or a pseudo-bill method. Researchers from the "Evaluating VA Costs" project will describe how they are using pseudo-bill methods to find the costs of six VA medical centers. They will describe the use of CPT codes to estimate the cost of physician services and the most common reasons why some codes recorded in VA are rejected by Medicare coding software. Researchers will discuss how pharmaceutical costs may be compared to RedBookä wholesale prices using VA pharmacy utilization files and how prosthetic costs may be estimated by combining the VA National Prosthetic Patient Database (NPPD) with HCFA data. They will discuss the practicalities and pitfalls of these methods. Workshop participants will receive handouts describing data sources, issues involved in merging the data with regionally adjusted price information, and a comparison of regionally-adjusted usual, customary and reasonable (UCR) charges at the 60th percentile to the appropriate Medicare fees.

Target: Health Services researchers, senior VHA administrators, VA clinicians.

Familiarity: None

155. Evidence-Based Healthcare

Eduardo Ortiz, MD, MPH, VA San Diego Healthcare System; University of California, San Diego, Southern California Evidence-Based Practice Center, San Diego, CA

Objectives: This workshop will introduce participants to the principles of evidence-based healthcare, teach them about evidence-based resources, and improve their knowledge and skills in searching the biomedical literature.

Activities: Using presentations and interactive, case-based searching demonstrations, participants will learn: 1) What is meant by evidence-based healthcare; 2) Why an evidence-based approach to healthcare is important and how it can improve clinical decision-making; 3) About the various biomedical databases; 4) How to more efficiently and completely search Medline and other NLM databases (searching demonstration); 5) About evidence-based resources, such as the Cochrane Library, ACP Journal Club, Evidence-Based Medicine, Evidence-Based Practice Centers, and others; 6) How to access and use the Cochrane Library (searching demonstration); and 7) Practical applications of evidence-based healthcare in their own setting.

Target Audience: Clinicians and clinician-educators involved in patient care whose level of knowledge about evidence-based healthcare and experience with biomedical databases and literature searching is at the beginner to intermediate level.

Audience Familiarity: Participants are likely to be familiar with the term "evidence-based medicine," but they may not know what is meant by the concept or its implications for making healthcare decisions. Participants are assumed to have little or no formal training in the principles and techniques of evidence-based healthcare. No prerequisite knowledge or skills in the subject areas are required.

156. Approaches to Guidelines Implementation: Primary Care Depression Treatment

Edmund F Chaney, PhD, VA PSHCS Seattle Division, Seattle, WA, LV Rubenstein, HSR&D COE, Sepulveda, CA, SC Hedrick, HSR&D COE, Seattle, WA, EM Yano, HSR&D COE, Sepulveda, CA

Objectives: To provide attendees with a framework for understanding guideline implementation for depression in primary care and to suggest solutions to implementation challenges. The workshop will be based on information from three state-of-the-art studies of primary care depression treatment effectiveness.

Activities: Review depression research within a translational research agenda: World Health Organization researchers have suggested that unipolar major depression is the leading cause of disability worldwide. Depression is prevalent in primary care and is usually suitable for treatment within that setting. Current primary care for depression often suffers from under-recognition, under-diagnosis, inadequate strength and duration of treatment and inadequate follow-up. Research on improving depression care suggests that quality improvement for this illness requires organizational change, patient activation and education, and provider education, as well as involvement by mental health specialists. We discuss how to analyze and modify barriers to guideline implementation in the system, provider and patient domains.

Review VA primary care systems and their strengths and weaknesses for depression care guideline implementation: We present findings from three survey waves (1993, 1996 and 1999) of all VA Primary Care Clinics to illustrate differences in key organizational factors among programs which may require different implementation approaches. Marketing, negotiation, training and monitoring are basic elements of system intervention. We use two experiments--the NIMH Mental Health Awareness Project (MHAP, currently in process) and the AHCPR Partners in Care study (PIC, recently completed with positive results)--to illustrate organizational change issues including practice redesign, and models for interaction between external experts and local leaders and staff. The MHAP study was carried out in Kaiser Permanente and VA sites, and the PIC study in a nationally diverse group of non-academic managed care organizations.

Implementation of Evidence-Based Depression Care Models in VA: We use Project Mood, the first randomized trial of collaborative care for depression in VA primary care, to illustrate effects of different internal organizational relationships among primary care providers and local mental health resource staff. Acute phase outcome data suggest patients in both study arms improved, with a team treatment intervention resulting in significantly more improvement in depression symptomatology than a consult-liaison care model. This successful model, as well as the PIC study, used methods to increase patient self-management skills by providing information, individualizing treatment, enhancing patient choice, and supporting change attempts through active follow-up systems, and these methods will be discussed.

Target Audience: Administrators, clinicians and health service researchers interested in guidelines implementation and depression treatment.

Audience Familiarity: Variable.

157. Using Qualitative Methods in Health Services Research: Describing and Understanding Health Care

Victoria A. Parker, CHQOER, Bedford, MA, BG Bokhour, CHQOER, JA Clark, CHQOER, DA Paterniti, HQCUS, James Tulsky, MD, Center for Health Services Research in Primary Care, Durham VA Medical Center

Objectives: Qualitative methods are increasingly requested in program announcements, and are appearing more frequently in health services research. Qualitative methods include many different research strategies, which share important features including intensive contact with clinical practices or informants, seeking access to the behaviors, perceptions, interpretations and understandings of those individuals involved, explicit inclusion of context as relevant to gaining understanding, and use of language as a primary medium for analysis. Qualitative methods are particularly appropriate to the study of poorly understood or under-theorized health care processes, and often complement quantitative findings. They are, however, difficult to use without adequate theoretical grounding and practical training. In this workshop, we will present an overview of qualitative research methodology. Using examples from current projects, we will discuss approaches to research problem definition, sampling strategies, data collection techniques, analytic procedures, and issues of reliability and validity. In addition, we will initiate the development of a collaborative network of investigators within VA HSR&D to promote well-designed, informative qualitative studies in the future.

Activities: The workshop will consist of several distinct segments, led by different presenters. We will discuss connections between question formulation and decisions about using qualitative methods, either alone or as part of a multi-method approach. Participants will be encouraged to share their own experiences with exploring such methods. We will provide an overview of the landscape of different kinds of qualitative data collection, including interview data, data from participant observation, interactional data, and data derived from archival materials such as written policies and meeting minutes. Using two recent projects, we will address issues of sample size, theoretical sampling, alternative approaches to the analysis of qualitative data, the use of computer-assisted text analysis and the assessment of reliability and validity. One project combines in-depth interviews, focus groups, and survey data to examine the impact of early-stage prostate cancer and its treatment on patients' quality of life.. The other uses interviews, observations of workgroup events, and observations of clinician-patient interaction to develop a model of ways in which organizational context influences clinician-patient relationships. To foster collaboration with colleagues, we will identify and discuss common issues encountered by researchers using qualitative methods. Participants will receive a packet of resources and suggested readings for furthering their explorations of qualitative methods. Participants can expect to become more discerning consumers of qualitative proposals and studies, and to know what questions to ask when qualitative methods are under consideration.

Target Audience: This workshop is designed for health services researchers who currently use qualitative methods as well as for those who are interested in using qualitative methods in the future. It will also be valuable to those who are responsible for reviewing proposals with significant qualitative components, and for consumers of research based on qualitative methods.

Audience Familiarity: The workshop design assumes little to no familiarity with qualitative methods, but would be enriched by the participation of researchers experienced with such methods.

158. Use of Diagnostic Cost Groups to Predict Resource Use by VHA Patients

William A. West, PhD, VA Center for Healthcare Quality, Outcomes and Economic Research, Bedford, MA, A Rosen, VA Center for Healthcare Quality, Outcomes and Economic Research, A Hendricks, VA Center for Healthcare Quality, Outcomes and Economic Research, S Loveland, VA Center for Healthcare Quality, Outcomes and Economic Research, J Anderson, VA Center for Healthcare Quality, Outcomes and Economic Research, J Rothlender, VA Center for Healthcare Quality, Outcomes and Economic Research

Objectives: Diagnostic Cost Groups (DCGs), a case-mix measurement system used by Medicare and managed care organizations for risk adjustment, is currently under consideration for use by the Veterans’ Health Administration (VHA). The purpose of this study was to determine how well DCGs predict resource use by VHA patients and potential impact on Veteran’s Integrated Service Networks (VISNs).

Methods: We identified a 1% random sample (n=26,095) of veterans receiving treatment in VHA medical centers during FY97. Inpatient, outpatient and extended care diagnosis files were used to find all diagnoses (from face-to-face clinician encounters) recorded for these patients. DxCG™ software was used to classify patients based on their diagnoses and to calculate their "relative risk" (expected resource use relative to the baseline population). VHA utilization data for these patients were extracted from FY97 outpatient procedure and inpatient treatment files. As a proxy for resource use, we used Medicare’s DRG relative weights, Relative Value Units (RVUs), and fee schedules to calculate the national average Medicare reimbursement for each inpatient stay and outpatient service. For inpatient stays, the DRG relative weight was multiplied by $3800 (national average DRG ). For outpatient physician services (CPT procedure codes), we multiplied the total RVU by the national conversion factor ($36.68). For lab tests and other non-physician services we used Medicare’s national fee limit (or state-wide average if no national limit). Approximately 8% of outpatient services did not have a corresponding RVU and therefore were not included in the reimbursement calculations. These "costs" were totaled for each patient and regression analysis was used to determine how well DCG relative risk scores predicted outpatient, inpatient, and total cost. Observed-to-expected ratios were aggregated to estimate the effect at the VISN level.

Results: Based on Medicare reimbursement rates, the average outpatient cost per patient was $993, average inpatient cost was $927, and average total cost was $1920. The DCGs predicted 13% of the variation in outpatient cost, 23% of the variation in inpatient cost, and 43% of the variation in total cost (p>.0001). At the VISN level, observed-to-expected ratios ranged from .88 to 1.17. An analysis of observed-to-expected ratios by cost level revealed DCGs over-predicted cost at the low end and under-predicted cost at the high end. The average observed-to-expected ratio for patients whose observed cost was less than $1000 was .47 and the average ratio above $1000 was 1.95.

Conclusions: While DCGs appear useful in predicting resource use, the large difference in observed-to-expected ratios by cost level suggests a need for modifications to increase accuracy. Improved accuracy at the patient level could have a substantial impact at the VISN level.

Impact: Efforts to control costs while maximizing efficiency in health care spending are dependent on accurate case-mix measurement and risk adjustment. The results of this study suggest DCGs offer promise as a method for risk adjustment in the VHA but need modification in order to avoid inaccurate measurement at aggregate levels.

HSR&D Funded: MPC97009.

159. Enrollment in the Medicare Program among Veterans Turning Age 65

Steven M. Wright, PhD, Massachusetts Veterans Epidemiology Research and Information Center (MAVERIC), West Roxbury, MA, M Hossain, Massachusetts Veterans Epidemiology Research and Information Center (MAVERIC), LA Petersen, Houston Center for Quality of Care and Utilization Studies

Objectives: As the veteran population ages, many veterans who use the VA (VA users) are approaching age 65 and will be initially enrolling in the Medicare Program. The extent to which VA users utilize Medicare and/or VA services upon turning age 65 is unknown. The purpose of this study was to determine the extent of Medicare enrollment among VA users turning age 65, identify factors that may influence enrollment status, and explore patterns of VA and Medicare use before and after age 65.

Methods: We established a cohort of all veterans who were users of either inpatient or outpatient VA services between 1992 and 1994 (VA users) and who had their 65th birthday in 1994. We linked this cohort with the Medicare Denominator file by SSN, gender and date of birth to obtain Medicare enrollment status. Demographic and utilization data were obtained from the VA Patient Treatment File (PTF) and Outpatient Clinic Files (OPC) and the Medicare Denominator and MEDPAR files. We identified birthdays in 1994 so that we had sufficient data to measure utilization of VA (inpatient and outpatient) and Medicare-financed (inpatient) services three years before and three years after VA users 65th birthday.

Results: There were 83,356 VA users who turned age 65 in 1994. 18,847 (22.6%) were enrolled in the Medicare program prior to turning age 65 based on disability entitlement. Of the remaining 64,509 VA users eligible for Medicare based on age, 79.1% were enrolled in Medicare after age 65. Overall, 83.8% (n=69,894) of VA users turning age 65 in 1994 enrolled in Medicare. There was no difference in enrollment status between VA users who were service connected versus non-service connected. VA users enrolled in Medicare were more likely to have a VA hospitalization in the prior year and a higher number of VA outpatient visits than non-enrollees. Among VA users initially eligible for Medicare based on age, VA utilization declined after age 65. VA outpatient care was used by a greater percentage of patients before versus after age 65 (92.8% vs 76.2%). VA inpatient care was used by a greater percentage of patients before versus after age 65 (27.1% vs 22.9%). Medicare inpatient care was received by 13.5% of patients after age 65.

Conclusions: Most VA users enrolled in the Medicare program upon turning age 65. However, nearly one quarter (23%) of VA users under age 65 were already enrolled in Medicare based on disability entitlement. We found no differences in Medicare enrollment status by VA service connected status. VA inpatient and outpatient services are used by more VA users before versus after age 65.

Impact: Medicare provides an alternative source of health care services for most VA users after their 65th birthday. The availability of Medicare services has serious implications for the use, costs and planning of healthcare services for disabled and elderly veterans. Understanding the full extent to which veterans utilize VA and/or Medicare services when turning age 65 requires more research, including measures of patient comorbidities and healthcare market factors.

160. Provider Profiling: An Example of Inconsistent Results When Using Aggregated Patient-level Data

James A. Rothendler, MD, Bedford VA Medical Center, Bedford, MA, A Rosen, Bedford VA Medical Center, D Berlowitz, Bedford VA Medical Center, Z Cong & W West, Bedford VA Medical Center, S Loveland & A Lee, Bedford VA Medical Center, L Kazis, Bedford VA Medical Center

Objectives: Providers, such as VISN's, are being profiled on a variety of measures that capture health outcomes and resource utilization. In order to understand observed differences in outcomes among providers, analyses may be performed that relate resource utilization to health outcomes. This study illustrates how patient-level vs. VISN-level analyses may result in different conclusions.

Methods: Because the quantity of medical services rendered for a given disease burden may be one factor that affects outcomes (i.e. more services yielding better results), we studied the relationship between outpatient resource utilization and health related quality of life (HRQoL). We analyzed data on 31,432 veterans who responded to a mailed HRQoL survey (Veterans SF-36) in January-February, 1998. The patients receiving the survey had been based on a national probability sample and randomly selected by hospital within each VISN from among those who utilized outpatient VA services. Physical and mental component summary scores (PCS and MCS) were calculated from the Veterans SF-36 for each patient and also aggregated at the VISN level. For each patient, outpatient utilization was calculated for FY 1998 by summing the CPT-4-based relative value units (RVU's) for all outpatient visits and procedures. An "expected" utilization score was determined by multiplying the utilization mean by a case-mix adjusted severity score based on patient-specific ICD-9 diagnoses and demographics. The case-mix adjuster employed was the Diagnostic Cost Group (DCG) model which has been utilized to predict resource use (i.e. costs) in both Medicare patients and managed care organizations. An observed to expected resource utilization ratio (OER) was then calculated on both a patient and VISN level.

