E H Blackstone1. 1. Department of Surgery, University of Alabama at Birmingham 35294-0007, USA.
Abstract
BACKGROUND: The hazard function is the instantaneous rate of occurrence of a time-related event, such as death. Methods exist to determine the hazard function from clinical outcome (follow-up) data, to identify risk factors for higher hazard, and to generate patient-specific predictions. In this article, hazard function methodology is extended to the arena of healthcare costs. METHODS: Using a compartmental analogue, a hazard function method is introduced to analyze patient-specific distributions of hospital costs, charges, or length of stay when these are truncated by the competing risk of in-hospital death. Another hazard function method is proposed to analyze time-related healthcare costs, exploiting weighted repeated events theory. RESULTS: Successful application of the hospital costs methodology in the hazard function domain is demonstrated for length-of-stay data after coronary artery bypass grafting. A previously reported analysis of variable morbidity from repeated thromboembolic events illustrates the weighted events methods. CONCLUSIONS: Existing hazard function methodology can be extended to analyze healthcare costs, either of a single episode of care or of multiple time-related episodes.
BACKGROUND: The hazard function is the instantaneous rate of occurrence of a time-related event, such as death. Methods exist to determine the hazard function from clinical outcome (follow-up) data, to identify risk factors for higher hazard, and to generate patient-specific predictions. In this article, hazard function methodology is extended to the arena of healthcare costs. METHODS: Using a compartmental analogue, a hazard function method is introduced to analyze patient-specific distributions of hospital costs, charges, or length of stay when these are truncated by the competing risk of in-hospital death. Another hazard function method is proposed to analyze time-related healthcare costs, exploiting weighted repeated events theory. RESULTS: Successful application of the hospital costs methodology in the hazard function domain is demonstrated for length-of-stay data after coronary artery bypass grafting. A previously reported analysis of variable morbidity from repeated thromboembolic events illustrates the weighted events methods. CONCLUSIONS: Existing hazard function methodology can be extended to analyze healthcare costs, either of a single episode of care or of multiple time-related episodes.