Modeling the Potential Economic Impact of the Medicare Comprehensive Care for Joint Replacement Episode-Based Payment Model

Racial Differences in Care and Outcomes After Total Hip and Knee Arthroplasties: Did the Comprehensive Care for Joint Replacement Program Make a Difference?

BACKGROUND

There is a paucity of literature on racial differences across a full total joint arthroplasty (TJA) "episode of care" and beyond. Given various incentives, the Comprehensive Care for Joint Replacement (CJR) program in the U.S. may have impacted preexisting racial differences across this care continuum. The purposes of the present study were (1) to assess trends in racial differences in care/outcome characteristics before, during, and after TJA surgery and (2) to assess if the CJR program coincided with reductions in these racial differences.

METHODS

This retrospective cohort study includes data on 1,483,221 TJAs (based on Medicare claims data, 2013 to 2018). Racial differences between Black and White patients were assessed for (1) preoperative characteristics (Deyo-Charlson comorbidity index, patient sex, and age), (2) characteristics during hospitalization (length of stay, blood transfusions, and combined complications), and (3) postoperative characteristics (90 and 180-day readmission rates and institutional post-acute care). Additionally, Medicare payments for each period were assessed. Racial differences (Black versus White patients) were expressed in terms of odds ratios (ORs) and 95% confidence intervals (CIs) per year. A "difference-in-differences" analysis (comparing before and after CJR implementation, with non-CJR hospitals being used as controls) estimated the association of the CJR program with changes in racial differences.

RESULTS

In both 2013 and 2018, Black patients (n = 74,390; 5.0%) were more likely than White patients to have a higher Deyo-Charlson comorbidity index (score of >0) (OR = 1.32 [95% CI = 1.28 to 1.36] and OR = 1.32 [95% CI = 1.28 to 1.37]), to require more transfusions (OR = 1.55 [95% CI = 1.49 to 1.62] and OR = 1.77 [95% CI = 1.56 to 2.01]), to be discharged to institutional post-acute care (OR = 1.40 [95% CI = 1.36 to 1.44] and OR = 1.49 [95% CI = 1.43 to 1.56]), and to be readmitted within 90 days (OR = 1.38 [95% CI = 1.32 to 1.44] and OR = 1.21 [95% CI = 1.13 to 1.29]) (p < 0.05 for all). Adjusted difference-in-differences analyses demonstrated that the CJR program coincided with reductions in racial differences in 90-day readmission (-1.24%; 95% CI, -2.46% to -0.03%) and 180-day readmission (-1.28%; 95% CI, -2.52% to -0.03%) (p = 0.044 for both).

CONCLUSIONS

Racial differences persist among patients managed with TJA. The CJR program coincided with reductions in some racial differences, thus identifying bundle design as a potential novel strategy to target racial disparities.

LEVEL OF EVIDENCE

Prognostic Level III. See Instructions for Authors for a complete description of levels of evidence.

Predictive capacity of four machine learning models for in-hospital postoperative outcomes following total knee arthroplasty

Background

Machine learning (ML) methods have shown promise in the development of patient-specific predictive models prior to surgical interventions. The purpose of this study was to develop, test, and compare four distinct ML models to predict postoperative parameters following primary total knee arthroplasty (TKA).

Methods

Data from the Nationwide Inpatient Sample was used to identify patients undergoing TKA during 2016-2017. Four distinct ML models predictive of mortality, length of stay (LOS), and discharge disposition were developed and validated using 15 predictive patient and hospital-specific factors. Area under the curve of the receiver operating characteristic curve (AUCROC) and accuracy were used as validity metrics, and the strongest predictive variables under each model were assessed.

Results

A total of 305,577 patients were included. For mortality, the XGBoost, neural network (NN), and LSVM models all had excellent responsiveness during validation, while random forest (RF) had fair responsiveness. For predicting LOS, all four models had poor responsiveness. For the discharge disposition outcome, the LSVM, NN, and XGBoost models had good responsiveness, while the RF model had poor responsiveness. LSVM and XGBoost had the highest responsiveness for predicting discharge disposition with an AUCROC of 0.747.

