Arthur J. Hartz

A comparison of observational studies and randomized, controlled trials.

BACKGROUND: For many years it has been claimed that observational studies find stronger treatment effects than randomized, controlled trials. We compared the results of observational studies with those of randomized, controlled trials. METHODS: We searched the Abridged Index Medicus and Cochrane data bases to identify observational studies reported between 1985 and 1998 that compared two or more treatments or interventions for the same condition. We then searched the Medline and Cochrane data bases to identify all the randomized, controlled trials and observational studies comparing the same treatments for these conditions. For each treatment, the magnitudes of the effects in the various observational studies were combined by the Mantel-Haenszel or weighted analysis-of-variance procedure and then compared with the combined magnitude of the effects in the randomized, controlled trials that evaluated the same treatment. RESULTS: There were 136 reports about 19 diverse treatments, such as calcium-channel-blocker therapy for coronary artery disease, appendectomy, and interventions for subfertility. In most cases, the estimates of the treatment effects from observational studies and randomized, controlled trials were similar. In only 2 of the 19 analyses of treatment effects did the combined magnitude of the effect in observational studies lie outside the 95 percent confidence interval for the combined magnitude in the randomized, controlled trials. CONCLUSIONS: We found little evidence that estimates of treatment effects in observational studies reported after 1984 are either consistently larger than or qualitatively different from those obtained in randomized, controlled trials.

Hospital Characteristics and Mortality Rates

The Health Care Financing Administration (HCFA) publishes hospital mortality rates each year. We undertook a study to identify characteristics of hospitals associated with variations in these rates. To do so, we obtained data on 3100 hospitals from the 1986 HCFA mortality study and the American Hospital Associations 1986 annual survey of hospitals. The mortality rates were adjusted for each hospitals case mix and other characteristics of its patients. The mortality rate for all hospitalizations was 116 per 1000 patients. Adjusted mortality rates were significantly higher for for-profit hospitals (121 per 1000) and public hospitals (120 per 1000) than for private not-for-profit hospitals (114 per 1000; P less than 0.0001 for both comparisons). Osteopathic hospitals also had an adjusted mortality rate that was significantly higher than average (129 per 1000; P less than 0.0001). Private teaching hospitals had a significantly lower adjusted mortality rate (108 per 1000) than private nonteaching hospitals (116 per 1000; P less than 0.0001). Adjusted mortality rates were also compared for hospitals in the upper and lower fourths of the sample in terms of certain hospital characteristics. The mortality rates were 112 and 121 per 1000 for the hospitals in the upper and lower fourths, respectively, in terms of the percentage of physicians who were board-certified specialists (P less than 0.0001), 112 and 120 per 1000 for occupancy rate (P less than 0.0001), 113 and 120 per 1000 for payroll expenses per hospital bed (P less than 0.0001), and 113 and 119 per 1000 for the percentage of nurses who were registered (P less than 0.0001).

Interstitial Pneumonitis After Bone Marrow Transplantation: Assessment of Risk Factors

Data from 932 patients with leukemia who received bone marrow transplants were analyzed to determine factors associated with an increased risk of developing interstitial pneumonitis. Interstitial pneumonitis developed in 268 patients for a 2-year actuarial incidence of 35 +/- 4% (SD) and with a mortality rate of 24%. Six factors were associated with an increased risk: use of methotrexate rather than cyclosporine after transplantation (relative risk, 2.3; p less than 0.0002); older age (relative risk, 2.1; p less than 0.0001); presence of severe graft-versus-host disease (relative risk, 1.9; p less than 0.003); long interval from diagnosis to transplantation (relative risk, 1.6; p less than 0.002); performance ratings before transplantation of less than 100% (relative risk, 2.1; p less than 0.0001); and high dose-rates of irradiation in patients given methotrexate after transplantation (relative risk, 3.2; p less than 0.03). The risk of developing interstitial pneumonitis ranged from 8% in patients with none of these adverse risk factors to 94% in patients with all six. These findings may help to identify patients at high risk for this complication.

