Summary

Changes to the preoperative disposition plan created by social workers utilizing the RAPT scoring system predict unanticipated extensions to patients’ length of stay after total joint arthroplasty.

Abstract

Introduction

Relative increases in both the demand and cost of joint replacement surgeries have brought increased attention and scrutiny to the hospital length of stay (LOS) patient's experience after total joint arthroplasty (TJA). In many cases, prolonged LOS is often a consequence of a lack of personnel, without which, institutions have a reduced capacity for complex discharge planning. In light of this, urban healthcare institutions have placed greater emphasis on the involvement of social workers in the care of those undergoing TJA. The integration of social workers into the multidisciplinary team has played a significant role in combating the increasing prolonged LOS, as well as, its corresponding increase in cost burden. Despite this, there remains limited evidence supporting the benefits of increased supportive counseling for patients undergoing TJA. This study analyzed preoperative factors that are components of preoperative social workers' evaluations to identify which factors resulted in an increased LOS after TJA. We hypothesize that there is a lack of uniformity resulting in a variation of practice patterns which influences hospital LOS after TJA.

Methods

Retrospective chart review included patients undergoing primary TJA at a single, urban center from January 2018 to July 2021. We examined patient outcomes as well as various components of preoperative planning by social workers including Risk Assessment and Prediction Tool (RAPT) scores, predicted and actual disposition, preoperative agreement with disposition plan, and LOS. A custom-made Python script was created to run a decision tree and feature importance to understand which RAPT factor was critical in predicting disposition. Brief interviews were conducted with 3 social workers to describe the process of creating and maintaining a disposition plan. Pearson’s Chi-Squared test was performed to calculate the significance between categorical variables. For analysis of continuous variables (LOS, Age, RAPT score) Kruskal Wallis and Wilcoxon rank sum tests were performed.

Results

679 patients met the inclusion criteria. 336 (49.6%) patients exceeded their anticipated LOS. RAPT score (OR 0.87 95% CI 0.80-0.95) and change in anticipated disposition (OR 2.83 95% CI 1.65-5.01) were significant predictors of exceeding the anticipated length of stay. On average, patients who required higher (4.8 ± 3.3 days) and lower (3.7 ± 2.3 days) levels of care had longer LOS than patients who had no change (2.5 ± 1.8 days) in anticipated disposition (p < 0.001). Patients requiring either higher or lower levels of care than anticipated had lower RAPT scores, 7.8 and 7.4 respectively, at preoperative evaluation compared to 8.6 for patients with no change in anticipated level of care (p = 0.002 and p = 0.040, respectively). Interviews with social workers revealed that the current RAPT threshold standard was 7, which would suggest the placement of those who score below 7 to be a Skilled Nursing Facility (SNF) and above 7 to be discharged home. Decision tree with feature importance analysis for RAPT factors resulted in home support as the strongest variable with a coefficient of 0.24 followed by age with a coefficient of 0.20 and ambulation with a coefficient of 0.16.

Conclusions

Changes to the preoperative discharge plan designed by social workers using the RAPT scoring system result in prolonged LOS regardless of if the change was to a higher or lower level of care than anticipated. The administrative burdens of determining a specific facility, acquiring appropriate rehabilitation documentation, and getting insurance approval increased the LOS in patients unexpectedly requiring SNF or Sub-Acute Rehabilitation (SAR) placement. The decision tree recommends that the availability of home support should be used to determine disposition for patients considered borderline between discharge to SNF or home. A less clear modifiable variable that predicted changes to the discharge plan was patient agreement with the plan. Our social workers reported that patients commonly have a different support system upon discharge than self-reported during preoperative assessment. Thus, longitudinal communication after coming to a mutually agreed upon discharge plan with the patient should lessen the predicted LOS considerably.