Arnold M. Epstein

The Relation between Health Insurance Coverage and Clinical Outcomes among Women with Breast Cancer

Background Women without private health insurance are less likely than privately insured women to be screened for breast cancer, and their treatment may differ after cancer is diagnosed. In this study we addressed two related questions: Do uninsured patients and those covered by Medicaid have more advanced breast cancer than privately insured patients when the disease is initially diagnosed? And, for each stage of disease, do uninsured patients and patients covered by Medicaid die sooner after breast cancer is diagnosed than privately insured patients? Methods we studied 4675 women, 35 to 64 years of age, in whom invasive breast cancer was diagnosed from 1985 through 1987, by linking New Jersey State Cancer Registry records to hospital-discharge data. We compared the stage of disease and stage-specific survival among women with private insurance, no insurance, and Medicaid coverage through June 1992. We also estimated the adjusted risk of death for these groups, using proportional-hazards regression analy...

RACIAL DISPARITIES IN ACCESS TO RENAL TRANSPLANTATION: Clinically Appropriate or Due to Underuse or Overuse?

BACKGROUND Despite abundant evidence of racial disparities in the use of surgical procedures, it is uncertain whether these disparities reflect racial differences in clinical appropriateness or overuse or underuse of inappropriate care. METHODS We performed a literature review and used an expert panel to develop criteria for determining the appropriateness of renal transplantation for patients with end-stage renal disease. Using data from five states and the District of Columbia on patients who had started to undergo dialysis in 1996 or 1997, we selected a random sample of 1518 patients (age range, 18 to 54 years), stratified according to race and sex. We classified the appropriateness of patients as data on candidates for transplantation and analyzed rates of referral to a transplantation center for evaluation, placement on a waiting list, and receipt of a transplant according to race. RESULTS Black patients were less likely than white patients to be rated as appropriate candidates for transplantation according to appropriateness criteria based on expert opinion (71 blacks [9.0 percent] vs. 152 whites [20.9 percent]) and were more likely to have had incomplete evaluations (368 [46.5 percent] vs. 282 [38.8 percent], P<0.001 for the overall chi-square). Among patients considered to be appropriate candidates for transplantation, blacks were less likely than whites to be referred for evaluation, according to the chart review (90.1 percent vs. 98.0 percent, P=0.008), to be placed on a waiting list (71.0 percent vs. 86.7 percent, P=0.007), or to undergo transplantation (16.9 percent vs. 52.0 percent, P<0.001). Among patients classified as inappropriate candidates, whites were more likely than blacks to be referred for evaluation (57.8 percent vs. 38.4 percent), to be placed on a waiting list (30.9 percent vs. 17.4 percent), and to undergo transplantation (10.3 percent vs. 2.2 percent, P<0.001 for all three comparisons). CONCLUSIONS Racial disparities in rates of renal transplantation stem from differences in clinical characteristics that affect appropriateness as well as from underuse of transplantation among blacks and overuse among whites. Reducing racial disparities will require efforts to distinguish their specific causes and the development of interventions tailored to address them.

Delayed Access to Health Care: Risk Factors, Reasons, and Consequences

OBJECTIVE To determine characteristics of patients reporting delays in care before hospitalization and the reasons for those delays. DESIGN Survey; personal interviews. SETTING Five hospitals in Massachusetts. PATIENTS Subjects were drawn from a consecutive sample of all adult patients (excluding obstetrics or psychiatry patients) hospitalized during the first 6 months of 1987 as part of a larger study of hospital costs. For the current study, if patients were re-admitted, we included in our analysis only data on the first admission during the study period. We obtained usable survey data from 12,068 of 17,231 eligible patients. RESULTS Delays in care were reported by 16% of patients. The odds of reporting delays in care among patients who were black, poor, uninsured, or without a regular physician were 40% to 80% greater than those for other patients (P less than 0.01). Most patients who reported delays thought that their problem was not serious (64%). Cost was an important factor in delaying care for patients in lower socioeconomic positions; the odds of delaying care because of cost for patients who were both poor and uninsured were 12 times greater than the odds for other patients (P less than 0.001). After controlling for diagnosis-related groups (DRGs) and severity, patients who reported delays had 9% longer hospital stays compared with others (P less than 0.001). CONCLUSIONS Patients generally thought to be disadvantaged are at especially high risk for delaying care for conditions that eventually lead to hospitalization. Because these delays are associated with longer hospital stays and potentially poorer health outcomes, interventions that reduce delays seem especially important.

