The sensitivity and specificity of forecasting high-cost users of medical care.

OBJECTIVES: This study compares the ability of 3 risk-assessment models to distinguish high and low expense-risk status within a managed care population. Models are the Global Risk-Assessment Model (GRAM) developed at the Kaiser Permanente Center for Health Research; a logistic version of GRAM; and a prior-expense model. GRAM was originally developed for use in adjusting Medicare payments to health plans. METHODS: Our sample of 98,985 cases was drawn from random samples of memberships of 3 staff/group health plans. Risk factor data were from 1992 and expenses were measured for 1993. Models produced distributions of individual-level annual expense forecasts (or predicted probabilities of high expense-risk status for logistic) for comparison to actual values. Prespecified "high-cost" thresholds were set within each distribution to analyze the models' ability to distinguish high and low expense-risk status. Forecast stability was analyzed through bootstrapping. RESULTS: GRAM discriminates better overall than its comparators (although the models are similar for policy-relevant thresholds). All models forecast the highest-cost cases relatively well. GRAM forecasts high expense-risk status better than its comparators within chronic and serious disease categories that are amenable to early intervention but also generates relatively more false positives within these categories. CONCLUSIONS: This study demonstrates the potential of risk-assessment models to inform care management decisions by efficiently screening managed care populations for high expense-risk. Such models can act as preliminary screens for plans that can refine model forecasts with detailed surveys. Future research should involve multiple-year data sets to explore the temporal stability of forecasts.

Modeling risk using generalized linear models.

Traditionally, linear regression has been the technique of choice for predicting medical risk. This paper presents a new approach to modeling the second part of two-part models utilizing extensions of the generalized linear model. The primary method of estimation for this model is maximum likelihood. This method as well as the generalizations quasi-likelihood and extended quasi-likelihood are discussed. An example using medical expense data from Washington State employees is used to illustrate the methods. The model includes demographic variables as well as an Ambulatory. Care Group variable to account for prior health status.

Health-based payment and computerized patient record systems.

Health care information technology is changing rapidly and dramatically. A small but growing number of clinicians, especially those in staff and group model HMOs and hospital-affiliated practices, are automating their patient medical records in response to pressure to improve quality and reduce costs. Computerized patient record systems in HMOs track risks, diagnoses, patterns of care, and outcomes across large populations. These systems provide access to large amounts of clinical information; as a result, they are very useful for risk-adjusted or health-based payment. The next stage of evolution in health-based payment is to switch from fee-for-service (claims) to HMO technology in calculating risk coefficients. This will occur when HMOs accumulate data sets containing records on provider-defined disease episodes, with every service linked to its appropriate disease episode for millions of patients. Computerized patient record systems support clinically meaningful risk-assessment models and protect patients and medical groups from the effects of adverse selection. They also offer significant potential for improving quality of care.

Cost-effectiveness of a hospital-based smoking cessation intervention.

OBJECTIVES: This study evaluated the cost-effectiveness of a smoking cessation and relapse-prevention program for hospitalized adult smokers from the perspective of an implementing hospital. It is an economic analysis of a two-group, controlled clinical trial in two acute care hospitals owned by a large group-model health maintenance organization. The intervention included a 20-minute bedside counseling session with an experienced health counselor, a 12-minute video, self-help materials, and one or two follow-up calls. METHODS: Outcome measures were incremental cost (above usual care) per quit attributable to the intervention and incremental cost per discounted life-year saved attributable to the intervention. RESULTS: Cost of the research intervention was $159 per smoker, and incremental cost per incremental quit was $3,697. Incremental cost per incremental discounted life-year saved ranged between $1,691 and $7,444, much less than most other routine medical procedures. Replication scenarios suggest that, with realistic implementation assumptions, total intervention costs would decline significantly and incremental cost per incremental discounted life-year saved would be reduced by more than 90%, to approximately $380. CONCLUSIONS: Providing brief smoking cessation advice to hospitalized smokers is relatively inexpensive, cost-effective, and should become a part of the standard of inpatient care.

Building health plan databases to risk adjust outcomes and payments.

