Clinical Characteristics and Outcomes of Medicare Patients Undergoing Total Hip Arthroplasty, 1991-2008 FREE

Total hip arthroplasty is a safe and effective therapy for patients with advanced degenerative joint disease.^1- 2 In recent years, there has been a dramatic increase in performance of this procedure both in the United States and abroad.^3- 6 There is a general assumption that increasing experience with total hip arthroplasty has resulted in improvements in patient outcomes, as has been observed in other procedures,^7- 9 but rigorous empirical data documenting such improvement are limited.^10- 14 This lack of data are striking given that an estimated 280 000 total hip arthroplasty procedures are performed annually at a cost of more than $12 billion.⁵

The lack of rigorous evaluation of total hip arthroplasty outcomes is somewhat surprising given ongoing efforts to regionalize surgical procedures to higher-volume and higher-quality hospitals.^15- 17 As a generally elective surgery, total hip arthroplasty is the type of procedure that should be amenable to regionalization. Moreover, the introduction of the Medicare prospective payment system in 1983 provided incentive to hospitals to control costs.^18- 19 A particularly important cost-control mechanism has been through reducing hospital length of stay (LOS). However, there is growing concern that hospitals may be reducing LOS by discharging patients prematurely, resulting in increased use of skilled-care facilities^20- 21 and increased patient readmissions, thus eliminating much of the cost savings originally envisioned by the prospective payment system.^22- 23

The objective of this study was to evaluate long-term trends in the outcomes of Medicare beneficiaries undergoing primary and revision total hip arthroplasty and to explore whether reductions in hospital LOS might be associated with increased discharge of patients to postacute care settings, increased readmission rates, or a combination of both outcomes.

We used Medicare Provider Analysis and Review (MedPAR) Part A data files to identify fee-for-service beneficiaries who underwent primary or revision total hip arthroplasty between 1991 and 2008. Patients were identified using International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) procedure codes (81.51 for primary total hip arthroplasty and 80.05, 81.53, 00.70, 00.71, 00.72, and 00.73 for revision total hip arthroplasty).^24- 27 The Part A files contain a range of data collected from discharge abstracts for all hospitalized fee-for-service Medicare enrollees including patient demographics, ICD-9-CM codes for primary and secondary diagnoses and procedures, admission source (eg, emergency department or transfer from outside hospital), admission and discharge dates, discharge disposition (coded as 25 separate categories), death occurring as many as 3 years after discharge, a unique patient identifier that allows for identification of patient readmissions, and each hospital's unique 6-digit identification number.

For purposes of this analysis, patient race was categorized as white, black, other, and not available. Comorbid illnesses present on the index admission were identified using algorithms described by Elixhauser et al^28- 29 that consider 30 specific conditions and exclude comorbid conditions that may represent complications of care or that are related to the primary reason for hospitalization.

To perform our 18-year longitudinal study, we used 2 separate Medicare Part A data files obtained from the Centers for Medicare & Medicaid Services (1 file detailed 1991-2005; 1 file 2006-2008). Each of these files contains an internally consistent unique patient identifier that allows for the tracking of individual patients across time within the file. However, because the unique patient identifiers differ across the 2 different data files, it was impossible to link individual patients across the 2 different files. Thus, for example, if a single patient were hospitalized in December 2005 and readmitted again in January 2006, the lack of a consistent unique identifier across the 2 files precluded us from determining whether a single patient were admitted and readmitted or whether 2 different patients were admitted (one in December 2005 and another in January 2006). From a practical standpoint, this had minimal impact on our results but did influence our methodological approach, as described further.

