Case-mix adjusting performance measures in a veteran population: pharmacy- and diagnosis-based approaches

Promising algorithms to perilous applications: a systematic review of risk stratification tools for predicting healthcare utilisation

OBJECTIVES

Risk stratification tools that predict healthcare utilisation are extensively integrated into primary care systems worldwide, forming a key component of anticipatory care pathways, where high-risk individuals are targeted by preventative interventions. Existing work broadly focuses on comparing model performance in retrospective cohorts with little attention paid to efficacy in reducing morbidity when deployed in different global contexts. We review the evidence supporting the use of such tools in real-world settings, from retrospective dataset performance to pathway evaluation.

METHODS

A systematic search was undertaken to identify studies reporting the development, validation and deployment of models that predict healthcare utilisation in unselected primary care cohorts, comparable to their current real-world application.

RESULTS

Among 3897 articles screened, 51 studies were identified evaluating 28 risk prediction models. Half underwent external validation yet only two were validated internationally. No association between validation context and model discrimination was observed. The majority of real-world evaluation studies reported no change, or indeed significant increases, in healthcare utilisation within targeted groups, with only one-third of reports demonstrating some benefit.

DISCUSSION

While model discrimination appears satisfactorily robust to application context there is little evidence to suggest that accurate identification of high-risk individuals can be reliably translated to improvements in service delivery or morbidity.

CONCLUSIONS

The evidence does not support further integration of care pathways with costly population-level interventions based on risk prediction in unselected primary care cohorts. There is an urgent need to independently appraise the safety, efficacy and cost-effectiveness of risk prediction systems that are already widely deployed within primary care.

An Aggregated Comorbidity Measure Based on History of Filled Drug Prescriptions: Development and Evaluation in Two Separate Cohorts

BACKGROUND

The ability to account for comorbidity when estimating survival in a population diagnosed with cancer could be improved by using a drug comorbidity index based on filled drug prescriptions.

METHODS

We created a drug comorbidity index from age-stratified univariable associations between filled drug prescriptions and time to death in 326,450 control males randomly selected from the general population to men with prostate cancer. We also evaluated the index in 272,214 control females randomly selected from the general population to women with breast cancer.

RESULTS

The new drug comorbidity index predicted survival better than the Charlson Comorbidity Index (CCI) and a previously published prescription index during 11 years of follow-up. The concordance (C)-index for the new index was 0.73 in male and 0.76 in the female population, as compared with a C-index of 0.67 in men and 0.69 in women for the CCI. In men of age 75-84 years with CCI = 0, the median survival time was 7.1 years (95% confidence interval [CI] = 7.0, 7.3) in the highest index quartile. Comparing the highest to the lowest drug comorbidity index quartile resulted in a hazard ratio (HR) of 2.2 among men (95% CI = 2.1, 2.3) and 2.4 among women (95% CI = 2.3, 2.6).

CONCLUSIONS

A new drug comorbidity index based on filled drug prescriptions improved prediction of survival beyond age and the CCI alone. The index will allow a more accurate baseline estimation of expected survival for comparing treatment outcomes and evaluating treatment guidelines in populations of people with cancer.

Examining Health Care Mobility of Transgender Veterans Across the Veterans Health Administration

Purpose: Transgender veterans are overrepresented in the Veterans Health Administration (VHA) compared with in the general population. Utilization of multiple different health care systems, or health care mobility, can affect care coordination and potentially affect outcomes, either positively or negatively. This study examines whether transgender veterans are more or less health care mobile than nontransgender veterans and compares the patterns of geographic mobility in these groups. Methods: Using an established cohort (n = 5,414,109), we identified 2890 transgender veterans from VHA electronic health records from 2000 to 2012. We compared transgender and nontransgender veterans on sociodemographic, clinical, and health care system-level measures and conducted conditional logistic regression models of mobility. Results: Transgender veterans were more likely to be younger, White, homeless, have depressive disorders, post-traumatic stress disorder (PTSD), and hepatitis C. Transgender veterans were more likely to have been health care mobile (9.9%) than nontransgender veterans (5.2%) (unadjusted odds ratio = 2.02, 95% confidence interval = 1.73-2.36). In a multivariable model, transgender status, being separated/divorced, receiving care in less-complex facilities, and diagnoses of depression, PTSD, or hepatitis C were associated with more mobility, whereas older age was associated with less mobility. For the top three health care systems utilized, a larger proportion of transgender veterans visited a second health care system in a different state (56.2%) than nontransgender veterans (37.5%). Conclusions: Transgender veterans were more likely to be health care mobile and more likely to travel out of state for health care services. They were also more likely to have complex chronic health conditions that require multidisciplinary care.

