Abstract
Aims: The anterior approach (AA) for total hip arthroplasty (THA) is associated with more rapid recovery when compared to traditional approaches. The purpose of this study was to benchmark healthcare resource utilization and costs for patients with THA via AA relative to matched patients.
Materials and methods: This study queried Medicare claims data (2012–2014) to identify patients who received THA via an AA from experienced surgeons, and matched these patients to a control cohort (all THA approaches). Direct and propensity-score matching were employed to maximize similarity between patients and hospitals in the two cohorts. Hospital length of stay (LOS), the proportion of patients discharged to home or home health, and post-acute claim payments during the 90-day episode were assessed. Generalized estimating equations were applied to control for imbalances between the cohorts and clustering of outcomes within hospitals.
Results: A total of 1,794 patients were included after patient matching. Patients who received AA had significantly lower mean hospital LOS vs patients in the control group (2.06 ± 1.36 vs 2.98 ± 1.58 days, p < .0001). The adjusted proportion of patients discharged to home was nearly 20 percentage points higher in the AA cohort vs the control cohort (87.3% vs 68.7%, p < .0001). Post-acute claim payments for AA patients were nearly 50% lower than those for control patients ($4,139 vs $7,465, p < .0001).
Conclusion: AA patients had significantly lower post-acute care resource use when compared to control patients. Further research is warranted to evaluate the cost effectiveness of AA among surgeons of varying experience levels.
Introduction
Total hip arthroplasty (THA) procedure volume in the US has more than doubled from ∼150,000 cases in 2000 to 326,100 cases in 2010, with continued growth anticipated due to the aging population and demand among younger, active patientsCitation1,Citation2. Notwithstanding the positive impact of this procedure on quality-of-life, it represents a significant and growing cost to payersCitation3. As such, THA has been the subject of payment reforms.
The Centers for Medicare and Medicaid Services’ (CMS) Comprehensive Care for Joint Replacement (CJR) program, implemented in 2016, has had a significant impact on economic incentives and care pathways for joint replacement. This program establishes penalties or rewards for hospitals relative to target payment amounts and quality thresholds for 90-day episodes of care after a lower-extremity joint replacement. This and similar bundled payment reforms across both private and public sectors underscore the need for hospitals and surgeons to focus on improving the quality and cost of care beyond the surgical admission.
THA in the US is most commonly performed via a posterolateral approachCitation4. Despite a successful track record of favorable outcomes, posterolateral and anterolateral approaches typically require detachment of stabilizing muscles of the hip, such as the short external rotators and gluteus medius. In contrast, the anterior approach utilizes an inter-neural and inter-muscular plane.
The anterior approach to THA has been associated with more rapid recovery and lower dislocation rates when compared to traditional approachesCitation5–7. However, the anterior approach has also been critiqued for a steep learning curve, and the potential for increased complications relative to traditional approachesCitation8. The full body of literature is mixed on the relative merits of each approach, with some studies indicating superiority of the anterior approachCitation7,Citation9–13, and others failing to identify differences in either the short- or the long-term outcomesCitation14–16.
To date there is a dearth of comparative literature evaluating economic outcomes after THA with the anterior approach in general, and no published studies are available on the economics of the anterior approach technique as described by Matta et al.Citation6. L’Hommedieu et al.Citation8 compared Medicare patients who received THA with either an anterior approach or posterior approach, and did not find significant cost differences between patient cohorts. This study did not focus on a standardized, homogeneous approach to anterior approach, as described herein. Furthermore, the inclusion and effect of surgeons’ learning curves is also not easily ascertained from prior work.
Our study, therefore, seeks to benchmark 90-day economic outcomes after THA among patients who have received a consistent direct anterior approach relative to those observed for the Medicare THA population at large after controlling for patient, physician, and hospital-level variables.
