Estimating log models: to transform or not to transform?

Health economists often use log models to deal with skewed outcomes, such as health utilization or health expenditures. The literature provides a number of alternative estimation approaches for log models, including ordinary least-squares on ln(y) and generalized linear models. This study examines how well the alternative estimators behave econometrically in terms of bias and precision when the data are skewed or have other common data problems (heteroscedasticity, heavy tails, etc.). No single alternative is best under all conditions examined. The paper provides a straightforward algorithm for choosing among the alternative estimators. Even if the estimators considered are consistent, there can be major losses in precision from selecting a less appropriate estimator.

Health insurance: tradeoffs revisited.

In two recent papers, [Journal of Health Economics 18(2), 141-152, Journal of Health Economics 18(6), 811-824] Nyman raised some questions about the welfare calculations and conclusions in our earlier paper [Manning and Marquis, Health insurance: the tradeoff between risk pooling and moral hazard, Vol. 15, 1996]. This note discusses the erroneous criticisms in his papers. First, although, we estimated a Marshallian demand curve, our calculations are based on compensating variations that incorporate the gains from risk pooling. Second, our estimates of second best insurance plans indicate that some cost sharing is optimal, in contradiction to his assertion that our results raise questions about the desirability of insurance coverage. The comment also deals with other issues raised by Nyman.

Cost-effectiveness of systematic depression treatment for high utilizers of general medical care.

BACKGROUND: Expanding access to high-quality depression treatment will depend on the balance of incremental benefits and costs. We examine the incremental cost-effectiveness of an organized depression management program for high utilizers of medical care. METHODS: Computerized records at 3 health maintenance organizations were used to identify adult patients with outpatient medical visit rates above the 85th percentile for 2 consecutive years. A 2-step screening process identified patients with current depressive disorders, who were not in active treatment. Eligible patients were randomly assigned to continued usual care (n = 189) or to an organized depression management program (n = 218). The program included patient education, antidepressant pharmacotherapy initiated in primary care (when appropriate), systematic telephone monitoring of adherence and outcomes, and psychiatric consultation as needed. Clinical outcomes (assessed using the Hamilton Depression Rating Scale on 4 occasions throughout 12 months) were converted to measures of "depression-free days." Health services utilization and costs were estimated using health plan-standardized claims. RESULTS: The intervention program led to an adjusted increase of 47.7 depression-free days throughout 12 months (95% confidence interval [CI], 28.2-67.8 days). Estimated cost increases were $1008 per year (95% CI, $534-$1383) for outpatient health services, $1974 per year for total health services costs (95% CI, $848-$3171), and $2475 for health services plus time-in-treatment costs (95% CI, $880-$4138). Including total health services and time-in-treatment costs, estimated incremental cost per depression-free day was $51.84 (95% CI, $17.37-$108.47). CONCLUSIONS: Among high utilizers of medical care, systematic identification and treatment of depression produce significant and sustained improvements in clinical outcomes as well as significant increases in health services costs.

Interpreting results in mental health research.

It is often difficult to interpret the clinical or policy significance of findings from mental health research when results are presented only in terms of statistical significance. Results expressed in terms of p values or as a metric corresponding to a mental health status scale are seldom intuitively meaningful. To help interpret the significance of research results, we demonstrate a social validity approach that relates scores on mental health status scales to four subsequent major life events. A logistic regression model is used to estimate the relation between mental health status scores and the probability of subsequent major life events, using data obtained on Medicaid beneficiaries with schizophrenia from an evaluation of the Utah Prepaid Mental Health Plan. Using this relatively simple approach will demonstrate to policy makers, clinicians, and researchers the social impact of an outcome, thereby aiding in the interpretation of the significance of results.

Effect of multiple-source entry on price competition after patent expiration in the pharmaceutical industry.

OBJECTIVE: To analyze the effect of multiple-source drug entry on price competition after patent expiration in the pharmaceutical industry. DATA SOURCES: Originators and their multiple-source drugs selected from the 35 chemical entities whose patents expired from 1984 through 1987. Data were obtained from various primary and secondary sources for the patents' expiration dates, sales volume and units sold, and characteristics of drugs in the sample markets. STUDY DESIGN: The study was designed to determine significant factors using the study model developed under the assumption that the off-patented market is an imperfectly segmented market. PRINCIPAL FINDINGS: After patent expiration, the originators' prices continued to increase, while the price of multiple-source drugs decreased significantly over time. By the fourth year after patent expiration, originators' sales had decreased 12 percent in dollars and 30 percent in quantity. Multiple-source drugs increased their sales twofold in dollars and threefold in quantity, and possessed about one-fourth (in dollars) and half (in quantity) of the total market three years after entry. CONCLUSION: After patent expiration, multiple-source drugs compete largely with other multiple-source drugs in the price-sensitive sector, but indirectly with the originator in the price-insensitive sector. Originators have first-mover advantages, and therefore have a market that is less price sensitive after multiple-source drugs enter. On the other hand, multiple-source drugs target the price-sensitive sector, using their lower-priced drugs. This trend may indicate that the off-patented market is imperfectly segmented between the price-sensitive and insensitive sector. Consumers as a whole can gain from the entry of multiple-source drugs because the average price of the market continually declines after patent expiration.