Use of Data-Driven Methods to Predict Long-term Patterns of Health Care Spending for Medicare Patients

Importance

Current approaches to predicting health care costs generally rely on a single composite value of spending and focus on short time horizons. By contrast, examining patients' spending patterns using dynamic measures applied over longer periods may better identify patients with different spending and help target interventions to those with the greatest need.

Objective

To classify patients by their long-term, dynamic health care spending patterns using a data-driven approach and assess the ability to predict spending patterns, particularly using characteristics that are potentially modifiable through intervention.

Design, Setting, and Participants

This cohort study used a retrospective cohort design from a random nationwide sample of Medicare fee-for-service administrative claims data to identify beneficiaries aged 65 years or older with continuous eligibility from 2011 to 2013. Statistical analysis was performed from August 2018 to December 2019.

Main Outcomes and Measures

Group-based trajectory modeling was applied to the claims data to classify the Medicare beneficiaries by their total health care spending patterns over a 2-year period. The ability to predict membership in each trajectory spending group was assessed using generalized boosted regression, a data mining approach to model building and prediction, with split-sample validation. Models were estimated using (1) prior-year predictors and (2) prior-year predictors potentially modifiable through intervention measured in the claims data. These models were evaluated using validated C-statistics. The relative influence of individual predictors in the models was evaluated.

Results

Among the 329 476 beneficiaries, the mean (SD) age was 76.0 (7.2) years and 190 346 (57.8%) were female. This final 5-group model included a minimal-user group (group 1, 37 572 individuals [11.4%]), a low-cost group (group 2, 48 575 individuals [14.7%]), a rising-cost group (group 3, 24 736 individuals [7.5%]), a moderate-cost group (group 4, 83 338 individuals [25.3%]), and a high-cost group (group 5, 135 255 individuals [41.2%]). Potentially modifiable characteristics strongly predicted these patterns (C-statistics range: 0.68-0.94). For groups with progressively increasing spending in particular, the most influential factors were number of medications (relative influence: 29.2), number of office visits (relative influence: 30.3), and mean medication adherence (relative influence: 33.6).

Conclusions and Relevance

Using a data-driven approach, distinct spending patterns were identified with high accuracy. The potentially modifiable predictors of membership in the rising-cost group represent important levers for early interventions that may prevent later spending increases. This approach could be adapted by organizations to target quality improvement interventions, particularly because numerous health care organizations are increasingly using these routinely collected data.

Case-mix for assessing primary care value (CPCV)

Measuring primary care (PC) performance and designing payment systems that reward value rather than volume have been a great challenge due in large part to lack of reliable risk adjustment mechanisms pertinent to primary care. Using risk scores designed for total resource needs to assess PC performance or set PC payment rates is inadequate because high-cost patients may not have high needs in PC and vice versa. The greatest challenge in developing a risk algorithm for PC is that significant components of PC providers' workload are unobservable but needed in the modeling. In this study, we sought to overcome this challenge by analyzing 5,172,773 patients in the U.S. Veterans Affairs (VA) healthcare system to identify potential proxies of the unobservable PC workload. By combining the number of PC visits and prescription drug classes, we formed a proxy for the expected PC workload, which enabled us to develop a case-mix algorithm pertaining to primary care. The resultant algorithm with high explanatory power (R² = 0.702) is based on a publicly available patient classification system to account for patient comorbidities and thus can be used by other health systems to compare PC performance, workload, staffing levels, and to set more equitable payment rates.

Longitudinal Patterns of Spending Enhance the Ability to Predict Costly Patients: A Novel Approach to Identify Patients for Cost Containment

BACKGROUND

With rising health spending, predicting costs is essential to identify patients for interventions. Many of the existing approaches have moderate predictive ability, which may result, in part, from not considering potentially meaningful changes in spending over time. Group-based trajectory modeling could be used to classify patients into dynamic long-term spending patterns.

OBJECTIVES

To classify patients by their spending patterns over a 1-year period and to assess the ability of models to predict patients in the highest spending trajectory and the top 5% of annual spending using prior-year predictors.

SUBJECTS

We identified all fully insured adult members enrolled in a large US nationwide insurer and used medical and prescription data from 2009 to 2011.

