With health care costs increasing and economic resources diminishing, substantial efforts have been directed toward improving the quality of care delivered in a cost-effective manner. For a total hip arthroplasty (THA) performed in the United States between 1997 and 2001, total hospital cost, including direct and indirect costs, was estimated as averaging $13,339.1 In 2012, this cost was estimated to be between $43,000 and $100,000.2 This overall cost estimate, along with the rate at which the procedure is performed, may present an opportunity for cost savings.
Length of hospital stay (LHS) is an important outcome measure that has been assessed for optimal health care delivery. Prolonged LHS implies increased resource expenditure. Therefore, it is crucial to identify factors associated with prolonged LHS in order to reduce costs. Investigations have identified factors shown to affect LHS after THA. These factors include advanced age, medical comorbidities, obesity, intraoperative time, anesthesia technique, surgical site infection, and incision length.3-7
We conducted a study to identify the patient and clinical factors that affect LHS and to determine whether the specific day of the week when primary THA is performed affects LHS at a large tertiary-care university-based medical center. This information may prove valuable to hospital planning committees allotting operating room time and floor staffing for elective surgical cases with the goal of delivering cost-efficient care.
Materials and Methods
After obtaining institutional review board approval for this study, we retrospectively analyzed all primary unilateral THAs (273 patients) performed at our institution, a tertiary-care teaching hospital, between January 2010 and May 2011. The majority of the surgeries were performed through a posterior approach, and a majority of the implants were uncemented. All patients followed the same postoperative clinical pathway; no fast-track pathway was used.
The combined effects of day of surgery, American Society of Anesthesiologists (ASA) grade, anesthesia type, intraoperative time, estimated blood loss (EBL), incision length, presence of complications, age, sex, body mass index (BMI), disposition (skilled nursing facility vs home), transfusion, hematocrit, and hemoglobin on LHS were analyzed using a multiple quasi-Poisson regression model that included a random effect for surgeon. A Poisson regression model (typically used for count data) was deemed appropriate, as LHS was reported in whole days; a quasi-Poisson model relaxes the Poisson model assumption that the variance in the data equals the mean. The random effect for surgeon adjusts for any correlation among data from surgeries conducted by the same surgeon.
All complications were recorded. Complications included excess wound drainage,8 wound hematoma (a case of excess wound drainage necessitated surgical irrigation and débridement), new-onset atrial fibrillation, non-ST-elevation myocardial infarction, atrial flutter, urinary tract infection, pulmonary embolism, disseminated intravascular coagulation, hepatic decompensation as manifested by elevated liver enzymes, pneumonia, gastroesophageal reflux disease, gastric ulcer, sepsis, delirium, hypotension, and dysphagia.
The parameter estimates reported from the quasi-Poisson regression model are incident rate ratios (IRRs). IRR represents the change in expected LHS for a 1-unit change in a continuous variable (eg, age) or between categories of a categorical variable (eg, sex). IRR higher than 1 indicates higher risk as the continuous variable increases or a higher risk relative to the comparator group for a categorical variable. IRR lower than 1 indicates lower risk.
Results
Table 1 summarizes patient characteristics by surgical day. Mean LHS ranged from a minimum of 3.7 days for patients who had surgery on a Monday to a maximum of 4.2 days for patients who had surgery on a Thursday.
Table 2 summarizes results of the multivariate quasi-Poisson regression analysis of LHS by surgical day, ASA grade, anesthesia type, intraoperative time, EBL, incision length, presence of complications, age, sex, and BMI. With all other variables included in the model adjusted for, each additional point in ASA grade was associated with a 12% increase in LHS (P = .019). In addition, with all other variables included in the model adjusted for, LHS was 33% longer for patients with complications than for patients without complications (P < .001) and 12% longer for patients who received transfusions than for patients who did not (P = .046). LHS did not differ significantly by the day of the week when the surgery was performed (P = .496). Disposition status (skilled nursing facility vs home) as a variable to determine LHS did approach statistical significance (P = .061). As the effect size we were interested in detecting was an approximate 1-day increase in LHS for patients who had surgery later in the week relative to patients who had surgery earlier in the week, our sample size was adequate (range of required sample size, 200-300 patients). This study had 99% power to detect a 27% increase in LHS (equivalent to 1 day or more).