Our intention was to examine changes in outcomes of patients undergoing primary and revision total hip arthroplasty procedures. Because primary total hip arthroplasty is most often an elective procedure while revision total hip arthroplasty can be either an elective or more urgent procedure, we applied separate exclusion criteria to the primary and revision total hip arthroplasty populations in accordance with prior studies.^27,30- 31 For primary total hip arthroplasty we sequentially excluded patients with acute fractures (136 887), patients admitted through the emergency department (19 721), and patients admitted after transfer from another acute-care hospital (2856); these exclusion criteria were developed to select a population of primary elective total hip arthroplasty patients. Our revision total hip arthroplasty population did not exclude these types of patients because revision total hip arthroplasty is often an emergent or unscheduled procedure. This study was approved by the University of Iowa Institutional Review Board.

Our outcomes of interest included hospital length of stay (LOS), mortality (in-hospital, within 30 and 90 days of admission), discharge disposition, and all-cause readmission within 30 days and 90 days of admission. Discharge disposition was stratified into 6 mutually exclusive categories: home; skilled care/intermediate care; transfer to another acute-care hospital; inpatient rehabilitation; dead; and other. We limited our 30-day readmission analysis to patients whose index hospitalization occurred prior to November 30, 2005, for procedures performed between 1991 and 2005, and prior to November 30, 2008, for procedures performed between 2006 and 2008, to allow for a full 30-day follow-up period. Similarly, our 90-day readmission analysis was limited to patients whose index hospitalization occurred prior to September 30, 2005, for procedures performed between 1991 and 2005, and prior to September 30, 2008, for procedures performed between 2006 and 2008. Patients with hospital LOS longer than 1 year were excluded from the LOS analysis in accordance with other recent studies.²²

We examined the demographic characteristics and prevalence of key comorbid illnesses for patients who underwent total hip arthroplasty during the study period; for simplicity, data are presented separately for each 2-year period (eg, 1991-1992, 1993-1994, etc). We used analysis of variance for comparisons of continuous variables and the Mantel-Haenszel χ² test for categorical variables. All analyses were performed separately for primary and revision total hip arthroplasty patients.

We compared rates of each of the outcomes described previously for each of the 2-year periods using similar statistical methods. We used graphical techniques to examine changes in LOS, mortality, discharge disposition, and readmission rates over the course of the study period. To account for the changing demographics of the primary and revision total hip arthroplasty populations over time, we calculated risk-standardized mortality ratios that adjusted for age, sex, and race.³² Inclusion of patient demographics in calculating these rates is important given that prior studies have demonstrated differential complication and use rates for joint arthroplasty among different demographic groups.^33- 35 All analyses for these variables were conducted separately for the primary and revision total hip arthroplasty cohorts.

All P values were 2-tailed, with P values less than .05 considered statistically significant. All statistical analyses were performed using SAS version 9.2 (SAS Institute Inc, Cary, North Carolina).

To assess the robustness of our results, we performed 2 supplemental analyses. First, we repeated our analyses while restricting our analysis to the first primary or first revision hip arthroplasty procedure performed on each patient during a 12-month period; this eliminates patients who undergo “staged” bilateral arthroplasty procedures or complicated revisions. Second, we repeated our analyses after adding back the excluded populations described previously (eg, fracture patients) (eFigure 1 and eFigure 2).

Our final study population included 1 453 493 elective primary total hip arthroplasty procedures and 348 596 revision hip arthroplasty procedures performed on Medicare beneficiaries between 1991 and 2008. For primary total hip arthroplasty comparing 1991-1992 and 2007-2008 (Table 1), mean (standard deviation [SD]) age increased from 74.1 (6.0) years (95% confidence interval [CI], 74.0-74.1) to 75.1 (6.5) years (95% CI, 75.1-75.2) (P = .01); and diabetes prevalence increased from 7.1% (95% CI, 7.0%-7.3%) to 15.5% (95% CI, 15.4%-15.7%), and obesity prevalence increased from 2.2% (95% CI, 2.2%-2.3%) to 7.6% (95% CI, 7.5%-7.7%) (P < .001 for each), respectively.

