Total Knee Arthroplasty Volume, Utilization, and Outcomes Among Medicare Beneficiaries, 1991-2010 FREE

Total knee arthroplasty (TKA) is a common and safe procedure typically performed for relief of symptoms in patients with severe knee arthritis. Available data suggest that approximately 600 000 TKA procedures are performed annually in the United States at a cost of approximately $15 000 per procedure ($9 billion per year in aggregate).^1- 4 While TKA does not typically reduce mortality, the procedure results in marked improvements in health-related quality of life and functional status and is highly cost-effective.^2,5 Total knee arthroplasty is now among the most common major surgical procedures performed in the United States.⁶

The increase in TKA can be viewed as an indication of the success of this procedure in safely reducing pain and improving functional status for an aging population.^7- 8 However, the increase in TKA can also be viewed as yet another source of strain on government, insurers, individuals, and businesses struggling with unremitting growth in health care spending.^9- 11 Despite the clinical and economic policy importance of TKA, there are few analyses evaluating recent trends over time in use of and outcomes associated with TKA.^1,12- 14

Thus, the primary objective of our study was to evaluate longitudinal trends in primary and revision TKA volume, per capita utilization, and outcomes in the US Medicare population. The secondary objective was to examine patient and hospital factors associated with increased risk for hospital readmission given the growing likelihood of bundled payments for orthopedics in the near future.^15- 16

We linked 2 sequential Medicare Provider Analysis and Review (MedPAR) Part A data files (the first covering the period from 1991-2005 and the second from 2006-2010), each containing a 100% sample of hospitalizations for fee-for-service beneficiaries. These data were used to identify all enrollees aged 65 years and older who underwent primary or revision TKA between 1991 and 2010. Patients were identified using International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) procedure codes 81.54 for primary and codes 80.06, 81.55, 00.80, 00.81, 00.82, 00.83, and 00.84 for revision TKA.^17- 20

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 (eg, home, nursing home, inpatient rehabilitation, transfer to another acute care hospital, dead); death occurring up to 3 years after discharge; each patient's unique Medicare beneficiary number allowing for identification of patient readmissions; and each hospital's unique 6-digit identification number. Comorbid conditions present on the index admission were identified using algorithms described by Elixhauser et al,^21- 22 which consider 30 specific conditions and exclude comorbid conditions that may represent complications of care or that are related to the primary reason for hospitalization. This project was approved by the University of Iowa Institutional Review Board.

Our intention was to examine changes in volume, utilization, and outcomes of patients undergoing primary and revision TKA procedures. To generate appropriate analytical cohorts, we applied several inclusion and exclusion criteria (eFigure 1 and eFigure 2). First, we excluded Medicare HMO enrollees because the MedPAR data are incomplete for enrollees in such plans.

Second, we limited our cohort to the first primary (or revision) TKA performed on a given patient during any 30-day period using methods we have described previously.²³ We also excluded bilateral or staged procedures that occurred within the 30-day window; this exclusion is necessary because Medicare data historically have not included sidedness for a specific procedure. Thus, for a patient who underwent 2 primary TKA procedures in close temporal proximity, it is impossible to know if this represented an initial primary procedure followed by an early complication requiring a second procedure or a planned bilateral (ie, staged) procedure.

Third, because primary TKA is most often an elective procedure whereas revision TKA can be either an elective or more urgent procedure, we applied separate exclusion criteria to the primary and revision TKA populations in accordance with prior studies as described below. For primary TKA (eFigure 1), we sequentially excluded patients admitted through the emergency department (n = 18 497) and patients admitted after transfer from another acute care hospital (n = 3295); these exclusion criteria were developed to select a population of primary elective TKA patients. The revision TKA population (eFigure 2) did not exclude these types of patients because revision TKA can be an emergent or unscheduled procedure and thus exclusion of these populations would not make sense.

We examined the demographic characteristics and prevalence of key comorbid conditions for patients who underwent TKA between 1991 and 2010; for simplicity, data are presented separately for each 4-year period (eg, 1991-1994, 1995-1998, etc). We examined changes in the mean number of comorbid conditions per patient during each 4-year period. We used analysis of variance for comparisons of continuous variables and the χ² test for categorical variables and tested for differences in linear trends. All analyses were performed separately for primary and revision TKA patients.