Randomized trial of a depression management program in high utilizers of medical care.

BACKGROUND: High utilizers of nonpsychiatric health care services have disproportionally high rates of undiagnosed or undertreated depression. OBJECTIVE: To determine the impact of offering a systematic primary care-based depression treatment program to depressed "high utilizers" not in active treatment. DESIGN: Randomized clinical trial. SETTING: One hundred sixty-three primary care practices in 3 health maintenance organizations located in different geographic regions of the United States. PATIENTS: A group of 1465 health maintenance organization members were identified as depressed high utilizers using a 2-stage telephone screening process. Eligibility criteria were met by 410 patients and 407 agreed to enroll: 218 in the depression management program (DMP) practices and 189 in the usual care (UC) group. INTERVENTION: The DMP included patient education materials, physician education programs, telephone-based treatment coordination, and antidepressant pharmacotherapy initiated and managed by patients' primary care physicians. MAIN OUTCOME MEASURES: Depression severity was measured using the Hamilton Depression Rating Scale (Ham-D) and functional status using the Medical Outcomes Study 20-item short form (SF-20) subscales. Outpatient visit and hospitalization rates were measured using the health plan's encounter data. RESULTS: Based on an intent-to-treat analysis, at least 3 antidepressant prescriptions were filled in the first 6 months by 151 (69.3%) of 218 of DMP patients vs 35 (18.5%) of 189 in UC (P < .001). Improvements in Ham-D scores were significantly greater in the intervention group at 6 weeks (P = .04), 3 months (P = .02), 6 months (P < .001), and 12 months (P < .001). At 12 months, DMP intervention patients were more improved than UC patients on the mental health, social functioning, and general health perceptions scales of the SF-20 (P < .05 for all). CONCLUSION: In depressed high utilizers not already in active treatment, a systematic primary care-based treatment program can substantially increase adequate antidepressant treatment, decrease depression severity, and improve general health status compared with usual care.

Lifetime costs and compensation for injuries.

Little is known about how well individuals are compensated for injuries. This study uses data from a 1989 survey to estimate both the lifetime costs and compensation for injuries. Our findings show that about 55% of the cost is compensated by public or private programs. Compensation rates are lower for disabling injuries and those of long duration. The results also suggest that compensation system reforms that would place stricter limits on maximum compensation might not be a distributionally fair solution. The reasons are that costs are highly skewed, and the share of costs recovered by compensation programs is currently lowest for injuries that are long term, disabling, and the most expensive.

Outcomes for Medicaid beneficiaries with schizophrenia under a prepaid mental health carve-out.

This study examines the impact of a mental health carve-out program in Utah on mental health status of Medicaid beneficiaries with schizophrenia. Three community mental health centers contracted to provide mental health care for all Medicaid beneficiaries in their service areas under managed care arrangements, while beneficiaries in the remainder of the state remained under traditional Medicaid. A pre-post evaluation was utilized, with a contemporaneous control group of Utah Medicaid beneficiaries with schizophrenia under traditional Medicaid. From 1991 to 1994, the average beneficiary's mental health status improved, but the improvement was less under the carve-out program than under traditional fee-for-service Medicaid. The difference was the greatest for beneficiaries with the worst mental health status at baseline, with effects growing over time. Medicaid beneficiaries with schizophrenia experienced less improvement in mental health status under a carve-out arrangement for mental health care compared to what would have happened under traditional Medicaid.

Depression among high utilizers of medical care.