Results: At the VISN level, there was a significant positive correlation between PCS and OER (R=0.60, p<.003), and a non-significant correlation between MCS and OER (R=0.06, p=.81). This would suggest that at this level of aggregation, higher than expected utilization is associated with better PCS HRQoL scores. However, at the patient level, there was a small negative correlation between PCS and OER (R= -0.04, p<.0002), and between MCS and OER (R= -0.12, p<.0002), suggesting that a higher level of care than expected is associated with slightly lower HRQoL scores.

Conclusions: Patient level data suggests that higher outpatient utilization (in relation to risk-adjusted predicted utilization) is associated with slightly lower levels of PCS. However, if these data are analyzed at the VISN level, different conclusions would be reached. While neither the patient nor VISN-level analyses directly address the issue of causal relationship, the patient level data suggests that it unlikely that higher VISN-level OER is causally related to better HRQoL. (This may be an example of "ecological fallacy").

Impact: Understanding variations among VISN's and other populations within the VA is an important issue in efforts to improve quality of care. This study suggests that in trying to understand causal associations, caution must be used in utilizing group-level data.

HSR&D Funded: MPC-970091.

161. An Examination of Factors Associated with Physician Treatment Recommendation for Coronary Artery Bypass Graft Surgery

Michelle van Ryn, PhD, MPH, Minneapolis HSR&D Center of Excellence, Center for Chronic Disease Outcomes Research, Minneapolis, MN, E Hannan, Professor and Chair, Department of Health Policy, Management, and Behavior, School of Public Health, University at Albany, Rensselaer, New York

Objectives: This study was intended to determine the degree to which physicians' ratings of patients' personality, ability, and behavioral likelihoods were associated with physicians' recommendations for Coronary Artery Bypass Graft surgery (CABG) while controlling for clinical appropriateness and necessity for CABG according to criteria established by RAND as well as payer, patient demographic characteristics, clinical characteristics and coronary anatomy, and physician age, sex, specialty, and race/ethnicity.

Methods: In the first study stage, the RAND criteria was applied to detailed clinical information abstracted from medical records to determine if CABG surgery was appropriate and/or necessary in a stratified random sample of 4905 patients receiving an angiogram in one of 8 randomly selected hospitals in N.Y. The RAND criteria utilizes medical record and angiography data to provide an appraisal of appropriateness/necessity by examining the intersection of numerous indications particular to each patient. In the second stage, survey data was collected from a weighted random sub-sample of 950 patients. For this sample, physicians identified by the hospital, patient, or another physician as having a major role in the treatment determination were sent questionnaires (N=1384). The questionnaires included questions about the physicians' recommendations, reasons for the recommendation, their perceptions of characteristics of the post-angiogram encounter, and their perceptions of patients' abilities, personality, and behavioral likelihoods, among others. Physicians returned questionnaires on 64% (886) of the encounters sampled. Logistic regression was used to develop models predicting physician recommendations for CABG among CAD patients who were classified as appropriate and necessary for CABG.

Results: Physicians' perceptions of patients' behavioral characteristics were significantly associated with their treatment recommendations among both patients who were rated by RAND as appropriate and necessary for CABG, controlling for clinical characteristics, co-morbidities, coronary anatomy, age, payer, race, sex, and physician demographic characteristics. Physicians' ratings of patients on these behavioral characteristics were associated with patient race/ethnicity and appear to mediate the observed relationship between race/ethnicity and access to revascularization in this sample.

Conclusions: Reports on this study published elsewhere indicated that African Americans rated by RAND as "appropriate" or "necessary" for CABG were less likely to receive CABG than their white counterparts, controlling for patient payer and demographic characteristics. In addition when patients who were appropriate or necessary for CABG didn't receive it, in 90% of the cases the physician reported not recommending CABG. These prior results indicated that a significant proportion of under-use of CABG is due to physician recommendation. The present study suggests that physicians' perceptions of patients' behavioral characteristics has a significant impact on their treatment recommendations, and in turn physician perceptions of patients on these characteristics are associated with patient race/ethnicity.

Impact: Numerous studies in the past decade have reported race/ethnicity differences in access to a wide variety of treatments that persist independent of payer, appropriateness and treatment site. The findings presented here have significant implications for understanding race differences in access to treatments, as well as for interventions to improve equity in access both within and outside the VA system.

162. Smoking Cessation Patterns Across 35 Years of Follow-up in the Normative Aging Study.

Arthur J. Garvey, PhD, Harvard School of Dental Medicine, Boston, MA, PS Vokonas, MD, Normative Aging Study, VAOPC, Boston, MA

Objectives: The prevalence of cigarette smoking in the United States has decreased over the past 35 years. Most results on smoking prevalence rates to date, however, have been based on cross-sectional surveys which are vulnerable to mis-classification errors that can distort estimates of smoking prevalence and rates of smoking cessation. We used prospective data collected for 35 years on a cohort of community-dwelling veterans to provide more reliable information on smoking cessation patterns over the period 1963-1998.

Methods: Subjects were from the Normative Aging Study (NAS), a longitudinal and interdisciplinary study of the aging process in men supported by the Department of Veterans Affairs since 1963. The 2,280 members of the NAS represent a middle-class sample of men who reside in the greater- Boston area. Mean age at entry to the NAS was 41 years (range 21-80 years). We calculated quit rates for each year from 1963-1998, and we also sub-divided the sample by age group, education status, and amount smoked, and calculated quit rates for these subgroups. The cessation rate for a given year was obtained by dividing the number of quitters in a given year by the total number of subjects who began that year as smokers.

Results: We found very substantial rates of smoking cessation in this cohort of smokers. The majority (64%) quit smoking over the 35-year period of follow-up. Yearly rates of cessation rose substantially in the period 1965-1969, after the publication of the landmark Surgeon General's Report in 1964, and remained high for more than 20 years, with a decline being obvious in the period beginning with the early 1990s. The most striking sub-group differences were between light (<= 20 cigarettes per day) and heavy smokers (> 20 cigarettes per day), with heavy smokers consistently less likely to quit in any given year. Age-group and education-level differences were time-period related, with both older smokers, and the highly educated more likely to quit in the early years after entry to the NAS, while younger smokers and less-educated smokers tended to "catch-up" in ceasing smoking after the mid-1980s.

Conclusions: Results suggest that the majority of male smokers will quit smoking as they become older, a finding that may give encouragement to those providing care to patients who currently smoke. The sole exception to this general trend were heavily-dependent smokers, who quit in much smaller proportions compared to lighter smokers. Because of the much higher risk of morbidity and mortality among heavy smokers, special pharmacologic and behavioral interventions should be considered to help them cease smoking.

Impact: Our study seems to provide the most compelling evidence to date that it is the heavy smokers who are less likely to quit smoking in any given year, and over a protracted period of years. A clear implication of these findings is that pharmacotherapies such as nicotine replacement or bupropion, and/or intensive behavioral therapies will almost certainly be needed to help these heavy smokers quit.

*163. A Conceptual Model of Clinical Structure

Myles D. Spar, MD, VA Greater Los Angeles, Sepulveda, CA, K Kahn, UCLA General Internal Medicine, D Carlisle, UCLA General Internal Medicine

Objectives: The Institute of Medicine concluded "widespread quality problems exist throughout American Medicine." A major effort in needed to remedy this.We propose a model of clinical structure which aims to add to the understanding of determinants of better quality. Donabedian’s traditional model of health care quality describes structure, process, and outcomes as the essential elements. Measures of process and outcomes have been well described, but the effects of changes in structure have been less well-studied. In order to understand how some provider groups are having success at improving quality, we need to define what can be done on the organizational level to influence clinical practice, resulting in better processes and outcomes. We define clinical structure as the mutable organizational characteristics which target providers and patients in order to facilitate optimal use of available resources. We feel a new model of clinical structure is needed for several reasons. The impact of larger organizations on clinical care is not included in existing measures of quality, that influence is predictive of process and outcomes, and a novel data set is needed to assess this clinical structure. We have the opportunity to measure the validity of such a new model through a project incorporating patient level data from patient surveys and medical record reviews from 56 different groups of providers. We propose a conceptual model of clinical structure based on a review of the literature of interventions showing significant impacts. We have grouped the interventions into five domains including patient support, provider support, continuity, locus of control, and provider-patient match. We developed these through careful consideration of possible agents, targets, settings, and tools related to the interventions described with the goal of developing a survey of medical directors from each medical group. In order to develop the survey, we carefully refined our model in terms of specific elements about which we could ask in a reliable way and which we felt were most likely to impact on process or outcomes as shown in the literature. For example, with regard to the domain of patient support, we subdivided this into subdomains of: patient education, reminders, communication, provider interaction, and access to services. Within patient education we considered actions taken by physicians, extenders, nurses, other staff or patients themselves; possible targets were all patients or subsets of patients; the possible setting includes before, during, or after the visit; possible tools include written materials or tapes (simply made available or actively handed out), classes or support groups or computerized and interactive. Studies show that written and taped materials, interactive educational programs, one-on-one self-management programs, group classes, and support groups all lead to improvement in symptoms or reduction in urgent visits. We therefore crafted items asking about these specific programs. We reiterated this process for each of the subcategories within each of our five domains in order to develop our survey which will be used to both describe the current state of clinical structure and to validate our model.

*164. A Method to Estimate Prescription Status Using a VA Pharmaceutical Administrative Database

Cuc H. Quach, PharmD, VA San Diego Healthcare System, San Diego, CA, BR Phillips, VA San Diego Healthcare System, RE Barber, VA San Diego Healthcare System, E Ortiz, VA San Diego Healthcare System, SA Bozzette, VA San Diego Healthcare System

Objectives: To automate assessment of antiretroviral (ARV) prescribing for quality improvement purposes using a VA pharmaceutical administrative database to estimate what prescriptions were active for a given patient on a given date.

Methods: The Immunology Case Registry (ICR) is a longitudinal comprehensive database containing demographic and clinical information on all HIV-infected veterans receiving care at the VHA. We developed a computer-generated algorithm (CGA) using the outpatient pharmaceutical file in the ICR to estimate active ARV prescriptions. The ICR includes information on medication name, fill and refill dates, quantity supplied, but it does not include information on whether a prescription is active. The study population included ICR patients with activity (outpatient, inpatient, pharmacy, or laboratory) 90 days before the index date (ID). We estimated active prescriptions under varying rules for an arbitrarily selected ID. The definition of active prescription was varied by two related factors. One factor was whether we included medications filled only in the period before the ID (upper limit) or in the period before and after the ID (lower limit). The other factor that we varied was the length of the period before the ID (and after the ID for consideration of common prescriptions). We considered 45 and 90-day intervals. To assess the various definitions, we compared the estimates of active prescriptions produced under each definition to the results of a case-by-case clinical review by a pharmacist specializing in HIV care for a random sample of 250 patients. All pharmacy records were reviewed for up to 180 days before and after the ID using agreed upon rules. Of these, 50 records were independently reviewed by a physician and pharmacist to test inter-rater agreement. To compare the pharmacist review and the CGA estimates under various definitions, we classified each of the 250 patients as on highly active antiretroviral therapy (HAART) or not. HAART is defined as three or more ARVs from specific ARV classes. We compared the pharmacist and the CGA classification under the four approaches using percent agreement and kappa statistic.

Results: Inter-rater agreement between physician and pharmacist was 100%. Agreement between pharmacist-generated profiles and 45-day lower, 45-day upper, 90-day lower, and 90-day upper limit CGA profiles was 60.4%, 72.4%, 78.8%, and 98.4%, respectively. Kappa statistics were 0.27, 0.39, 0.48, and 0.94, respectively.

Conclusions: The resulting kappa statistics indicate that CGA profiles of patients with activity and prescriptions 90 days prior to the index date agreed with the pharmacist-generated profiles to a very high extent.

Impact: This method provides a potentially useful and cost-effective tool for using pharmaceutical records to classify ARV profiles of large numbers of patients. Comparison of CGA classification to guideline based indicators may be a useful tool for continuous quality improvement.

HSR&D Funded: HIV 98-000

*165. A Multidisciplinary Approach with Sexual Trauma Survivors

Arlene Bradley, MD, VA Roseburg Healthcare System, Roseburg, OR, B Thacker, Eugene Veterans' Center

Objectives: To determine if a multidisciplinary approach to managing female sexual trauma survivors correlates with improved health status and decreased utilization of expensive health services.

Methods: A retrospective case series study of 15 female veterans with complex histories of childhood and military sexual assault. VA and Vet Center medical and mental health providers (physician, psychologist, nurse trauma therapist, social worker) began collaborating January 1997, providing multidisciplinary care in a team case-management model to these survivors. Data about patients’ qualitative aspects of health (e.g. employment status, suicidality, destructive behavior, adherence to medical treatment) and healthcare service utilization (e.g. hospitalizations, clinic visits) was collected for the adjusted time period prior (minimum of 9 months) to collaborative multidisciplinary treatment. This was then compared to similar data for the time period when collaboration occurred.

Results: Despite limited resources and geographic separation, qualitative improvement in patients’ health and lifestyle occurred. In addition, utilization of expensive healthcare resources diminished. Hospital admissions and bed-days-of-care were much lower. Interestingly, mental health clinic visits occurred much more frequently during the time period that collaboration occurred.

Conclusions: Limitations of case series reports curtail the conclusions that can be derived. However, this case series seems to suggest that a multidisciplinary approach to managing this group of patients correlates with decreased utilization of expensive health services, decreased healthcare expenditures, more effective psychotropic prescribing, decreased patient pain, increased patient satisfaction and overall improved health status. Limitations of this report include Case series study design; Small sample size; Lack of standardized measurement health status; Non-VARHS services (unless reported to providers) may not be included in report; Cause and effect can’t definitively be determined; Cost analysis not performed; Statistical analysis of data limited.

Impact: The tantalizing differences should prompt more rigorous investigation and clarification of the relationship between multidisciplinary care and health status, health service utilization, and health care expenditures in complex female trauma survivors.

*166. A National Program Initiative to Evaluate Spinal Cord Medicine Clinical Practice Guidelines

Audrey Nelson, PhD, RN, Tampa VAMC, Tampa, FL, BN Simmons, Tampa VAMC, HT Bosshart, Augusta VAMC, SP Burns, Seattle VAMC, ML Guihan, Hines VAMC

Objectives: This project includes phase one of a larger study designed to develop, implement, and evaluate a model to integrate clinical practice guidelines into the routine operation of spinal cord injury (SCI) centers. Specifically, we have designed a model for implementing two SCI practice guidelines into practice, which is being implemented and evaluated at six VA SCI centers.

Methods: We will evaluate the model using a time-series design, with data collected over three points in time: pre-publication of the guidelines, pre-implementation of the guideline integration model, and post-implementation of the guideline integration model. Data will be collected from chart reviews, using the patient as the unit of analysis. Additionally, we will use a nested factorial design to compare the proportion of veterans who experience adverse outcomes in the six intervention sites (N=1,350) with veterans in the 17 non-participating VA SCI Centers (N=5,150). In addition to testing the two hypotheses, we will also conduct a process evaluation, describing the process of disseminating and implementing these two guidelines at six facilities. Specific data collection strategies for the process evaluation include focus groups, medical record reviews, interviews, extraction from administrative databases, process logs, and surveys.