We used graphical methods to plot the annual primary and revision TKA Medicare volume over time. We calculated per capita TKA utilization rates by dividing the number of procedures performed each year by the number of beneficiaries enrolled in the fee-for-service Medicare program and plotted these results graphically.

We compared linear trends in several important outcomes of interest for primary and revision TKA: hospital length of stay (LOS); discharge disposition; selected arthroplasty complications resulting in readmission within 30 days of discharge; and all-cause readmission rates within 30 days of discharge. Discharge disposition was categorized as home, skilled or intermediate care (which also incorporated outpatient rehabilitation), inpatient rehabilitation, and other. We examined changes in the rates of 6 separate adverse outcomes occurring during the index admission (mortality) or readmission within 30 days of discharge (mortality, pulmonary embolism, deep vein thrombosis, wound infection, postoperative sepsis, and myocardial infarction) that have been examined in prior studies of arthroplasty using administrative data.^20,24- 25

We also examined changes in rates of a composite outcome representing the occurrence of one or more of the individual adverse outcomes as well as all-cause readmission within 30 days of discharge. To evaluate the reasons for readmission among the primary and revision TKA cohorts, we applied the Agency for Healthcare Research and Quality (AHRQ) Clinical Classification Software (CCS).²⁶ This software synthesizes more than 14 000 ICD-9-CM codes into 231 mutually exclusive clinically meaningful disease categories. For each 4-year study period, we examined the 5 most common categories associated with readmission and the proportion of all readmissions during each period that were associated with each category; this allowed us to examine how the causes of readmission have changed over time.

We used standard logistic regression to calculate risk-adjusted 30-day readmission rates and composite outcome [(observed/adjusted) × unadjusted 20-year rates] and used standard linear regression to calculate risk-adjusted hospital LOS [(observed −adjusted) + unadjusted 20-year LOS].²⁷ These models adjusted for age (categorized as 65-69 years, 70-74, 75-79, and ≥80 years), sex, race (categorized as white, black, and other), and comorbidities to account for the changing demographics of the TKA populations over time.²⁸ Race was included in these models to allow us to account for previously documented racial disparities in joint arthroplasty when calculating standardized utilization rates for our analysis.^29- 30 We used graphical methods to plot discharge disposition, hospital LOS, readmission rates, and composite outcome between 1991 and 2010. All analyses were conducted separately for the primary and revision TKA cohorts.

We conducted several supplemental analyses of interest. Focusing on the most recent 4 years of data (2007-2010), we examined the relationship between patient and hospital factors and hospital readmission; as in prior analyses, primary and revision TKA were examined separately. We used bivariate methods to compare differences in patient and hospital factors among patients who did and did not experience readmission within 30 days of discharge. We then examined both patient-level and hospital-level factors that may have affected the 30-day readmission rate by employing a series of 4 standard logistic regression models that progressively adjusted for an increasing array of factors. In all models, the dependent variable was a binary variable with the value of 1 if a given patient was readmitted and 0 if not.

Model 1 adjusted for patient demographics alone (ie, age, race, sex); model 2 added adjustment for the number of comorbidities; model 3 added adjustment for hospital teaching status (major, minor, and nonteaching) and hospital procedural volume (calculated separately for the primary and revision TKA cohort and categorized by hospital volume quartiles); and model 4 added additional adjustment for each patient's hospital LOS, modeled in its log-transformed state. In all 4 models, we also included calendar year (2007, 2008, 2009, and 2010) to account for underlying temporal trends. We conducted several sensitivity analyses. In particular, we repeated our analyses after adding back excluded populations (eg, primary TKA cases admitted through the emergency department). We also repeated our analyses looking at 90-day outcomes rather than 30-day outcomes.

All P values are 2-tailed, with P <.05 deemed statistically significant. All statistical analyses were performed using SAS version 9.2.

Our final study population included 3 271 851 elective primary TKAs and 318 563 revision TKAs between 1991 and 2010. The total number of fee-for-service Medicare enrollees increased from 29 892 351 in 1991 to 39 250 746 in 2010, whereas the number of primary TKA procedures increased from 93 230 in 1991 to 243 802 in 2010 (an increase of 161.5%) (Figure 1). The number of revision TKA procedures increased from 9650 in 1991 to 19 871 in 2010 (an increase of 105.9%) (Figure 1).