OBJECTIVE: To determine the prevalence of unrecognized or unsuccessfully treated depression among high utilizers of medical care, and to describe the relation between depression, medical comorbidities, and resource utilization. DESIGN: Survey. SETTING: Three HMOs located in different geographic regions of the United States. PATIENTS: A total of 12,773 HMO members were identified as high utilizers. Eligibility criteria for depression screening were met by 10,461 patients. MEASUREMENTS AND MAIN RESULTS: Depression status was assessed with the Structured Clinical Interview for Diagnostic and Statistical Manual of Mental Disorders. Fourth Edition. Depression screening was completed in 7,203 patients who were high utilizers of medical care, of whom 1,465 (20.3%) screened positive for current major depression or major depression in partial remission. Among depressed patients, 621 (42.4%) had had a visit with a mental health specialist or a diagnosis of depression or both within the previous 2 years. The prevalence of well-defined medical conditions was the same in patients with and patients without evidence of depression (41.5% vs 41.5%, p = .87). However, high-utilizing patients who had not made a visit for a nonspecific complaint during the previous 2 years were at significantly lower risk of depression (13.1% vs 22.4%, p < .001). Patients with current depression or depression in partial remission had significantly higher numbers of annual office visits and hospital days per 1,000 than patients without depression. CONCLUSIONS: Although there was evidence that mental health problems had previously been recognized in many of the patients, a large percentage of high utilizers still suffered from active depression that either went unrecognized or was not being treated successfully. Patients who had not made visits for nonspecific complaints were at significantly lower risk of depression. Depression among high utilizers was associated with higher resource utilization.

The unreliability of individual physician "report cards" for assessing the costs and quality of care of a chronic disease.

CONTEXT: Physician profiling is widely used by many health care systems, but little is known about the reliability of commonly used profiling systems. OBJECTIVES: To determine the reliability of a set of physician performance measures for diabetes care, one of the most common conditions in medical practice, and to examine whether physicians could substantially improve their profiles by preferential patient selection. DESIGN AND SETTING: Cohort study performed from 1990 to 1993 at 3 geographically and organizationally diverse sites, including a large staff-model health maintenance organization, an urban university teaching clinic, and a group of private-practice physicians in an urban area. PARTICIPANTS: A total of 3642 patients with type 2 diabetes cared for by 232 different physicians. MAIN OUTCOME MEASURES: Physician profiles for their patients' hospitalization and clinic visit rates, total laboratory resource utilization rate and level of glycemic control by average hemoglobin A1c level with and without detailed case-mix adjustment. RESULTS: For profiles based on hospitalization rates, visit rates, laboratory utilization rates, and glycemic control, 4% or less of the overall variance was attributable to differences in physician practice and the reliability of the median physician's case-mix-adjusted profile was never better than 0.40. At this low level of physician effect, a physician would need to have more than 100 patients with diabetes in a panel for profiles to have a reliability of 0.80 or better (while more than 90% of all primary care physicians at the health maintenance organization had fewer than 60 patients with diabetes). For profiles of glycemic control, high outlier physicians could dramatically improve their physician profile simply by pruning from their panel the 1 to 3 patients with the highest hemoglobin A1c levels during the prior year. This advantage from gaming could not be prevented by even detailed case-mix adjustment. CONCLUSIONS: Physician "report cards" for diabetes, one of the highest-prevalence conditions in medical practice, were unable to detect reliably true practice differences within the 3 sites studied. Use of individual physician profiles may foster an environment in which physicians can most easily avoid being penalized by avoiding or deselecting patients with high prior cost, poor adherence, or response to treatments.

The effects of group size and group economic factors on collaboration: a study of the financial performance of rural hospitals in consortia.

STUDY QUESTIONS: To determine factors that distinguish effective rural hospital consortia from ineffective ones in terms of their ability to improve members' financial performance. Two questions in particular were addressed: (1) Do large consortia have a greater collective impact on their members? (2) Does a consortium's economic environment determine the degree of collective impact on members? DATA SOURCES AND STUDY SETTING: Based on the hospital survey conducted during February 1992 by the Robert Wood Johnson Hospital-Based Rural Health Care project of rural hospital consortia. The survey data were augmented with data from Medicare Cost Reports (1985-1991), AHA Annual Surveys (1985-1991), and other secondary data. STUDY DESIGN: Dependent variables were total operating profit, cost per adjusted admission, and revenue per adjusted admission. Control variables included degree of group formalization, degree of inequality of resources among members (group asymmetry), affiliation with other consortium group(s), individual economic environment, common hospital characteristics (bed size, ownership type, system affiliation, case mix, etc.), year (1985-1991), and census region dummies. PRINCIPAL FINDINGS: All dependent variables have a curvilinear association with group size. The optimum group size is somewhere in the neighborhood of 45. This reveals the benefits of collective action (i.e., scale economies and/or synergy effects) and the issue of complexity as group size increases. Across analyses, no strong evidence exists of group economic environment impacts, and the environmental influences come mainly from the local economy rather than from the group economy. CONCLUSION: There may be some success stories of collaboration among hospitals in consortia, and consortium effects vary across different collaborations. RELEVANCE/IMPACT: When studying consortia, it makes sense to develop a typology of groups based on some performance indicators. The results of this study imply that government, rural communities, and consortium staff and steering committees should forge the consortium concept by expanding membership in order to gain greater financial benefits for individual hospitals.

