Statistical Distribution of VA (Veterans Administration) Hospital Stay Data: Average or Median Length of Stay

摘要

Hospital length of stay is a widely used quantitative, operational measure of health care practice and policy, is used as a monitor of resource consumption and clinical efficacy of treatments, and serves as a tool in utilization review. The use of the traditional average value of a set of hospitalizations as a summary descriptor for length-of-stay information is contingent upon the meeting of the statistical assumption that the values fit a normal distribution. Using fiscal year 1983 Patient Treatment File records from five affiliated VA Medical Centers in Medical District 20 (Texas and Oklahoma), analyses of lengths-of-stay data show that the distributions are in fact highly skewed (asymmetrical) because of the presence of a (small) percentage of extremely long stays which offset a core frequency of smaller values. This asymmetry was independent of how the data were grouped -- that is,individual diagnosis, diagnosis related groupings, major diagnostic category, individual hospital, or group of hospitals. The long stays tend to 'inflate' the average such that it is not the best representative value of the data set. The findings were similar when the analyses were run on just those records which fit the Commission on Professional and Hospital Activities' definition (stay of less than 100 days, 'regular' source of admission and discharge).