RESEARCH DESIGN

Group-based trajectory modeling was used to classify patients by their spending patterns over a 1-year period. We assessed the predictive ability of models that categorized patients in the top fifth percentile of annual spending and in the highest spending trajectory, using logistic regression and split-sample validation. Models were estimated using investigator-specified variables and a proprietary risk-adjustment method.

RESULTS

Among 998,651 patients, in the best-performing model, prediction was strong for patients in the highest trajectory group (C-statistic: 0.86; R: 0.47). The C-statistic of being in the top fifth percentile of spending in the best-performing model was 0.82 (R: 0.26). Approaches using nonproprietary investigator-specified methods performed almost as well as other risk-adjustment methods (C-statistic: 0.81 vs. 0.82).

CONCLUSIONS

Trajectory modeling may be a useful way to predict costly patients that could be implementable by payers to improve cost-containment efforts.

Profiling Patients' Healthcare Needs to Support Integrated, Person-Centered Models for Long-Term Disease Management (Profile): Research Design

Background:

This article presents the design of PROFILe, a study investigating which (bio)medical and non-(bio)medical patient characteristics should guide more tailored chronic care. Based on this insight, the project aims to develop and validate ‘patient profiles’ that can be used in practice to determine optimal treatment strategies for subgroups of chronically ill with similar healthcare needs and preferences.

Methods/Design:

PROFILe is a practice-based research comprising four phases. The project focuses on patients with type 2 diabetes. During the first study phase, patient profiles are drafted based on a systematic literature research, latent class growth modeling, and expert collaboration. In phase 2, the profiles are validated from a clinical, patient-related and statistical perspective. Phase 3 involves a discrete choice experiment to gain insight into the patient preferences that exist per profile. In phase 4, the results from all analyses are integrated and recommendations formulated on which patient characteristics should guide tailored chronic care.

Discussion:

PROFILe is an innovative study which uses a uniquely holistic approach to assess the healthcare needs and preferences of chronically ill. The patient profiles resulting from this project must be tested in practice to investigate the effects of tailored management on patient experience, population health and costs.

Association of medical home team-based care functions and perceived improvements in patient-centered care at VHA primary care clinics

INTRODUCTION

Team-based care is central to the patient-centered medical home (PCMH), but most PCMH evaluations measure team structure exclusively. We assessed team-based care in terms of team structure, process and effectiveness, and the association with improvements in teams׳ abilities to deliver patient-centered care.

MATERIAL AND METHODS

We fielded a cross-sectional survey among 913 VA primary care clinics implementing a PCMH model in 2012. The dependent variable was clinic-level respondent-reported improvements in delivery of patient-centered care. Independent variables included three sets of measures: (1) team structure, (2) team process, and (3) team effectiveness. We adjusted for clinic workload and patient comorbidity.

RESULTS

4819 surveys were returned (25% estimated response rate). The highest ratings were for team structure (median of 89% of respondents being assigned to a teamlet, i.e., a PCP working with the same clinical associate, nurse care manager and clerk) and lowest for team process (median of 10% of respondents reporting the lowest level of stress/chaos). In multivariable regression, perceived improvements in patient-centered care were most strongly associated with participatory decision making (β=32, P<0.0001) and history of change in the clinic (β=18, P=0008) (both team processes). A stressful/chaotic clinic environment was associated with higher barriers to patient centered care (β=0.16-0.34, P=<0.0001), and lower improvements in patient-centered care (β=-0.19, P=0.001).

CONCLUSIONS

Team process and effectiveness measures, often omitted from PCMH evaluations, had stronger associations with perceived improvements in patient-centered care than team structure measures.

IMPLICATIONS

Team process and effectiveness measures may facilitate synthesis of evaluation findings and help identify positive outlier clinics.

[Morbidity observed in a health area: Impact on professionals and funding]

Objetivo

Analizar, en el contexto de un área sanitaria, la morbilidad desagregada por centro de salud de los pacientes que entran en contacto con los servicios asistenciales para proponer un ajuste a la financiación en el pago per cápita.

Diseño

Estudio descriptivo, retrospectivo, de la morbilidad observada en los ciudadanos asignados a un área de salud durante el año 2010.

Emplazamiento

Área 9 de salud de la Comunidad Autónoma de Madrid, que comprende los municipios de Fuenlabrada, Humanes y Moraleja de Enmedio. Incluyendo todos los niveles de atención sanitaria.