Discussion

The ML models tested demonstrated a range of poor to excellent responsiveness and accuracy in the prediction of the assessed metrics, with considerable variability noted in the predictive precision between the models. The continued development of ML models should be encouraged, with eventual integration into clinical practice in order to inform patient discussions, management decision making, and health policy.

Pain intensity in total hip arthroplasty patients: how communication influences satisfaction

INTRODUCTION

An important global measure of health care quality is patient satisfaction. Patient satisfaction partially determines hospital reimbursement for procedures such as total hip arthroplasty (THA). Press Ganey (PG) survey responses assess patient satisfaction, and impact reimbursement. Current efforts to maximise repayment for THA include reducing postoperative pain. The "Pain Management" survey domain is considered a significant factor in patient ratings, but other studies have highlighted staff communication domains as determinants of satisfaction. Therefore, the purpose of this study is to compare PG survey responses to inpatient pain intensity.

METHODS

We queried the PG database for all patients who underwent a THA between November 2012 and January 2015. This yielded a total of 302 patients. Descriptive statistics were performed to analyse patient-level demographics. A multivariate regression model was constructed utilising pain intensity as the dependent variable.

RESULTS

Patients rating of "Communication with Doctors" (B = -25.534; p < 0.001) and "Communication about Medicines" (B = -31.49; p = < 0.001) domains were representative of patient pain intensity. No other factors demonstrated a significant relationship to pain intensity.

CONCLUSIONS

Patient satisfaction continues to be important in care quality. Surrogate markers, such as the PG survey, can guide institutions looking to improve care. Our study revealed scores for "Communication with Doctors" and "Communication about Medicines" best represented true pain intensity levels for THA recipients during the postoperative period. The "Pain Management" domain did not display a relationship to pain intensity. The current method of measuring patient satisfaction should be reassessed to better represent patient responses and outcomes.

Deep Learning Preoperatively Predicts Value Metrics for Primary Total Knee Arthroplasty: Development and Validation of an Artificial Neural Network Model

BACKGROUND

The objective is to develop and validate an artificial neural network (ANN) that learns and predicts length of stay (LOS), inpatient charges, and discharge disposition before primary total knee arthroplasty (TKA). The secondary objective applied the ANN to propose a risk-based, patient-specific payment model (PSPM) commensurate with case complexity.

METHODS

Using data from 175,042 primary TKAs from the National Inpatient Sample and an institutional database, an ANN was developed to predict LOS, charges, and disposition using 15 preoperative variables. Outcome metrics included accuracy and area under the curve for a receiver operating characteristic curve. Model uncertainty was stratified by All Patient Refined comorbidity indices in establishing a risk-based PSPM.

RESULTS

The dynamic model demonstrated "learning" in the first 30 training rounds with areas under the curve of 74.8%, 82.8%, and 76.1% for LOS, charges, and discharge disposition, respectively. The PSPM demonstrated that as patient comorbidity increased, risk increased by 2.0%, 21.8%, and 82.6% for moderate, major, and severe comorbidities, respectively.

CONCLUSION

Our deep learning model demonstrated "learning" with acceptable validity, reliability, and responsiveness in predicting value metrics, offering the ability to preoperatively plan for TKA episodes of care. This model may be applied to a PSPM proposing tiered reimbursements reflecting case complexity.

Bundled Care for Hip Fractures: A Machine-Learning Approach to an Untenable Patient-Specific Payment Model

OBJECTIVES

With the transition to a value-based model of care delivery, bundled payment models have been implemented with demonstrated success in elective lower extremity joint arthroplasty. Yet, hip fracture outcomes are dependent on patient-level factors that may not be optimized preoperatively due to acuity of care. The objectives of this study are to (1) develop a supervised naive Bayes machine-learning algorithm using preoperative patient data to predict length of stay and cost after hip fracture and (2) propose a patient-specific payment model to project reimbursements based on patient comorbidities.