Results: A combination of expert panels, focus groups, and surveys were used to design the model and target guideline implementation strategies. Eight focus groups were held nationally with providers. Perceived barriers included time for documentation, lack of provider cues, patient compliance, and level of evidence to support specific recommendations. Based on these barriers four implementation strategies are being implemented nationally: Computerized Clinical Decision Support Systems, Patient-Focused Interventions, Local Opinion Leaders, and Social Marketing/Outreach Visits. Baseline data on provider adherence, provider attitudes about clinical practice guidelines, and knowledge of guidelines will be available 12/99.

Conclusions: The newly designed Model for Implementing Clinical Practice Guidelines in Spinal Cord Injury is being implemented at six SCI Centers in the VA. The model has received support from the Chief Consultant for Spinal Cord Injury and Disorders Strategic Healthcare Group, Paralyzed Veterans of America, National Consortium for Spinal Cord Injury Medicine Clinical Practice Guidelines, and SCI QUERI. Efforts are underway to evaluate the impact of this model on provider adherence, clinical outcomes, and cost.

Impact: Although VHA has made progress in the implementation of clinical practice guidelines in recent years, significant work is needed to evaluate the outcomes of using clinical practice guidelines. Have we improved patient care? Have we reduced costs and eliminated unnecessary procedures and treatments? How have staff accepted the use of clinical practice guidelines in their daily treatment of patients? Are patients satisfied with their care when providers follow clinical practice guidelines? This program is specifically designed to address these difficult questions in one of the "special emphasis programs" in VHA, spinal cord dysfunction.

*167. A Randomized Trial to Compare Methods of Eliciting Utility Measures for Cost Utility Analyses

Julianne Souchek, Houston VAMC HSRD - Baylor College of Medicine, Houston, TX, J Stacks, Houston VAMC HSRD - Baylor College of Medicine, NP Wray, Houston VAMC HSRD - Baylor College of Medicine, CM Ashton, Houston VAMC HSRD - Baylor College of Medicine, B Brody, Baylor College of Medicine, Center for Ethics, Medicine, and Public Issues, MM Byrne, Houston VAMC HSRD - Baylor College of Medicine

Objectives: When cost-utility analysis is used to determine the value of treatment, it is important that the difference between treatments does not depend on the method used to obtain the utilities of the resulting health states. One of the objectives of this study was to determine whether utility scores obtained by 3 different techniques converged. The 3 techniques were the standard gamble, time trade-off, and rating scale. Furthermore, did the different techniques when used by the same individuals yield equal mean utility scores for several different health states?

Methods: We asked individuals with a recent diagnosis of advanced prostate cancer at 3 large hospitals to evaluate treatment choices and the health states that would result from the disease and treatment choice. There were 8 health states, 3 containing orchiectomy, 3 with medical castration, and 2 with no treatment. Participants were randomized to 2 techniques to be administered in one of 2 orders. One research technician trained by a social psychologist administered the evaluation techniques. The participants ranked the 8 states, then assigned utilities to the states by one method and then the other method. Three 2x2x8 repeated-measures analyses of variance were used to compare the means of the utilities of the states between techniques. Pearson correlation of the 2 sets of utility scores per person was computed and the mean correlation was computed for each pair of techniques.

Results: Approximately 60 individuals were randomized to each technique, 20 to each pair of techniques. The profiles of the mean utility scores were parallel but not equal. The time trade-off preference scores were higher than the rating scale values and the standard gamble scores. The rating scale and standard gamble scores were not different. Sequence effects were not significant; that is, the order of the techniques did not matter. The mean utility scores of the health states were different. Pearson correlation indicated there was convergence between techniques.

Conclusions: The differences between the 8 health states were preserved across the three techniques for obtaining utility scores, but the mean utility scores assigned to the states differed between techniques. The standardized utility scores for the 8 health states did not differ by technique.

Impact: Any of the 3 techniques will yield approximately the same difference in quality-adjusted life-years between two treatment programs.

HSR&D Funded: SDR 93-007.

*168. Impact of HSR&D Development Grant in VISN 8

Judith A. Laughlin, RN, PhD, Bay Pines VAMC, Bay Pines, FL, A Nelson, Tampa VAMC, M Rittman, Gainesville VAMC

Objectives: The purpose of the VA HSR&D Service Developmental Project Program is to provide venture capital that will encourage investigators in VA medical facilities without an existing HSR&D Service Field Program to explore innovative approaches for resolving problems that have direct relevance for VA's clinical mission. In 1997, nurse researchers located at three VISN 8 VAMCs (Tampa, Bay Pines, and Gainesville) were funded for a project entitled: "Nurse Staffing, Care Delivery, & Outcomes Across the Rehabilitation Care Continuum." The project's goals are: (1) Develop an infrastructure to enhance the capacity for health services research in VISN 8; (2) Establish a health services research and development training and support program; and (3) Submit fundable health services research projects.

Methods: An interdisciplinary Steering Committee chaired by the VISN 8 Network Director meets quarterly. These meetings serve as an opportunity to set specific goals, review progress, identify HSR&D resources, resolve problems and determine a strategy for continued growth. The three nurse researchers and various workgroups from the Steering Committee meet frequently via electronic communications to design, implement, and evaluate projects.

Results: The grant was expanded from 3 VAMCs to all facilities in the VISN. It has been successful in developing a network of researchers within VISN 8; focusing VA funded research into core areas plus substantially increasing VA funded proposals; and disseminating research findings. *VISN 8 developed and funded three positions and provided funds for training in Evidence-Based Practice. In addition, a total of 3 Post-Doctoral and 2 Pre-Doctoral VA Nursing Research Fellowships have been funded. Experts from HSR&D Field Programs, ERIC, VERDICT, MDRC, VA Health Economists, and National and International Health Care Leaders have presented seminars, conferences, mentoring experiences and proposal reviews. * In the past 2.5 years, three core areas of funded research in the VISN have been developed. They are Health Services and Systems; Rehabilitation; and Evidence-Based Practice. Between 1997 and October 1999, the VISN has increased its newly funded VA projects by 850% from $1,820,000 to $9,927,000. 54% or 11 of these projects are HSR&D projects and 46% or 6 are RR&D projects. Recently, the VISN was awarded funding for a 3-year HSR&D/VISN Collaborative Health Services Targeted Research Project. Eligibility to apply under this solicitation was limited to VISNs that do not currently have a national peer-reviewed and funded, HSR&D Center of Excellence geographically located within their network. *Dissemination of research findings includes national and international presentations and publications. To date, 50+ presentations have been made and 25+ research manuscripts have been published.

Conclusions: The Nurse Staffing, Care Delivery and Outcomes Across the Rehabilitation Continuum grant has successfully achieved all three project goals.

Impact: HSR&D Developmental Project Program Grants can be successful if there is VISN-wide and VISN Director support, a strong steering committee, and committed project leaders. Developmental projects are the first step toward establishing a core of experienced health services researchers, increasing funded projects, and eventually obtaining a HSR&D Center of Excellence.

HSR&D Funded: DEV 97-017.

*169. Accuracy of ICD_9-CM Codes in Identifying Stroke Patients

Dean M. Reker, PhD, RN, VAMC, Kansas City, MO and Center on Aging, University of Kansas Medical Center, Kansas City, KS, Kansas City, MO, BB Hamilton, Center on Aging, Duke University Medical Center, Durham, NC, PW Duncan, VAMC, Kansas City, MO and Center on Aging, University of Kansas Medical Center, Kansas City, KS, SJ Yeh, Harvard Extension School, Harvard University, Cambridge, MA, A Rosen, Center for Health Quality, Outcomes, and Economic Research, Bedford VAMC, Bedford, MA

Objectives: Stroke patients are often selected for research and reporting using ICD-9-CM diagnostic codes (430-438.xx). This is a study of the accuracy of these codes in identifying new (incident) stroke patients.

Methods: A sample of 415 stroke patients and 240 non-stroke patients were selected from 11 Veterans Affairs Medical Centers. A valid stroke diagnosis was determined by medical record review. Identification of the study sample was not wholly based on ICD-9-CM in order to estimate false negative cases. Stroke patients and non-stroke patients from administrative files were matched and merged with the diagnosis-validated sample. Accuracy estimates were calculated for individual ICD-9-CM codes and two coding algorithms, one of which maximized sensitivity while the other maximized specificity.

Results: Individual stroke ICD-9-CM codes appearing in the admission diagnosis field (cause of admission) and discharge diagnosis field (length of stay related) were analyzed for accuracy of identifying stroke patients. Two coding algorithms were identified with the following performance statistics: 1) 98% sensitivity; 7% specificity, and 2) 49% sensitivity, 79% specificity.

Conclusions: This study identifies significant variation in the accuracy of cerebrovascular ICD-9-CM codes in identifying stroke patients and proposes two coding algorithms for maximizing sensitivity or specificity. Policy analysts and investigators are urged to take caution when using administrative data to identify stroke patients.

Impact: This research 1) quantifies the accuracy of individual ICD-9-CM codes in the identification of new stroke cases and 2) proposes, depending on sample needs, two ICD-9-CM algorithms to identify stroke patients.

HSR&D Funded: ACC97-1141.

*170. African Americans with HIV-related Pneumocystis Carinii Pneumonia Receive Better Care than Whites and Survive More Often

Jorge P. Parada, Edward Hines Jr. VA Hospital, Hines, IL, JS Chmiel, Northwestern University, C Chan, Northwestern University, MB Goetz, VA Greater Los Angeles Healthcare System, AM Sipler, Northwestern University, CL Bennett, VA Chicago Health Care System - Lakeside Division

Objectives: African Americans are more likely than whites to receive high quality care for many medical illnesses and die less often as a result. The HIV-epidemic has had a disproportionately severe impact on racial/ethnic minorities, raising concern about poor AIDS care and outcomes for African Americans. We evaluated if HIV-infected African Americans received poorer care for Pneumocystis carinii pneumonia (PCP), one of the most common serious opportunistic AIDS-related infections.

Methods: Data on sociodemographics, severity of illness, processes of care, and outcomes were obtained from medical records for white and African American HIV-related PCP patients from 78 hospitals in Chicago, Los Angeles, Miami, New York, Seattle, Phoenix, Tucson, Memphis, and Nashville for the years 1995-1997.

Results: In comparison to PCP patients who were white (n=709) or Hispanic (n=445), African Americans (n=701) were most likely to: be injection drug users, to be unemployed, be homeless, and lack private health insurance (p<0.001 for each comparison). Sites of care differed, with African Americans and Hispanics being most likely to receive care at county hospitals, hospitals with teaching affiliations, and hospitals with high PCP experience levels (p<0.001 for each comparison). Patterns of outpatient care prior to admission differed, with African Americans being less likely to have received any antiretroviral therapy, PCP prophylaxis, or MAC prophylaxis (p<0.001 for each comparison). After adjustment for relevant clinical and sociodemographic factors, African Americans were more likely than whites to receive timely PCP medications (OR= 1.52, 95% CI= 1.13,2.05) and timely corticosteroids for moderate to severe PCP cases, (OR= 1.39, 95% CI= 1.14,1.69) and least likely to die (OR= 0.70, 95% CI= 0.51,0.97), while patterns of care and survival were similar between Hispanic and white PCP patients.

Conclusions: Compared to white PCP patients, African Americans were 39% to 52% more likely to receive timely inpatient therapeutic care, and died 30% less often. Differential use of county and not-for-profit hospitals, most of which had high AIDS-related PCP experience and high rates of timely use of PCP medications, may account for some of our findings.

Impact: AIDS-related PCP appears to be the first medical illness identified where African Americans are more likely than white persons to receive good medical care and die less often as a result.

HSR&D Funded: HSR&D Epidemiology.

*171. An Innovative Model (GIFT -Geriatric Interdisciplinary Faculty Training) for Formal CQK/TQM Training in Long Term Care

Denise K. A. Duffy, Department of VA Social Work/GRECC, Miami, FL, C. Williams, Department of VA GRECC/University of M School of Nursing, M. Silverman, Department of VA GRECC/U of Miami School of Medicine, Geriatric Medicine, E. J. Olsen, Department of VA GRECC/Psychiatry/U of M Geropsychiatry, N. Musson, Department of VA Speech GRECC/, L. Alfred, NOVA School of Pharmacy

Objectives: To increase MDS completion rates on schedule including all appropriate team members. Sub objectives: increase team understanding of improvement, development mission, understand current MDS process, find root causes for low completion rates

Methods: Percent MDS completed on time, pre and post test of team training in TQM/CQI.Patient/resident records were examined for all residents on one Nursing Home unit (n=47). Completion was determined for all quarterly (follow-up) MDS due from January 1998 to December 1998 (n+114). MDS completion tabulated separately for quarterly MDS due: a) prior to training (n=12), b) during training (n=35) and c) after training (n=67)

Results: Due pre-training (12 MDS): 11 (91.7%) missing. Due during training (35 MDS): 15 completed (42.9%), 9 partially completed (25.7%), 11 missing (31.4%). Due post-training (67 MDS), 5 completed (7.5%), 35 partially completed (52.2%) and 27 missing (40.3%). Improvement occurred during raining period. Lasting changed did not occur. Many barriers to completion were discovered. Additional PDCA cycles are needed.

Conclusions: JACHO visit confounds results. Assignment changes complicated the intervention. MDS software needed (handwritten method fraught with problems). Teaching needed re: MDS process and significance. Quality advisor needed to hold gains.

Impact: The far reaching effects of training professional, faculty and student professionals within the Department of Veterans Affairs Medical Center community on total quality management and continuous quality improvement in care to the veterans we serve. Our immediate goal is for geriatric patients but other age groups may also be affected in the training process.

*172. Are Veterans Willing to Pay for Lifestyle Medications?

Emily I. Yu, Center for the Study of Healthcare Provider Behavior, VA Greater Los Angeles Healthcare System, Los Angeles, CA, SM Asch, Center for the Study of Healthcare Provider Behavior, VA Greater Los Angeles Healthcare System, PA Glassman, Center for the Study of Healthcare Provider Behavior, VA Greater Los Angeles Healthcare System, NM Paige, VA Greater Los Angeles Healthcare System, LJ Passman, Center for the Study of Healthcare Provider Behavior, VA Greater Los Angeles Healthcare System, PG Shekelle, Center for the Study of Healthcare Provider Behavior, VA Greater Los Angeles Healthcare System

Objectives: Healthcare providers have grappled with balancing patient and provider demand for lifestyle drugs with the increased cost that providing coverage would entail. As with many managed care plans, at Veteran Affairs facilities, finasteride (Propecia) and sildenafil (Viagra) are currently "non-formulary" medications for alopecia and erectile dysfunction, respectively. We studied willingness to pay from the patient perspective for these two lifestyle medications.

Methods: Between December 1998 and March 1999, male veterans from two VA Greater Los Angeles Healthcare System facilities who were in waiting rooms for primary care or impotence clinic appointments were recruited to complete an 18-item self-administered, anonymous survey. For patients with self-reported need of a lifestyle medication, we assessed willingness to pay using t-tests and linear regression models. For all participants, we examined how perception of VA care would change if lifestyle medications were made available.

