Poor Recovery of Activities-of-Daily-Living Function Is Associated With Higher Rates of Postsurgical Hospitalization After Total Joint Arthroplasty

Abstract

Objective

Medicare beneficiaries are increasingly using home health (HH) as the first postacute care setting after hospital discharge following total joint arthroplasty (TJA). Yet, prior research has shown that changes in payment models for TJA may negatively influence functional outcomes for Medicare beneficiaries. The purpose of this study was to evaluate the impact of poor functional outcomes during an HH episode of care on hospitalization risk for older recipients of TJA.

Methods

For this study, 5822 Medicare beneficiaries who underwent elective TJA and subsequently participated in HH care following hospital discharge were identified using Medicare hospitalizations records and HH claims. Recovery of activities-of-daily-living (ADL) function was evaluated using patient assessment data completed at HH admission and discharge from the Medicare Outcomes and Assessment Information Set (OASIS). Hospitalization outcomes were captured from Medicare hospital claims. Cox proportional hazards regression was used to evaluate the hazard ratio for hospitalization after HH discharge.

Results

The 5822 Medicare beneficiaries who received a TJA and subsequently were discharged to HH were evaluated (n = 3989 [68.6%] following total knee replacement, n = 1883 [31.4%]) following total hip replacement). Nearly 9% (n = 534) of patients did not improve their ability to perform ADLs during the HH episode; this lack of improvement was associated with a more than 2-fold increase in hospital readmission rate following HH discharge (2.3% vs 4.9%). In adjusted models, there was a significant 77% increase (hazard ratio = 1.77; 95% CI = 1.14–2.74) in hospitalization risk during the 90-day postsurgical period.

Conclusion

Poor recovery of ADL function in HH settings following TJA is strongly associated with elevated risk of future hospitalizations.

Impact

Medicare beneficiaries who fail to make substantive improvements in basic ADL function during HH care episodes following TJA may need intensive monitoring from interdisciplinary team members across the continuum of care, especially during transitions from home care to outpatient care.

Lay Summary

An increasing number of patients receive home health care after joint replacement surgery, but outcomes after home health are unclear. These findings suggest that improvements in basic tasks such as walking or bathing are associated with a lower likelihood of hospitalization.

Introduction

Use of home health (HH) care after total joint arthroplasty (TJA)is increasing for Medicare beneficiaries.^1–3 This increase is driven in part by enactment of alternative payment models for elective joint replacement surgeries that bundle together the costs of hospital and postacute care, thereby incentivizing use of lower-cost home care as an alternative to skilled nursing facility or rehabilitation hospital admissions.³

However, concomitant with the push toward greater use of HH services for Medicare beneficiaries after TJA was significant downward pressures on other service use during the 90-day bundle period, including skilled rehabilitation services to reduce overall episode costs.⁴ Recent research has shown that the shift toward higher use of HH services after TJA is associated with lower rehabilitation volume⁵; this decrease in rehabilitation volume has been shown to negatively impact recovery of key activities of daily living (ADLs) for patients in HH settings.⁶ Failure to recover ADL function in the first 1 to 3 months after hospitalization period is associated with disability and nursing home admissions; however, little research has evaluated the implications of failure to recover function on hospitalization risk for older TJA recipients. This is important because functional recovery is a potentially modifiable risk factor, and could be improved with more targeted care delivery.

The implications of poor ADL recovery for older TJA recipients have not been well characterized, but other medically complex older adult populations who do not improve functionally in the early posthospitalization period have elevated risk of hospital readmissions.⁷ Therefore, the purpose of this article is to determine whether failure to improve ADL function is associated with hospitalization risk for older adults recovering from total joint replacement surgery in the HH setting.