Cost-effectiveness and cost-benefit analyses in the medical literature. Are the methods being used correctly?

OBJECTIVE To determine whether published cost-effectiveness and cost-benefit analyses have adhered to basic analytic principles. DESIGN Structured methodologic review of published articles. STUDY SAMPLE Seventy-seven articles published either from 1978 to 1980 or from 1985 to 1987 in general medical, general surgical, and medical subspecialty journals. MAIN OUTCOME MEASUREMENTS Articles were reviewed to assess the use and reporting of six fundamental principles of analysis. These principles were derived by reviewing widely cited textbooks and articles describing the methods for performing economic analyses and by selecting the methods universally recommended. MAIN RESULTS Overall performance was only fair. Three articles adhered to all six principles, and the median number of principles to which articles adhered was three. Among the problems noted were failure to make underlying assumptions explicit and, therefore, verifiable, and failure to test assumptions with sensitivity analyses. No improvement in performance was observed between 1978 and 1987. Articles in general medical journals, however, were more likely to use analytic methods appropriately than articles in the general surgical or medical subspecialty literature. CONCLUSIONS Greater attention should be devoted to ensuring the appropriate use of analytic methods for economic analyses, and readers should make note of the methods used when interpreting the results of economic analyses.

Results of the National Initiative for Cancer Care Quality: How Can We Improve the Quality of Cancer Care in the United States?

PURPOSE In 1999, the National Cancer Policy Board called attention to the quality of cancer care in the United States and recommended establishing a quality monitoring system with the capability of regularly reporting on the quality of care for patients with cancer. METHODS Using data from a patient survey 4 years after diagnosis and review of medical records, we determined the percentage of stage I to III breast cancer and stage II to III colorectal cancer survivors in five metropolitan statistical areas (MSAs) across the United States who received recommended care specified by a comprehensive set of explicit quality measures. RESULTS Two thousand three hundred sixty-six (63%) of 3,775 eligible patients responded to the survey, and 85% consented to have their medical records reviewed. Our final analytic sample (n = 1,765) included 47% of the eligible patients. Patients with breast and colorectal cancer received 86% of recommended care (95% CI, 86% to 87%) and 78% of recommended care (95% CI, 77% to 79%), respectively. Adherence to quality measures was less than 85% for 18 of the 36 breast cancer measures, and significant variation across MSAs was observed for seven quality measures. The percent adherence was less than 85% for 14 of the 25 colorectal cancer measures, and one quality measure demonstrated statistically significant variation across the MSAs. CONCLUSION Initial management of patients with breast and colorectal cancer in the United States seemed consistent with evidence-based practice; however, substantial variation in adherence to some quality measures point to significant opportunities for improvement.

Using proxies to evaluate quality of life. Can they provide valid information about patients' health status and satisfaction with medical care?

Instruments using interview data to measure health status have been increasingly used to measure patient outcomes. To assess the potential utility of proxy responses about health status when subjects are unable to respond, the authors compared the responses of 60 subject and proxy pairs on instruments measuring overall current health, functional status, social activity, emotional health, and satisfaction with medical care. Proxies were asked to respond as they thought the subject would. Subject and proxy responses were strongly correlated with each other for overall health, functional status, social activity, and emotional health (P < .001), and moderately correlated for satisfaction (P < .005). Proxies reported lower emotional health and satisfaction than did subjects (P < .005). Proxy and subject mean responses were generally similar for overall health, functional status, and social activity. However, those proxies who spent more time per week helping the subject rated the subjects functional status and social activity as more impaired than did the subject (P <.05). Subjects who had poorer overall health tended to rate their health relatively lower than did the proxies (P < .05). These results suggest that use of proxies intermingled with subjects to measure health status through interview may lead to biased results.

A causal model of health status and satisfaction with medical care.

Patients with better health status have often been shown to be more satisfied with their medical care, but the causal factors in this relation have not been determined. In this study, a longitudinal assessment of these two constructs was undertaken in which older patients in a health maintenance organization were interviewed at baseline (N=590) and again 1 year later (N=526) about their health status and satisfaction with their medical care. Structural equation modeling using LISREL procedures revealed that the predominant direction of causation went from earlier self-perceived overall health and functional ability to later levels of satisfaction. There was no evidence for causal paths going from satisfaction to later health. In addition, a test of spuriousness indicated that for self-perceived overall health, the significant longitudinal path was unlikely to be explained by unmeasured confounding variables.