Results: Of 402 eligible patients approached, 339 (84%) agreed to participate. The participant group consisted of mostly low-income adult men with racially and ethnically diverse backgrounds. Participants with self-reported need for each lifestyle medication were analyzed separately for primary care (PC) and impotence clinic (IC) groups. Willingness to pay for finasteride was not assessed in the impotence clinic group due to small numbers, leaving three groups for analysis: finasteride-PC, sildenafil-PC, and sildenafil-IC. Finasteride-PC participants had a mean willingness to pay of $5.10 for a one-month prescription of daily finasteride (wholesale cost=$27). Sildenafil-PC and sildenafil-IC groups both had a mean willingness to pay of $6.60 for four sildenafil pills (wholesale cost=$20). Higher income (p=0.004) and increasing self-reported degree of need (p=0.008) were associated with willingness to pay for finasteride-PC when controlling for age, race/ethnicity, insured status, and co-morbid conditions. For sildenafil-PC, higher income (p=0.04) and younger age (p=0.006) were associated with willingness to pay. No included variables were associated with willingness to pay for sildenafil-IC. For 50%-83% of all patients surveyed, VA care would be perceived more positively, regardless of need, if these lifestyle medications were made available.

Conclusions: Although patients were willing to pay more than their usual expected costs, their willingness to pay did not extend to cover the full cost of the medications, even at discounted wholesale prices. In this low-income veteran population, patients with self-reported need of a lifestyle medication are willing to pay 19%-33% of the wholesale cost or 2-3 times their maximum co-payment of $2 to acquire it. In our population, patient satisfaction with care would increase if either of these medications, and sildenafil in particular, were offered as part of the basic prescription drug benefit. Whether this potential increase in patient satisfaction and perhaps quality of life are worth the increased pharmacy costs remains a topic for further research.

Impact: This study demonstrates that making lifestyle medications available with moderately increased co-payments may be one mechanism for a healthcare system to increase patient satisfaction while controlling costs.

*173. Assessment of Spirituality in Drug Addiction

P. Bart Vrtunski, PhD, VARC, Cleveland Department Veterans Affairs Medical Center (DVAMC), Brecksville, OH, FW Harvey, VARC, Cleveland DVAMC, BJM Bradley, VARC, Cleveland DVAMC

Objectives: A major impetus for approaching the spirituality in drug addicts came from the Joint Commission on Accreditation of Healthcare Organizations (JCAHO) in 1994. The JCAHO mandated assessment of "the spiritual orientation of the patient". This is an archival study of an 88-item questionnaire constructed for assessment of spirituality in alcohol and drug users.

Methods: Patients participating in the study were 1250 male veterans (mean age 45 yrs, mean education 12 yrs). The spirituality questionnaire was admini-stered as one of measures used for intake asses-sment. All questions were statements related to various aspects of life and higher power. Response to each was a degree of agreement on scale 1 to 7. Principal component analysis was the method of data reduction. Correlation between components and clinical assessment tests was examined.

Results: The statistical analysis generated three principal components and they were named relation to higher power (HP), negative emotional responsivity (NE) and positive emotional responsivity (PE). For the three, the reliability index (alpha) was from .84 to .95. Four of the seven variables of Addiction Severity Index were significantly assosciated with NE component, while the HP was inversely related to Beck Depression and Barratt Impulsivity scales.

Conclusions: A large overlap between various clinical asses-sment tools (a number of which were not mentioned here) and our questionnnare, clearly points to a higly complex nature of spirituality. This points to a need for design and examination of different instruments, which would illuminate more precisely specific features of the targeted function. Even more important questions are to see how spirituality changes over time, how is it affected by treatments and/or whether it can predict outcome of treatment.

Impact: While the spirituality belongs in many different areas of study, concerning the clinical work with addiction the surface was hardly scratched. VA Health System with its numerous facilities may represent an ideal environment for significant contribution to this field.

*174. Association of Intern Workload with Length of Stay and with Patient Satisfaction

Marvin J. Bittner, VA Medical Center, Omaha, NE, EC Rich, VA Medical Center, Omaha, RL Recker, VA Medical Center, Omaha, PD Turner, Creighton University, RL Rospond, Drake University, MW Lubeley, VA Medical Center, Omaha

Objectives: Previous studies suggest that intern workload affects both resource utilization and patient satisfaction. We looked for these findings in a 12-month study of patients with a variety of diagnoses.

Methods: We studied patients of Creighton medical interns assigned to non-intensive care unit inpatient floors at the Omaha VA Medical Center July 6, 1998 through June 30, 1999. We used length of stay (LOS) as a measure of resource utilization. We studied satisfaction with the American Board of Internal Medicine’s Patient Satisfaction Questionnaire (PSQ). This ten-item instrument assesses physicians’ interpersonal skills. Each item has five responses, from 1 (excellent) to 5 (poor). We analyzed responses from patients who completed the questionnaire in a face-to-face interview the morning after admission; however, for patients admitted after 5 pm, the interview occurred the second day after admission. We noted severity of illness (Fiscal Year 99 DRG weight). We noted these intern factors: gender, census size, number of admissions during the 24-hour admitting period, experience (days since start of internship), faculty ratings of humanistic qualities, and in-training examination score. Using multiple linear regression, with natural logarithm of LOS as the dependent variable, we assessed the effect of intern workload (census, admissions) and controlled for DRG weight and intern factors (gender, experience, humanistic qualities rating, examination score). We performed a similar regression analysis with PSQ as the dependent variable.

Results: Among 434 patients, mean LOS was 6.1 days (standard deviation 7.28) and skewed to the right (skewness 4.0). Satisfaction scores (mean 1.90, standard deviation 0.76) resembled those in the literature. Mean census was 4.70 (s.d. 1.94) and admissions 2.01 (s.d. 1.20). Of the admissions, 33.8% were to women interns. Only 11.5% of the admissions involved interns with more than three admissions in the 24-hour period. In the LOS model, adjusted R-squared was 0.138 (p < 0.001). DRG weight was associated with longer stay (p < 0.001), but male intern gender was associated with shorter stay (mean 7.2 days for women, 5.5 for men, p = 0.048). In the satisfaction model, adjusted R-squared was 0.015 (p = 0.061). Male intern gender (p = 0.015) and a larger census (p = 0.038) were associated with greater patient satisfaction, but more admissions (p = 0.023) was associated with less satisfaction. There was a trend for more satisfaction among patients of interns with better ratings of humanistic qualities (p = 0.054).

Conclusions: In a setting with controlled intern workload, workload variables were not significant in explaining resource utilization. A model of satisfaction incorporating workload variables explained only 1% of variance in satisfaction. Factors associated with satisfaction included male intern gender, fewer admissions, and larger censuses. There was a trend for more satisfaction with interns who had better ratings of humanistic qualities.

Impact: An academic VA medical center can achieve a consistently high level of patient satisfaction with physician interpersonal skills when intern workload is controlled. In such a setting, workload effects on resource utilization may be undetectable. Opportunities for improvement in satisfaction may include tighter control of admissions and training in communication skills.

*175. Bed Closures and Incarceration among VA Mental Health Service Users in Upstate New York

Rani A. Desai, NEPEC, West Haven VA, West Haven, CT, RA Rosenheck, NEPEC, West Haven VA, S Banks, The Bristol Observatory, J Pandiani, New York State Office of Mental Health, RA Desai, NEPEC, West Haven, VA

Objectives: To examine incarceration of VA mental health service users in 16 counties in Northeastern New York State from 1994 to 1997, a period of extensive VA bed closures.

Methods: Data on service use were obtained on male patients who used VA mental health services, community hospitals and State hospitals. Public records were also obtained of men who were incarcerated in local jails. Using Probabilistic Population Estimation, the frequency distribution of date of birth in each sample was employed to determine the degree of overlap between clinical and incarcerated samples, without having to rely on personal identifiers.

Results: Altogether 12.3% of VA mental health inpatients and outpatients were incarcerated between 1994 and 1997: 26.5% of those aged 18-39 and 7.9% of those aged 40 or older. Dually diagnosed patients had the highest rates of incarceration (25%) followed by those with substance abuse alone (21%), and those with psychiatric problems alone (11%). Rates of incarcerationamong patients hospitalized in VA (11.7%) were lower than among those in General hospitals (23%)or State hospitals (21.7%), especially for those younger than 40 years old. There was no significant increase in the annual rate of incarceration from 1994 (3.7%) to 1997 (4.0%), in spite of extensive VA bed closures during that time.

Conclusions: Substantial proportions of mental health system users are incarcerated, especially younger patients with dual diagnoses, and these rates are generally similar across systems. The closure of substantial mental health inpatient beds in recent years does not appear to have increased the rate of incarceration in this population.

Impact: As VA has moved from a primarily inpatient mental health care model to more community-oriented treatment, there has been concern that vulnerable veterans might be more likely to be incarcerated instead of treated. These data do not support this concern.

*176. Characteristics and Outcomes of Women Undergoing Hysterectomies in VA Facilities

Frances Weaver, PhD, Hines VAMC, Hines, IL, D Hynes, Hines VAMC, D Ippolito, Hines VAMC, B Thakkar, Hines VAMC

Objectives: The objectives of this research are to: describe the patient sociodemographic, preoperative risk, structure and process characteristics and outcomes of women undergoing hysterectomies at VAMCs; and examine outcomes using multivariate modeling techniques.

Methods: This study is a secondary analysis of the National Surgical Quality Improvement Program (NSQIP) database. For the present analyses, all hysterectomy procedures collected by the NSQIP between FY92 and FY97 were selected. The NSQIP contains data on 62 preoperative risk and demographic characteristics, four process, and 23 outcome variables. We supplemented the NSQIP with data from the VA’s Patient Treatment File for readmission data, pathology reports from sites, and facility characteristic data. Any hospitalizations that occurred within one year of the procedure were captured. Outcomes include: postoperative length of stay, morbidity within 30 days (i.e., one or more of 21 complications monitored by the NSQIP), and any readmissions that occurred as a result of a complication of the procedure within one year post surgery. An expert panel of surgeons identified relevant diagnostic codes and assigned a time period within which specific readmission diagnoses would be considered a complication.

Results: Over a six year period, VA performed 1,758 hysterectomies. The majority were abdominal hysterectomies (75%), whereas 22% were vaginal, and less than 4% were laproscopic-assisted. 36 cases were radical procedures. The most frequent indications for surgery included leiomyomas (31%), bleeding problems (16%), and endometriosis (10%). Women were predominately white (64%), with an average age of 42.8 years, and a minority were married (37%). Preoperatively, 40% of these women were smokers, 14% had a history of hypertension requiring medications, and 5% were frequent alcohol users. The mean length of stay was 4.2 days (sd=3.7) and almost 9% experienced one or more complications within 30 days. Logistic regression modeling of 30-day morbidity identified 6 significant predictors. Preoperative patient variables, including impaired functional status, current smoker, alcohol use within two weeks of admission, dyspnea, and infection, as well as, longer operation times were associated with greater probability of morbidity (c-index=0.673). Readmissions due to complications of surgery occurred in 3.1% of women. Most frequent reasons included intestinal obstruction (17%), post-op infection (13%), and disruption of operative wound (11%). Modeling of postoperative length of stay and readmissions due to complications is in process and will be completed by March 2000.

Conclusions: Research on hysterectomy outcomes have focused on single outcomes and limited examination of preoperative characteristics. This work examining multiple preoperative characteristics and outcomes indicates that risky health behaviors and poorer health status prior to surgery were significant predictors of 30-day morbidity. Contrary to expectation, uterine size, indication for surgery, and surgical approach were not predictors of complications within 30 days. Complications resulting in readmission are comparable to that found in the private sector.

Impact: Women veterans account for less than 5% of all VA health care users. The rate of female-specific procedures at any one facility is very low. Given these infrequent events, this study will help to inform the system about factors that may affect women’s hysterectomy outcomes and point to areas for further research.

HSR&D Funded: GEN 97-016

*177. Chronic Obstructive Pulmonary Disease (COPD) Healthcare Costs for Veterans at the Atlanta VA Medical Center

Francesca M. Cordi, Atlanta VAMC and Emory University, Decatur, GA, DJ Bower, Atlanta VAMC and Emory University, TL McDermott, Atlanta VAMC and Emory University, NM Mencke, Atlanta VAMC, JL Stephenson, Atlanta VAMC, KAS Cannella, Atlanta VAMC and Emory University, WN Kelly, Atlanta VAMC, SM Aguayo, Atlanta VAMC and Emory University and Morehouse University

Objectives: The fourth leading cause of death in the United States (1995), chronic obstructive pulmonary disease (COPD) is an increasing burden for our nation’s healthcare systems. Our study focused on three questions: 1. What was the resource consumption at the Atlanta VA for COPD in 1997? 2. How do these costs compare to overall Atlanta VA cost estimates? 3. How do these costs compare to national cost estimates?

Methods: Using Decision Support Service (DSS) software from Transition Systems, Inc., we obtained data on all patients and patients with an ICD9 or a DRG code for COPD seen at the Atlanta VA in 1997. We compared our data with the US National Hospital Discharge Survey (NHDS) of 1997, and with the national VA data reported for 1994.

Results: At the Atlanta VA, there were 343 total admissions for 264 patients with COPD in 1997, at a total cost of 2.9 million dollars, and an average cost of $8,590 per admission. Of the 264 patients admitted for COPD at the Atlanta VA in 1997, there were 12 deaths at a total cost of $240,000, and an average cost of $20,000 per admission. However, without including those 12 admissions, the average cost for a COPD admission was $8,182 per admission on patients 65 years old or older (n = 197), and $8,163 per admission on patients 64 years old or younger (n = 134). For outpatient expenditures, there were 2,681 outpatient encounters for 1,103 patients with COPD in 1997, with an average cost of $127 per encounter on patients 65 years old or older, and an average cost of $123 per encounter on patients 64 years old or younger. Compared to all admissions at this facility, COPD admissions were 6 percent, and costs were 7 percent of the inpatient budget. In contrast, COPD outpatients accounted for only 0.25% of the workload and 0.002% of the outpatient costs. Of note, at our VA, 11 of the top 16 most costly DRGs also have COPD as a major comorbidity. In comparison, the NHDS reported 391,747 total admissions for COPD in 1997, at a total cost of 3.32 billion dollars, and an average cost of $8,475 per admission. Similarly, the national VA database reported 24,157 total admissions for COPD in 1994, at a total cost of 232.2 million dollars, and an average cost of $9,612 per admission. Acknowledged limitations of our data include possible misdiagnoses, the possibility of inappropriate coding, difficulty locating comparable data, a gender skew to males (only 5 admissions for women veterans with COPD), and the fact that VA costs incorporate professional fees.

Conclusions: Our data suggest that our expenses for COPD admissions are similar to those in other facilities. In addition, our data raise the possibility that by increasing ambulatory care interventions in an effort to prevent admissions we may be able to reduce our total expenditures for COPD.

Impact: Future studies will examine these costs in more detail, as well as the unknown differentials between smokers and nonsmokers, and between smokers with and without COPD.

*178. Clinical Characteristics and Treatment Preferences of Veterans with Prostate Cancer

Timothy J. Wilt, MD, MPH, Veterans Affairs/VISN 13 Center for Chronic Diseases Outcomes Research, Minneapolis, MN, Minneapolis, MN, MK Brawer, MD, Northwest Prostate Institute, Seattle, WA, ICV Netto, MD, Hampton VA Medical Center, Hampton, VA, K. Jones, M.S., Veterans Affairs Cooperative Studies Program Coordinating Center, Perry Point, MD, J Collins, ScD, Veterans Affairs Cooperative Studies Program Coordinating Center, Perry Point, MD

Objectives: To determine clinical characteristics and treatment preferences of a national cohort of veterans with newly diagnosed prostate cancer.