OBJECTIVES: To highlight the types and sources of data on medical risk and outcomes routinely collected by managed care organizations over time; to summarize the quality and consistency of these data; and to describe some of the difficulties that arise in collecting, pooling, and using these data. DESIGN: Synthesis of the experiences of two risk-adjustment modeling projects in assembling large volumes of demographic, diagnostic, and expense data from several health maintenance organizations (HMOs) over multiple years. SETTING: Six large HMOs from the Northwest, North Central, and Northeast regions of the USA. INTERVENTIONS: Health plans were approached to participate in a risk-adjustment study, presented with an extensive variable-by-variable data request, and, if willing to participate, asked to specify a desired process for extracting, copying, and transferring selected variables to the study site for purposes of research. Depending on local circumstances, three different approaches were used: (i) health plan staff obtained the data and organized them into the requested study format; (ii) study staff were provided access to health plan data systems to perform the extractions directly; and (iii) health plans hired contract programmers to perform the extractions under the direction of the study team. Key measures of risk and cost were extracted and merged into analysis files. MAIN OUTCOME MEASURES: Complete and consistent eligibility maps, demographic information, inpatient and outpatient diagnoses, and total health plan expense for each enrollee. RESULTS: We have been successful in collecting and integrating complete utilization, morbidity, demographic, and cost data on total memberships of five large HMOs as well as a subset from a sixth HMO, all for multiple years. CONCLUSION: While HMOs vary greatly in the quality and comprehensiveness of their data systems, these attributes have been improving across the board over time. Automated health plan data systems represent potentially valuable sources of data on health risks and outcomes and can be used to benchmark disease management programs and risk adjust capitation payments and medical outcomes.

The effect of increased prescription drug cost-sharing on medical care utilization and expenses of elderly health maintenance organization members.

OBJECTIVES: The nature and extent of prescription drug benefits for the elderly are a continuing concern for health-care managers and policy makers. This study examined the impact of increased prescription drug cost-sharing on the drug and medical care utilization and expenses of the elderly. METHODS: Two groups of well-insured Medicare risk-based members of a large health maintenance organization (HMO) had their copayments increased in different years during a 3-year period. Four 2-year analysis periods were established for comparing these elderly groups. During one analysis period, copayments did not change in either group. RESULTS: Moderate increases of from $1 to $3, from $3 to $5 per copayment, and from 50% per dispensing to 70% per dispensing with a maximum payment per dispensing resulted in lower annual per capita prescription drug use and expenses. No consistent annual changes were observed in either medical care utilization (office visits, emergency room visits, home health-care visits, hospitalizations) or total medical care expenses across analysis periods. CONCLUSIONS: No consistent relationships were observed between increased copayments per dispensing and medical care utilization and expense. Future research needs to address the impact on the classes of medications received and related health status, and the impact of larger increases in copayments per dispensing on medical care and health-related factors.

Analysis of the costs of NSAID-associated gastropathy. Experience in a US health maintenance organisation.

Clinicians recognise nonsteroidal anti-inflammatory drugs (NSAIDs) as valuable first-line agents in the treatment of rheumatic disorders and as dangerous irritants to the gastrointestinal tract. This has led to questions about the economic impact of NSAID-induced gastropathy in populations. This study estimated the 1992 costs of NSAID-associated gastropathy episodes, and calculated an iatrogenic cost factor for NSAID-associated gastropathy among elderly members of a health maintenance organisation (HMO), the Northwest Region of Kaiser Permanente. Using data retrieved from automated databases and from medical records, NSAID and antiulcer drug costs were calculated, and estimates were made of the incidence rates of inpatient and outpatient NSAID-associated gastropathies, the services provided to treat them, and the cost of those services. Kaiser Permanente Northwest spent $US0.35 for each $US1.00 spent on NSAID therapy for the elderly, an iatrogenic cost factor of 1.35. The estimated average treatment per NSAID-associated gastropathy episode was $US2172. The average outpatient pharmacy cost per elderly NSAID user was $US80 and estimated average NSAID-associated treatment cost per elderly NSAID user was $US43. Although the findings were specific to the HMO because of the databases used, the methodology employed and the drug formulary influence on NSAID selection, they show that a substantial amount of resources were used to treat NSAID-induced gastropathies in the elderly, underscoring the risk of prescribing NSAIDs and reinforcing the need for their prudent use in elderly patients.

The impact of increasing patient prescription drug cost sharing on therapeutic classes of drugs received and on the health status of elderly HMO members.

