The Main Predictors of Length of Stay After Total Knee Arthroplasty

Background: Often, differences in length of stay after total knee arthroplasty are solely attributed to patient factors. Therefore, our aim was to determine the influence of patient-related and procedure or structural-related risk factors as predictors of length of stay after total knee arthroplasty. Methods: A prospective cohort of 4,509 patients (54.6% of whom had Medicare for insurance) underwent primary total knee arthroplasty across 4 facilities in a single health-care system (from January 1, 2016, to September 30, 2017). Risk factors were categorized as patient-related risk factors (demographic characteristics, smoking status, Veterans RAND 12 Item Health Survey Mental Component Summary score [VR-12 MCS], Charlson Comorbidity Index, surgical indication, Knee injury and Osteoarthritis Outcome Score [KOOS], deformity, range of motion, and discharge location probability assessed by a nomogram predicting location after arthroplasty) or as procedure or structural-related risk factors (hospital site, surgeon, day of the week when the surgical procedure was performed, implant type, and surgical procedure start time). Multivariable cumulative link (proportional odds logistic regression) models were built to identify significant predictors from candidate risk factors for 1-day, 2-day, and ≥3-day length of stay. Performance was compared between a model containing patient-related risk factors only and a model with both patient-related and procedure or structural-related risk factors, utilizing the Akaike information criterion (AIC) and internally validated concordance probabilities (C-index) for discriminating a 1-day length of stay compared with >1-day length of stay. Results: Patient-related risk factors were significant predictors of length of stay (p < 0.05). A longer length of stay was predicted by older age, higher body mass index (BMI), higher Charlson Comorbidity Index, lower VR-12 MCS, and female sex. However, when the procedure or structural factors were added to the patients’ risk factors, the AIC decreased by approximately 1,670 units. This indicates that procedure or structural-related risk factors provide clinically relevant improvement in explaining length of stay in addition to patient-related risk factors. Conclusions: Despite patient-related factors such as age, sex, and comorbidities providing substantial predictive value for length of stay after total knee arthroplasty, the main driving predictors of single-day length of stay after total knee arthroplasty were procedure or structural-related factors, including hospital site and surgeon. Understanding the risk factors that affect outcomes after total knee arthroplasty provides the opportunity to influence and potentially modify them favorably to optimize care.