Methods

This is a retrospective analysis of 2012 Medicare HH administrative and claims data, linked to hospitalization records and Medicare demographic files. The methods for developing the surgical cohort have been described previously.⁸ Briefly, we first used the 5% Medicare Provider and Analysis Review (MEDPAR) file to identify Medicare fee-for-service beneficiaries who were discharged to HH from a short-stay hospital after their first elective total knee arthroplasty (International Classification of Diseases [ICD]-9 procedure code 81.54) or total hip arthroplasty (ICD-9 procedure code 81.51) in 2012. The 5% MEDPAR file is a random sample of hospitalization claims for Medicare beneficiaries that contains patient-level and payment variables for each beneficiary stay, including comorbidities, length of stay, and Medicare reimbursements. We eliminated those with a hip replacement secondary to a fracture, identified by associated ICD-9 codes (820.XX) in any position on the index hospitalization claim for patients with a primary diagnosis-related group (DRG) code for hip replacement. We limited the sample to those who had at least 1 ADL impairment at HH admission, and completed a Medicare episode of HH care, consisting of an evaluation and a discharge Outcome and Assessment Information Set (OASIS) assessment. The OASIS file includes a comprehensive evaluation of medical complexity and ADL disability conducted at entrance into HH care. Those who were readmitted to the hospital or who died during the HH episode were excluded, because these patients do not have a recorded functional assessment at discharge. Lastly, the Medicare Master Beneficiary Summary File was linked to the surgical cohort by a common Medicare identification number to determine Medicare and Medicaid enrollment, patient race, and death dates for patients in the study sample. Development of the sample cohort is shown in Figure 1. Ethics approval for this study was granted by the University of Colorado Multiple Institutional Review Board and completed under a data use agreement with the Centers for Medicare & Medicaid Services.

Figure 1.

Development of the analytic sample.

Outcomes

The primary study outcome was all-cause hospitalization to a short-stay hospital following HH discharge that occurred within 90 days of surgery, chosen to coincide with the end of a 90-day postsurgical period that has relevance for patients participating in later and current Comprehensive Care for Joint Replacement (CJR) bundles. In brief, CJR bundles tested by Medicare incentivized hospitals to keep costs of a total joint replacement episode spanning from 3 days prior to surgery to 90 days postsurgery at or below target prices based on regional averages. Thus, any costly outcomes, such as readmissions, would have increased relevance over this period. To evaluate costs associated with readmission, associated Medicare payment to the hospital was also tabulated.

Clinical Characteristics

Patient age, gender, race data, and dual eligibility for Medicaid were extracted from the Master Beneficiary Summary File. We used the 31-item Elixhauser Comorbidity Index to characterize multimorbidity in the sample population.⁹ Elixhauser comorbidities were extracted from the ICD-9 diagnosis codes present in the MEDPAR file and coded using a validated algorithm.¹⁰ Hospital length of stay was also extracted from the MEDPAR set. HH length of stay was calculated as the dates between the admission and discharge OASIS assessments, and cost data were extracted from the HH claims file.

The primary predictor of interest was improvement in ADL function during the HH episode, as assessed by changes between the admission OASIS and the discharge OASIS tool. Whereas the OASIS assesses 9 different ADL tasks (bathing, upper body dressing, lower body dressing, grooming, walking, transferring, toileting, toilet hygiene, and feeding), we chose to include the 3 tasks endorsed by the National Quality Forum and publicly reported for all HH agencies: transferring (from bed to chair), bathing, and ambulation.¹¹ Each ADL task was scored dichotomously as able to perform without human assistance (independent) or unable to perform task without human assistance (dependence). For transfers, independence was categorized as a score of “0” on the 6-point scale ranging from 0 to 5; all other scores were considered dependence because they included use of human assistance as a criterion to complete the task. For ambulation, we considered scores of “0” or “1” (able to walk inside, over uneven ground, and navigate stairs with or without the assistance of a device) as independence, whereas all other scores on the 7-point scale were considered dependence with ambulation. Lastly, for bathing, we considered scores of “0” (able to perform without human assistance and without a device) or “1” (able to perform without human assistance and with a device) as independence, whereas all other scores on the 7-point scale were considered dependence. These anchors were chosen because any change from dependence to independence would indicate clinically meaningful improvements. Patients were categorized as “No ADL Improvement” or “Any ADL Improvement” between evaluation and HH discharge. Patients who failed to improve ADL function, or experienced a decline between baseline and discharge, were categorized as “No Improvement.”