Panel on cost-effectiveness in health and medicine recommendations: identifying costs.

The assignment of costs in a cost-effectiveness analysis is a complex and disputed issue. The Panel on Cost-Effectiveness in Health and Medicine was convened to discuss standards that could be applied across a range of areas of cost-effectiveness. Additionally, the Panel had a mandate to resolve some controversial issues about the practice of cost-effectiveness that created difficulty in making comparisons across studies. The Panel proposed these guidelines: (1) Do at least some of the analysis from a social perspective; (2) Assign values to resources that reflect their opportunity costs; (3) Avoid zero counting of resources; (4) Avoid double counting of resources; (5) Make analyses only as exacting as necessary in a study. Difficulties in data collection were discussed. Among other questions considered by the panel were how to assign a value to the patient's time and which productivity costs to include in a cost-effectiveness analysis. With tools and suggestions from the Panel on Cost-Effectiveness in Health and Medicine, the cost analyst can report costs accurately and provide accurate comparisons of cost performance across states, trial modalities, or diseases.

Patterns of outpatient use of mental health services for Medicaid beneficiaries under a prepaid mental health carve-out.

This study examines the impact of a mental health carve-out, the Utah Prepaid Mental Health Plan (UPMHP), on use of outpatient mental health services by Medicaid beneficiaries with schizophrenia. Data were collected through interviews with the same group of Medicaid schizophrenic beneficiaries. A pre/post comparison with a contemporaneous control group examined the impact of the program on type of outpatient services used by beneficiaries. The results indicate a greater reliance on medically-oriented outpatient mental health services in treatment of beneficiaries under the UPMHP. Medicaid beneficiaries with schizophrenia in the UPMHP group received relatively fewer day treatment visits, but relatively more medication visits and individual therapy visits over the first 3 1/2 years of the program.

The effect of the Utah Prepaid Mental Health Plan on structure, process, and outcomes of care.

Capitation reduced Medicaid costs but had limited effects on most measures of process and outcome. Clients under capitation with the poorest mental health at baseline performed more poorly over time on some measures.

Can specialists reduce costs? The case of referrals to orthopaedic surgeons.

This study investigates whether timely referral to specialists, in this case orthopaedic surgeons, potentially can reduce the costs of a health care episode. Five musculoskeletal diagnoses were chosen, and the diagnostic and treatment history of approximately 2500 persons with these five diagnoses was traced to determine when in the course of their treatment episode they were referred from nonorthopaedist to orthopaedist care and how much their episode of care cost. It was found that the average episodic costs for those who were referred earlier in the episode of care was lower than for those referred later. The implication is that there are likely to be numerous identifiable health conditions that should be flagged for early referral to specialists to reduce costs. It also questions the validity of the presumption that specialist care is necessarily more expensive.

A longitudinal study of hospitalization rates for patients with chronic disease: results from the Medical Outcomes Study.

OBJECTIVE: To prospectively compare inpatient and outpatient utilization rates between prepaid (PPD) and fee-for-service (FFS) insurance coverage for patients with chronic disease. DATA SOURCE/STUDY SETTING: Data from the Medical Outcomes Study, a longitudinal observational study of chronic disease patients conducted in Boston, Chicago, and Los Angeles. STUDY DESIGN: A four-year prospective study of resource utilization among 1,681 patients under treatment for hypertension, diabetes, myocardial infarction, or congestive heart failure in the practices of 367 clinicians. DATA COLLECTION/EXTRACTION METHODS: Insurance payment system (PPD or FFS), hospitalizations, and office visits were obtained from patient reports. Disease and severity indicators, sociodemographics, and self-reported functional status were used to adjust for patient mix and to compute expected utilization rates. PRINCIPAL FINDINGS: Compared to FFS, PPD patients had 31 percent fewer observed hospitalizations before adjustment for patient differences (p = .005) and 15 percent fewer hospitalizations than expected after adjustment (p = .078). The observed rate of FFS hospitalizations exceeded the expected rate by 9 percent. These results are not explained by system differences in patient mix or trends in hospital use over four years. Half of the PPD/FFS difference in hospitalization rate is due to intrinsic characteristics of the payment system itself. CONCLUSIONS: PPD patients with chronic medical conditions followed prospectively over four years, after extensive patient-mix adjustment, had 15 percent fewer hospitalizations than their FFS counterparts owing to differences intrinsic to the insurance reimbursement system.