During the same period, the per capita utilization of primary TKA increased by 99.2% (Figure 2) and the per capita utilization of revision TKA increased by 56.8% (Figure 2). For primary TKA, the mean (SD) age increased from 73.8 (5.8) years (95% CI, 73.8-73.8 years) in 1991-1994 to 74.2 (6.2) years (95% CI, 74.2-74.2 years) in 2007-2010, (P <.001). The prevalence of diabetes increased from 10.5% (95% CI, 10.4%-10.6%) to 21.7% (95% CI, 21.6%-21.7%) and the prevalence of obesity increased from 4.0% (95% CI, 3.9%-4.0%) to 11.5% (95% CI, 11.4%-11.6%; P < .001 for each). Trends were similar for revision TKA (Table 1). In particular, the mean (SD) age increased from 74.2 (5.9) years (95% CI, 74.1-74.3 years) in 1991-1994 to 74.8 (6.5) years (95% CI, 74.7-74.8 years) in 2007-2010 (P <.001). The prevalence of diabetes increased from 11.0% (95% CI, 10.7%-11.3%) to 24.2% (95% CI, 23.9%-24.5%) and the prevalence of obesity increased from 3.7% (95% CI, 3.5%-3.8%) to 10.1% (95% CI, 9.9%-10.3%) (P < .001 for each).

For primary TKA, the mean hospital LOS declined from 7.9 days (95% CI, 7.8-7.9) in 1991-1994 to 3.5 days (95% CI, 3.5-3.5) in 2007-2010 (Table 2), a relative decline of 55.7% (P < .001). The percentage of patients discharged home after primary TKA declined from 67.5% (95% CI, 67.3%-67.6%) in 1991-1994 to 39.9% (95% CI, 39.8%-40.0%) in 1999-2002 before increasing to 56.2% (95% CI, 56.1%-56.3%) in 2007-2010 (Table 2 and eFigure 3); alternatively the percentage of patients discharged to inpatient rehabilitation increased from 14.6% (95% CI, 14.5%-14.7%) in 1991-1994 to 29.4% (95% CI, 29.3%-29.5%) in 1999-2002 before declining to 11.4% (95% CI, 11.3%-11.4%) in 2007-20120 and discharge to outpatient rehabilitation facilities increased steadily throughout the study period. Unadjusted mortality within 30 days after discharge decreased from 0.5% (95% CI, 0.4%-0.5%) in 1991-1994 to 0.3% (95% CI, 0.3%-0.3%) in 2007-2010, a 40% relative reduction (P < .001). Unadjusted rates of most other adverse outcomes remained relatively stable over the study period as did the rate of the composite outcome (Table 2).

In contrast, all-cause 30-day readmission rates increased from 4.2% (95% CI, 4.1%-4.2%) in 1991-1994 to 5.0% (95% CI, 4.9%-5.0%) in 2007-2010 (P < .001)(Table 2). In adjusted analyses, we found that although hospital LOS for primary TKA declined throughout the study period (eFigure 4), both 30-day all-cause readmission rates (eFigure 5) and rates of the composite outcome (eFigure 6) declined initially, but have been increasing in recent years. In an analysis of the diagnoses and conditions associated with readmission after primary TKA, we observed relatively little change over time (eTable 1) with surgical and cardiac complications being relatively common as well as gastrointestinal hemorrhage and infection, particularly in recent years.

For revision TKA, the mean hospital LOS declined from 8.9 days (95% CI, 8.8-8.9) in 1991-1994 to 5.0 days (95% CI, 5.0-5.0) in 2007-2010, a relative decline of 43.8% (P < .001; Table 3). Trends in discharge disposition after revision TKA (Table 3 and eFigure 3) demonstrated a similar pattern to that which was observed for primary TKA, a decline in discharges to home or inpatient rehabilitation and an increase in discharge to skilled care and outpatient rehabilitation. Mortality within 30 days of discharge increased modestly from 0.7% (95% CI, 0.6%-0.7%) in 1991-1994 to 0.9% (95% CI, 0.8%-0.9%) in 2007-2010 (a 28.6% relative increase) and readmission for wound infection, hemorrhage, sepsis, and myocardial infarction each increased by more than 100% (P < .001 for each). For revision TKA, the unadjusted rate of the composite outcome increased from 2.7% (95% CI, 2.6%-2.9%) in 1991-1994 to 5.3% (95% CI, 5.2%-5.5%) in 2007-2010 (P < .001). All-cause unadjusted readmission rates within 30 days of discharge increased from 6.1% (95% CI, 5.9%-6.4%) to 8.9% (95% CI, 8.7%-9.2%) during the study period (eFigure 5). The most common causes of readmission after revision TKA are displayed in eTable 2.

