One-Year Readmissions Following Total Joint Arthroplasty May Be Associated With Failure to Achieve the Minimal Clinically Important Difference of Patient-Reported Outcomes Measurement Information System Physical, Mental, and Physical-Short Form-10a

The primary aims of our study were to determine if hospital readmissions within one year following primary total joint arthroplasty (TJA) and their relative timing influence patients’ ability to achieve the two-year Patient-Reported Outcomes Measurement Information System (PROMIS) physical, PROMIS mental, and PROMIS Physical-Function-Short-Form-10a (SF-10a) minimal clinically important difference (MCID). This is a retrospective study conducted using data from a multi-institutional, arthroplasty registry. Only patients with paired patient-reported outcome measure (PROM) assessments (preoperatively and two years postoperatively) were included. Five separate readmission cohorts were formed: (1) any-cause readmission within one year, (2) any-cause readmission within 90 days, (3) non-index-surgery-related readmission within 90 days, (4) index-surgery-related readmission within one year, and (5) index-surgery-related readmission within 90 days. A propensity score match was used to match each of the patients to one of the 972 patients (1:1 basis) in the non-readmission group. The association between failure to achieve each of the three two-year MCIDs and Readmission status was analyzed using logistic regression. We found that all readmissions within one year and index-surgery-related readmissions within one year resulted in an increased risk of failure to achieve the two-year MCID across all three collected PROMs. Index surgery-related readmissions within 90 days (OR 3.24; 95% CI 1.05-11.05; p=0.048) sustained significantly different rates of two-year PROMIS physical MCID achievement compared to matched controls. Postoperative complications requiring readmission, particularly those related to the joint arthroplasty and those within 90 days of index surgery, significantly impact the ability to achieve the two-year MCID of PROMs.


Introduction
The increasing demand and cost for arthroplasty care have necessitated a need to monitor outcomes [1][2][3]. Metrics such as infection rates, re-operation rates, and hospital readmission rates are measures of postoperative outcomes, but are relatively infrequent events [4]. Patient-reported outcome measures (PROMs), on the other hand, allow patients to categorize a subjective analysis of their health state on a variety of dimensions, including general health, pain, mental health, and overall physical function. The Patient-Reported Outcomes Measurement Information System (PROMIS), developed by the National Institute of Health, is a 10-question survey that measures both the domains of physical function (PROMIS physical) and mental health (PROMIS mental) and is useful in a variety of diverse treatments and procedures, including total joint arthroplasty (TJA) [5][6][7][8][9]. There is also an abbreviated version, the Physical Function Short Form 10a (SF-10a), which also consists of 10 questions and determines a patient's physical function level. To aid in the interpretation of raw PROM scores, various PROM metrics have been developed, such as the Patient Acceptable Symptom State (PASS) and the Minimal Clinically Important Difference (MCID) [10][11]. These metrics evaluate the pre-to postoperative difference in PROM scores that are deemed clinically significant for the average patient [12][13][14][15].
There is a growing understanding of how adverse surgical events, such as periprosthetic infections, revisions, and hospital readmissions may affect patient satisfaction, PROM scores, and the ability to achieve MCID. While some of these events are discrete and easy to categorize, hospital readmissions are quite diverse and require a deeper analysis due to their frequency and lack of uniformity. For example, readmission for a periprosthetic joint infection or a myocardial infarction constitutes two vastly different pathophysiologic events and therefore can impact postoperative outcomes in varying ways [16][17][18]. Furthermore, many analyses on readmissions focus on those events within the first 90-day postoperative period, which may be driven by insurer-based definitions of the "global" period of care. There is limited work reporting on readmission beyond the first 90 days and its implications for patient outcomes.
The primary aim of this study was to investigate whether hospital readmission within one year following primary TJA and its timing relative to the index surgery influenced patients' ability to achieve the two-year PROMIS physical, PROMIS mental, and SF-10a MCID.