METHODS

Using the New York Statewide Planning and Research Cooperative System database, we studied 98,562 Medicare patients who underwent operative management for hip fracture from 2009 to 2016. A naive Bayes machine-learning model was built using age, sex, ethnicity, race, type of admission, risk of mortality, and severity of illness as predictive inputs.

RESULTS

Accuracy was demonstrated at 76.5% and 79.0% for length of stay and cost, respectively. Performance was 88% for length of stay and 89% for cost. Model error analysis showed increasing model error with increasing risk of mortality, which thus increased the risk-adjusted payment for each risk of mortality.

CONCLUSIONS

Our naive Bayes machine-learning algorithm provided excellent accuracy and responsiveness in the prediction of length of stay and cost of an episode of care for hip fracture using preoperative variables. This model demonstrates that the cost of delivery of hip fracture care is dependent on largely nonmodifiable patient-specific factors, likely making bundled care an implausible payment model for hip fractures.

A Weighted Index of Elixhauser Comorbidities for Predicting 90-day Readmission After Total Joint Arthroplasty

BACKGROUND

Evolving reimbursement models increasingly compel hospitals to assume costs for 90-day readmission after total joint arthroplasty. Although risk assessment tools exist, none currently reach the predictive performance required to accurately identify high-risk patients and modulate perioperative care accordingly. Although unlikely to perform adequately alone, the Elixhauser index is a set of 31 variables that may lend value in a broader model predicting 90-day readmission.

METHODS

Elixhauser comorbidities were examined in 10,022 primary unilateral total joint replacements, of which 4535 were hip replacements and 5487 were knee replacements, all performed between June 2013 and January 2018 at a single tertiary referral center. Data were extracted from electronic medical records using structured query language. After randomizing to derivation (80%) and validation (20%) subgroups, predictive models for 90-day readmission were generated and transformed into a system of weights based on each parameter's relative performance.

RESULTS

We observed 497 90-day readmissions (5.0%) during the study period, which demonstrated independent associations with 14 of the 31 Elixhauser comorbidity groups. A score created from the sum of each patient's weighted comorbidities did not lose substantial predictive discrimination (area under the curve: 0.653) compared to a comprehensive multivariable model containing all 31 unweighted Elixhauser parameters (area under the curve: 0.665). Readmission risk ranged from 3% for patients with a score of 0 to 27% for those with a score of 8 or higher.

CONCLUSIONS

The Elixhauser comorbidity score already meets or exceeds the predictive discrimination of available risk calculators. Although insufficient by itself, this score represents a valuable summary of patient comorbidities and merits inclusion in any broader model predicting 90-day readmission risk after total joint arthroplasty.

LEVEL OF EVIDENCE

III.

Association of Medicare's Bundled Payments for Care Improvement initiative with patient-reported outcomes

OBJECTIVE

To determine whether the Bundled Payments for Care Improvement (BPCI) initiative affected patient-reported measures of quality.

DATA SOURCES

Surveys of Medicare fee-for-service beneficiaries discharged from acute care hospitals participating in BPCI Model 2 and comparison hospitals between October 2014 and June 2017. Variables from Medicare administrative data and the Provider of Services file were used for sampling and risk adjustment.

STUDY DESIGN

We estimated risk-adjusted differences in patient-reported measures of care experience and changes in functional status, for beneficiaries treated by BPCI and comparison hospitals.

DATA COLLECTION

We selected a stratified random sample of BPCI and matched comparison beneficiaries. We fielded nine waves of surveys using a mail and phone protocol, yielding 29 193 BPCI and 29 913 comparison respondents.

PRINCIPAL FINDINGS

Most BPCI and comparison survey respondents reported a positive care experience and high satisfaction. BPCI respondents were slightly less likely than comparison respondents to report positive care experience or high satisfaction. Despite these differences in care experience, there was no difference between BPCI and comparison respondents in self-reported functional status approximately 90 days after hospital discharge.

CONCLUSIONS

These findings reduce concerns that BPCI may have unintentionally harmed patient health but suggest room for improvement in patient care experience.