Materials and methods
Data sources
Medicare is the largest health insurance program in the US, providing coverage to persons 65 years or older and persons younger than 65 years who have end-stage renal disease or who are disabled. Medicare claims and enrolment (denominator) data from the 2012–2014 Medicare Standard Analytic File (SAF) were used for this study. All acute inpatient, home health, skilled nursing, and hospital outpatient department claims were included. The claim files include: International Classification of Diseases, Ninth Revision Clinical Modification, (ICD-9-CM) diagnoses associated with the service provided; characterization of the treatment received encoded by ICD-9-CM; Current Procedural Terminology (CPT)/Healthcare Common Procedure Coding System (HCPCS) or facility revenue center codes; service dates; and the amounts charged and paid for each service. Per CMS’s requirements, any cell counts below 11 are withheld from disclosure.
Sample selection
Patients 65 years of age or older who underwent a THA in the inpatient setting between January 1, 2012 and October 1, 2014 were identified through Medicare Severity-Adjusted Diagnosis Related Groups (MS-DRGs) 469 or 470 accompanied by primary ICD-9-CM procedure code 81.51. The “Index” date for each patient was defined as the day of discharge from initial THA. All patients were required to have continuous availability of data for at least 90 days after the Index date. Patients were excluded if they (i) had a non-Medicare primary payer, (ii) were eligible for Medicare due to end-stage renal disease (ESRD), (iii) had a hip fracture diagnosis during the index hospitalization (ICD-9-CM diagnosis codes: 733.10, 733.14, 733.15, 733.19, 820.xx, 821.xx), (iv) lacked 90-day Medicare Part A and Part B enrollment after index discharge, or (v) died anytime during the 90 days after the index discharge.
The anterior approach cohort consisted of the sub-set of the THA sample for six surgeons who used the anterior approach for THA as described by Matta et al.Citation6 for all of their THA patients between 2012 and 2014. All participating surgeons confirmed that their approach featured the use of Hana® table (an orthopedic table from Mizuho OSI, Inc., Union City, CA) and intra-operative fluoroscopy.
All six surgeons had performed a minimum of 500 anterior approach THAs prior to 2012, and were, therefore, well beyond the initial learning curve for the procedure. The CORAIL Hip System and PINNACLE Acetabular Cup System were the most commonly used implants for these procedures. THA patients who were not included in the anterior approach cohort were eligible for inclusion in the control cohort.
Medical claims in the database were used to record demographics, the presence or absence of selected comorbid conditions, and hospital and surgeon characteristics for each THA patient. The demographic and other general descriptive variables included age, gender, race, year of THA, and eligibility for Medicaid. Patient comorbidities included obesity (ICD-9-CM codes: V85.35, V85.36, V85.31, V85.39, V85.32, V85.33, V85.3, V85.37, V85.34, V85.38, 278.00), morbid obesity (ICD-9-CM codes: V85.44, V85.42, V85.45, V85.41, V85.42, V85.43, V85.4, V85.45, 278.01), diabetes with or without complications (ICD-9-CM codes: 250.0–250.4, 250.7-250.9), osteoarthritis of hip (ICD-9-CM codes: 715.15 or 715.95, 715.96), rheumatoid arthritis (ICD-9-CM codes: 714.0-714.2, 714.4, 714.8, 714.9), osteoporosis and osteopenia (ICD-9-CM codes: 733.01, 733.03, 733.09, 733.90), and the Charlson Comorbidity Index (CCI). The CCI is an aggregate measure of comorbidity representing 17 medical conditions (e.g. heart disease, cancer, diabetes), with higher scores indicative of greater comorbid burdenCitation17. The THA procedure volume for each surgeon comprised THA procedure counts between 2012 and 2014 for each national provider identifier (NPI) associated with the index claim. Hospital characteristics included the number of beds, geographic class (large urban, small urban, other urban, rural), volume of THA, teaching status (resident-to-bed ratio), and disproportionate share percentage.