In adjusted analyses, revision TKA demonstrated a steady decrease in hospital LOS (eFigure 4) accompanied by an initial decline in readmissions that has reversed in recent years (eFigure 5) and by an increase in both unadjusted and adjusted rates of the composite outcome (eFigure 6).

In bivariate comparison of patients who were and were not readmitted within 30 days after primary TKA (Table 4), we found that patients who were readmitted were older than those who were not readmitted (mean age, 75.6 vs 74.1 years, P < .001), less likely to be women, more likely to be black, and had a higher number of comorbid conditions (mean number of conditions 2.1 among nonreadmitted vs 2.5 among readmitted, P < .001). Findings were generally similar for the revision TKA cohort (Table 4). In our regression analyses focusing on primary TKA readmissions (Table 5), several patient-level factors were associated with increased odds of hospital readmission including older age, black race, male sex, greater number of comorbid conditions, and longer hospital LOS during the index admission. Hospital factors including minor teaching and nonteaching status (as compared with major teaching) as well as greater primary TKA hospital volume were associated with decreased patient readmission rates (Table 5). Results for revision TKA were generally similar (Table 5). Results of sensitivity analyses (available by request from the authors) were similar to the main results described above.

In an analysis of Medicare administrative data from 1991-2010, we identified a number of interesting trends related to TKA. First, we found a marked increase in the volume of primary TKA procedures being performed, an increase that appeared to be driven not only by an increase in the number of Medicare enrollees but also a substantial increase in the per-capita utilization of TKA procedures. Second, we observed changes in patients' discharge disposition over time with a decline in the use of inpatient rehabilitation and an increase in the use of outpatient rehabilitation. Third, we found a significant decrease in hospital LOS that was accompanied by increasing readmission rates over the past decade.

Primary Medicare TKA volume increased approximately 162% from 93 230 in 1991 to 243 802 in 2010 and revision volume increased 106% from 9650 in 1991 to 19 871 in 2010. These figures suggest that growth in primary and revision TKA volume is being driven by both an increase in the number of Medicare enrollees and an increase in per capita arthroplasty utilization. Our findings extend those of a limited body of prior research that has demonstrated increasing volume and per capita utilization of knee arthroplasty.^1,31- 32 This growth is likely driven by a combination of factors including an expansion in the types of patients considered likely to benefit from TKA, an aging population, and an increasing prevalence of certain conditions that predispose patients to osteoarthritis, most notably obesity.³³

It is important to note the apparent stabilization of joint arthroplasty utilization in recent years. Our findings extend the work of Bini and colleagues³⁴ who found evidence of slowing growth in joint arthroplasty utilization within the Kaiser-Permanente health care system between 2000 and 2009. It is unclear whether this slowing of joint arthroplasty growth is the result of the protracted US economic downturn, saturation of patient demand for arthroplasty, changes in reimbursement, or changes in provider beliefs about the risks and benefits of arthroplasty.^35- 36

The growth in TKA should prompt consideration of whether too many (or too few) of these procedures are being performed both in aggregate and among key patient subgroups defined by race, sex, or age.^30,37- 38 Any effort to answer this question raises the issues of TKA indications and appropriateness. A number of clinical practice guidelines for TKA have been developed to guide clinicians and policy makers in evaluating appropriateness.^39- 43 These guidelines typically suggest consideration of TKA for patients with severe functional limitation unresponsive to conservative management (ie, medications and physical therapy).

While Cobos et al⁴⁴ estimated that as many as 25% of TKA procedures performed in Spain might be considered inappropriate, few such studies have been performed in the United States.¹² Conducting studies investigating appropriateness has historically been difficult because of a lack of a national joint arthroplasty registry, although there have been encouraging developments recently to suggest that this may change.^45- 46 Thus, it is difficult to determine the extent to which the growth in TKA utilization represents growth in appropriate use of a highly effective procedure or overuse of a highly reimbursed procedure for which indications still depend on clinical judgment. It is likely that both factors are at play.