Medium-Term Outcomes of Digital Versus Conventional Home-Based Rehabilitation After Total Knee Arthroplasty: Prospective, Parallel-Group Feasibility Study

Background

Physical rehabilitation is recommended after total knee arthroplasty (TKA). With the expected increase in TKA over the next few decades, it is important to find new ways of delivering cost-effective interventions. Technological interventions have been developed with this intent, but only preliminary evidence exists regarding their validity, with short follow-up times.

Objective

This study aimed to present the follow-up results of a feasibility study comparing two different home-based programs after TKA: conventional face-to-face sessions and a digital intervention performed through the use of an artificial intelligence-powered biofeedback system under remote clinical monitoring.

Methods

The digital intervention uses a motion tracker allowing 3D movement quantification, a mobile app and a Web portal. This study presents the results of the previous single-center, prospective, parallel-group, feasibility study including an 8-week active treatment stage and further assessments at 3 and 6 months post-TKA. Primary outcome was the Timed Up and Go score, and secondary outcomes were the Knee Osteoarthritis Outcome Scale (KOOS) score and knee range of motion.

Results

A total of 59 patients completed the study (30 in the digital intervention group and 29 in the conventional rehabilitation group) and follow-up assessments. During the active treatment stage, patients in the digital intervention group demonstrated high engagement and satisfaction levels, with an 82% retention rate. Both groups attained clinically relevant improvements from baseline to 6 months post-TKA. At the end of the 8-week program, clinical outcomes were superior in the digital intervention group. At the 3- and 6-month assessments, the outcomes remained superior for the Timed Up and Go score (P<.001) and all KOOS subscale scores (at 3 months, P<.001 overall; at 6 months, KOOS Symptoms: P=.006, Pain: P=.002, Activities of Daily Living: P=.001, Sports: P=.003, and Quality of Life: P=.001). There was progressive convergence between both groups in terms of the knee range of motion, which remained higher for standing flexion in the digital intervention group than the conventional group at 6 months (P=.01). For the primary outcome, at 6 months, the median difference between groups was 4.87 seconds (95% CI 1.85-7.47), in favor of the digital intervention group.

Conclusions

The present study demonstrates that this novel digital intervention for independent home-based rehabilitation after TKA is feasible, engaging, and capable of maximizing clinical outcomes in comparison to conventional rehabilitation in the short and medium term; in addition, this intervention is far less demanding in terms of human resources.

Trial Registration

ClinicalTrials.gov NCT03047252; https://clinicaltrials.gov/ct2/show/NCT03047252

An Analysis of Centers for Medicare & Medicaid Service Payment in Maryland: Can a Global Budget Revenue Model Save Money in Lower Extremity Arthroplasty?

BACKGROUND

Maryland is the only state utilizing the Global Budget Revenue (GBR) model to reduce costs. The purpose of this study is to evaluate whether the GBR payment model effectively reduced the following: (1) costs of inpatient hospital stays; (2) post-acute care costs; (3) lengths of stay (LOS); (4) readmission rates; and (5) discharge disposition in patients who underwent primary total hip and knee arthroplasty (THA and TKA).

METHODS

We evaluated the Maryland Centers for Medicare & Medicaid Service database for THAs and TKAs performed at 6 hospitals 1 year prior to (2012) and after the initiation of GBR (2015). We compared differences in costs for each inpatient care episode, post-acute care periods (total costs, acute rehabilitation, short-term nursing facility, home health, durable medical equipment), readmissions, LOS, and discharge disposition.

RESULTS

Hospitals had a significant reduction in mean inpatient care costs for THA and TKA (P < .0001). There was a significant reduction in total post-acute care costs following THA (P < .001). Home healthcare had a significant increase in cost following THA and TKA (P < .0001). There was a significant reduction in durable medical equipment costs for THA (P < .0001). There was a significant decrease in LOS for THA and TKA (P < .0001). There was a significant increase in patients discharged home (THA, P = .0262; TKA, P = .0058).

CONCLUSION

The Maryland healthcare model may be associated with a reduction in inpatient and post-acute care costs. Furthermore, implementation of GBR may result in reductions in LOS and readmission rates.