Trends were similar for revision total hip arthroplasty comparing 1991-1992 and 2007-2008 (Table 2). In particular, mean (SD) age increased from 75.8 (6.9) years (95% CI, 75.7-75.9) to 77.3 (7.2) years (95% CI, 77.2-77.4), diabetes prevalence increased from 7.2% (95% CI, 6.9%-7.4%) to 15.7% (95% CI, 15.3%-16.0%), and obesity prevalence increased from 1.4% (95% CI, 1.3%-1.5%) to 4.7% (95% CI, 4.5%-4.9%) (P < .001 for each), respectively.

For primary total hip arthroplasty comparing 1991-1992 and 2007-2008, mean hospital LOS decreased from 9.1 days (95% CI, 9.1-9.2) to 3.7 days (95% CI, 3.7-3.7) (Table 3 and Figure 1), a 59.3% relative decline (P < .001); in-hospital mortality after primary total hip arthroplasty decreased from 0.5% (95% CI, 0.5%-0.5%) to 0.2% (95% CI, 0.2%-0.2%), a 60.0% relative reduction (P < .001); 30-day mortality decreased from 0.7% (95% CI, 0.7%-0.7%) to 0.4% (95% CI, 0.4%-0.4%), a 42.9% relative reduction (P = .004); and 90-day mortality decreased from 1.2% (95% CI, 1.2%-1.3%) to 0.8% (95% CI, 0.7%-0.8%) (P < .001), respectively. After adjustment for patient characteristics, risk-adjusted 30-day mortality over the study period decreased from 0.7% (95% CI, 0.7%-0.8%) to 0.3% (95% CI, 0.3%-0.4%) and 90-day mortality decreased from 1.3% (95% CI, 1.2%-1.3%) to 0.7% (95% CI, 0.7%-0.7%) (P < .001 for each).

The proportion of primary total hip arthroplasty patients discharged to home decreased from 68.0% (95% CI, 67.8%-68.3%) (P < .001) in 1991-1992 to 48.2% (95% CI, 48.0%-48.5%) in 2007-2008, while the proportion of patients discharged to skilled or intermediate care increased from 17.8% (95% CI, 17.6%-18.1%) to 34.3% (95% CI, 34.1%-34.5%) (P < .001) (Table 3 and Figure 1). The 30-day all-cause readmission rate decreased from 5.9% (95% CI, 5.8%-6.1%) in 1991-1992 to 4.6% (95% CI, 4.5%-4.7%) in 2001-2002 (P < .001), before increasing to 8.5% (95% CI, 8.4%-8.6%) in 2007-2008 (P < .001)—results were similar for 90-day readmission rates.

For revision total hip arthroplasty, mean hospital LOS decreased from 12.3 days (95% CI, 12.2-12.4) in 1991-1992 to 6.0 days (95% CI, 6.0-6.1) in 2007-2008 (Table 4 and Figure 2), a 51.2% relative decline (P < .001). Unadjusted in-hospital mortality after revision total hip arthroplasty decreased from 1.8% (95% CI, 1.6%-1.9%) in 1991-1992 to 1.2% (95% CI, 1.1%-1.3%) in 2007-2008, a 33.3% relative reduction (P < .001). Conversely, during the study period, unadjusted 30-day mortality increased from 2.0% (95% CI, 1.8%-2.1%) in 1991-1992 to 2.4% (95% CI, 2.2%-2.5%) in 2007-2008, a 20.0% relative increase (P = .004), and 90-day mortality increased from 4.0% (95% CI, 3.8%-4.2%) to 5.2% (95% CI, 5.0%-5.4%), a 30.0% relative increase (P < .001). However, after adjustment for patient characteristics, risk-adjusted 30-day mortality remained stable within a narrow range between 1.9% and 2.3% (P = .22), while adjusted 90-day mortality remained near 4.5% throughout the study period (P = .16).

