Early antidepressant therapy for elderly patients.

PURPOSE: We studied factors affecting the management of depression in older patients, especially the use of early antidepressant therapy. METHODS: We recruited 128 primary care physicians to view one version of a 5-minute videotape of an elderly patient with somatic symptoms that were suggestive of depression, and to complete an interview that assessed decision making. Using an experimental factorial design, 16 versions of the videotape were produced, holding constant the clinical features of the case, while varying the patient's age, race, sex, and socioeconomic status. Dependent variables were the physicians' probability assessment of depression and the recommendation of antidepressant medication after the first visit. RESULTS: Depression was considered a possible diagnosis by 121 physicians (95%) and the most likely diagnosis by 69 (54%). Sixteen physicians (13%) recommended antidepressant therapy after the first visit, and they were less likely than other physicians to order initial laboratory tests to assess the possibility of other conditions. Recommendations for antidepressant therapy was not associated with patient age, sex, race, or socioeconomic status, or with physician sex, race, or experience. Family physicians were more likely than internists to recommend an antidepressant (19% [12/64] vs. 6% [4/64], P = 0.04). CONCLUSION: Based on a 5-minute vignette, physicians were likely to recognize depression, independent of patient characteristics. Those recommending early antidepressant therapy were more likely to be in family medicine and less likely to investigate other diagnoses initially.

A meta-analysis of quality-of-life estimates for stroke.

BACKGROUND: Researchers performing cost-effectiveness analyses often incorporate quality-of-life (QOL) estimates. OBJECTIVE: To aid analysts, we performed a meta-analysis to estimate quality of life for minor, moderate, and major stroke and assessed the relative importance of study design characteristics in predicting the quality of life of patients with stroke. METHODS: Through a systematic search we identified 20 articles reporting 53 unique QOL weights for stroke. Each article was read and QOL weights and study characteristics were recorded. We used a hierarchical linear model (HLM) to perform a meta-regression. The model included severity of stroke, elicitation method, respondents, and QOL scale bounds as explanatory variables. RESULTS: Severity of stroke (p < 0.0001) and the bounds of the scale (p = 0.0015) were significant predictors of quality of life, while the elicitation method and respondents were not. Pooling QOL weights using the HLM model, we estimated a quality of life of 0.52 for major stroke, 0.68 for moderate stroke, and 0.87 for minor stroke if the time trade-off method is used to assess quality of life from community members when the scale bounds range from death to perfect health. CONCLUSIONS: We found no systematic difference in stroke QOL weights depending on elicitation method or respondents. However, quality of life is sensitive to the bounds of the scale. Because the pooled QOL estimates reported here are based on a comprehensive review of the QOL literature for stroke, they should be of great use to researchers performing cost-utility analyses of interventions designed to prevent or treat stroke, or where stroke is a possible side effect of therapy.

A meta-analysis of utility estimates for HIV/AIDS.

The authors performed a meta-analysis to derive pooled utilities for HIV/AIDS and to assess the relative importance of study design characteristics in predicting utilities. Twenty-five articles were identified reporting 74 unique utilities elicited from 1956 respondents. The authors used a hierarchical linear model to perform the meta-analysis, with disease stage, elicitation method, respondent type, and the upper-bound and lower-bound labels for the utility scale as the independent variables. Disease stage (P = 0.016) and respondent type (P = 0.014) were significant predictors of utility. Elicitation method was of marginal significance (P = 0. 052). Bounds were not significant. Pooling utilities, the authors estimate a utility of 0.70 for AIDS, 0.82 for symptomatic HIV and 0.94 for asymptomatic HIV when the time tradeoff method is used to elicit utilities from patients and the scale ranges from death to perfect health. The pooled utilities reported here should be of great use to researchers performing cost-utility analyses of interventions for HIV/AIDS.