Association of Total Knee Replacement Removal From the Inpatient-Only List With Outpatient Surgery Utilization and Outcomes in Medicare Patients

Key Points Question Was total knee replacement (TKR) removal from the Medicare inpatient-only (IPO) list in 2018 associated with changes in outcomes in Medicare patients? Findings In this cohort study of 37 588 Medicare fee-for-service procedures in a US state administrative database, older, Black, and female patients, as well as patients treated at safety-net hospitals, were less likely to undergo outpatient TKR. When compared with total hip replacements, the IPO policy was not associated with changes in outcomes after TKRs, except for an increase of $770 per encounter. Meaning These results suggest that there may be disparities in access to outpatient TKRs; TKR removal from the IPO list resulted in a modest increase in cost for surgical encounters but no difference in postoperative health care utilization.


Introduction
Total knee replacement (TKR) is a common procedure for end-stage degenerative disease of the knee, with annual volume in the US estimated at over 680 000 per year in 2014, with a projected increase to 935 000 by 2030. 1,2Patients consistently report satisfaction, pain relief, and improved function after TKR. 3,4e Centers for Medicare & Medicaid Services (CMS) has historically considered TKR to be an inpatient procedure; however, improvements in perioperative protocols have resulted in shorter hospital stays. 5,6Prior studies have demonstrated that outpatient TKR with stays less than 48 hours is safe in carefully selected patients, with no difference in postoperative readmissions, emergency department (ED) visits, or failure to rescue [7][8][9][10][11][12][13][14][15][16] and with lower risk for medical complications 14,17 compared with inpatient TKR.These findings contributed to the removal of TKR from the CMS inpatient-only (IPO) list in January 2018, allowing TKR to be performed in the inpatient or outpatient hospital setting 18,19 and the addition of TKR to the Ambulatory Surgery Center (ASC) Covered Procedures List in January 2020. 20The proportion of outpatient TKRs for Medicare beneficiaries increased from 0.2% in 2017 to 36.4% in early 2019, 21 with savings of $355 million over the first 18 months following policy implementation. 22Prior work has shown that this policy has also led to increases in outpatient TKR among privately insured patients. 23ere are few reports on the use and outcomes of TKR among Medicare patients after IPO policy implementation, especially when compared with a control group to whom the policy did not apply.The study objectives were to evaluate (1) the demographics of patients undergoing outpatient TKR after IPO policy implementation and (2) whether IPO policy was associated with an overall change in postoperative outcomes (eg, 30-day or 90-day readmissions and ED visits, discharge destination, and encounter cost) for patients undergoing TKRs (treatment group) compared with those undergoing total hip replacements (THRs) (control group).

Methods
The University of Rochester institutional review board approved the study, with a waiver granted for informed consent due to the deidentified nature of the administrative data used.Data analysis was performed from October 2021 to May 2022.This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

Data Sources and Study Cohort
We used the inpatient, outpatient, ambulatory surgery, and emergency department files from the New York Statewide Planning and Research Cooperative System (SPARCS) database between 2016 and 2019.SPARCS is a comprehensive all-payer data reporting system that collects patient-level data on patient characteristics, diagnoses and treatment, services, and charges for care in New York State (NYS) health care facilities.The surgical codes used to identify patients undergoing TKR (intervention group) or THR (control group) are in eTable 1 in Supplement 1.Because IPO policy is applicable to Medicare fee-for-service (Medicare) patients only, we limited the study cohort to these patients using primary source of payment typology variable in SPARCS.
We linked the SPARCS facility identifier with American Hospital Association (AHA) facility identifier using the Healthcare Cost and Utilization Project AHA linkage files 24 25 We used the 2015 NYS Annual Medicaid Disproportionate Share Hospital (DSH) Report file to obtain DSH data, 26 the CMS impact file for cost-to-charge ratios needed for estimating costs from total hospital charges, 27 and Federal Reserve data to convert costs to 2019 US dollars. 28e study included Medicare TKR or THR encounters from January

Outcome Variables
To determine characteristics of patients who were more likely to undergo outpatient TKRs following IPO policy implementation (objective 1), the outcome was a binary indicator of outpatient (outpatient or ambulatory service claim type in SPARCS) or inpatient (inpatient claim type in SPARCS)TKR.
To assess the association of IPO policy implementation with post-TKR outcomes (objective 2), the outcomes were binary indicators for 1 or more readmissions within 30 or 90 days of discharge after the index encounter; 1 or more ED visits within 30 or 90 days of discharge after the index encounter, defined as ED encounter (SPARCS ED claim type) or admission from ED within the time frame of interest after hospital discharge; discharge to a facility vs home; and a continuous indicator for total cost (adjusted to 2019 dollars) for the index encounter estimated from facility charges (chrg_tot_amt variable in SPARCS) using the cost-to-charge ratio.We used 2020 SPARCS data to evaluate 30-day and 90-day outcomes for those operations in 2019.