Statistical Methods

Demographics and clinical characteristics comparing hospitalized and nonhospitalized patients were calculated. Raw event rates were calculated and compared using a χ² test, and then Kaplan-Meir survival curves were generated to evaluate the bivariate relationship between failure to recover ADL function and hospitalization in TJA recipients while accounting for differing exposure times. We used the log-rank test to evaluate differences in survival curves. We then performed multivariable Cox proportional hazards regression to estimate the 90-day readmission hazard ratio (HR) for those without ADL improvement relative to those with ADL improvement. Schoenfeld residuals were used to evaluate for any violation of the proportional hazards assumption. Adjusted models controlled for baseline ADL disability, age, sex, Elixhauser score, Medicaid dual eligibility, surgical type, and HH length of stay. Models additionally controlled for race to account for potential inequities in delivery of postdischarge care. There was no evidence of a proportional hazards violation on examination of the residuals, and thus, additional time by variable interactions were not explored. A P-value less than .05 was used to determine statistical significance. Because claims data and mandatory Medicare HH evaluations are used for billing, there were no missing data in any variables used in the analysis. We also conducted a sensitivity analysis on the data, repeating the proportional hazards regression after removing the 14 hospitalizations with a primary admission diagnosis for revision of the total joint replacement because these admissions likely have a different relationship with functional recovery than hospitalization for other medical reasons such as syncope. All statistical analysis was conducted with SAS 9.4 using the PROC PHREG procedure (SAS Institute Inc., Cary, NC, USA).

Role of the Funding Source

The funders played no role in the design, conduct, or reporting of this study.

Results

Overall, 5822 patients successfully completed an episode of HH care within 90 days after a total joint replacement and were included in the analysis. No patients entered HH with independence in all key ADL tasks. Patients started HH within a median of 4 days (interquartile range [IQR]: 4–6 days) after surgery, and HH care was initiated 1 day (IQR: 1.0–2.0 days) after hospital discharge. Descriptive statistics for the sample are presented in Table 1. Overall, 145 patients (2.6%) were hospitalized during the 90-day care bundle period following HH discharge, with a median (IQR) time to readmission of 29 (15–47) days. Hospitalized patients were similar in age, gender, and surgical type, but had higher comorbidity burden and were more likely to be dual eligible for Medicare and Medicaid.

Table 1.

Sample Demographics^a

Variable	Hospitalized (n = 145)	No Hospitalization (n = 5677)
Age, y, median (IQR)	71.5 (66.8–78.6)	71.3 (67.1–76.3)
Female sex, n (%)	92 (63.5)	3487 (61.4)
Race, n (%)
White	132 (91.0)	5035 (91.0)
Black	<11b	360 (6.3)
Other	<11b	279 (4.9)
Medicaid beneficiary, n (%)	28 (19.3)	599 (10.5)
Elixhauser score, median (IQR)	2.5 (2.0–4.0)	2.0 (1.0–3.0)
TKA surgery, n (%)	101 (69.7)	3889 (68.5)
Hospital LOS, d, median (IQR)	3.0 (3.0–4.0)	3.0 (3.0–3.0)
Hospital discharge to HH admission, d, median (IQR)	1.0 (1.0–2.0)	1.0 (1.0–2.0)
Baseline publicly reported ADL disabilities, median (IQR)	3.0 (3.0-3.0)	3.0 (3.0–3.0)
HH LOS, d, median (IQR)	17.0 (12.0–25.0)	19.0 (13.0–26.0)

^a ADL = activities of daily living; HH = home health; IQR = interquartile range; LOS = length of stay; TKA = total knee replacement.

^b Data are suppressed as a result of Medicare data use requirements for cells with fewer than 11 observations.

Over 92% (5361/5822) of the sample started the HH episode requiring human assistance with all 3 publicly reported ADL tasks, and 7% (416/5822) required assistance with 2 of the 3 ADLs. After the HH episode, human assistance was required by 25% (1467/5822) for ambulation either over uneven ground or climbing stairs, 29% (1682/5822) required human assistance or supervision for transfers from bed to chair, and 19% (1085/5822) required assistance from another person for taking a shower or bath.