Outcomes
Primary outcomes of the study included the percentage of patients discharged home (with or without home health services) after index surgery and total wage-adjusted 90-day post-acute care costs (Medicare payment amounts) after discharge. Secondary outcomes included length of stay and costs for the index hospitalization, as well as 90-day post-acute care cost and utilization in the following settings of care: home health, inpatient rehabilitation facility, skilled nursing facility, hospital outpatient department, and readmissions.
Statistical analysis
We took a multi-step approach to maximize similarity between patients and hospitals in the anterior approach cohort and those in the control group. First, we sought to determine if the ratio of anterior approach patients in the hospitals for each of the participating surgeons was 1:4 to control patients within the same hospital. If sufficient counts were unavailable for within-hospital controls, we identified hospitals from the nearest census divisions with characteristics similar to those of anterior approach hospitals from which to obtain control patientsCitation18. Hospitals were matched to anterior approach hospitals on the following, which were categorized by tertiles: (i) THA volume, (ii) disproportionate share percentage, and (iii) resident-to-bed ratio.
We then developed propensity-score models for each anterior approach hospital using demographic factors, comorbid conditions, year of THA, and Medicaid eligibility status. For hospitals with sufficient counts for within-hospital controls, direct matching was performed using the influential variables identified from the propensity score models: age, osteoarthritis of hip, obesity, year of hip arthroplasty, and Medicaid eligibility. For patients who could not be matched within the same hospitals, propensity scores were used to identify suitable control patients from matched hospitals. We used a greedy-matching technique for this purpose, wherein the matching algorithm proceeded in a hierarchical sequence until matches up to one digit of the propensity score were no longer possibleCitation19.
All matched patient data (direct and propensity-score) as described above were then used to develop a Generalized Estimating Equation (GEE) model, which sought to account for clustering within hospitals and to further control for imbalances between the anterior approach cohort and matched controls. A logit function was used to assess the proportion of patients discharged to home (with or without home health) vs other settings, while a log-link function with a gamma distribution was used to assess 90-day post-acute wage-adjusted Medicare payments. All regression models were adjusted for 18 characteristics including age, sex, race, year of surgery, CCI, obesity, morbid obesity, diabetes, osteoarthritis, rheumatoid arthritis, osteoporosis, Medicaid eligibility, surgeon THA volume, hospital resident-to-bed ratio, teaching status, disproportionate share percentage (DSH), hospital volume and number of hospital beds. Based on regression results, marginal outcomes of the anterior approach cohort relative to the control cohort were estimated using the method of recycled predictionsCitation20.
Between-cohort differences on baseline characteristics and outcome measures were compared using Student’s t-tests, Chi-square tests, and Wilcoxon rank-sum tests for continuous, categorical, and cost variables, respectively, before and after matching. Standardized differences and tests of significance were used to assess for between-group imbalances (). A standardized difference below 0.1 was concluded to indicate a negligible difference between compared groups for each measureCitation21.
Table 1. Baseline patient, hospital, and surgeon characteristics before and after matching.
Results
Unmatched cohorts: baseline characteristics
A total of 288,314 patients with THA between January 1, 2012 and October 1, 2014 were identified after applying inclusion/exclusion criteria. Baseline (pre-match) patient, hospital, and surgeon characteristics for the anterior approach and control cohorts are presented in . Patients were similar in terms of mean (± SD) age (72.12 ± 8.42 years vs72.50 ± 8.86 years, p = .176) and gender (59.7% females vs 61.9% females, p = .164). However, patients who received an anterior approach were more likely to be white (94.8% vs 92.0%, p = .005), more likely to have received THA in 2014 (36.6% vs 30.7%, p < .0001), and less likely to be eligible for Medicaid (8.3% vs 10.5%, p < .0001) than patients in the control cohort. Comorbidities were also less common among pre-matched anterior approach patients relative to control patients, respectively: obesity (4.3% vs 9.4%, p < .0001); rheumatoid arthritis (0.1% vs 3.5%, p < .0001); and osteoporosis/osteopenia (1.7% vs 2.9%, p = .040). In addition, patients who received an anterior approach were more likely than unmatched controls to have a CCI score of zero (65.1% vs 56.9%, p < .0001).
