METHODS We reviewed the cost-accounting records of all surviving patients discharged from our hospital during fiscal year 1998 with LOS of 4 days or more (n = 12,365). Costs were measured using the University of Michigan's general ledger and activity-based cost-accounting system, which is a conventional framework that uses software from market leader TSI (Eclipsys; Delray Beach, FL). Patient costs were tracked on a daily basis and further decomposed into three categories. Variable direct costs measure expenditures that can be identified directly with the care of individual patients on a particular day, such as laboratory tests, radiographs, and disposable supplies. Fixed direct costs capture expenditures that can be identified with a specific hospital department but not with a particular patient, and indirect costs lie entirely outside individual departments. Examples of fixed direct costs include equipment and medical devices used to care for trauma patients. Examples of indirect costs include the admissions area and the chief executive officer's salary. Taken together, fixed direct and indirect costs are collectively referred to as "hospital overhead.'' Indirect overhead costs are allocated as follows. After being identified as such, they are grouped into pools of like funds (eg, buildings, facilities, administrative). Hospital finance proceeds to determine an allocation metric (eg, square footage in the case of buildings) and the direct departments designated to receive this overhead. The indirect costs are then allocated to individual departments by a simultaneous equations method, so that the order of allocation is immaterial. Once the departmental allocation is determined, that pool of expenses is allocated to specific patient-billed services based upon direct costs. In the chemical laboratory, for example, if the direct cost for a glucose test is $1 and the direct cost for a liver enzyme test is $6, then the liver enzyme service is assigned six times more overhead expense than the glucose. This method of activity-based cost accounting is conventional. This study scrutinizes the physician-controllable variable direct costs per patient because in the short run, overhead is beyond physicians' immediate control. Unlike the Cook County Hospital analysis,8 our study categorizes nursing as a variable direct cost. We include nursing among the categories that physicians can influence. Because nursing constitutes the majority of end-of-stay variable direct costs, treating nursing expenditures as a fixed cost (which they may be over the short term) would effectively halve the results that follow. Individual patient costs were broken out on a daily basis. We omitted all patients who died or were discharged against medical advice (on the grounds that physicians did not determine the timing of these patients' discharges) and those patients with LOS of 3 days or less. The population that remained was then stratified by LOS and subsequently divided into subgroups of patients who had major surgery and patients who did not. To gain a more focused perspective, we also examined all 665 patients discharged from the hospital's level I trauma center. Within this group, nine activities together accounted for more than 99% of total costs: nursing (42.8% of total cost), surgical services (12.7%), laboratory (9.4%), radiology (8.4%), pharmacy (8.3%), emergency services (7.9%), respiratory and pulmonary (5.1%), rehabilitation services (3.3%), and supplies (1.3%). The entire patient population was first analyzed by looking at the first 3 days and the last 3 days of the patients' hospital stays. We then dropped patients who died or were discharged against medical advice. Within this group, separate analyses were conducted for patients who spent at least 7 days in the hospital and had a minimum of 3 days in the ICU, and for patients with LOS of 4 days or more. Statistical analyses were performed using the two-tailed Student's t-test with p values denoting levels of statistical significance (p < 0.05). Mean ± SD is reported. Introduction | Methods | Results | Discussion | References   JACS