Key Independent Variables
To examine association of IPO policy with a change in TKR outcomes, the key independent variables were a binary indicator for procedure (THR vs TKR), a binary indicator for policy phase (whether a discharge occurred before or after the January 1, 2018, policy implementation), and an interaction between these variables.We chose patients undergoing THRs as a control group because the recovery and rehabilitation protocols are similar and standardized, 5,6 and many lower-extremity total joint replacement policies include both procedures. 29THR was not removed from the IPO list until 2020.

Statistical Analysis Descriptive Statistics
We report trends in outpatient TKR use from 2016 to 2019.We used χ 2 and Mann-Whitney U tests to compare characteristics of patients undergoing outpatient vs inpatient TKRs.

Multivariable Analysis
To determine characteristics of patients who had higher likelihood of undergoing outpatient TKRs, we estimated a multivariable generalized linear mixed model 31  We estimated multivariable generalized linear mixed models with logit link (for binary outcomes) or log link (for total cost outcome) and a difference-in-differences strategy at the encounter level to examine the association of IPO policy with the overall change in post-TKR outcomes.Difference-in-differences is an econometric method that is commonly used for policy evaluation and isolates the independent association of the policy with outcomes among the treatment group (TKRs) after controlling for changes in the control group (THRs). 32It does this by computing the difference in end points before and after policy implementation in the intervention group and compares this to the difference in the control group.Before estimating the difference-indifferences models, we checked for the parallel trends assumption.This assumption does not require the treatment (TKR) and control (THR) groups to be identical, but that the end points for the treatment group would have evolved in the same way as that in the control group in the absence of the policy.In case of violation of the assumption as evidenced by the covariates for year or procedure-year interaction being statistically significant with P < .05,4][35] All models controlled for patient-and facility-level covariates and facility-level random effects, including whether surgery was inpatient or outpatient.We used the -margins-command in Stata to compute adjusted estimates.Statistical analysis was conducted using Stata version 17 (StataCorp).A 2-tailed P < .05 was considered statistically significant.

Sensitivity Analysis
To test for robustness of our findings, we conducted the following sensitivity analyses.First, we estimated generalized linear mixed models with an identity (instead of logit or log) link to evaluate the association of IPO policy with post-TKR outcomes.Second, we reestimated our original models with facilitylevel fixed effects rather than random effects.Third, we included encounters for Asian, American Indian or Alaska Native, Native Hawaiian or other Pacific Islander, other, or missing race and ethnicity.Fourth, we reclassified all Medicare TKR encounters after policy implementation with stays less than 2 nights as outpatient due to potential differences in the coding of these variables based on billing codes.

Association of IPO Policy With Post-TKR Outcomes Relative to Post-THR Outcomes
The parallel trends assumption was violated for 90-day readmissions, 30-day and 90-day ED visits, and non-home discharge end points and outcomes (eTable 4 in Supplement 1).We addressed this violation by including interactions between year and procedure in the difference-in-differences models for these outcomes. 33,34The overall TKR rates of postoperative readmissions and ED visits were significantly lower after IPO policy implementation compared with before, while cost was significantly higher (Table 3; Figure 2; full model estimates available in eTable 5 in Supplement 1).95% CI, −1.12% to 2.71%; P = .76).However, these changes in the TKR cohort were not significantly different from the changes in the THR cohort other than an increase in cost of $770 per encounter (AD, $770; 95% CI, $83 to $1457; P = .03)(Table 3).Our findings from the sensitivity analysis were overall similar to those seen in the main analysis (eTables 6, 7, 8, and 9 in Supplement 1).