Hospital characteristics and surgeon volume for THA also differed between the two unmatched cohorts. Patients in the anterior approach cohort were less likely than controls to have received their THA in a teaching hospital (19.8% vs 43.3%, p < .0001) or a hospital in a large urban area (40.2% vs 46.1%, p < .0001). Mean surgeon volume for hip arthroplasty was significantly higher in the anterior approach cohort compared to the control cohort (275.82 ± 146.18 vs 83.85 ± 79.77, p < .0001).
Matched cohorts: baseline characteristics
After applying direct (486 patients) and propensity-score (1308 patients) matching techniques, 1,794 patients (897 anterior approach and 897 controls) remained available for comparative analysis. No significant between-group differences in baseline patient characteristics were observed for these matched cohorts. Patients who received an anterior approach were similar to control patients with respect to mean age (72.11 ± 8.5 vs 72.12 ± 7.8, p = .968), gender (59.4% females vs 59.5% females, p = .962), race (94.8% vs 94.7%, p = .699), year of THA (36.2% vs 37.4% in 2014, p = .863) and Medicaid eligibility (8.5% vs 7.8%, p = .604). Overall comorbidity burden, as measured by the CCI, was also similar for the two matched cohorts (all p > .05). However, significant differences persisted in hospital teaching status (20% vs 27.0%, p = .0004), large urban hospitals (40.6% vs 52.6%, p < .0001) and mean surgeon volume for THA (273.90 ± 145.69 vs 172.44 ± 198.08, p < .0001).
Matched cohorts—univariate analysis for direct medical resource utilization
Patients in the anterior approach cohort had significantly lower hospital length of stay for the index THA procedure when compared to matched patients in the control group (2.06 ± 1.36 days vs 2.98 ± 1.58 days, p < .0001) (). Utilization of any 90-day post-acute services () was also significantly lower among patients in the anterior approach cohort vs those in the control cohort: home health agency (58.3% vs 68.1%, p < .0001), skilled nursing facility (12.2% vs 28.5%, p < .0001), and inpatient rehabilitation facility (1.2% vs 6.4%, p < .0001). Readmissions were also statistically lower in the anterior approach cohort (5.0% vs 7.9%, p = .013).
Figure 1. Mean (95% CI) hospital length of stay for total hip arthroplasty.
Figure 2. Healthcare utilization during 90-day post-acute care in each setting.
Mean wage-adjusted payments for index hospitalizations were significantly lower for anterior approach patients vs control patients ($10,873 vs $12,825, p < .0001). Excluding small differences in mean per-patient hospital outpatient payments ($360 vs $330, p = .017), 90-day post-acute care wage-adjusted payments in all other settings of care were also significantly lower for the anterior approach cohort vs the control cohort: home health agency ($1,688 vs $2,095, p < .0001), skilled nursing facility ($1,244 vs $2,952, p < .0001), inpatient rehabilitation facility ($178 vs $965, p < .0001), and readmissions to inpatient care settings ($561 vs $1,269, p < .0001) ().
Figure 3. Mean (95% CI) index hospitalization and 90-day post-acute costs by setting.
Matched cohorts—multivariate analysis for primary end-points
Adjusted results from the multivariate GEE model indicate that patients in the anterior approach cohort were significantly more likely than controls to be directly discharged to home (with or without a home health agency) after the index THA than controls (87.3% vs. 68.7%, p < .0001). Adjusted, per-patient post-acute care payments during the 90-day period were nearly 50% lower for anterior approach patients vs controls ($4,139 vs $7,465, p < .0001), resulting in per-patient 90-day post-acute savings of $3,326 ().
Table 2. Predicted outcomes based on regression.
Discussion
To our knowledge, this is the first study to have benchmarked economic outcomes for a well-defined technique for an anterior approach in THA. Patients in the anterior approach cohort had economically important and statistically significantly lower post-acute utilization when compared to THA patients in the control cohort. These differences were associated with a $3,326 reduction in adjusted mean per-patient spending for the Medicare program, which represents 13% of the total episode payments as reported by the CMSCitation22.