Our finding of significant changes in patient's discharge dispositions following TKA over the 20-year study period is important and hints at the complexities of restraining cost growth. The increase in the percentage of TKA patients discharged to inpatient rehabilitation and skilled care during the 1990s is consistent with prior reports.⁴⁷ These reports typically relate the increased use of post–acute care to the implementation of the prospective payment system for acute care hospitals in 1983, which in turn created a powerful incentive for hospitals to reduce hospital LOS by rapidly discharging patients to post–acute care settings when patients were too ill to safely be discharged home.^48- 49 However, the rapid increase in Medicare post–acute care spending in the 1990s prompted passage of the Balanced Budget Act (BBA) of 1997 and implementation of a prospective payment system for outpatient skilled care in 1998 and inpatient rehabilitation in 2002.^48,50 Our results are consistent with the anticipated effects of these policy changes, a reduction in the use of post–acute care and an increase in the percentage of patients being discharged home after TKA since 2004.

The finding of declining hospital LOS accompanied by increasing readmission rates mirrors results of a number of recent studies.^51- 52 The results of our study as well as other publications suggest that there are limitations to what extent LOS can be reduced and that cost savings from further LOS reductions are unlikely to materialize.^51- 52 In particular, there is an inherent tradeoff between shorter hospital LOS, greater need for post–acute care, and higher readmission rates.

A number of other findings merit mention. Our finding of increased comorbidity over time likely reflects a combination of factors including increasingly aggressive coding practices and increasing prevalence of certain comorbidities (eg, diabetes and obesity).^53- 54 The increasing rates of many surgical complications including myocardial infarction, infection, and hemorrhage particularly after revision TKA accompanied by a much smaller increase in mortality is interesting. It seems likely that many of these increases reflect more aggressive testing combined with detection bias resulting from newer more sensitive diagnostic tests (eg, troponin for myocardial infarction or D-dimer for deep vein thrombosis) rather than a true increase in surgical complications.⁵⁵ However it is also possible that the incidence of certain complications such as myocardial infarction may be increasing, perhaps as a consequence of a greater burden of obesity and diabetes.

Arguably, the most concerning complication is the increase in readmissions for infection in the revision TKA cohort. While there are well recognized limitations in administrative data for identifying surgical site infections,^56- 57 our findings should not be discounted prematurely.^58- 59 There are several potential explanations for increasing infection rates in the revision TKA population. One possibility is that the increase in infections represents an increase in revision TKAs being performed specifically to treat infected prostheses. If this were the case, the increase in revision TKA procedures performed would constitute infections that were “present on admission.”^60- 61 Alternatively, it is possible that the increase in infections represents a real increase in postoperative surgical infections after revision TKA perhaps as a consequence of the increasingly resistant organisms colonizing hospitals. It is also possible that reduced hospital LOS may lead to reduced vigilance for early signs of superficial wound infection in the postoperative period resulting in higher rates of serious infectious complications. In either case, the increase in infection rates associated with revision TKA warrants close attention.

In addition, the increase in primary TKA utilization (99%) has been larger than the increase in revision TKA utilization (51%) over the past 20 years. It is possible that this reflects the durability of modern implants and improved surgical technique resulting in a reduced likelihood that patients undergoing primary TKA will require a revision procedure in the future.⁶² Alternatively, it is possible that the rapid increase in primary TKA over the past 20 years will eventually result in a substantial increase in demand for revision TKA procedures as prosthetic devices wear over time, a possibility that would have significant clinical and economic implications.³⁵

Our study has a number of limitations. First, our study was limited to fee-for-service Medicare beneficiaries who constitute approximately 60% of the TKA population.^14,31 Our findings should be extrapolated with caution to other populations including younger patients and Medicare managed care enrollees. Second, our study relied upon administrative data and thus we were unable to evaluate a number of important arthroplasty outcomes including functional status and patient satisfaction. Third, we lacked clinical detail and therefore were unable to determine the indications for TKA at the level of the individual patient. Likewise, we lacked chart review data for identification of TKA outcomes and complications. Fourth, we focused our analysis on TKA adverse outcomes resulting in hospital readmission within 30 days of discharge. We were unable to identify complications that did not result in an inpatient admission and our 30-day follow up interval would not capture certain late complications (eg, infection) that may become apparent over a longer time period.²⁵

In summary, over the past 20 years, increases in TKA volume have been driven by both an increase in the number of Medicare enrollees and increase in per capita utilization. We also observed decreased hospital LOS that was accompanied by increased hospital readmission rates and rising rates of infections complications.