The top DRG code for rehospitalization was revision of joint replacement (DRG 467 and 468; n = 14). The next most common DRG codes, all comprising fewer than 11 admissions each, were cardiac arrhythmias (DRG 310), gastrointestinal illness (DRG 392), and syncope and collapse (DRG 312). The median (IQR) Medicare cost of hospital readmissions at the 60-day period was $6340 ($4387–$12,633) in 2012 US dollars.

At HH discharge, 91% (5288/5822) had improved in function. Only 0.2% (11/5822) had declined in function, and the remainder (523/5822) had unchanged levels of disability at discharge as at the admission OASIS assessment. The 534 patients with no ADL improvement (declined or with unchanged function between admission and discharge) during the HH episode of care had significantly higher rates of readmission over the post-HH period both in terms of raw event rates (2.3% vs 4.7%; P < .001) and in unadjusted Kaplan-Meier models (log-rank test P < .001; Fig. 2). After adjusting for baseline ADL disability, age, sex, race, Elixhauser score, Medicaid dual eligibility, surgical type, and HH length of stay, those with no ADL improvement had a 77% increase in hospitalization risk over the 90-day postsurgical window (HR = 1.77; 95% CI = 1.14–2.74). Relationships between other covariates and readmissions are shown in Table 2. In the sensitivity analysis, our findings were similar to those of the primary analysis and are not reported.

Figure 2.

Kaplan-Meier curves showing time-to-hospitalization across those who either failed to improve (solid red line) or those who made any improvement (dashed green line) in basic ADL function during the HH episode of care. ADL = activities of daily living; HH = home health.

Table 2.

Hazard Ratios for 90-Day Readmission After Home Health Episodes^a

Variable	Hazard Ratio (95% CI)
Primary Predictor of Interest
No improvement in ADL function	1.77 (1.14–2.74)
Covariates
Age, y	1.01 (0.99–1.03)
Sex, M	1.00 (0.71–1.41)
Race
Black vs White	0.85 (0.44–1.65)
Other vs White	0.71 (0.22–2.27)
Medicaid dual eligibility	2.00 (1.25–3.22)
Elixhauser score	1.35 (1.23–1.49)
Home health length of stay, d	1.00 (0.98–1.02)
TKA surgery (vs THA)	1.00 (0.70–1.42)
Baseline disability count	1.22 (0.67–2.22)

^a ADL = activities of daily living; M = male; THA = total hip arthroplasty; TKA = total knee arthroplasty.

Discussion

The results from this study are the first to highlight the potential prognostic value of poor ADL recovery trajectory after TJA. Nearly 1 in 10 Medicare beneficiaries are discharged from HH after elective TJA without improvement in any key ADL function. Older adults who are discharged without improving in key ADL tasks are at elevated risk of downstream hospitalizations. Because these hospitalizations occur within the 90-day window during which hospitals are now held accountable for patient outcomes, there is significant policy relevance to these findings as well.

The relationship between poor recovery of physical function and higher readmission rates is consistent with research from other postacute care settings. For patients both within and discharged from skilled nursing facilities, higher physical disability is associated with all-cause and potentially avoidable readmission risk.^12–14 Similarly, for those discharging to or from rehabilitation hospitals, higher ADL disability is associated with higher rates of readmission.^15,16 However, these studies have evaluated largely heterogeneous, and more medically complex populations. Our findings in the healthier elective total joint replacement population add to this body of literature, and support that recovery of physical function serves as a highly sensitive, and potentially modifiable, independent biomarker for vulnerability to hospitalizations in Medicare beneficiaries recovering from surgery in HH settings. This is important because disability in these tasks is linked to downstream adverse events^17–19 if they persist.