Participantes

La totalidad de ciudadanos con tarjeta sanitaria asignada a un centro de salud del área que haya mantenido contacto con los servicios públicos de salud del propio área.

Mediciones

Se obtienen y agrupan los contactos codificados de los pacientes mediante el agrupador poblacional 3MTM Clinical Risk Grouping Software (CRG) cada paciente resulta incluido en un grupo homogéneo y excluyente con una morbilidad numérica y sentido clínico. A través de la tarjeta sanitaria se conoce centro de salud, médico de atención primaria, edad y sexo.

Resultados

Se estratifica la morbilidad por centro de salud, médico de atención primaria, edad y sexo y analizando las diferencias entre cada una de ellas y sus diferentes combinaciones.

Conclusiones

Se comprueba cómo los valores promedio de morbilidad de la población presentan valores distintos en cada zona básica de salud. Para mantener el principio de equidad sería necesario ajustar pago per cápita y número de tarjetas asignadas en función de la morbilidad observada de la población.

An analytics approach to designing patient centered medical homes

Recently the patient centered medical home (PCMH) model has become a popular team based approach focused on delivering more streamlined care to patients. In current practices of medical homes, a clinical based prediction frame is recommended because it can help match the portfolio capacity of PCMH teams with the actual load generated by a set of patients. Without such balances in clinical supply and demand, issues such as excessive under and over utilization of physicians, long waiting time for receiving the appropriate treatment, and non-continuity of care will eliminate many advantages of the medical home strategy. In this paper, by using the hierarchical generalized linear model with multivariate responses, we develop a clinical workload prediction model for care portfolio demands in a Bayesian framework. The model allows for heterogeneous variances and unstructured covariance matrices for nested random effects that arise through complex hierarchical care systems. We show that using a multivariate approach substantially enhances the precision of workload predictions at both primary and non primary care levels. We also demonstrate that care demands depend not only on patient demographics but also on other utilization factors, such as length of stay. Our analyses of a recent data from Veteran Health Administration further indicate that risk adjustment for patient health conditions can considerably improve the prediction power of the model.

Racial, Income, and Marital Status Disparities in Hospital Readmissions Within a Veterans-Integrated Health Care Network

Hospital readmission is an important indicator of health care quality and currently used in determining hospital reimbursement rates by Centers for Medicare & Medicaid Services. Given the important policy implications, a better understanding of factors that influence readmission rates is needed. Racial disparities in readmission have been extensively studied, but income and marital status (a postdischarge care support indicator) disparities have received limited attention. By employing three Poisson regression models controlling for different confounders on 8,718 patients in a veterans-integrated health care network, this study assessed racial, income, and martial disparities in relation to total number of readmissions. In contrast to other studies, no racial and income disparities were found, but unmarried patients experienced significantly more readmissions: 16%, after controlling for the confounders. These findings render unique insight into health care policies aimed to improve race and income disparities, while challenging policy makers to reduce readmissions for those who lack family support.

Work Limitations and Their Relationship to Morbidity Burden among Academic Health Center Employees with Diabetes

The objective of this study was to determine the prevalence of work limitations and their relationship to morbidity burden among academic health center employees with diabetes. Employees with diabetes were surveyed via Internet and mail using the Work Limitations Questionnaire. Morbidity burden was measured using the Adjusted Clinical Groups methodology. Seventy-two percent of the employees with diabetes had a work limitation. Adjusted odds ratios for overall, physical, time, and output limitations were 1.81, 2.27, 2.13, and 2.14, respectively. Morbidity burden level is an indicator of work limitations in employees with diabetes and can be used to identify employees who may benefit from specialized services aimed at addressing their work limitations associated with diabetes.

Use of outpatient care in VA and Medicare among disability-eligible and age-eligible veteran patients

Background

More than half of veterans who use Veterans Health Administration (VA) care are also eligible for Medicare via disability or age, but no prior studies have examined variation in use of outpatient services by Medicare-eligible veterans across health system, type of care or time.

Objectives

To examine differences in use of VA and Medicare outpatient services by disability-eligible or age-eligible veterans among veterans who used VA primary care services and were also eligible for Medicare.

Methods

A retrospective cohort study of 4,704 disability- and 10,816 age-eligible veterans who used VA primary care services in fiscal year (FY) 2000. We tracked their outpatient utilization from FY2001 to FY2004 using VA administrative and Medicare claims data. We examined utilization differences for primary care, specialty care, and mental health outpatient visits using generalized estimating equations.