The proportion of revision total hip arthroplasty patients discharged to home decreased from 57.4% (95% CI, 56.8%-57.9%) in 1991-1992 to 35.4% (95% CI, 34.9%-35.8%) in 2007-2008 (P < .001), while the proportion of patients discharged to skilled or intermediate care increased from 26.7% (95% CI, 26.2%-27.1%) to 42.4% (95% CI, 42.0%-42.9%) (P < .001) (Table 4 and Figure 2). The 30-day all-cause readmission rate decreased from 8.7% (95% CI, 8.3%-9.0%) in 1991-1992 to 8.2% (95% CI, 7.9%-8.5%) in 1999-2000 (P < .001), before increasing to 14.1% (95% CI, 13.8%-14.5%) in 2007-2008 (P < .001), with similar results for 90-day readmission rates.

In an analysis of 1991-2008 Medicare administrative data, 3 trends were identified. First, we found that despite increasing patient complexity, both unadjusted and adjusted mortality for primary total hip arthroplasty showed substantial improvement over time. Conversely, our second finding was that for revision total hip arthroplasty, unadjusted mortality appeared to increase modestly but this increase was largely explained by increasing patient complexity. Third and most importantly, marked declines in hospital LOS for both primary and revision total hip arthroplasty seemed to correspond with an increase in the proportion of patients who were discharged to postacute care and an increase in patient readmissions.

A number of our findings warrant further discussion. First, we found increasing complexity of both primary and revision total hip arthroplasty patients. Although a number of studies have documented increasing complexity of patients in the cardiovascular disease literature,^22- 23,36 few studies have evaluated trends in patient complexity in orthopedics and most have focused on knee arthroplasty.^37- 40 Our finding of an increase in the mean age of both primary and revision total hip arthroplasty patients suggests that at least some of the increase in patient complexity is real and is not simply an artifact of more aggressive coding practices (ie, upcoding).⁴¹

Our finding of increased patient complexity is particularly important when considering the changes in patient mortality that were observed. We found clinically and statistically significant reductions in primary total hip arthroplasty mortality, albeit from very low baseline levels. Conversely, the finding of an increase in revision total hip arthroplasty mortality in unadjusted analyses was somewhat unexpected; the finding that this increase in mortality was no longer significant in adjusted analyses is reassuring. The absence of similar large-scale studies of the US hip arthroplasty population with which to compare our results highlights the need for more rigorous study of total hip arthroplasty outcomes.

Second, the marked decrease in hospital LOS for both primary and revision total hip arthroplasty may have policy implications. The motivation for hospitals to reduce LOS under the Medicare prospective payment system has been well described,^42- 43 but the impact of declining LOS is much less clear. While Bueno et al²² noted that reductions in LOS for Medicare beneficiaries hospitalized with congestive heart failure were accompanied by an increase in readmission rates, a prior study by Baker et al⁴⁴ found that reductions in LOS were not associated with an increase in readmissions. To the best of our knowledge, no prior studies have evaluated associations with reduction in LOS in regard to joint arthroplasty.

We found that the reduction in hospital LOS in both primary and revision total hip arthroplasty patients was accompanied by a significant increase in the proportion of patients discharged to postacute care facilities (ie, nursing homes and rehabilitation centers), and a significant reduction in the proportion of patients discharged directly home. Moreover, we found that while the reduction in hospital LOS was not associated with an increase in readmission rates between 1992 and 1999, in more recent years, readmission rates have risen markedly. These findings reinforce the potential wisdom of moving to bundled payments, reimbursing for episodes of care, or a combination of both as a way for incentivizing the correct LOS, rather than perpetual reductions in LOS that seem to be occurring.

Our study has limitations that warrant brief mention. First, our study was limited to fee-for-service Medicare beneficiaries and thus extrapolation to other populations should be done with caution. That said, more than 60% of all total hip arthroplasty procedures are performed on Medicare enrollees, making this an appropriate data set for studying total hip arthroplasty. Second, our study relied upon administrative data and thus we were unable to evaluate some important arthroplasty outcomes including functional status and patient satisfaction.

In conclusion, mortality after primary total hip arthroplasty has declined moderately over time and mortality after revision total hip arthroplasty has remained stable despite substantial increases in patient complexity. There have also been marked reductions in hospital LOS and rising readmission rates.