OBJECTIVE: To assess the impact of increased prescription drug copayments on the therapeutic classes of drugs received and health status of the elderly. HYPOTHESES TESTED: Increased prescription drug copayments will reduce the relative exposure to, annual days use of, and prescription drug costs for drugs used in self-limiting conditions, but will not affect drugs used in progressive chronic conditions and will not reduce health status. STUDY DESIGN: Each year over a three-year period, one or the other of two well-insured Medicare risk groups in an HMO setting had their copayments per dispensing increased. Sample sizes ranged from 6,704 to 7,962. DATA SOURCES/DATA COLLECTION: Automated administrative data systems of the HMO were used to determine HMO eligibility, prescription drug utilization, and health status. ANALYSIS DESIGN: Analysis of variance or covariance was employed to measure change in dependent variables. FINDINGS: Relative exposure, annual days of use, and prescription drug costs for drugs used in self-limiting conditions and in progressive chronic conditions were not affected in a consistent manner across years by increases in prescription drug copayment. Health status may have been adversely affected. Larger increases in copayments appeared to generate more changes. CONCLUSIONS: Small changes in copayments did not appear to substantially affect outcomes. Large changes in copayments need further examination.

Treatment refusal procedures and service utilization: a comparison of involuntarily hospitalized populations.

This article examines treatment refusal in a large group of hospitalized civilly committed patients. Comparison is made between those subjects whose refusal was reviewed by Oregon's administrative procedures for treatment refusal (override group) and those committed patients who more readily accepted treatment and were not evaluated by this procedure. The objective was to examine the override process and to explore potential differences between these groups in their utilization of hospital and community mental health services before and after the index hospitalization. We reviewed hospital charts on all subjects who went through the administrative override procedure and collected state hospital and community mental health services information from the statewide computerized information system on all subjects in the study. Several key differences were found between the groups. The override sample had significantly more women, and these patients spent significantly more time in the index hospitalization and had had more past hospitalizations. There were no differences between the groups in their utilization of community services before or after the index hospitalization and no difference in hospitalization rates after the index hospitalization. The conclusion is that the Oregon override procedure is functioning consistently, without undue delay in decision making. More investigation is necessary to determine whether override subjects represent a distinct subpopulation within the larger group of chronically mentally ill patients.

Enrollment duration, service use, and costs of care for severely mentally ill members of a health maintenance organization.

BACKGROUND: The rapid growth of prepaid health care and the increasing enrollment of Medicaid clients in health maintenance organizations (HMOs) raise concerns about the adequacy of services for persons with severe mental illness in capitated health plans. Uncontrolled studies have suggested that enrollment of HMO members with mental illness may be prematurely terminated. METHODS: We identified 250 adult Kaiser Permanente Northwest Region (Portland, Ore) members who were enrolled during 1986 or 1987 and had chart diagnoses of schizophrenia or bipolar disorder. Severely mentally ill subjects were matched by age and sex with control HMO members with and without diabetes mellitus. Records of the HMO and the state mental health agency were reviewed to determine HMO enrollment duration, private and public service utilization, and HMO costs of care during the 4-year follow-up period. RESULTS: The severely mentally ill subjects had 42 months of HMO enrollment during the follow-up period compared with 37 months for the controls without diabetes mellitus and 47 months for the patients with diabetes mellitus (P < .001). When HMO enrollment prior to the study was taken into account, the severely mentally ill subjects and those with diabetes mellitus had similar membership duration. Among the severely mentally ill subjects, community mental health service use was related to longer duration of HMO enrollment (P < .05) but HMO costs of care per member per month were not related to retention. The severely mentally ill subjects were high users of mental health services but their use of general medical care was similar to that of the controls without diabetes mellitus. CONCLUSIONS: This controlled study found no evidence for early termination of HMO members with costly mental illness. Use of community mental health care was associated with longer duration of HMO enrollment.

Development of use and cost measures in a nursing intervention for family caregivers and frail elderly patients.

The purpose of this pilot study was to refine and evaluate methods of measuring costs of an innovative home-health nursing intervention designed to support frail, older persons and their family caregivers. We evaluated a multifaceted strategy to collect a detailed utilization profile from 22 caregiver/care receiver dyads for hospital, ambulatory, home health, nursing home, and community services. The strategy was feasible for most participants, maximized accuracy of cost data, and minimized research burden on study participants. Lower overall costs were found in the intervention group, but the difference was not significant. Approaches to the measurement of costs in this study can serve as models for evaluating other innovations in nursing, home care, and long-term care.

Chronic disease, functional health status, and demographics: a multi-dimensional approach to risk adjustment.

