L A Green1, M T Ruffin. 1. Department of Family Practice, University of Michigan Medical School, Ann Arbor 48104.
Abstract
BACKGROUND: Past studies have conflicted regarding the existence of sex bias in the treatment of women with ischemic cardiac disease. This study explored the effect of different analytic models on conclusions about sex bias. METHODS: A retrospective analysis of medical records was performed on 787 patients evaluated for potential acute cardiac ischemia in the emergency departments of two nonteaching community hospitals. The Acute Coronary Ischemia Time Insensitive Predictive Instrument (ACI-TIPI) was used to estimate the likelihood of ischemic disease. The decisions to admit to hospital, not to admit to hospital, and to discharge with diagnosis of myocardial infarction were the outcome variables. RESULTS: Logistic regression models of increasing levels of detail were applied and evaluated. Analysis using summary data (similar to discharge abstracts or claims data) revealed that patient sex affected admission decisions, but an analysis of clinically detailed data by hospital was required to reveal the nature of the effect. There was disparity in admission decisions by sex at one hospital but not at the other. The odds ratio for admission (women vs men) was 0.546 (95% CI, 0.33 to 0.91) at Hospital A, and 1.22 (95% CI, 0.72 to 2.05) at Hospital B. This disparity appeared to be related to a high rate of admission (67%) among men with low (< 10%) probability of acute ischemia. CONCLUSIONS: Differences in treatment of suspected acute cardiac ischemia by sex may be a practice variation phenomenon rather than a uniform bias. When these differences occur, they may represent overtreatment of men rather than inadequate treatment of women. Because summary or billing datasets lack clinical detail, they are inadequate for the study of physician decision-making.
BACKGROUND: Past studies have conflicted regarding the existence of sex bias in the treatment of women with ischemiccardiac disease. This study explored the effect of different analytic models on conclusions about sex bias. METHODS: A retrospective analysis of medical records was performed on 787 patients evaluated for potential acute cardiac ischemia in the emergency departments of two nonteaching community hospitals. The Acute Coronary Ischemia Time Insensitive Predictive Instrument (ACI-TIPI) was used to estimate the likelihood of ischemic disease. The decisions to admit to hospital, not to admit to hospital, and to discharge with diagnosis of myocardial infarction were the outcome variables. RESULTS: Logistic regression models of increasing levels of detail were applied and evaluated. Analysis using summary data (similar to discharge abstracts or claims data) revealed that patient sex affected admission decisions, but an analysis of clinically detailed data by hospital was required to reveal the nature of the effect. There was disparity in admission decisions by sex at one hospital but not at the other. The odds ratio for admission (women vs men) was 0.546 (95% CI, 0.33 to 0.91) at Hospital A, and 1.22 (95% CI, 0.72 to 2.05) at Hospital B. This disparity appeared to be related to a high rate of admission (67%) among men with low (< 10%) probability of acute ischemia. CONCLUSIONS: Differences in treatment of suspected acute cardiac ischemia by sex may be a practice variation phenomenon rather than a uniform bias. When these differences occur, they may represent overtreatment of men rather than inadequate treatment of women. Because summary or billing datasets lack clinical detail, they are inadequate for the study of physician decision-making.