When patients fail to recover function after total joint replacement, the optimal next steps are somewhat unclear. Making no ADL improvement after TJA could be the result of inadequate or poorly dosed rehabilitation.⁶ Indeed, inadequate rehabilitation dose after total joint replacement is a rising concern. The race to minimize costs during care bundles has led many hospital systems and surgical practices to explore reducing or eliminating rehabilitation use after elective total joint replacements^4,20 For some patients, a longer HH rehabilitation stay or additional services may be warranted, because it is unlikely that patients requiring human assistance with bathing, ambulation, or transfers can easily access rehabilitation services outside the home or participate in other restorative interventions. Yet, weighing the importance of functional recovery against overuse of health care resources is a delicate balancing act for hospitals, orthopedic surgeons, and policymakers. However, a growing body of evidence suggests strong short-term and long-term economic value for promoting improvements in mobility for Medicare beneficiaries.²¹ Research is needed to determine the optimal content and frequency of postoperative interventions to promote improvements in physical function after elective total joint replacement.

However, making no ADL improvement could also be the result of an unresolved medical concern that may necessitate coordination between postacute care clinicians, orthopedic surgeons, and primary care physicians to address. Older adults undergoing elective joint replacement have an increased risk of complications compared with younger cohorts.²² These complications may manifest during the preoperative or perioperative period, but our findings suggest that early failure to recover function postoperatively may represent a novel biomarker for hospitalization risk during the recovery process. Overall, poor recovery of function during an HH episode should necessitate a comprehensive reassessment of patients, and consideration of extended rehabilitation episodes. Our findings suggest that the patients who do not recover are still routinely discharged from HH care and would be expected to continue rehabilitation in the community—a daunting challenge for those with unchanged or worsened disability in key activities of daily living.

Limitations

This work is not without limitations. First, this research is only applicable to Medicare fee-for-service total joint replacement recipients. Thus, the results may not be generalizable to other payers. Secondly, these results only evaluated outcomes during HH, and do not consider other rehabilitation services that may have been received after HH discharge. However, this is consistent with Medicare readmission metrics, which do not generally consider service use in risk-adjustments. There are also additional factors, such as living environment, social support, socioeconomic status, and availability of medical resources that are not accounted for in these datasets but may be related to readmission risk. Lastly, these data span 2012 and 2013, and thus total joint replacement rehabilitation has changed since that time period. However, the relationship between functional recovery and readmissions likely has not changed, and perhaps has increased urgency in the face of downward pressure on rehabilitation use after surgery.

Conclusions

Among Medicare beneficiaries recovering from a total joint replacement, an early failure to recover independence in bathing, transfers, and ambulation is an independent risk factor for hospital readmissions during the 90-day postsurgical period. Research is needed to determine the optimal rehabilitation strategies to improve ADL function in the immediate posthospitalization period for this vulnerable population.

Author Contributions

Concept/idea/research design: J. Falvey, M.J. Bade, J.E. Forster, J.E. Stevens-Lapsley

Writing: J. Falvey, M.J. Bade, J.E. Forster

Data collection: J. Falvey

Data analysis: J. Falvey, M.J. Bade, J.E. Forster, J.E. Stevens-Lapsley

Project management: J. Falvey

Fund procurement: J. Falvey, J.E. Stevens-Lapsley

Providing institutional liaisons: J. Falvey, J.E. Stevens-Lapsley

Consultation (including review of manuscript before submitting): J.E. Forster, J.E. Stevens-Lapsley

Funding

One or more of the authors has received support from the National Institute on Aging (grant numbers T32AG000279, F31AG056069, and T32AG019134), the Foundation for Physical Therapy Research (PODS II Award to Dr Falvey), the Center on Health Services Training and Research (CoHSTAR), and the American Physical Therapy Association Home Health Section. Support for Veterans Affairs (VA)/Center for Medicare and Medicaid Services (CMS) data provided by the Department of Veterans Affairs, VA Health Services Research and Development Services, VA Information Resource Center (Project Numbers SDR-02-237 and 98–004). The views expressed in this article are those of the authors and do not necessarily reflect the position or policy of the Department of Veterans Affairs or the United States government. This article was submitted in partial fulfillment of Dr Falvey’s Doctor of Philosophy degree dissertation at the University of Colorado, Anschutz Medical Campus.

Disclosures

Dr Falvey receives royalties for courses developed on hospital readmissions for Medbridge Inc. Drs Falvey and Stevens-Lapsley are members of the PTJ Editorial Board. The authors completed the ICMJE Form for Disclosure of Potential Conflicts of Interest and reported no other conflicts of interest.