Methods: Veterans with newly diagnosed prostate cancer from 45 VAMC were evaluated for inclusion into the VA/NCI/AHCPR CSP #407: Prostate cancer Intervention Versus Observation Trial (PIVOT) a randomized trial comparing radical prostatectomy versus expectant management for men with clinically localized prostate cancer. As part of a larger longitudinal cohort study, clinical and tumor characteristics as well as treatment preferences wererecorded on all veterans regardless of PIVOT eligibility or participation.

Results: From November 1994 to October 1999 a total of 8820 men with newly diagnosed prostate cancer were identified and included on the registry. Men were predominately elderly (mean age = 66.1 years), Caucasian (69%) and had clinically localized prostate cancer (83%). Over half (54%) had Prostate Specific Antigen (PSA) levels < 10.0 ng/ml and 76% had either well or moderately differentiated prostate cancer. Initial treatment selection included radical prostatectomy (25%), expectant management (31%), radiation (19%), hormone therapy (18%), brachytherapy (1%). In men with clinically localized prostate cancer 10.7% were eligible and randomized into PIVOT. Men rolled in this randomized trial (n = 579) were similar to the national cohort of veterans with clinically localized prostate cancer who were eligible but declined participation (n = 3163) as measured by: age, race, PSA levels, histologic grade and health status. Overall, > 85% of men with clinically localized prostate cancer reported their health status as excellent, very good or good. Men declining PIVOT enrollment sited "not willing to leave decision for treatment to chance" as the primary reason for refusal to participate. Treatment preferences in men declining enrollment included: radical prostatectomy (37%); expectant management (32%); radiation (20%) and early hormone therapy (6%). Men choosing radical prostatectomy were younger, had a lower PSA, more likely to be of white race and less likely to have poorly differentiated prostate cancer than men choosing other therapies (all p values < 0.01).

Conclusions: Most veterans with prostate cancer have clinically localized disease. They are predominately elderly and Caucasian, report good-excellent health status and select a wide range of treatment options. Men enrolled in a randomized trial to determine the preferred treatment option for clinically localized prostate cancer are representative of a larger cohort of veterans with this disease. Enhancing enrollment in randomized trials of prostate cancer treatment requires addressing patients’ concerns about leaving cancer treatment decisions to chance.

Impact: Understanding patient characteristics and treatment preferences will improve prostate cancer health care delivery. Completion of the Prostate cancer Intervention Versus Observation Trial will provide the necessary information to determine whether early intervention with radical prostatectomy or expectant management provides superior length and quality of life in men with clinically localized prostate cancer.

HSR&D Funded: Yes (CCDOR Grant).

*179. Comparing Depression Screening Methods in Primary Care

Edmund F Chaney, PhD, VA PSHCS, Seattle Division MHS Beh Med, Seattle, WA, NM Hasenberg, Seattle VA HSR&D COE, SC Hedrick, Seattle VA HSR&D COE, GR Paden, VA PSHCS, Seattle Division Med GIMC, B Felker, VA PSHCS, Seattle Division MHS

Objectives: Despite increased focus on the importance of improving depression care in primary care, depression remains under-detected and under-treated. Quality improvement for chronic illnesses such as depression requires comprehensive organizational change. A necessary element is identification and screening to establish a patient registry. Project Mood, the first randomized trial of collaborative care for depression in VA primary care, used multiple methods of identifying patients with unipolar depression and/or dysthymia. We present relative strengths and weaknesses of patient identification through provider referral, waiting-room interview, self-administered waiting-room screen, and mailed questionnaires.

Methods: From January, 1998 through February 1999, we identified depressed patients in the Seattle VA General Internal Medicine Primary Care Clinic (GIMC) using four methods. 1. During the recruitment period we evaluated all possibly depressed patients referred to us by primary care providers. 2. The Ambulatory Care Quality Improvement trial (ACQUIP) was surveying the clinic population by mail, using the SF-36 Mental Health and Illness Scale to identify suspect affective disorder, then sending out the Symptom Checklist 20 depression symptomatology measure (SCL-20) . We contacted those patients positive on the SCL-20 for serious depression. 3. For the first six months of our study another study had been conducting waiting-room interviewing in the GIMC to identify patients with minor depression. We further evaluated those patients with four or more symptoms of depression. 4. For the last eight months of our study, the GIMC implemented a two-item self-administered depression screen as part of a health promotion survey. We evaluated all patients positive on this screen. Our evaluations consisted of a chart review and a 30-40 minute structured computer-assisted telephone interview by a trained interviewer. The interview was based on the Prime-MD with additional questions about previous treatment history and current barriers to treatment. The chart and interview information was assessed by a clinical psychologist to establish diagnoses and exclusion criteria.

Results: Project Mood evaluated 1458 primary care patients at risk for depression and found 24% already in treatment. An additional 26% met criteria for major depression and/or dysthymia and were interested in treatment. Factors reducing detection efficiency were patients leaving the clinic (3%), difficulty contacting patients despite extensive efforts (23%) and patients unable to understand the telephone interview (1%). Exclusion factors were not depressed (10%), another psychiatric problem primary (4%) and refusal (9%). Sensitivity of the four methods ranged from 68% for provider referral to 18% for self-administered screening.

Conclusions: Sensitivity of different depression screening methods is related to the method's intrusiveness, timeliness, effort and sustainability. Different methods would be appropriate for different organizational conditions and goals.

Impact: This study provides information about implementation issues and comparative efficiency of four depression identification strategies in primary care that can be used by planners in apportioning treatment resources to design effective collaborative depression treatment interventions.

HSR&D Funded: 95-097.B

*180. Costs of Healthcare for Women Veterans in Atlanta: a Look at Gender Differences

Francesca M. Cordi, Atlanta VAMC and Emory University, Decatur, GA, KAS Cannella, Atlanta VAMC and Emory University, JS Noblin, Atlanta VAMC, JL Stephenson, Atlanta VAMC, SM Aguayo, Atlanta VAMC and Emory & Morehouse Schools of Medicine

Objectives: Women have joined the military services in increasing numbers since World War Two. As a result, women are becoming veterans and using VA healthcare resources in increasing numbers. Since the early nineties, the National Health Institutes has required a statement in all NIH-funded research describing the inclusion or exclusion of women and minorities. In 1994, Congress mandated that military women be included in all Department of Defense-funded research. While there is some research to indicate that women using VA facilities receive high standards of care, little has been reported on total resource consumption by gender. Our preliminary study focused on this general question: Is there a difference in costs and outcomes for male and female patients in Fiscal Year (FY) 1998?

Methods: Using Decision Support Service (DSS) software from Transition Systems, Inc., and Premier’s PERSPECTIVE ONLINE software, we obtained information on clinical indicators (complications rates, readmission rates, mortality) and financial indicators (average length of stay (ALOS), costs) for men and women admitted to the Atlanta VAMC in FY 1998.

Results: We had a total of 5,871 admissions, of which 5,660 (96.4 percent) were male and 211 (3.6 percent) were female. Women’s readmission rate was 4.2 percent (9), their complication rate was 4.3 percent (9), and their mortality rate was 2.8 percent (6). Men’s readmission rate was 3.2 percent (180), their complication rate was 4.5 percent (253), and their mortality rate was 5.4 percent (307). The ALOS for women was 12.5 days. For men, it was 13.6 days. Atlanta VA expenditures for a woman averaged $6,227 per admission, only 68 percent of the cost for men ($9,097 per admission). The Atlanta VAMC expended a total of $52,805,043 on inpatients. Two and half percent ($1,313,831) went to the care of women, while 97.5 percent ($51,491,212) percent went to men’s care.

Conclusions: We conclude that the gender differences are larger for the number of hospital admissions, mortality, and costs than for ALOS, readmission rates, and complication rates, at the Atlanta VAMC for FY 1998.

Impact: The possibility of some gender bias in hospital admissions exists. Other limitations include inappropriate coding and questions regarding potential differences between genders in ICU stays. Future studies will investigate the lower number of hospital admissions and costs in the medical treatment of women as inpatients. Men’s increased costs and mortality will be addressed. Outpatient records will be examined to determine what, if any, gender differences exist.

*181. Department of Veterans Affairs Spinal Cord Injury Quality Enhancement Research Initiative (SCI QUERI)

Charlesnika Mayfield, MPH, Edward Hines VA Medical Center, SCI QUERI Research Coordinating Center, Hines, Illinois, F Weaver, PhD, Edward Hines VA Medical Center, SCI QUERI Research Coordinating Center, M Hammond, MD, VA Puget Sound Health Care System, SCI QUERI Clinical Coordinating Center, SCI QUERI Group, Edward Hines VA Medical Center, VA Puget Sound Health Care System

Objectives: The Veterans Affairs Spinal Cord Injury Quality Enhancement Research Initiative (SCI QUERI) mission is to translate research discoveries and innovations into better patient care and systems improvement. The primary task of the SCI QUERI is to support, enhance, and integrate research that demonstrates a positive relationship to improved health and quality of life for veterans with SCI. Completion of the tasks of SCI QUERI will result in the documentation of the current state of knowledge in SCI care, identification of gaps in knowledge from both provider and veteran perspectives, and development of research initiatives to close these gaps.

Methods: The SCI QUERI has focused on several tasks. They include: 1. Documenting data resources available for SCI research including the National Spinal Cord Dysfunction Registry and other VA databases. 2. Convening focus groups with health providers and veterans with SCI to identify areas of concern and knowledge gaps. 3. Examining screening tools and components of the annual SCI evaluation to see if they are preventive approaches and improve quality of life. A survey will be sent to health care providers concerning the use of screening tools and whether the diagnostic yield and/or outcomes data is stored and accessible. The next step would be to assess diagnostic yield, subsequent treatment changes, and impact on patient outcomes. 4. Identifying best practices through evaluation of the implementation of clinical practice guidelines (CPGs) for Prevention of Thromboembolism and Neurogenic Bowel Management in Adults. SCI QUERI staff are also considering research questions pertaining to recommendations in other CPGs to initiate and promote more evidence-based recommendations. 5. Development and dissemination of an RFA to address pressure ulcer prevention and management.

Results: Documentation of the registry and VA databases is ongoing. As a result of the focus groups, one RFA has been written and disseminated on Pressure Ulcer Prevention and Management and one is being developed on the Preoperative Risks and Outcomes of Surgery for Individuals with SCI. A manuscript on the focus groups is also in development. The survey for the annual evaluation components is being revised and will be sent to providers soon. Tools to implement the CPGs are being developed and their impact on outcomes are being assessed.

Conclusions: The ultimate goal of the SCI QUERI is to enhance the quality of life of veterans with SCI. This goal will be supported through the tasks QUERI has undertaken to provide data-driven, evidence-based best practices for the care of veterans with SCI and linking these with patient outcomes.

Impact: As a result of each of these tasks, several actions will be accomplished. One is the development and dissemination of RFAs to investigate research topics. Second is the building of an evidence-based preventive health program for SCI patients. Third is the generation of hypotheses for future research into patient care management and clinical practice. The final action will be evaluating the effectiveness of the dissemination and implementation of clinical practice guidelines.

HSR&D Funded: SCI 98-000.

*182. Depressive Symptoms and Relapse to Substance Use

Geoffrey M. Curran, PhD, HSR&D Center for Mental Healthcare Research, North Little Rock, AR, BM Booth, HSR&D Center for Mental Healthcare Research, JE Kirchner, HSR&D Center for Mental Healthcare Research

Objectives: To examine, in a VA substance use treatment population, if depressive symptoms at 3-weeks post-detoxification are a risk factor for relapse.

Methods: The sample includes 100 consecutive "discharges" from intensive outpatient treatment for substance use disorder(s) at a mid-Southern VAMC. Participants in this observational study were interviewed at treatment entry, treatment exit (3 weeks post-detoxification), and three months post-treatment.

Results: In multivariate logistic regression controlling for age, education, marital status, medical severity presenting at treatment entry, number of prior treatments, and receipt of prescription(s) of anti-depressants either during or post-treatment, depressive symptoms (BDI) measured at 3-weeks post-detoxification were significantly associated with relapse at 3 months post-treatment. Subjects with a BDI score of 20+, the common cut-point of "severe" symptomatology in this population, had an OR of 4.1 (comparison group = BDI<20). Subjects’ BDI score at treatment entry was not related to subsequent drug or alcohol use at 3 months post-treatment. Our data also showed that subjects with severe symptoms at treatment exit who received an anti-depressant prescription in the during-or-post-treatment phase (as a result of usual care) had better substance use outcomes at 3 months post-treatment than the severely symptomatic subjects who did not receive a prescription for antidepressants.

Conclusions: Our data indicate that assessment for depressive symptoms several weeks post-detoxification is most useful for identifying an at-risk population. While the "care as usual" at this VAMC resulted in the prescription of antidepressant medication to about half of the subjects who reported severe symptoms at treatment exit (clinicians did not view BDI scores), our data indicate that a formal assessment of symptoms as part of usual care would identify a larger population of at-risk individuals. Impact: Our data suggest that formal assessment of depressive symptoms at treatment exit (approximately 3-weeks post-detoxification), in conjunction with an enhanced clinical intervention for depression, could substantially improve care for this at-risk population.

*183. Determinants of Technology-related Decisions Among VISN Clinical Managers

Elizabeth J Adams, Management Decision and Research Center, Boston, MA, EC Alligood, Management Decision and Research Center, K Flynn, Management Decision and Research Center, MP Charns, Management Decision and Research Center, Boston, MA

Objectives: The MDRC Technology Assessment (TA) Program provides critical information to senior VHA managers to support evidence-based decisions used to advance veterans health care. This study examines VISN decision-making processes for using or acquiring health care interventions and the information used to support these decisions.

Methods: TA Program staff conducted structured telephone surveys of all VISN Clinical Managers (CM). The survey examined decision-making processes in using or acquiring health care technologies and information gathering processes employed by each VISN. The survey explored information sources used in decision making, satisfaction with quality and timeliness of information and preferences in modes of communication.

Results: Preliminary findings of the 7 (33%) CMs surveyed thus far show that the vast majority of technology decisions are for replacement of existing equipment and acquisition of new technologies. The most frequent decisions in the last quarter of FY 99 were for replacing imaging equipment and purchasing telemedicine systems. There is considerable variation in decision-making processes across VISNs. The locus of decision making depends largely on the degree of facility integration within the VISN. Networks with integrated patient care lines make all decisions at the VISN level, whereas less integrated networks, for example those with traditional service structures, decentralize lower cost technology decisions to the facilities. All networks have internal multidisciplinary committee structures in place. CMs prefer access to efficacy/effectiveness and cost data and rely heavily on clinical experts to synthesize evidence and make recommendations. Many consult colleagues’ experiences with the technology. In most cases, the methods of evidence synthesis were unknown and objectivity had been questioned. A minority of CMs had used external resources to supplement information gathering. Satisfaction with these resources was mixed, as the quality of information and the timeliness were often inadequate. CMs who have used TA Program reports were very satisfied with the quality and, to a lesser extent, the timeliness. Most CMs preferred summarized technology assessments or evidence syntheses to full reports but wanted access to the full reports when needed. For modes of communication CMs most preferred MS Exchange and least preferred the Internet, citing unreliability and sluggishness of the internal system as the main causes.