Materials And Methods
Level of Evidence III. This retrospective study was conducted with Institutional Review Board approval using data from a regional, multi-institutional, arthroplasty registry. This registry is populated from the electronic medical records of patients treated in a healthcare network comprised of seven hospitals. All cases of primary total hip arthroplasty (THA) and total knee arthroplasty (TKA) performed since 2016 within this hospital network are captured in the registry as well as all postoperative readmissions with the same network, with an approximate 10% lost-to-follow-up rate within one year of surgery [19]. The PROMIS physical, PROMIS mental, and SF-10a scores are collected preoperatively and at yearly postoperative intervals. Only patients with paired PROM assessments (preoperatively and two years postoperatively) were included in this analysis. In order to maximize the number of patients for analysis and to account for inherent variations in the timing of patient follow-up visits, preoperative scores were defined as those captured within six months prior to the index procedure and the two-year postoperative period as those collected between 1.5 and 2.5 years from the index surgery.
A registry query for all primary THA and primary TKA patients with complete preoperative and two-year postoperative PROM score sets resulted in 1,302 patients. Variables collected for each patient included: demographics (age at index surgery, sex, body mass index [BMI]), Charlson Comorbidity Index (CCI), preoperative diagnoses (myocardial infarction, cancer, cerebrovascular accident, diabetes mellitus, hypertension, or congestive heart failure), discharge disposition, length of stay, and identity of the institution where the THA or TKA was performed. PROM scores for these patient populations were used to derive the three distinct MCIDs at the two-year postoperative timepoint using the distribution-based method, which defines the MCID as one-half of the standard deviation of the preoperative to postoperative change in each PROM for the entire population [20][21][22]. Of the 1,302 patients who completed preoperative and two-year PROM assessments, two-year MCIDs of 4.02, 4.46, and 6.19 for the PROMIS physical, PROMIS mental, and SF-10a, respectively, were determined.
A rigorous chart review was performed to classify the primary reason for the initial hospital readmission. Patients were then categorized into those having any readmission within one year of the index TJA and those having no readmissions. Patients undergoing subsequent elective procedures within 90 days of index TJA (n=34) were excluded from further analysis because of the inherent difficulty in attributing the complication readmission as related to the first elective procedure or the second elective procedure. Patients with multiple readmissions were categorized by only the readmission related to index TJA. If none of the readmissions pertained to index TJA, only the first readmission was categorized. With respect to timing and indication, readmission events were classified according to timing as early (within 90 days of index surgery) or any (0-365 days) and the reason for the readmission (surgical or other). The index-surgery-related readmissions were further classified according to the chief complaint identified by the consulted orthopaedic surgeon in the discharge summary. For non-index-surgery-related readmissions, the chief complaint as identified in the patient's discharge summary was recorded and further classified by organ system. A complete list of readmission events is included in Appendix A. Readmission events are further stratified by readmission group in Appendix B.

Statistical analysis
A propensity score match was used to match each patient in the five subgroups to one of the 972 patients (1:1 basis) in the non-readmission group. A separate propensity score match was performed for each PROM. Patients were matched on age, BMI, history of myocardial infarction, cancer, or cerebrovascular accident, CCI, congestive heart failure, type II diabetes, discharge disposition, hospital, length of stay, sex, and joint. Chi-squared test of independence and student's t-test were used to compare categorical and continuous demographic variables respectively, between all patients (unmatched) as well as the matched cohorts. For continuous variables used in the propensity score matching, matching was performed for values within 0.5 standard deviations of the mean of the variable. For each patient in the non-readmission and readmission cohorts, the change from preoperative to two-year postoperative PROM assessments was calculated and used to determine whether that patient achieved the two-year MCID. The three primary outcomes of interest were failure to achieve PROMIS physical, PROMIS mental, and SF-10a MCIDs at two years. Readmission status was interpreted as a binary variable in regression analysis. For each matched readmission cohort, every candidate predictor variable was included in multivariable logistic regression so as to not omit potential interactions. For only the variables that were significantly associated (p<0.05) with the outcome of interest after regression analysis, odds ratios (OR) and corresponding 95% confidence intervals (CI) were reported. All statistical analyses were performed using R (The R Foundation, Vienna, Austria) and RStudio (RStudio, Boston, MA, USA).