Results

Among Medicare-eligible veterans who used VA primary care, disability-eligible veterans had more VA primary care visits (p < 0.001) and more VA specialty care visits (p < 0.001) than age-eligible veterans. They were more likely to have mental health visits in VA (p < 0.01) and Medicare-reimbursed visits (p < 0.01). Disability-eligible veterans also had more total (VA+Medicare) visits for primary care (p < 0.01) and specialty care (p < 0.01), controlling for patient characteristics.

Conclusions

Greater use of primary care and specialty care visits by disability-eligible veterans is most likely related to greater health needs not captured by the patient characteristics we employed and eligibility for VA care at no cost. Outpatient care patterns of disability-eligible veterans may foreshadow care patterns of veterans returning from Afghanistan and Iraq wars, who are entering the system in growing numbers. This study provides an important baseline for future research assessing utilizations among returning veterans who use both VA and Medicare systems. Establishing effective care coordination protocols between VA and Medicare providers can help ensure efficient use of taxpayer resources and high quality care for disabled veterans.

Association of incident dementia with hospitalizations

CONTEXT

Dementia is associated with increased rates and often poorer outcomes of hospitalization, including worsening cognitive status. New evidence is needed to determine whether some admissions of persons with dementia might be potentially preventable.

OBJECTIVE

To determine whether dementia onset is associated with higher rates of or different reasons for hospitalization, particularly for ambulatory care-sensitive conditions (ACSCs), for which proactive outpatient care might prevent the need for a hospital stay.

DESIGN, SETTING, AND PARTICIPANTS

Retrospective analysis of hospitalizations among 3019 participants in Adult Changes in Thought (ACT), a longitudinal cohort study of adults aged 65 years or older enrolled in an integrated health care system. All participants had no dementia at baseline and those who had a dementia diagnosis during biennial screening contributed nondementia hospitalizations until diagnosis. Automated data were used to identify all hospitalizations of all participants from time of enrollment in ACT until death, disenrollment from the health plan, or end of follow-up, whichever came first. The study period spanned February 1, 1994, to December 31, 2007.

MAIN OUTCOME MEASURES

Hospital admission rates for patients with and without dementia, for all causes, by type of admission, and for ACSCs.

RESULTS

Four hundred ninety-four individuals eventually developed dementia and 427 (86%) of these persons were admitted at least once; 2525 remained free of dementia and 1478 (59%) of those were admitted at least once. The unadjusted all-cause admission rate in the dementia group was 419 admissions per 1000 person-years vs 200 admissions per 1000 person-years in the dementia-free group. After adjustment for age, sex, and other potential confounders, the ratio of admission rates for all-cause admissions was 1.41 (95% confidence interval [CI], 1.23-1.61; P < .001), while for ACSCs, the adjusted ratio of admission rates was 1.78 (95% CI, 1.38-2.31; P < .001). Adjusted admission rates classified by body system were significantly higher in the dementia group for most categories. Adjusted admission rates for all types of ACSCs, including bacterial pneumonia, congestive heart failure, dehydration, duodenal ulcer, and urinary tract infection, were significantly higher among those with dementia.

CONCLUSION

Among our cohort aged 65 years or older, incident dementia was significantly associated with increased risk of hospitalization, including hospitalization for ACSCs.

An in-depth assessment of a diagnosis-based risk adjustment model based on national health insurance claims: the application of the Johns Hopkins Adjusted Clinical Group case-mix system in Taiwan

BACKGROUND

Diagnosis-based risk adjustment is becoming an important issue globally as a result of its implications for payment, high-risk predictive modelling and provider performance assessment. The Taiwanese National Health Insurance (NHI) programme provides universal coverage and maintains a single national computerized claims database, which enables the application of diagnosis-based risk adjustment. However, research regarding risk adjustment is limited. This study aims to examine the performance of the Adjusted Clinical Group (ACG) case-mix system using claims-based diagnosis information from the Taiwanese NHI programme.

METHODS

A random sample of NHI enrollees was selected. Those continuously enrolled in 2002 were included for concurrent analyses (n = 173,234), while those in both 2002 and 2003 were included for prospective analyses (n = 164,562). Health status measures derived from 2002 diagnoses were used to explain the 2002 and 2003 health expenditure. A multivariate linear regression model was adopted after comparing the performance of seven different statistical models. Split-validation was performed in order to avoid overfitting. The performance measures were adjusted R2 and mean absolute prediction error of five types of expenditure at individual level, and predictive ratio of total expenditure at group level.