Conclusions: VISNs use traditional multidisciplinary processes to gather information about the use and/or purchase of health care technologies. Decision making across VISNs continues to emphasize evidence from clinical expert opinion, and the degree of bias in their opinions can vary substantially. There is a essential need to provide VISN leadership with objective, valid information from scientific research in a timely manner to promote evidence-based decision making in veterans health care.

Impact: The TA Program will use this information to expand its outreach to support VISN decision-making needs. In addition to producing high quality systematic reviews of the peer-reviewed literature, the TA Program will disseminate periodic electronic bulletins of technology-related information and will conduct regular horizon scanning to support rigorous evidence-based improvements to veterans’ health care.


*184. Developing an Illustrated Patient Satisfaction Evaluation Tool for Low Literacy Veterans

David A. Asch, MD, MBA, Philadelphia VA Medical Center, Philadelphia, PA, JJ Murphy, MD, Philadelphia VA Medical Center, EE Reynolds, MD, University of Pennsylvania Health System, AC Aguirre, MPA, University of Pennsylvania, JA Shea, Ph.D., University of Pennsylvania

Objectives: Both inside and outside the VA, tremendous effort is being directed toward measuring health care quality, with the ultimate goal of improving the care delivered to veterans and others. Patient satisfaction remains an essential dimension of health care quality, but most assessments of patient satisfaction are self-administered and so tend to exclude patients whose literacy skills are not up to the task of the instrument. These patients, who represent up to one third of veterans and perhaps larger proportion of patients outside the VA, are therefore systematically excluded when patient satisfaction is assessed using written materials. Moreover, their low literacy may create special needs that will remain underrepresented if text-based instruments are used. Our objective is to develop and evaluate an easy to administer patient satisfaction instrument that does not require English literacy and has the potential to reintroduce these stakeholders into the assessment of patient satisfaction.

Methods: We have begun translating established and tested patient satisfaction instruments (e.g., the VA Clinic Patient Satisfaction Form and the University of Pennsylvania Health System Outpatient Satisfaction Survey) into illustrations that communicate individual domains of patient satisfaction using pictures rather than words. To do this, we have auditioned and hired artists and have gone through several iterations to develop preliminary images. They will be refined on the basis of feedback from patient focus groups. The processes used to develop these images demonstrate the difficulties in but importance of communicating effectively with this substantial minority of veterans.

Results: At this writing, preliminary sets of images have been developed illustrating provider timeliness, provider continuity, provider clarity, timeliness of scheduling, and several other domains of patient satisfaction. For each of these, critical concepts can be communicated with pictures, although slightly different illustration styles or pictorial elements can have unexpected implications. Pictures that convey the emotion of the patient in a situation improve clarity. Pairs of pictures that express both the best and worst example of a domain (e.g., a provider looking with interest at a patient vs. a provider looking impatiently at his/her watch to express provider interest) improve clarity. Other essential elements in the image, for example, patient or provider ethnicity, may alter the message.

Conclusions: Effective communication of important patient satisfaction domains using pictures is feasible, although challenging. Expression of patient emotion may be critical in conveying a consistent message through images. Gender and ethnic information implied by images may alter their interpretation.

Impact: This project is aimed at providing a voice to low literate veterans in the assessment of quality of care in VA facilities.

HSR&D Funded: PCC 98-071.

*185. Development of Outcome Measures for Evaluating an Organizational Intervention

Nelda P. Wray, Houston Center for Quality of Care and Resource Utilization Studies, Houston, TX, MM Byrne, Houston Center for Quality of Care and Resource Utilization Studies, MP Charns, Management and Decision Research Center, CM Ashton, Houston Center for Quality of Care and Resource Utilization Studies

Objectives: Veterans Affairs' (VA) health care administration officials have encouraged VA medical centers to implement new organizational structures to improve quality of care and resource utilization. The Management Decision and Research Center (MDRC) in Boston and the Houston Quality of Care and Utilization Studies (HQCUS) Center for Excellence were commissioned by VA Headquarters to evaluate the effectiveness of one type of organizational structure, service lines (SLs), in improving quality of care in VA. SLs organize health care provision around the type of care provided, e.g. primary care, mental health, rather than around the type of health care provider, e.g. clinician, nursing services. The purpose of this study was to develop a conceptual framework that would drive the development of specific outcome measures to assess the performance of SLs in improving quality of care.

Methods: We developed the following hypothesis for our inquiry: In comparison with other organizational structures, SLs are more successful in achieving the patient care goals of a health care system. We reviewed the patient care goals of the VA health care system as manifested in the Network Directors performance standards and also considered other important patient outcomes. We used two factors to evaluate potential outcome measures. First, we assessed whether there were clear mechanisms by which organizational structure would affect these measures. Second, we identified outcome measures that are obtainable from administrative databases.

Results: A survey of 143 facility directors in VA health care system showed two clearly prevalent service areas in which SLs had been introduced: primary care SLs (PCSLs) implemented in 110 facilities, and mental health SLs (MHSLs) implemented in 108 facilities. For PCSLs, nine facility-level outcomes were obtainable and could be affected by organizational structure, and thus could be used to assess PCSL performance: percentage enrollment in primary care, acute bed day rate, proportion of users with at least one hospitalization, discharge rate, multi-stay rate, ambulatory care sensitive hospitalization rate, specialty visit rate, urgent care visit rate, urgent care visits per total visits. Likewise, six facility-level outcomes were judged appropriate for use in assessing the performance of the MHSLs: acute psychiatric bed day rate, total acute bed day rate, proportion of psychiatric admissions without 30 day previous outpatient clinic visit, 30 day readmission rate following index psychiatric hospitalization, proportion of psychiatric hospitalizations with an outpatient visit within 30 days, and urgent care visit rate for psychiatric patients.

Conclusions: Analysis of management or organizational structure performance should be guided by scientific methods. An investigator should pose a hypothesis, and develop valid and reliable measures that allow a testing of the hypothesis. Evaluation of organizational or management strategies should not solely be descriptive in an anthropologic sense - it should be analytic.

Impact: This study shows the feasibility of applying rigorous analytic methods to the design and conduct of research to evaluate changes in organizational structure and management practice. Adopting the above approach may have a dramatic effect on the quality of such evaluations.

HSR&D Funded: MRR 97-010.

*186. Disability after Fracture in Postmenopausal Osteoporotic Women: the Fracture Intervention Trial

Howard A. Fink, MD, MPH, Geriatric Research Education and Clinical Center, Section of General Internal Medicine, Minneapolis, MN, KE Ensrud, MD, MPH, Section of General Internal Medicine, VA Medical Center, Minneapolis Center for Chronic Disease Outcomes Research, VA Medical Center, Minneapolis, RM Pieper, MPH, Division of Epidemiology, School of Public Health, University of Minnesota, Minneapolis, MC Nevitt, PhD, Department of Epidemiology and Biostatistics, University of California, San Francisco, PJ Schreiner, PhD, Division of Epidemiology, School of Public Health, University of Minnesota, Minneapolis, DB Nelson, PhD/DC Bauer, MD/SR Cummings, MD, Center for Chronic Disease Outcomes Research, VA Medical Center, Minneapolis/Department of Epidemiology and Biostatistics, University of California, San Francisco/Department of Epidemiology and Biostatistics, University of California, San Francisco

Objectives: Osteoporotic fractures are a frequent cause of morbidity in the elderly. In the current study, we prospectively measured disability following fracture at various individual skeletal sites, including non-hip, non-spine locations, in a population of postmenopausal women with low bone mineral density (BMD).

Methods: Participants were women in the Fracture Intervention Trial (FIT), a randomized, placebo-controlled trial designed to determine whether treatment with alendronate reduced fracture incidence. The FIT population comprised 6459 women aged 55 to 81 years, with a femoral neck BMD of 0.68 g/cm2 or less (QDR-2000 Hologic Inc). During follow-up (mean 3.8 years), women reported all incident clinical fractures. Fractures were radiographically confirmed; those secondary to malignancy or excessive trauma were excluded. For assessment of disability, subjects with a clinical fracture were asked if, because of the fracture: (1) "did you cut down on the things that you usually do, such as going to work or working around the house"; (2) "how many days did you cut down on the things that you usually do; and (3) "how many days did you stay in bed more than half the day (include days while an overnight patient in a hospital)." If a woman reported a hospitalization resulting from the fracture, hospitalization length was confirmed by review of medical records. Days hospitalized or confined to bed are termed ‘bed days’ and the sum of other days during which participants cut down on usual activities plus the number of ‘bed days’ are termed ‘limited activity days.’ Mean ‘bed days’ and ‘limited activity days’ for the 8 most common confirmed fracture sites were estimated using linear models.

Results: During follow-up, 909 women experienced 1124 confirmed incident fractures. The most common fracture sites were wrist (n=235), spine (n=143; thoracic spine=78, lumbar spine=65), foot (n=115), humerus (n=85), hip (n=78), toes (n=74) and ankle (n=73). Of the 909 confirmed first fractures, 822 fractures met eligibility criteria for disability analysis. Women with lumbar spine fractures experienced 20.9 mean ‘bed days’ (95%CI=15.8,26.0) due to the fracture. Those with hip fractures experienced 19.9 mean ‘bed days’ (95%CI=15.9,23.9), while thoracic spine fractures were associated with 9.9 mean ‘bed days’ (95%CI=5.7,14.2) and ankle fractures with 8.8 mean ‘bed days’ (95%CI=5.1,12.5). The number of mean ‘bed days’ at other sites was not statistically different from zero. With regard to limitation of usual activities, women who experienced a lumbar spine fracture had a mean of 137.3 ‘limited activity days’ after fracture (95%CI=111.8,162.8). Hip fractures resulted in 98.1 mean ‘limited activity days’ (95%CI=77.9,118.4), followed by fractures of the ankle (73.5; 95%CI=55.0,92.0), humerus (73.0; 56.2,89.9), thoracic spine (64.7; 43.4,86.0) and wrist (52.7; 42.2,63.2).

Conclusions: In a population of community-dwelling postmenopausal women with low BMD, we found that osteoporotic fractures at numerous skeletal sites, including non-hip, non-spine locations, commonly result in prolonged and sometimes severe disability.

Impact: It is important to identify potential interventions which not only may prevent osteoporotic fractures, but may have distinct effects on fracture-related disability in individuals who experience fractures.

*187. Do Clinicians and Non-clinicians Agree on Proposals to Reduce the Costs of Care?

Lisa M. Winterbottom, MD, Portland VAMC, Portland, OR, GL Noel, MD, Portland VAMC, JA Jackson, RN, Portland VAMC, C Borrego, MD, Portland VAMC, LS Snodgrass, Portland VAMC, DC Smith, RRA, Portland VAMC

Objectives: 1) To determine to what extent three employee groups agree with proposed efficiency and cost reduction measures 2) To assess whether employee groups are more or less likely to agree with cost reductions that may result in staff reductions.

Methods: Each service of the medical center was presented with DSS cost information for procedures, clinics, hospitalizations, individual diagnoses, pharmacy expenses, and diagnostic imaging. The clinicians in each section were asked to prioritize 25% of their total costs as least essential to maintain the overall health of the enrolled patients. Two hundred and ninety six items for potential cost reductions were proposed in the following categories: drug elimination, drug restriction, drug substitution, service elimination, service restriction, catchment restriction, improved resource efficiency, new services, lab restriction, imaging restriction, revenue generation, and additional unspecified cuts. The savings from each proposal were projected and all 2000 employees of the medical center were invited to provide feedback on the proposed items. The feedback was used by the services in preparing their final rank lists of proposed cuts.

Results: Five hundred and sixty five employees (n=28%) participated in the poster feedback session and indicated their agreement or disagreement with the items; they also gave comments on as many items as they desired. Nineteen percent of the responders were clinicians (MDs, PhDs, NPs and PAs), 55% were employees who provide direct patient care (RN, secretaries, therapists and other staff in direct contact with patients) and 26% were non-clinical support staff and administrators. Overall members of each of the employee groups agreed with 88% of the proposed cost reduction measures. The proportion of measures with which a majority of employees disagreed were as follows: drug elimination (1/24), drug restriction (2/39), drug substitution (1/8), service elimination (8/39), service restriction (10/55), catchment restriction (1/20), improving resource efficiency (4/46), new service (0/16), lab restriction (0/12), imaging restriction (3/26), revenue generating (0/4), additional unspecified cuts (7/7).

Conclusions: The responders agreed with most of the proposed measures. However, non-clinician employees who work directly with patients were less likely to agree as stongly with all types of reductions as clinicians. Employees disagreed with service reduction items and unspecified cuts more than non-service reduction items (25% vs. 6 %, P<. 0001). Items with which most employees disagreed were primarily for palliation of chronic and terminal conditions and changes in screening for women’s health issues. Service reduction items that could result in position reductions did not receive more negative feedback than those that would not. Items that were nonspecific or not recommended were not supported. Employees’ responses were consistent for similar items. Using DSS information, clinicians can identify inefficient practices and non-obligatory services they are willing to change in order to maintain adequate funding for essential care and these items can achieve employee consensus.

Impact: Clinicians and hospital employees can collaborate in a process to reevaluate the costs of a medical center Clinician effort in item construction is validated by consistent and discriminating employee responses.

*188. Do Doctors and Patients Agree on the Benefits of Coronary Revascularization?

Jeffrey Whittle, VA Pittsburgh Healthcare System/University of Pittsburgh, Pittsburgh, PA, J Conigliaro, VA Pittsburgh Healthcare System/University of Pittsburgh, CB Good, VA Pittsburgh Healthcare System/University of Pittsburgh, KC Goldberg, Durham VAMC/Duke University Medical Center, P Sankar, University of Pennsylvania, M Skanderson, VA Pittsburgh Healthcare System/University of Pittsburgh

Objectives: Little is understood about how patients receive information about the potential benefits and risks of medical procedures. We compared physician and patient assessments of the benefits of coronary revascularization, as well as what the patients reported they had been told.

Methods: We prospectively enrolled patients who were offered or received percutaneous angioplasty (PTCA), coronary artery bypass grafting (CABG) or minimally invasive coronary artery bypass (midCAB)) at a VAMC or its university affiliate. We attempted to survey patients prior to the procedure when possible, but interviewed them post procedure if necessary. Patients were asked if they thought the procedure was likely to improve survival, and if "the doctor" had told them that it would. Similar questions were asked regarding symptom control in general, pain control in particular and functional status. A cardiologist or cardiology fellow involved in decision making about that revascularization indicated whether a procedure would be expected to increase survival or decrease symptoms.