Discussion
As PROMs continue to enhance patient care by incorporating a patient's subjective appraisal of a surgical procedure, metrics such as the MCID may help in interpreting the clinical significance of these outcomes. In investigating the association between readmissions and achievement of MCID, the authors found that patients with any hospital readmission within one year of TJA were less likely to achieve two-year MCID across PROMIS physical, PROMIS mental, and SF-10a compared to patients without a readmission event. It is important to note that these lower patient-reported outcomes were driven primarily by index surgeryrelated readmissions as opposed to readmissions for non-index surgery-related reasons. Furthermore, after stratifying by time, index surgery-related readmissions within 90 days of TJA were more strongly associated with failure to achieve PROMIS physical MCID as compared to matched controls. These findings can provide guidance for physicians when managing patient expectations and when providing postoperative care following TJA. To the authors' knowledge, the current study is the first to distinguish between readmissions related to the arthroplasty procedure and those unrelated as it pertains to impact on three commonly used PROMs. The data supports the notion that readmissions following TJA are heterogeneous and can impact the ultimate outcome differently based on the indication for readmission and time relative to arthroplasty. It also provides reassurance to surgeons when asked about the impact of non-index surgery-related readmissions on ultimate recovery after TJA.
Previous studies have focused on pre-operative factors that can predict satisfaction following joint replacement [23]. Recent efforts to utilize machine learning algorithms to predict those patients who would achieve MCID pre-operatively produced models with poor-to-good accuracy [24]. These metrics, however, may be influenced by adverse surgical events such as hospital readmissions. The results herein offer some explanation for the performance of the algorithm as postoperative events can impact satisfaction and the ability to achieve MCID. Previous authors have established that readmissions can influence patient satisfaction and subjective outcome following TJA [25][26]. Bourne et al. noted that patients who had postoperative complications requiring readmission after primary TKA were 1.9 times more likely to be dissatisfied with their outcome [25]. Similarly, Friebel et al. were able to demonstrate improvements in functional and quality of life scores with reductions in readmission rates [26]. The current results support these previous findings as all-cause readmissions within one year of index surgery were associated with a lower rate of achieving MCID in the study cohort.
Beyond the actual measures, it is also critical to consider the time point at which these measurements are taken. Previous evidence has suggested that TJA patients continue to experience gains for up to one year after TJA [27]. However, the largest improvements in patient-reported outcome scores have been observed to occur within the first three months of surgery with smaller-scale improvements at six and 12 months [28]. This may explain the strong negative predictive value of complications that occurred strictly within 90 days of surgery as compared to complications occurring between postoperative days 0 and 365. Our results align with those of Neuprez et al., who were able to demonstrate that THA patients with early complications were three times less likely to report WOMAC scores that achieved MCID [28]. Interestingly, early complications were not similarly predictive in TKA patients [28]. Patient-reported complications within three and six months have also been shown to be predictive of lower functional and quality of life scores at those same time points [29]. Similarly, complications within one year of surgery have previously been shown to negatively impact the likelihood of achieving a minimally important difference in five-year WOMAC scores in TKA patients but not THA patients [28].
Conceptually, it is logical to infer that any interruption in a patients' rehabilitation caused by a complication necessitating readmission during the acute postoperative period may impede them from achieving critical milestones, thus compromising the ultimate function of the joint. Similarly, the majority of readmissions unrelated to the index procedure may not be predictive of the patients' ability to achieve MCID, with the rarer exceptions of more debilitating reasons for readmission such as unrelated lower extremity fractures, stroke with motor deficits, and intensive care unit admissions. Conversely, those readmissions occurring after 90 days are less impactful as the majority of the functional gains and recovery may have already been achieved and the patient has had an opportunity to experience the benefits of their replaced joint thus buoying their subjective assessment of the surgery.
This study, however, is not without limitations and its results should be interpreted within the context of its strengths and weaknesses. Given its retrospective nature, the current study is limited in identifying additional potentially confounding variables that were not collected in the registry. Additionally, readmissions to facilities outside of the institutional network were unable to be included and could impact the analysis as a source of measurement bias. It is worth noting that the authors' institution does include a multitude of community and tertiary referral hospitals in the area, so the effects of uncaptured readmissions were theoretically minimized. Furthermore, loss-to-follow-up is a limitation inherent to any PROM analysis and this study is no exception. However, the arthroplasty registry utilized reports an approximate 10% lostto-follow-up rate within one postoperatively, thus helping to mitigate such biases [19]. Another possible source of selection bias stems from only using preoperative and two-year postoperative PROMs; it is possible that different findings may result from PROMs completed longer after index TJAs.
In this study, we derived our own MCID values using a distribution-based method as opposed to using an anchor-based method given the retrospective nature of our study and lack of anchor questionnaire available [30]. The distribution has been used in numerous studies evaluating PROMs in TJA. We found that our distribution-based calculation of the MCID of PROMIS Physical, Mental and SF-10a were similar to previously described calculations [31][32][33]. We also chose to calculate a combined MCID of the selected PROMs for TJA, pooling the data of THA and TKA. As THA and TKA are different procedures, it is possible that the MCID of our selected PROMs could differ between the procedures, although previous studies have chosen to combine THA and TKA to calculate a TJA MCID as the values of the MCID for the individual procedure are similar [33].
Additionally, the current study includes patients from a variety of surgeons at multiple hospitals lending credence to the applicability of the results. Another possible limitation involves the omission of one-year postoperative PROM scores when assessing readmissions that occurred in the first postoperative year. Only the two-year postoperative period was assessed as it would account for time-dependent completion of the postoperative assessment and allow for an adequate period of recovery prior to assessment completion. We also chose to classify patients with multiple readmissions according to the readmission related to index TJA, which may be a source of bias as patients with multiple readmissions are expected to have worse outcomes. The number of patients included, and the two-year follow-up scores should be considered strengths of this study.

Conclusions
In conclusion, postoperative complications requiring readmission, particularly those related to joint arthroplasty, significantly impact patient-reported outcomes and especially the ability to achieve the twoyear MCID. Moreover, complications within 90 days are more impactful than complications that occur later on during the first postoperative year. This information is important as measures to mitigate index surgeryrelated complications and readmissions will have a significant impact in improving the proportion of patients achieving MCID on PROMs.

Appendices
Appendix A