Starting insulin therapy in patients with type 2 diabetes: effectiveness, complications, and resource utilization.

CONTEXT: Although experimental studies show that insulin therapy can be safe and efficacious in improving glycemic control in type 2 diabetes under optimal conditions (ie, using patient volunteers with close monitoring under strict study protocols), little is known about its effectiveness, complication rates, and associated resource utilization in actual clinical practice. DESIGN: Cohort study. SETTING: Large staff-model health maintenance organization. PARTICIPANTS: A total of 8668 patients with type 2 diabetes cared for by generalist physicians from 1990 through 1993. OUTCOME MEASURES: Resource use (hospitalizations, outpatient visits, laboratory testing, and home glucose monitoring) and glycemic control were evaluated using combined clinical, survey, and administrative information systems data. Detailed clinical case-mix data, including a newly validated case-mix method, the Total Illness Burden Index, were collected on a subsample of 1738 patients. RESULTS: Among patients starting insulin therapy, hemoglobin A1c (HbA1c) decreased by 0.9 percentage point (95% confidence interval, 0.7-1.0) at 1 year compared with those receiving stable medication regimens; however, 2 years after starting insulin therapy, 60% still had HbA1c levels of 8% or greater. There was no evidence that some primary care physicians achieved better results than other primary care physicians when starting insulin therapy in their patients. Patients with the poorest baseline glycemic control achieved substantially greater HbA1c reductions; those with a baseline HbA1c level of 13% had a 3-fold greater decline in HbA1c than those whose baseline HbA1c level was 9%. For a subset of all patients for whom detailed clinical case-mix data were obtained, those taking insulin had higher resource use than those taking sulfonylureas, independent of illness severity. After adjusting for age, sex, race, socioeconomic status, disease duration, and severity of diabetes and comorbidities, insulin users had slightly more laboratory tests performed, 2.4 more outpatient visits per year, and almost 300 more fingersticks for home glucose testing per year compared with sulfonylurea users (all P<.01). Although 15% of patients receiving insulin therapy reported weekly symptoms of hypoglycemia, insulin therapy was not associated with an increase in emergency department visits (after case-mix adjustment) and resulted in only 0.5 hypoglycemia-related hospitalizations per 100 patient-years. CONCLUSIONS: For patients with type 2 diabetes who were cared for by generalist physicians, starting insulin therapy was generally safe and effective in achieving moderate glycemic control in patients who initially had poor glycemic control. However, insulin therapy was associated with increases in resource use and was rarely effective in achieving tight glycemic control, even for those with moderate control.

The cost to health plans of poor glycemic control.

OBJECTIVE: We tested the hypothesis that level of glycemic control is related to medical care costs in adults with diabetes. RESEARCH DESIGN AND METHODS: Regression analysis was used to estimate the relationship between glycemic control and medical care charges for 3,017 adults with diabetes who were continuously enrolled in a large health maintenance organization (HMO) over a 4-year period. Diagnosis of diabetes was ascertained from diagnostic and pharmaceutical databases using a method with an estimated sensitivity of 0.91 and an estimated specificity of 0.99. Charges for care included defined outpatient and inpatient services. Patients who disenrolled or who died during the 4-year period were excluded from the main analysis. RESULTS: Charges for medical care for patients with diabetes from 1993 to 1995 were closely related to HbA1c level in 1992 before and after adjustment for age, sex, coronary heart disease, and hypertension. Standardized 3-year estimates of charges ranged from $10,439 for patients without comorbid conditions to $44,417 for those with heart disease and hypertension. Medical care charges increased significantly for every 1% increase above HbA1c of 7%. For a person with an HbA1c value of 6%, successive 1% increases in HbA1c resulted in cumulative increases in charges of approximately 4, 10, 20, and 30%. The increase in charges accelerated as the HbA1c value increased. For patients with diabetes only, or with diabetes plus other chronic conditions, the rate of increase in charges with HbA1c was consistent. CONCLUSIONS: HbA1c provides useful information to providers and patients regarding both health status and future medical care charges. Economic data suggest that clinicians should assign high importance to low HbA1c results and aggressively maintain the HbA1c status of patients who have low HbA1c values. For economic as well as clinical reasons, it may be beneficial to lower HbA1c when it is > 8% and to reduce cardiovascular risk factors. The medical charge data suggest that investment in clinical systems to improve diabetes care may benefit both payers and patients.