Identifying high-cost episodes in lower extremity joint replacement

OBJECTIVES

To evaluate the ability of claims-based risk adjustment and incremental components of clinical data to identify 90-day episode costs among lower extremity joint replacement (LEJR) patients according to the Centers for Medicare & Medicaid Services (CMS) Comprehensive Care for Joint Replacement (CJR) program provisions.

DATA SOURCES

Medicare fee-for-service (FFS) data for qualifying CJR episodes in the United States, and FFS data linked with clinical data from CJR-qualifying LEJR episodes performed at High Value Healthcare Collaborative (HVHC) and Mayo Clinic in 2013. HVHC and Mayo Clinic populations are subsets of the total FFS population to assess the additive value of additional pieces of clinical data in correctly assigning patients to cost groups.

STUDY DESIGN

Multivariable logistic models identified high-cost episodes.

DATA COLLECTION/EXTRACTION METHODS

Clinical data from participating health care systems merged with Medicare FFS data.

PRINCIPAL FINDINGS

Our three populations consisted of 363 621 patients in the CMS population, 4881 in the HVHC population, and 918 in the Mayo population. When modeling per CJR specifications, we observed low to moderate model performance (CMS C-Stat = 0.714; HVHC C-Stat = 0.628; Mayo C-Stat = 0.587). Adding CMS-HCC categories improved identification of patients in the top 20% of episode costs (CMS C-Stat = 0.758, HVHC C-Stat = 0.692, Mayo C-Stat = 0.677). Clinical variables, particularly functional status in the population for which this was available (Mayo C-Stat = 0.783), improved ability to identify patients within cost groups.

CONCLUSIONS

Policy makers could use these findings to improve payment adjustments for bundled LEJR procedures and in consideration of new data elements for reimbursement.

Machine Learning and Primary Total Knee Arthroplasty: Patient Forecasting for a Patient-Specific Payment Model

BACKGROUND

Value-based and patient-specific care represent 2 critical areas of focus that have yet to be fully reconciled by today's bundled care model. Using a predictive naïve Bayesian model, the objectives of this study were (1) to develop a machine-learning algorithm using preoperative big data to predict length of stay (LOS) and inpatient costs after primary total knee arthroplasty (TKA) and (2) to propose a tiered patient-specific payment model that reflects patient complexity for reimbursement.

METHODS

Using 141,446 patients undergoing primary TKA from an administrative database from 2009 to 2016, a Bayesian model was created and trained to forecast LOS and cost. Algorithm performance was determined using the area under the receiver operating characteristic curve and the percent accuracy. A proposed risk-based patient-specific payment model was derived based on outputs.

RESULTS

The machine-learning algorithm required age, race, gender, and comorbidity scores ("risk of illness" and "risk of morbidity") to demonstrate a high degree of validity with an area under the receiver operating characteristic curve of 0.7822 and 0.7382 for LOS and cost. As patient complexity increased, cost add-ons increased in tiers of 3%, 10%, and 15% for moderate, major, and extreme mortality risks, respectively.

CONCLUSION

Our machine-learning algorithm derived from an administrative database demonstrated excellent validity in predicting LOS and costs before primary TKA and has broad value-based applications, including a risk-based patient-specific payment model.

Characterizing Patient Preferences Surrounding Total Knee Arthroplasty

Background:

Episode-based bundled payments for total knee arthroplasty emphasize cost-effective patient-centered care. Understanding patients’ perceptions of components of the total knee arthroplasty care episode is critical to achieving this care. This study investigated patient preferences for components of the total knee arthroplasty care episode.

Methods:

Best-worst scaling was used to analyze patient preferences for components of the total knee arthroplasty care episode. Participants were selected from patients presenting to 2 orthopaedic clinics with chronic knee pain. They were presented with descriptions of 17 attributes before completing a best-worst scaling exercise. Attribute importance was determined using hierarchical Bayesian estimation. Latent class analysis was used to evaluate varying preference profiles.