Discussion
TKR was removed from the CMS IPO list in 2018 due to evidence that these surgeries would be feasible to perform in the outpatient setting for Medicare patients.The association of this policy with TKR utilization and outcomes is largely unknown.We showed that use of outpatient TKR for Medicare patients is on the rise; that older, Black, and female patients, patients with more medically complex conditions, and patients treated in safety-net hospitals were less likely to be selected for outpatient TKR; and that the TKR IPO policy implementation was associated with an overall increase in TKR cost but no difference in postoperative readmissions, ED visits, or non-home discharge.
We found that outpatient TKR utilization has been increasing since policy implementation, and that patient-level and hospital-level factors including patient race and ethnicity are associated with outpatient TKR selection.The increase in outpatient TKR use was expected because TKR was considered an inpatient procedure by CMS prior to IPO policy implementation.Our inferences are consistent with the prior work by Barnes et al 21 showing that the rate of outpatient TKR in Medicare patients has increased from 0.2% in 2017 to 36.4% in the second quarter of 2019.We found that non-Hispanic Black and female patients were less likely to undergo outpatient TKR after controlling for medical comorbidities and other potential confounding factors.These findings raise questions Abbreviations: DSH, disproportionate share hospital payments; NR, not reported; OR, odds ratio; TKR, total knee replacement.
a Odds of undergoing outpatient surgery after policy implementation was modeled using a generalized linear mixed model with logit link adjusting for patient-level covariates, hospital-level covariates, and hospital random effects.Adjusted rate of outpatient TKR and the difference in the rates were estimated from the same model using the -marginswith contrast command in Stata 17.
b The age and year variables were treated as continuous variables in the analysis.The odds ratios for these variables can be interpreted as an increase/ decrease in odds of undergoing outpatient TKRs for each unit increase in the variable (ie, age or year).
a Adjusted outcomes (as percentages for binary outcomes and 2019 US dollars for total cost) from encounter-level models.Binary outcomes were modeled using generalized linear mixed models with a logit link adjusting for patient-level covariates, hospital-level covariates, and hospital random effects.Total cost was modeled using a generalized linear mixed model with clustering by facility, gamma distribution, and log link.The key independent variables were procedure (THR vs TKR), policy phase (before vs after IPO policy implementation), and an interaction between these 2 variables.Full model estimates (odds ratios for binary outcomes and coefficients for total cost models, both with 95% CIs) are included in eTable 5 in Supplement 1.The results of the test for parallel trends are included in eTable 4 in Supplement 1.The outcomes and changes in outcomes were obtained using the Stata margins and contrast commands.Models for 90-day readmissions, 30-day and 90-day ED visits, and non-home discharge outcomes included an interaction of year with procedure (TKR) because of violation of the parallel trends assumption in the preintervention period as shown in eTable 4 in Supplement 1.
b Adjusted outcomes (as percentages for binary outcomes and dollars for total cost) for THRs before implementation (2016-2017), after implementation (2018-2019), and the differences with IPO policy implementation.
c Adjusted outcomes (as percentages for binary outcomes and dollars for total cost) for TKRs before implementation (2016-2017), after implementation (2018-2019), and the differences with IPO policy implementation.
d Difference-in-differences estimates (as percentages for binary outcomes and dollars for total cost) for the association of the policy with outcomes after TKR (treatment) group after controlling for changes in outcomes after THR (control).This is considered the policy effect.regarding whether these patients may have reduced access to outpatient TKR and whether there may be a potential disincentive to care for these patients.The reason for this finding is unclear and likely multifactorial.Surgeons may be carrying out inadequate risk assessment in these patients, or these patients may be treated by surgeons who less commonly perform outpatient TKR or are treated at facilities not equipped to support outpatient TKR.This finding raises questions of whether IPO policy implementation may have worsened racial and ethnic and gender disparities in access to TKR. [36][37][38] Few studies have specifically evaluated the association of IPO implementation with outcomes

Figure 1 .
Figure 1.Trends in Total Hip Replacement (THR) and Total Knee Replacement (TKR) from 2016 to 2019 of TKR in Medicare patients.DeMik et al 39 reported no change in postoperative readmissions following TKR after IPO implementation compared with the period before implementation in the National Surgical Quality Improvement Program database; however, hospitals self-select into the program and the authors did not focus specifically in Medicare patients, to whom the policy directly applies.Our study fills an important gap by evaluating the policy association with utilization and cost outcomes specifically on Medicare TKRs compared with a control group, and we observed an overall increase in cost for TKRs but no association between IPO implementation and other TKR outcomes in the Medicare population after controlling for secular trends in the THR group.Because outpatient TKRs were shown to have lower overall cost than inpatient TKRs, further increases in the proportion of outpatient TKRs may help to reduce overall increases in per-encounter spending for TKRs.

Figure 2 .
Figure 2. Adjusted Trends in Outcomes Before and After Policy Implementation
1, 2016, to December 31, 2019.We excluded cases not financed by Medicare fee-for-service, age under 65 years, nonelective admission for fractures or degenerative changes (presumably with severe acute or subacute worsening of symptoms), with non-NYS residency, and unknown gender. 30

JAMA Network Open | Health Policy Removal
of Total Knee Replacements from the Inpatient-Only List and Outcomes in Medicare Patients

Table 1 .
Characteristics of Medicare Patients Undergoing Total Knee Replacement in the Inpatient and Outpatient Cohorts After the Inpatient-Only Policy Implementation (2018-2019) a (continued)

Table 2 .
Odds of Undergoing TKR in the Outpatient Setting in Medicare Patients After Inpatient-Only Policy Implementation (2018-2019) a (continued)