Public Reporting of Discharge Planning and Rates of Readmissions

BACKGROUND A reduction in hospital readmissions may improve quality and reduce costs. The Centers for Medicare and Medicaid Services has initiated a national effort to measure and publicly report on the conduct of discharge planning. We know little about how U.S. hospitals perform on the current discharge metrics, the factors that underlie better performance, and whether better performance is related to lower readmission rates. METHODS We examined hospital performance on the basis of two measures of discharge planning: the adequacy of documentation in the chart that discharge instructions were provided to patients with congestive heart failure, and patient-reported experiences with discharge planning. We examined the association between performance on these measures and rates of readmission for congestive heart failure and pneumonia. RESULTS We found a weak correlation in performance between the two discharge measures (r=0.05, P<0.001). Although larger hospitals performed better on the chart-based measure, smaller hospitals and those with higher nurse-staffing levels performed better on the patient-reported measure. We found no association between performance on the chart-based measure and readmission rates among patients with congestive heart failure (readmission rates among hospitals performing in the highest quartile vs. the lowest quartile, 23.7% vs. 23.5%; P=0.54) and only a very modest association between performance on the patient-reported measure and readmission rates for congestive heart failure (readmission rates among hospitals performing in the highest quartile vs. the lowest quartile, 22.4% vs. 24.7%; P<0.001) and pneumonia (17.5% vs. 19.5%, P<0.001). CONCLUSIONS Our findings suggest that current efforts to collect and publicly report data on discharge planning are unlikely to yield large reductions in unnecessary readmissions.

Stroke in children and sickle-cell disease Baltimore-Washington Cooperative Young Stroke Study

Background/Purpose: The Baltimore-Washington Cooperative Young Stroke Study is the largest biracial urban-suburban population-based study to examine the etiology of strokes in children. Methods: We identified all children aged 1 to 14 years discharged from all 46 hospitals in central Maryland and Washington, DC with a diagnosis of ischemic stroke and intracerebral hemorrhage in the years 1988 and 1991. Each medical record was reviewed by two neurologists for appropriateness of the diagnosis of stroke and for information on the patients history, clinical presentation, pertinent investigations, hospital stay, and outcome at time of discharge. Results: Eighteen children with ischemic infarction and 17 with intracerebral hemorrhage were identified. The most common cause of ischemic stroke was sickle-cell disease (39%), followed by vasculopathic (33%) and indeterminate(28%) causes. Causes of intracerebral hemorrhages were arteriovenous malformation (29%), hematologic (23%), vasculopathy (18%), surgical complication (12%), coagulopathy (6%), and indeterminate (12%). The overall incidence for childhood stroke was 1.29 per 100,000 per year, with ischemic stroke occurring at a rate of 0.58 per 100,000 and intracerebral hemorrhage occurring at a rate of 0.71 per 100,000. The incidence of stroke among children with sickle-cell disease was estimated to be 0.28% or 285 per 100,000 per year. Conclusion: Sickle-cell disease plays a disproportionately high role in childhood stroke when a biracial population is surveyed.

Quality of care by race and gender for congestive heart failure and pneumonia.

BACKGROUND Variations in the rates of major procedures by race and gender are well described, but few studies have assessed the quality of care by race and gender for basic hospital services. OBJECTIVE To assess quality of care by race and gender. RESEARCH DESIGN Retrospective review of medical records. SUBJECTS Stratified random sample of 2,175 Medicare beneficiaries hospitalized for congestive heart failure or pneumonia in Illinois, New York, and Pennsylvania during 1991 and 1992. MEASURES Explicit process criteria and implicit review by physicians. RESULTS In adjusted analyses, black patients with congestive heart failure or pneumonia received lower quality of care overall than other patients with these conditions by both explicit process criteria and implicit review (P < 0.05). On explicit measures, overall quality of care did not differ by gender for either condition, but significant differences were noted on explicit subscales. Women received worse cognitive care than men from physicians for both conditions, better cognitive care from nurses for pneumonia, and better therapeutic care for congestive heart failure (P < 0.05). Women received worse quality of care than men by implicit review (P = 0.03) for congestive heart failure but not pneumonia. CONCLUSIONS Consistent racial differences in quality of care persist in basic hospital services for two common medical conditions. Physicians, nurses, and policy makers should strive to eliminate these differences. Gender differences in quality of care are less pronounced and may vary by condition and type of provider or service.