Results: Patient surveys were available for 348 patient – physician pairs. 230 physicians provided useable surveys. Patients were likely to report that the physician had told them that the procedure would improve survival (47.3% of patients), symptoms (48.6%), pain (53.5%), or functional status (60.4%). Even more commonly, they believed the procedures improved these outcomes (96%, 93.3%, 97.7% and 93.4%, respectively). These responses were modestly correlated – for example, patients were slightly less likely to believe there would be a survival benefit if they had not heard their physician say this than if they had (93.3% vs 98.5%, p = 0.04). For improved symptoms and functional status, the difference was greater, though the vast majority of patients who had not heard their physician say there would be a benefit still expected one (85% vs 99%, p < 0.001). Patients were similarly likely to expect a survival benefit for all three procedures. Physicians listed improved survival as a benefit for 12% of PTCA patients, 83% of CABG patients and 67% of midCAB patients. These responses were not correlated with what the patient thought the physician had said. For example, there was no significant difference in the proportion of physicians who said there was a survival benefit among patients who did and did not expect such a benefit for PTCA (12.5% vs 10.6%), CABG (79.7% vs 89.1%) or midCAB (66.7% vs 66.7%), all p > 0.2). Physicians more frequently predicted a symptomatic benefit (PTCA 90%, CABG 69.6%, midCAB 89%), but there was no agreement between physician response and patient report of what the doctor had said.

Conclusions: Most persons considering coronary revascularization believe it will increase survival and improve symptoms, though many acknowledge their doctor did not tell them this. Moreover, even when the patients did recall the doctor telling them there was a survival benefit, this did not correspond to what doctor reported.

Impact: More accurate communication regarding benefits of medical therapies may require assessment of baseline patient understanding. We conjecture that this may help patients to more clearly understand the risks and benefits of these procedures.

*189. Do For-Profit Hospice Patients Have Longer, Less Complex Stays?

Karl Arthur Lorenz, West Los Angeles VAMC, Los Angeles, CA, KE Rosenfeld, West Los Angeles VAMC, SL Ettner, UCLA, DM Carlisle, UCLA, LJ Blackhall, Assisted Home Care and Hospice, SM Asch, West Los Angeles VAMC

Objectives: For profit hospices have become increasingly common since the initiation of the Medicare hospice benefit. We hypothesized that for-profit hospices would seek to maximize revenue by achieving longer average lengths of stay (LOS), and that to achieve longer LOS, for-profit hospices derive a greater percentage of referrals from long term care and serve a greater percentage of patients with noncancer diagnoses. We further hypothesized that to minimize costs, for-profit hospices utilize fewer highly skilled visits and offer less availability of costly palliative interventions chemotherapy and radiotherapy

Methods: Our preliminary analysis evaluated responses from the California Office of State Health Planning and Development Home Health Agency and Hospice survey. The survey includes aggregate information on organizational characteristics, service availability, staff visits, LOS, and patient demographic information for all 176 hospices licensed in California in 1997. Unadjusted and regression-adjusted differences by profit status were calculated for LOS, total nursing visits, percentage of skilled nursing visits, and availability of chemotherapy and radiotherapy. Reduced models controlled for profit status, size, urbanicity, type of hospice (inpatient/outpatient), and proportions of patients that were female, non-white, and noncancer diagnoses. Complete models additionally controlled for LOS.

Results: For-profit hospices were more likely to be urban (p=0.002) established after 1983 (p=0.001) and served a higher percentage of patients with noncancer diagnoses (46% vs 28%, p=0.0001) referred from long term care facilities (19% vs 4%, p=0.0000). For-profit hospices served a higher percentage of female patients (57% vs 51%, p=0.005), nonwhite patients (37% vs 27%, p=0.0224), patients over the age of 80 (69% vs 63%, p=0.0014), and patients with MediCal insurance (11% vs 7%, p=0.0191) than not-for-profit hospices. LOS averaged 50 days among for-profit and 45 days among not-for-profit hospices but did not differ significantly (p=0.1031). For-profit hospices reported more visits per patient (43 vs. 28, p=0.0195) although for-profits reported a lower percentage of skilled visits as a percentage of total nursing visits (55% vs 62%, p=0.0007). There was no difference between chemotherapy and radiotherapy availability between for-profit and not-for-profit hospices. Multivariate models of LOS demonstrated an association only with percentage of patients with noncancer diagnoses (beta = 32, p=0.0005). Log total nursing visits were associated with the log size of the hospice (beta = 0.87, p=0.0001), LOS (beta = 0.01, p=0.0001), and percentage of patients over 60 years old (beta = 1.6, p=0.0001). Multivariate analysis of the percentage of skilled nursing visits demonstrated association with only the percentage of patients greater than 60 years of age (beta= -0.43, p=0.0001).

Conclusions: For-profit hospices provide services to a greater percentage of elderly long term care patients with noncancer diagnoses. Differences in the complexity of services delivered are explained by these patient characteristics.

Impact: Differences in the intensity and mix of nursing services delivered by for-profit and not-for-profit hospices are related to differences in patient populations. Further research is needed to understand how differences in the intensity and mix of nursing services relate to clinically important outcomes at the end of life.

*190. Dysphagia Intervention: A Model Qualtiy Improvement Program

Nan D. Musson, VAMC, GRECC, GIFT, Miami, FL, MA Silverman, VAMC, GRECC, GIFT, L Granville, VAMC, GRECC, VA Abbott, VAMC

Objectives: The major goal of intervention for individuals with dysphagia intervention is to assure safe, efficient nutrition and hydration. An interdisciplinary dysphagia team targeted four major areas for quality improvement with an overall goal to facilitate preventative approaches to the care of adults with feeding and swallowing disorders. A 12 year history of continuous quality improvement projects facilitated by the dysphagia team will be reviewed and may be implemented in a variety health care settings. A retrospective review of the projects demonstrated a transition from quality improvement to clinical research.

Methods: Fourteen (14) different continuous quality improvement projects were initiated during a 12 year interval. The interdisciplinary team included representatives from speech pathology, dietary, dental, nursing, geriatrics, gastroenterology, otolaryngology, surgery, radiology, voluntary service and administration. Non-professional consumers were invited to participate on several of the projects and experts were included as needed to better define the problems. The team utilized the scientific approach to systematically solve problems and improve processes. Data was collected using a variety of quality improvement tools, for example: checksheets, flowcharts, control charts and frequency plots. Theories were periodically tested to determine the most efficient process to assure adequate nutrition and hydration for patients with feeding and swallowing disorders. As problems were resolved, the teams recommended for new procedures or programs for the medical center.

Results: Results: A retrospective review of the data revealed that the mean age of the patients identified with feeding and swallowing disorders was 75.5 years and the primary etiology for the feeding or swallowing disorder was commonly neurogenic in origin (80%). Of the patients monitored, 72% required feeding assistance in order to maintain oral intake and 62% presented with oral preparatory stage dyspahgia. Cognitive decline was identified in 68%. Food modifications were required for 59%, liquid modification for 49%, and 29% required enteral tube feeding. Patients fed meals in their room without assistance had an average weight loss of 1.38 lbs. After the initiation of the quality improvement programs, there was an average weight gain of 2.06 lbs per quarter. Initially 21% of the long-term care population had enteral tube feeding; the average has now decreased to 7% of the patients. A tube feeding support group and counseling program were initiated for patients and families. In addition, a cost-savings was identified when the projects identified over utilization of liquid nutritional supplements and strategies to bulk order afternoon snacks.

Conclusions: A continuum of continuous quality improvement data demonstrated improved nutrition and hydration status of patients with dysphagia, along with weight gain of nursing home residents, less food waste and less need for nutritional supplements when an appropriate diet is supplied and consumed. Educational programs have assisted in early identification and intervention of high-risk elderly individuals, and initiated a preventative approach to the management of patients with feeding and swallowing disorders. Outcome data have supported the need for new procedures and programs within the medical center and a database has resulted in several descriptive studies that served as a stepping stone for future prospective research projects.

Impact: Continuous quality improvement processes demonstrated improved nutrition, hydration, and efficient utilization of resources and staff for patients with feeding and swallowing disorders.

*191. Education-enhanced Rehabilitation for Homeless Veterans in Atlanta: Early Results

Samuel M. Aguayo, Atlanta VAMC, Emory and Morehouse Schools of Medicine, Decatur, GA, KAS Cannella, Atlanta VAMC and Emory University, MB Holt, Atlanta VAMC, LH Wright, Atlanta VAMC, J Santagata, Atlanta VAMC, J Reynolds; FM Cordi, Atlanta VAMC; Atlanta VAMC and Emory University

Objectives: Military veterans comprise one-third of the adult homeless population and forty percent of homeless males of the estimated 250,000 homeless Americans. Homeless veterans number an estimated 6,000-8,000 in Atlanta alone on any given night. Major causes of homelessness among veterans include mental illness and substance abuse. Our VA homeless program for veterans provides 1) assessment and stabilization, and 2) a supervised, structured rehabilitation program of treatment, housing, and work. For the past three years, veterans who successfully complete this rehabilitation program have been offered the opportunity to undertake an additional education-enhanced rehabilitation program. The enhanced program consists of a supervised program of formal education in conjunction with a continuation of the structured treatment, housing, and work rehabilitation program. Our initial objective was to evaluate the effect of this education-enhanced rehabilitation program on the following two outcomes: maintenance of gainful employment and freedom from relapse.

Methods: We made no changes in our regular assessment, stabilization, and rehabilitation program for homeless veterans. But following completion of this program we offered veterans the choice of 1) receiving usual outpatient care or 2) entering the education-enhanced rehabilitation program. The formal education consisted of career optimizing academic/vocational educational tracks that include minimal academic core curriculum. Successful rehabilitation was defined as the maintenance of gainful employment and freedom from relapse (psychiatric hospitalizations and/or substance abuse) for at least one year.

Results: About 25 percent of homeless veterans completing the usual rehabilitation chose to apply for the education-enhanced rehabilitation program. To date, 74 percent (23 of 31) of our homeless veterans who completed the education-enhanced rehabilitation program have been able to remain gainfully employed and have remained free from relapse. In contrast, our most optimistic estimate is that only 25 to 41 percent (50 to 82 of approximately 200) of homeless veterans who underwent only the usual rehabilitation program have been able to achieve these goals. Attrition tends to be about 50 percent during initial assessment and stabilization, about 33 percent during the structured treatment, housing, and work rehabilitation program, and 5 percent during the education-enhanced rehabilitation program. Limitations of our data include the lack of an appropriate comparison group and possible motivational bias evidenced by self-selection to the education-enhanced rehabilitation program.

Conclusions: Education-enhanced rehabilitation provides a greater likelihood of success for the comparatively small proportion of homeless veterans who choose this option. This group of homeless veterans appears different from those who either do not complete the usual rehabilitation program or who complete it but do not choose to enter the education-enhanced rehabilitation program.

Impact: Future studies will examine the effectiveness of this intervention under more structured conditions and explore specific differences between those who choose education and those who do not.

*192. Effects of Interactive Demonstration of Exercise Among Arthritics

Jatinder Kaur, MCHSPR Hines VA Hospital, Hines, IL, IA Elicier, Hines VA Hospital, E Budiman-Mak, MCHSPR Hines VA Hospital, K Kolbusz, Hines VA Hospital

Objectives: To demonstrate the practical and easy nature of exercise, increase patient awareness of the benefits of exercise while providing a practical way for patients to learn about methods to alleviate pain and discomfort caused by arthritis. Thus, improving their quality of life through life-style modifications.

Methods: In a period of two years (1997-1999), 73 male veteran patients with a variety of arthritis conditions were asked to voluntarily attend a 20 minute interactive exercise demonstration conducted by a licensed physical therapist on Mondays in the clinic waiting area. Subjects, age range 40-80 years, were recruited from the arthritis clinic waiting room while they were waiting to be seen by a physician. The importance of movement and flexibility were explained to the participants prior to the demonstration. The exercise sessions consisted of thera-band resistive strengthening exercises which involved the use of one-meter rubber bands of different colors and resistances. Subjects were educated on the various strengths and exercise techniques while learning various stretching exercises. Furthermore, range-of-motion techniques were illustrated in an easy relaxing manner using a ROM Dance Tai Chi video. During the video, the demonstrator performed each step and encouraged subjects to follow along and practice each technique. Finally, general conditioning and energy conservation techniques. Simple posture adjustments and joint protection techniques were discussed and demonstrated. The three demonstration techniques were routinely rotated. A staff rheumatologist, a physical therapist, and a patient education coordinator were available to answer patients' questions and concerns. Participants then completed a survey about the effectiveness of the demonstration and a week later to answer a phone interview regarding the frequency and benefits of the demonstration.

Results: The survey and phone interview was based on yes or no answers. Responses to the survey, n of 73, showed that patients reported the following: 97.3 percent learned a lot of new information, 98.6 percent felt the information would be useful in helping them perform daily living activities, 94.5 percent would definitely use the exercises, and 95.9 percent would recommend this presentation to a friend with arthritis. Phone interviews, response rate of 59 percent (43/73), showed that 79.1 percent of the respondants were still using the exercises one week later based on the leaflet provided and 62.8 percent reported feeling better as a result of exercising. However, only 46.5 percent reported sharing the information from the exercise demonstration, and 51.2 percent stated that they would return for another session.

Conclusions: These demonstrations allowed participants to experience a new form of convenient and practical exercise in an interactive manner. This increased patients’ self-confidence and knowledge to exercise on their own. There is evidence that there is dissemination of knowledge about exercise and arthritis to the participants and their peers.

Impact: The interactive, and user-friendly exercise techniques make the low budget demonstrations an effective and worthwhile approach for educating subjects about arthritis and pain management. The informal nature of the demonstrations made them enjoyable, thus increasing motivation for elderly arthritis patients to exercise.

*193. Epidemiology of Irritable Bowel Syndrome in United States Veterans

Mitchel Hoffman, Bay Pines VA Medical Center, Bay Pines, Florida, TW Butler, Bay Pines VA Medical Center, Bay Pines, FL

Objectives: To describe the epidemiology of irritable bowel syndrome (IBS) in a national cohort of United States hospitalized veterans. Approximately 3% of the United States population are afflicted with IBS. Although IBS is a common disorder there is limited information about its epidemiology.

Methods: The Department of Veterans Affairs maintains a computerized national database for patients hospitalized in VA Hospitals. We accessed the Patient Treatment Files (PTF) from fiscal years 1992-1994 to identify all veterans discharged with a primary or any of the nine secondary diagnoses of IBS. A data file was created for each veteran from the PTF database containing the veteran’s age, sex, race, marital status, state of residence, date of hospital discharge, principal diagnoses and up to nine secondary diagnoses. The occurrence of IBS was calculated and expressed as a proportionate morbidity ratio standardized for age, sex, race, marital status and geographic locality.

Results: 1,136,200 veterans accounted for 2,734,075 discharges from VA medical centers. We identified 820 individuals discharged with a primary diagnosis of IBS and 3,308 veterans with either a primary or secondary diagnosis. Female gender was strongly associated with IBS, a rate 4.2 times greater than for males (p<0.001). The proportional morbidity ratio for whites were 3.0-fold higher than for blacks (p<0.001). Rates for IBS were greater for those married (p<0.05) and lower rates were observed among those separated, never married or widowed. There was considerable geographic variation with a 2.3 gradient in risk for IBS between Maine with the highest and Hawaii with the lowest rates. Rates were highest in the Midwest, followed by the Northeast, Southeast and West. Disease rates were seasonal with highest rates in the spring and fall.

Conclusions: We present the results of the first national study analyzing the occurrence of IBS by age, race, sex, marital status, seasonal variation and geographic distribution. The disease is four times more common in women than in men, has ethnic variation with rates for whites three-fold greater than for blacks, and is more frequent among those married. Additionally, we noted seasonal and geographic variation in disease occurrence.