Relationship of obesity to diabetes: Influence of obesity level and body fat distribution

The relationship of clinical diabetes to body fat distribution and obesity level was examined in 15,532 women. After adjusting for relative weight, all upper body segment girth measurements (neck, bust, and waist) had strong positive associations with diabetes. In contrast, the lower body segment girth measurement (hips) had an equally strong but inverse association with diabetes. Based upon waist-to-hip girth ratio, women were divided into four subgroups. The prevalence of diabetes increased with increasing values of this ratio. Women in the upper quartile had about three times the prevalence of diabetes as women of comparable obesity level in the lowest quartile. Women with both upper body fat predominance and severe obesity had a relative risk of diabetes 10.3 times as great as nonobese subjects with lower body fat predominance. The results suggest that localization of fat in the upper body segment and severe obesity are two distinct additive risks for diabetes.

Hospitalization Costs Associated with Homelessness in New York City

BACKGROUND Homelessness is believed to be a cause of health problems and high medical costs, but data supporting this association have been difficult to obtain. We compared lengths of stay and reasons for hospital admission among homeless and other low-income persons in New York City to estimate the hospitalization costs associated with homelessness. METHODS We obtained hospital-discharge data on 18,864 admissions of homeless adults to New York Citys public general hospitals (excluding admissions for childbirth) and 383,986 nonmaternity admissions of other low-income adults to all general hospitals in New York City during 1992 and 1993. The differences in length of stay were adjusted for diagnosis-related group, principal diagnosis, selected coexisting illnesses, and demographic characteristics. RESULTS Of the admissions of homeless people, 51.5 percent were for treatment of substance abuse or mental illness, as compared with 22.8 percent for the other low-income patients, and another 19.7 percent of the admissions of homeless people were for trauma, respiratory disorders, skin disorders, and infectious diseases (excluding the acquired immunodeficiency syndrome [AIDS]), many of which are potentially preventable medical conditions. For the homeless, 80.6 percent of the admissions involved either a principal or a secondary diagnosis of substance abuse or mental illness -- roughly twice the rates for the other patients. The homeless patients stayed 4.1 days, or 36 percent, longer per admission on average than the other patients, even after adjustments were made for differences in the rates of substance abuse and mental illness and other clinical and demographic characteristics. The costs of the additional days per discharge averaged

Clinical signs of androgen excess as risk factors for coronary artery disease

4,094 for psychiatric patients,

The association of obesity with joint pain and osteoarthritis in the HANES data

3,370 for patients with AIDS, and

A Cluster Randomized Trial to Evaluate Physician/Pharmacist Collaboration to Improve Blood Pressure Control

2,414 for all types of patients. CONCLUSIONS Homelessness is associated with substantial excess costs per hospital stay in New York City. Decisions to fund housing and supportive services for the homeless should take into account the potential of these services to reduce the high costs of hospitalization in this population.

The association of body weight with recurrent cancer of the breast

Women with androgen excess have been found to have higher triglycerides and lower high-density lipoprotein cholesterol concentrations than nonhirsute women and are presumed to be at increased risk of coronary artery disease. However, definitive data linking androgen excess with coronary artery disease is lacking. We evaluated 102 women coming to coronary artery catheterization for signs and symptoms of androgen excess. Hirsutism was found more commonly in those women with confirmed coronary artery disease (chi 2 = 10.11). Waist:hip ratio (an index of android fat distribution) was associated with hirsutism (by ANOVA, F-test) and with coronary artery disease (t-test). The strongest associations were found in older women (aged greater than or equal to 60). Androgen excess in women may signal risk for coronary artery disease.

Mechanisms of Intraocuular Pressure Elevation after Pars Plana Vitrectomy

Data from 4225 persons from the National Health and Nutrition Examination Survey (HANES) was used to determine whether obesity was associated with osteoarthritis (OA) or joint pain. Subjects were divided into four groups on the basis of sex and race. We found that obesity was associated with OA of the knee for each sex/race group (p less than 0.01). The association was strongest for women, and it was present even for subjects without evidence of knee pain on physical examination. Frame size was not significantly associated with OA of the knee. Relative weight was weakly associated with OA of the hips in white women and nonwhite men but not significantly associated with OA of the sacroiliac joint. Diabetes did not seem to be an important risk factor for OA. These results suggest that the additional mechanical stress resulting from obesity is the principal reason for the association between obesity and OA.