RESULTS

The more comprehensive models performed better when used for explaining resource utilization. Adjusted R2 of total expenditure in concurrent/prospective analyses were 4.2%/4.4% in the demographic model, 15%/10% in the ACGs or ADGs (Aggregated Diagnosis Group) model, and 40%/22% in the models containing EDCs (Expanded Diagnosis Cluster). When predicting expenditure for groups based on expenditure quintiles, all models underpredicted the highest expenditure group and overpredicted the four other groups. For groups based on morbidity burden, the ACGs model had the best performance overall.

CONCLUSIONS

Given the widespread availability of claims data and the superior explanatory power of claims-based risk adjustment models over demographics-only models, Taiwan's government should consider using claims-based models for policy-relevant applications. The performance of the ACG case-mix system in Taiwan was comparable to that found in other countries. This suggested that the ACG system could be applied to Taiwan's NHI even though it was originally developed in the USA. Many of the findings in this paper are likely to be relevant to other diagnosis-based risk adjustment methodologies.

Line authority for nurse staffing and costs for acute inpatient care

There is little empirical evidence evaluating the effects of recent, widespread changes in nurse executive roles and nursing management structures on the costs of patient care. This retrospective cross-sectional study examined the relationship between line authority for nurse staffing and patient care costs (total, nursing, and non-nursing cost) using data from 124 Department of Veterans Affairs (VA) medical centers. After controlling for patient, facility, and market characteristics, nursing line authority was significantly associated with lower nursing cost per admission. Our results provide some evidence that a reduction in nursing line authority may adversely impact nursing costs.

Performance of comorbidity, risk adjustment, and functional status measures in expenditure prediction for patients with diabetes

OBJECTIVE

To compare the ability of generic comorbidity and risk adjustment measures, a diabetes-specific measure, and a self-reported functional status measure to explain variation in health care expenditures for individuals with diabetes.

RESEARCH DESIGN AND METHODS

This study included a retrospective cohort of 3,092 diabetic veterans participating in a multisite trial. Two comorbidity measures, four risk adjusters, a functional status measure, a diabetes complication count, and baseline expenditures were constructed from administrative and survey data. Outpatient, inpatient, and total expenditure models were estimated using ordinary least squares regression. Adjusted R(2) statistics and predictive ratios were compared across measures to assess overall explanatory power and explanatory power of low- and high-cost subgroups.

RESULTS

Administrative data-based risk adjusters performed better than the comorbidity, functional status, and diabetes-specific measures in all expenditure models. The diagnostic cost groups (DCGs) measure had the greatest predictive power overall and for the low- and high-cost subgroups, while the diabetes-specific measure had the lowest predictive power. A model with DCGs and the diabetes-specific measure modestly improved predictive power.

CONCLUSIONS

Existing generic measures can be useful for diabetes-specific research and policy applications, but more predictive diabetes-specific measures are needed.

Predictors of physical therapy clinic performance in the treatment of patients with low back pain syndromes

BACKGROUND AND PURPOSE

Little is known about organizational and service delivery factors related to quality of care in physical therapy. This study sought to identify characteristics related to differences in practice outcomes and service utilization.

SUBJECTS

The sample comprised 114 outpatient clinics and 1,058 therapists who treated 16,281 patients with low back pain syndromes during the period 2000-2001. Clinics participated with the Focus on Therapeutic Outcomes, Inc (FOTO) database.

METHODS

Hierarchical linear models were used to risk adjust treatment outcomes and number of visits per treatment episode. Aggregated residual scores from these models were used to classify each clinic into 1 of 3 categories in each of 3 types of performance groups: (1) effectiveness, (2) utilization, and (3) overall performance (ie, composite measure of effectiveness and utilization). Relationships between clinic classification and the following independent variables were examined by multinomial logistic regression: years of therapist experience, number of physical therapists, ratio of physical therapists to physical therapist assistants, proportion of patients with low back pain syndromes, number of new patients per physical therapist per month, utilization of physical therapist assistants, and setting.

RESULTS

Clinics that were lower utilizers of physical therapist assistants were 6.6 times more likely to be classified into the high effectiveness group compared with the low effectiveness group, 6.7 times more likely to be classified in the low utilization group compared with the high utilization group, and 12.4 times more likely to be classified in the best performance group compared with the worst performance group. Serving a higher proportion of patients with low back pain syndromes was associated with an increased likelihood of being classified in the lowest or middle group. Years of physical therapist experience was inversely associated with being classified in the middle utilization group compared with the highest utilization group.