Results:

One hundred and seventy-four patients completed the survey, and 117 patients (67%) were female. The mean age was 62.71 years. Participants placed the highest value on surgeon factors, including level of experience, satisfaction rating, and complication rates. Latent class analysis provided a 4-segment model of the population.

Conclusions:

This study demonstrated differences in patient preferences for the components of a total knee arthroplasty care episode and characterized distinct preference profiles among patient subsets. Stakeholders can use this information to focus efforts and policy on high-value components and to potentially create customized bundles guided by preference profiles.

Clinical Relevance:

This study is clinically relevant because the patient preferences identified here may help providers to design customized bundles for total knee arthroplasty care.

Skilled Nursing Facility Star Rating, Patient Outcomes, and Readmission Risk After Total Joint Arthroplasty

BACKGROUND

This study examined the correlation between publicly reported indicators of skilled nursing facility (SNF) quality and clinical outcomes after primary total knee arthroplasty (TKA) and total hip arthroplasty (THA).

METHODS

This retrospective analysis used Medicare claims from the Centers for Medicare and Medicaid Services 100% Standard Analytic File (2014-2015) that were linked to SNF quality star ratings from the Centers for Medicare and Medicaid Services Nursing Home Compare database. Overall SNF rating and subcomponents of the rating were evaluated for correlation to 30-day and 90-day risk of readmission. Ratings were based upon a 5-star rating system (1 representing the lowest quality). Cox proportional hazards regressions controlled for age, race, census division, hospital location, comorbidities, and SNF length of stay.

RESULTS

A total of 9418 SNFs, 58,064 TKA patients, and 26,837 THA patients met criteria. As SNF overall star rating increased from 1 to 5, incidence of all-cause 30-day readmission decreased from 6.4% to 5.0% for TKA (relative reduction [RR] 22%; P < .001) and from 9.1% to 6.2% for THA (RR 32%; P < .001). As nurse staffing rating increased, incidence of all-cause readmission decreased from 6.8% to 4.7% for the TKA cohort (30.9% RR; P < .001), and from 7.7% to 6.0% for the THA cohort (22.1% RR; P = .003). Regression analysis demonstrated that a higher star rating was associated with decreased risk of readmission (both cohorts P < .05).

CONCLUSIONS

For patients undergoing TKA or THA, the overall SNF star rating, nurse staffing ratios, and physical therapy intensity were significantly correlated with risk of readmission within 30 days of SNF admission.

Patients With Major Depressive Disorder Experience Increased Perception of Pain and Opioid Consumption Following Total Joint Arthroplasty

BACKGROUND

Pain in the immediate postoperative period following total joint arthroplasty is influenced by various patient factors, including major depressive disorder (MDD). Therefore, this study aimed to compare the patient perception of pain and opioid consumption between patients with and without MDD who received either a total knee arthroplasty (TKA) or total hip arthroplasty (THA). Specifically, we compared (1) pain intensity, (2) lengths of stay, (3) opioid consumption, and (4) patient perception of pain control.

METHODS

We reviewed our institutional Press Ganey database to identify patients with a diagnosis of MDD who received a THA (n = 48) and TKA (n = 68) between 2012 and 2016. An independent samples t-test and chi-square analyses were conducted to assess continuous and categorical variables, respectively. Analysis of covariance assessed the effects of depression on postoperative pain intensity. Mixed-design analysis of variance assessed the difference in opioid consumption between groups.

RESULTS

Patients with MDD who received THA or TKA demonstrated a higher mean pain intensity score when compared to those without MDD; however, this was not statistically different (235.6 vs 207.7; P = .264 and 214.8 vs 185.1; P = .055, respectively). Patients with MDD who received THA or TKA consumed more opioids when compared to those without MDD (P = .048 and P = .038, respectively).

CONCLUSION

Patients with MDD undergoing total joint arthroplasty consume more opioids compared to their matched cohort during the immediate postoperative period. Identifying patient-specific factors, such as MDD, could help arthroplasty surgeons modulate patients' course of recovery. These findings warrant more cooperation between arthroplasty surgeons and primary care providers to optimize outcome.