OBJECTIVE: The goal of this study was to develop unbiased risk-assessment models to be used for paying health plans on the basis of enrollee health status and use propensity. We explored the risk structure of adult employed HMO members using self-reported morbidities, functional status, perceived health status, and demographic characteristics. DATA SOURCES/STUDY SETTING: Data were collected on a random sample of members of a large, federally qualified, prepaid group practice, hospital-based HMO located in the Pacific Northwest. STUDY DESIGN: Multivariate linear nonparametric techniques were used to estimate risk weights on demographic, morbidity, and health status factors at the individual level. The dependent variable was annual real total health plan expense for covered services for the year following the survey. Repeated random split-sample validation techniques minimized outlier influences and avoided inappropriate distributional assumptions required by parametric techniques. DATA COLLECTION/EXTRACTION METHODS: A mail questionnaire containing an abbreviated medical history and the RAND-36 Health Survey was administered to a 5 percent sample of adult subscribers and their spouses in 1990 and 1991, with an overall 44 percent response rate. Utilization data were extracted from HMO automated information systems. Annual expenses were computed by weighting all utilization elements by standard unit costs for the HMO. PRINCIPAL FINDINGS: Prevalence of such major chronic diseases as heart disease, diabetes, depression, and asthma improve prediction of future medical expense; functional health status and morbidities are each better than simple demographic factors alone; functional and perceived health status as well as demographic characteristics and diagnoses together yield the best prediction performance and reduce opportunities for selection bias. We also found evidence of important interaction effects between functional/perceived health status scales and disease classes. CONCLUSIONS: Self-reported morbidities and functional health status are useful risk measures for adults. Risk-assessment research should focus on combining clinical information with social survey techniques to capitalize on the strengths of both approaches. Disease-specific functional health status scales should be developed and tested to capture the most information for prediction.

Improving the AAPCC (adjusted average per capita cost) with health-status measures from the MCBS (Medicare Current Beneficiary Survey).

Using data from the 1991 Medicare Current Beneficiary Survey (MCBS), multiple regression-based models predicting 1992 Medicare costs are developed and compared. A comprehensive model incorporating demographic, diagnostic, perceived health, and disability variables is shown to be stable and to fit the data well over the full range of Medicare-covered annual per capita expenses and for a variety of beneficiary subgroups defined by their health and functional status. This model produces stable unbiased estimates of expenditures on validation samples. A variant of this model is being considered for use in setting Medicare capitation payments for the second phase of the social/health maintenance organization (S/HMO) demonstration.

The effects of exercise on falls in elderly patients. A preplanned meta-analysis of the FICSIT Trials. Frailty and Injuries: Cooperative Studies of Intervention Techniques.

OBJECTIVE: To determine if short-term exercise reduces falls and fall-related injuries in the elderly. DESIGN: A preplanned meta-analysis of the seven Frailty and Injuries: Cooperative Studies of Intervention Techniques (FICSIT)--independent, randomized, controlled clinical trials that assessed intervention efficacy in reducing falls and frailty in elderly patients. All included an exercise component for 10 to 36 weeks. Fall and injury follow-up was obtained for up to 2 to 4 years. SETTING: Two nursing home and five community-dwelling (three health maintenance organizations) sites. Six were group and center based; one was conducted at home. PARTICIPANTS: Numbers of participants ranged from 100 to 1323 per study. Subjects were mostly ambulatory and cognitively intact, with minimum ages of 60 to 75 years, although some studies required additional deficits, such as functionally dependent in two or more activities of daily living, balance deficits or lower extremity weakness, or high risk of falling. INTERVENTIONS: Exercise components varied across studies in character, duration, frequency, and intensity. Training was performed in one area or more of endurance, flexibility, balance platform, Tai Chi (dynamic balance), and resistance. Several treatment arms included additional nonexercise components, such as behavioral components, medication changes, education, functional activity, or nutritional supplements. MAIN OUTCOME MEASURES: Time to each fall (fall-related injury) by self-report and/or medical records. RESULTS: Using the Andersen-Gill extension of the Cox model that allows multiple fall outcomes per patient, the adjusted fall incidence ratio for treatment arms including general exercise was 0.90 (95% confidence limits [CL], 0.81, 0.99) and for those including balance was 0.83 (95% CL, 0.70, 0.98). No exercise component was significant for injurious falls, but power was low to detect this outcome. CONCLUSIONS: Treatments including exercise for elderly adults reduce the risk of falls.

Assessing relative health plan risk with the RAND-36 health survey.

Unbiased risk assessment models base health plan payments on enrollee health care needs. We explored the risk structure of employed adult health maintenance organization (HMO) members using the RAND-36 health survey. We used multivariate techniques to estimate risk weights on demographic and health status factors. The dependent variable was annual real total health plan expense for covered services for the year following the survey. Repeated random-split-sample validation techniques minimized outlier influences. Five scales improved prediction over simple demographic factors, but demographic factors still were required to achieve unbiased forecasts. Self-reported health status is a useful and powerful risk measure for adults.