DISCUSSION AND CONCLUSION

These findings suggest that, in the treatment of patients with low back pain syndromes, clinics that are low utilizers of physical therapist assistants are more likely to provide superior care (ie, better patient outcomes and lower service use).

The impact of contract primary care on health care expenditures and quality of care

The Department of Veterans Affairs (VA) established community-based outpatient clinics to improve veterans' access to primary care. This article compares VA use and expenditures among primary care users at 76 VA-staffed community clinics (n = 17,060) and 32 non-VA contract community clinics receiving capitation (n = 6,842) using VA administrative databases. It estimates utilization using negative binomial models and expenditures using generalized linear one-part or two-part models. Contract community clinic patients are less likely to use all types of outpatient services than VA-staffed community clinic patients but had similar quality of care. For patients seeking care, contract community clinic patients had similar specialty care expenditures but lower primary care, outpatient, and overall expenditures. Results suggest that capitated contract clinics did not shift costs to specialty care and appeared to be an economically efficient mechanism for improving veterans' access to primary care while meeting VA quality of care standards.

Pharmacy- and Diagnosis-Based Risk Adjustment for Children With Medicaid

BACKGROUND

Risk adjustment is useful for adjusting health care payments based on patients' health status.

OBJECTIVE

This work seeks to examine how well pharmacy- and diagnosis-based risk adjusters predict child health expenditures in Medicaid populations.

RESEARCH DESIGN

We used 1994-1995 Medicaid claims files for all children ages 0-18 years who were not covered by managed care in 3 states: Georgia, New Jersey, and Wisconsin. We examined separately 6 risk adjustment methods, 2 pharmacy-based and 4 diagnosis-based. We compared predictive accuracy of the methods for the whole sample and stratified by state and Medicaid enrollment category.

FINDINGS

Models with risk adjustment (either diagnosis- or pharmacy-based) had better predictive accuracy than demographic models. The pharmacy and diagnosis-based models had similar predictive accuracy. Risk adjuster performance differed by Medicaid enrollment category and state. Risk-adjusted models generally underpredict expenditures in populations with worse health status (eg, those in the Supplemental Security Income program [SSI]). The pharmacy-based models performed well for children in SSI relative to children in foster care.

CONCLUSIONS

Both pharmacy- and diagnosis-based risk adjustment improved the prediction of health expenditures compared models without risk adjustment. No single risk adjuster performed best in all situations, suggesting that optimal choices of risk adjusters may differ by purpose and context.

The performance of administrative and self-reported measures for risk adjustment of Veterans Affairs expenditures

OBJECTIVE

To evaluate the performance of different prospective risk adjustment models of outpatient, inpatient, and total expenditures of veterans who regularly use Veterans Affairs (VA) primary care.

DATA SOURCES

We utilized administrative, survey and expenditure data on 14,449 VA patients enrolled in a randomized trial that gave providers regular patient health assessments.

STUDY DESIGN

This cohort study compared five administrative data-based, two self-report risk adjusters, and base year expenditures in prospective models.

DATA EXTRACTION METHODS

VA outpatient care and nonacute inpatient care expenditures were based on unit expenditures and utilization, while VA expenditures for acute inpatient care were calculated from a Medicare-based inpatient cost function. Risk adjusters for this sample were constructed from diagnosis, medication and self-report data collected during a clinical trial. Model performance was compared using adjusted R2 and predictive ratios.

PRINCIPAL FINDINGS

In all expenditure models, administrative-based measures performed better than self-reported measures, which performed better than age and gender. The Diagnosis Cost Groups (DCG) model explained total expenditure variation (R2=7.2 percent) better than other models. Prior outpatient expenditures predicted outpatient expenditures best by far (R2=42 percent). Models with multiple measures improved overall prediction, reduced over-prediction of low expenditure quintiles, and reduced under-prediction in the highest quintile of expenditures.

CONCLUSIONS

Prediction of VA total expenditures was poor because expenditure variation reflected utilization variation, but not patient severity. Base year expenditures were the best predictor of outpatient expenditures and nearly the best for total expenditures. Models that combined two or more risk adjusters predicted expenditures better than single-measure models, but are more difficult and expensive to apply.