Impact: The epidemiological distribution of disease suggests exogenous or environmental factors may be important determinants for the pathogenesis of IBS. Further studies are needed to ascertain the role of smoking, alcohol, medication, diet or endogenous hormone levels in IBS causation.

*194. Functional Status Prior to Hospitalization in Acutely Ill Elders: Validity and Clinical Importance of Retrospective Reports

Kenneth Covinsky, MD, MPH, SF VAMC, San Francisco, CA, RM Palmer, Cleveland Clinic, SR Counsell, Indiana University, ZM Pine, SF VAMC, LC Walter, SF VAMC, MM Chren, SF VAMC

Objectives: Patients' retrospective reports of their functional status prior to hospital admission are often used in longitudinal studies and by clinicians caring for hospitalized patients. However, the validity and clinical importance of these reports has not been established. Our aim was to examine the validity of retrospective reports by testing hypotheses about the relationships these measures would have with other clinical measures if they were valid.

Methods: This was a prospective cohort study of 1953 older patients (mean age 81, 64% women) hospitalized on the general medical service at two hospitals. Shortly after hospital admission, patients reported whether or not they were independent in each of 5 ADLs on admission, and at baseline two weeks prior to admission. Outcome measures included reported independence in each ADL 3 months after the hospitalization, and survival to one year.

Results: Patients' retrospective reports of their ADL function two weeks prior to admission had a clinically plausible relationship with admission ADL function, in that patients independent in an ADL on admission rarely reported they were dependent in that ADL 2 weeks prior to admission (range 2-6%). Patients' retrospective reports of pre-hospitalization ADL function demonstrated strong evidence of predictive validity. For example, among patients dependent in bathing on admission, patients who were reported as independent two weeks prior to admission were much more likely than those reported as dependent two weeks prior to admission to be independent 3 months after hospitalization (68% vs. 20%, p<.001). Similarly, among patients dependent in bathing on hospital admission, survival one year after hospitalization was much higher in patients who were independent in bathing two weeks prior to admission than patients who were dependent two weeks prior to admission (76% vs 59%, P<.001). Results were similar for each of the other four ADLs. In a logistic regression model controlling for the number of ADLs reported as dependent on admission, the number of ADLs reported as dependent two weeks before admission was significantly associated with one year mortality (OR=1.39 per dependent ADL, 95% CI= 1.26-1.54).

Conclusions: Hospitalized patients' assessments of their ability to perform ADLs prior to their hospitalization have evidence of face and predictive validity. These measures are strong predictors of important health outcomes such as functioning and survival. In particular, among patients dependent in ADL function on hospital admission, these results highlight the prognostic importance of inquiring about the patient's functional status prior to the onset of the acute illness.

Impact: These results demonstrate the importance of understanding prehospitalization functional status in Veterans hospitalized with medical illnesses. Among Veterans with similar admission functional status, the prognosis for independent functioning and survival is much greater among Veterans who were independent in ADL function prior to the acute illness than in Veterans who were dependent.



*195. Gender Differences in HIV, HBV, HCV and Associated Risks in Severely Mentally Ill Persons

Marian I. Butterfield MD, MPH, Durham VAMC, HSR&D, Duke University, Dept. of Psychiatry, Durham, NC, HB Bosworth, Durham VAMC, HSR&D, Duke University, Depts. of Psychiatry and Medicine, KG Meador, Durham VAMC, HSR&D, Duke University, Dept. of Psychiatry, KM Stechuchak, Durham VAMC, HSR&D, MS Swartz , Duke University, Dept. of Psychiatry, LA Bastian, Durham VAMC, HSR&D, Duke University, Dept. of Medicine

Objectives: There is a pressing need to study HIV, Hepatitis B (HBV) and Hepatitis C (HCV) prevalence and risks in persons, including veterans, with severe mental illness (SMI).

Methods: From June 1997 - December 1998, SMI persons from 5-sites (n = 969, 28.7% veterans) were enrolled in the study. HIV, HBV, and HCV risk behaviors and serostatus were assessed. SMI diagnoses include schizophrenia, schizoaffective disorder, bipolar disorder, depression, and posttraumatic stress disorder (PTSD).

Results: Observed prevalence rates of HIV (2.7%), HBV (18.8.4%), and HCV (16.2%) among SMI persons were much higher than the overall estimated population rates for these infections. Prevalence rates among SMI men vs SMI women were similar for HIV (3.0% vs 2.1%) and HBV (19.5% vs 17.1%). However, HCV prevalence rates were higher among SMI men (19.3%) than SMI women (9.8%). Gender differences in risk behaviors were observed. SMI women were more likely to have sex for drugs than SMI men were (14.8% vs 11.4%). SMI men were more likely than SMI women to have used crack cocaine (45.0% vs 27.5%) and/or injection drugs (20.9% vs 11.2%). Adjusting for these three risk behaviors, SMI men who used needles, crack cocaine, or had sex for drugs had higher rates of HCV infection than SMI women who had these risk behaviors. In contrast, for HBV there were not gender differences in using crack cocaine and sex for drugs. However, SMI men who used injection drugs had higher rates of HBV infection than SMI women.

Conclusions: In this study, SMI men and SMI women have high prevalence rates of HIV, HBV, and HCV. SMI men were more likely than SMI women to be HCV infected. HIV and HBV prevalence rates were similar among SMI men and SMI women. Routes and risks transmission of these infections may differ in SMI men and SMI women.

Impact: Persons with SMI, including veterans, are at considerable risk for numerous comorbid illnesses including HIV, HBV, and HCV. SMI veterans are chronic users of VA health services and are likely to be at risk or infected with these infections. VAMCs should consider targeted screening for these infections in SMI persons. Gender differences in risk behaviors and patterns of viral transmission of these infections may exist. SMI veterans are understudied, further research inclusive of SMI women veterans, is warranted.

HSR&D Funded: CSP706-D.





*196. Gender Differences in Satisfaction with Participation in Decision Making

Mary A. Jacobs, PhD, RN, Virginia Commonwealth University, Richmond, VA, Richmond, Virginia, MC Corley, Virginia Commonwealth University, Richamond, VA, MA Jacobs, McGuire VA Medical Center, Richmond, VA, P Minick, Virginia Commonwealth University, Richmond, VA

Objectives: The aims of this study were to investigate the patient’s desire for and satisfaction with participation in decision making and to assess the influence of race, gender, and age on satisfaction with shared decision making.

Methods: One hundred sixty-eight inpatients at two facilities were surveyed using the 24 item Patient Decision Making Instrument (PDMI). The PDMI, developed by the first author, was designed to determine the patient’s desire for and satisfaction with participation in treatment decisions and for information about treatment options. Validity was established using the cognitive assessment with 23 patients in a preliminary study. A composite "Patient Satisfaction with Participation" score was computed using three items about satisfaction with participation

Results: The sample consisted of patients with an average age of 56, 55 percent were white, and 73 percent were male. Patients’ self-reported level of illness was Very Sick (15 percent), Sick (29 percent), and Getting Better (56 percent). Our findings indicate that the majority of patients (96 percent) wanted to share in the decision making process with their physician. Fewer patients, but still a substantial number (83 percent), wanted to involve their families in the treatment choices. Only 35 percent were satisfied with their participation in the decision making process, 46 percent were somewhat satisfied, and 14 percent not satisfied. We did find gender differences in satisfaction with decision making. Female patients were more satisfied with their participation in treatment decisions than males (t=-3.25, p .001). Our findings suggest that the majority of patients wanted information about their treatment, treatment options, and consequences of that treatment. No other demographic variables were significant.

Conclusions: Most patients desired to share in the decision making process. Nearly all wanted information about their treatment. It is interesting to note that a gender gap was found in satisfaction with shared decision making. Men were less satisfied than females with their role in that decision process.

Impact: This study supports other findings that patients want to share in the decision making process. The gender differences we found have particular relevance for the VHA because of the predominately male population. Since shared decision making has been found to improve health outcomes and patient satisfaction, these findings have further implications for 1) patient education in terms of delivering information to our population, 2) self-management of chronic diseases, 3) satisfying the patient’s desire for participation by involving them in their care, and 4) educating providers about incorporating the shared decision making model in their practice and developing a more collaborative style. Future research is warranted to investigate gender differences of both patients and providers in shared decision making.



*197. Herbal Medicine in the United States: A Systematic Review of the Quantity and Quality of Information Available to Health Care Providers

Stephen Bent, MD, San Francisco VAMC and the University of California, San Francisco, San Francisco, CA, B Jacobs, San Francisco VAMC and the University of California, San Francisco

Objectives: Although approximately one-third of Americans are using herbal remedies, physicians rarely discuss this topic with their patients, which may lead to judgement errors or drug-herb interactions. We hypothesized that this lack of discussion may be due to limited information about herbs available to clinicians. This study examines the quantity and quality of information about herbs available to health care providers in the United States.

Methods: We performed a MEDLINE search (1966 - March 1998) for English-language, randomized controlled trials and systematic reviews for each of the ten most commonly used herbal remedies in this country. Foreign language articles were excluded since it was felt that the majority of health care providers in this country do not read articles not published in English. All identified randomized placebo-controlled trials and systematic reviews were included. Two reviewers independently rated the quality of each of the randomized controlled trials using an established scale. We prospectively hypothesized that in order for a physician to be able to discuss the efficacy of an herbal remedy for a specific indication, he or she must be able to locate either 2 randomized controlled trials of at least moderate quality (score of 3 on a 0-5 scale), and with a result in the same direction, or one systematic review. These criteria were applied to the studies identified for each of the herbs.

Results: Our search strategy identified randomized controlled trials for six of the ten most commonly used herbs, and systematic reviews for only one herb (garlic). The number of indications studied for these six herbs ranged from one for Astragalus, to twelve for Ginkgo biloba. The number of trials for each indication studied ranged from one to eleven, and the average quality of trials by indication varied from 1 to 4.5 (0-5 scale). As prospectively defined, there was adequate evidence for a physician to be able to discuss efficacy for only three of the ten most commonly used herbs.

Conclusions: Despite the widespread use of herbal products in this country, there is limited evidence about the efficacy of these agents available to US providers. This limited amount of information may discourage physicians from discussing this topic with their patients.

Impact: A lack of discussion about herbs may impair a provider's ability to adequately monitor treatment effects and side effects to both standard therapies and herbal products. Publication of systematic reviews of the safety and efficacy of herbs (based on studies in both English and non-English language) may provide physicians with sufficient evidence to promote informed discussion of this common therapy.

*198. How do we connect? Congruence between Perceptions of Coordination Mechanisms in Cardiology

Anne E. Sales, VA Puget Sound Health Care System, Seattle, WA, NR Every, VA Puget Sound Health Care System, A Keane, VA Puget Sound Health Care System, D Caldwell, VA Puget Sound Health Care System, L Parsons, VA Puget Sound Health Care System, Seattle, WA

Objectives: We examine congruence between chiefs of cardiology, staff cardiologists, and primary care providers (PCPs) in 31 VAMCs in five VISNs in their perceptions of the referral process between cardiology referral services and primary care services. One of the persistent concerns expressed by both PCPs and specialists in the VA is difficulty coordinating referrals. This report is part of a larger study examining the relationship between clinical integration and patient outcomes.

Methods: We conducted mailed surveys of all PCPs and cardiologists in 31 VAMCs in five VISNs, Of the 31 VAMCs, 10 are tertiary, providing cardiology referral services to the other 21 VAMCs as well as to primary care in their own VAMC. We asked each group about primary methods of communicating between primary care and cardiology about patient care and referrals.

Results: Response rates to the mailed surveys were 57% for PCPs and 58% for cardiologists. Roughly 25% of each group stated that they used email contact, and approximately 25% each stated that they used a dedicated transfer coordinator for primary contact. The correlation between PCP and cardiologist responses for these two items were .50 (p=.004) and -.21 (p=.26) respectively. There was even less agreement for all other items (telephoning the cardiology office, the attending, or a paging operator).

Conclusions: We found relatively little agreement between PCPs and cardiologists on the mechanics of how a referral is made. This supports the general perception that there are serious problems with the process of referring patients from primary care to one of the largest specialty services offered by the VA. While there is more to explore in determining where the problems in the referral process are, the lack of agreement on the relatively straightforward issue of how a referral is made is particularly troubling. It suggests that better information dissemination is needed between cardiology services and primary care.

Impact: This study will be of interest to clinician managers of cardiology services in particular and specialty services generally, as well as to primary care providers. In addition, it suggests the need for further research into some of the basic mechanics of processes such as referral services.

HSR&D Funded: ACC97-0792.

*199. How Patients Perceive Physician Communication Regarding Cardiac Testing

Tracie C. Collins, Houston Center for Quality of Care and Utilization Studies, Houston, TX., L Petersen, Houston Center for Quality of Care and Utilization Studies, N Kressin, Bedford VA Medical Center, J Clark, Bedford VA Medical Center

Objectives: Recent studies documenting racial variation in the use of cardiac procedures highlight the need to understand processes of communication and decision making. We investigated patients' perceptions of their communication with providers regarding cardiac testing.

Methods: We convened four focus groups with thirteen patients who had undergone cardiac stress testing with positive results, stratified by race (white vs. black) and whether or not they had undergone coronary angioplasty or bypass. Verbatim transcripts were analyzed qualitatively by a team of behavioral scientists and general internists in order to identify significant dimensions of communication and patient-provider relationships.

Results: We identified three prinicipal domains of communication regarding cardiac testing: the substance of the information that was communicated, the meaning of that information for the patient, and the quality of the patient's relationship with the physician. In reviewing the substance of the communication, patients revealed ambiguity in the information received. One patient stated, "They said the heart was fine but there was some blockage in the artery." One patient who was advised to have a heart catheterization described the communication as "They just said that they felt like that it needed to be done. I asked questions that morning when I was there. I wasn't satisfied, but I had to get this done. It was real vague." Focusing on the meaning of the information, patients felt that the terms used to describe their heart disease lacked clarity. A patient discussed being told that he should have a catheterization "because there was something going on, he described it then as a leaky heart. I'm not aware just what a leaky heart is or why its called a leaky." While there was no apparent variation by race in patients' perceptions of the information received, black patients repeatedly expressed a preference for building a relationship with physicians before agreeing to an invasive cardiac procedure. Statements made by black patients were "to make a decision on a major operation, even to angioplasty for me, I maybe would have agreed to that if I had known the physician doing the treatment," "They didn't know too much about me and my health status."

Conclusions: We found problematic aspects of the patients' experiences regarding communication of cardiac testing. Our findings suggest that although patients desire clarity from physicians, they are often confused regarding the information received. Both a lack of substance and vagueness of the information received may be linked to feelings of mistrust toward physicians when considering further diagnostic testing. While comments made within the groups were similar between the two races, blacks appeared more concerned with building a relationship with physicians before agreeing to invasive testing.

Impact: This study provided qualitative information regarding patients' perceptions of physician-patient communication and racial differences in the perception of the physician-patient relationship. We are incorporating these data into surveys of patients for a larger prospective study to assess racial variation in the utilization of cardiac procedures.

About VA Research | Research Programs | Publications and Reports | For Researchers | Site Map | VHA Home