L’Hommedieu et al.Citation8 compared anterior approach vs posterior approach patients using Medicare claims data, and did not observe differences in economic outcomes between approaches. In particular, the proportion of patients using SNF services—a key driver of post-acute costs—was approximately one-third for both anterior approach and posterior approach patients in that study. While a similar proportion of control subjects in our study used SNF services, this was true for only 12% of anterior approach patients. It is possible that the stark differences between outcomes for these two studies arise from differences in methodology (e.g. the use of multivariate modeling in our study), the experience of surgeons with the anterior approach technique, and/or variation in operative technique among represented surgeons.
Maratt et al.Citation23 conducted a retrospective analysis of 2,147 patients within the Michigan Arthroplasty Registry Collaborative Quality Initiative to compare dislocation rates between the anterior approach and a posterior approach. No statistical differences were observed between cohorts for 90-day dislocation rates, which were below 1% for both groups. Re-admission rates of ∼5% were observed for both patient cohorts, similar to that observed for the anterior approach cohort in our study. Mean length of stay was slightly lower for the anterior approach vs the posterior approach (2.54 vs 2.37 days, respectively), but this difference did not achieve statistical significance. Maratt et al.Citation23 did not control for surgeon volume, technique (e.g. intra-operative visualization and/or use of an orthopedic table) or experience with each approach. In contrast, our study included six high-volume surgeons experienced with a homogenous anterior approach technique.
Our study has important strengths and limitations. In particular, it is unique for the use of rigorous patient-matching techniques and for control of covariates spanning patient, surgeon, and hospital domains. An additional strength arises from the participation of six experienced surgeons who use the same anterior approach technique, as this eliminates learning-curve effects and minimizes technique-related heterogeneity within the anterior approach cohort. However, these aspects of the study do not allow for generalization of our findings to surgeons using a different surgical technique for anterior approach. This study is further limited by having included a mix of surgical approaches within the control cohort, a factor that precludes conclusions about the anterior approach vs any other specific technique, including alternative contemporary less invasive approaches to THA. We also limited our analysis to a Medicare fee-for-service population, and therefore cannot draw conclusions about results for patients with other payers, including commercial insurers. We used both direct and propensity score matching techniques to maximize similarity between patients, hospitals, and surgeons. However, post-matching differences persisted in terms of hospital teaching status, hospital geographic class and surgeon THA volume. We controlled for these unbalanced factors using a multivariate GEE model, thus providing doubly robust estimates. As with any retrospective study, unmeasured factors and selection/performance biases (e.g. patient and/or surgeon expectations) may have contributed to between-group differences. In particular, the represented surgeons may have adopted the anterior approach in concert with individual protocols designed to accelerate recovery from surgery.
Conclusions
Patients who received THA from one of six experienced surgeons who used a consistent anterior approach technique had significantly lower post-acute medical resource utilization when compared to matched patients in the overall Medicare population. These findings suggest that the anterior approach technique, as described in a standardized fashion after appropriate training, may be well suited to achieving success in a value-based payment environment.
Transparency
Declaration of funding
This study was funded by Johnson & Johnson (JnJ).
Declaration of financial/other relationships
AC, CH, and JL are employees of JnJ and own stocks in the company. JME peer reviewers on this manuscript have no relevant financial or other relationships to disclose.
Acknowledgments
The authors would like to thank David Wei and Sashi Yadalam for their help in the analysis of the study.
References
- Wolford ML, Palso K, Bercovitz A. Hospitalization for total hip replacement among inpatients aged 45 and over: United States, 2000–2010. NCHS data brief 2015:186;1-8
- McKee J. Younger patients demanding new joints. American Association of Orthopaedic Surgeons (AAOS). 2009. Available at: http://www.aaos.org/AAOSNow/2009/Feb/cover/cover3/?ssopc =1
- Losina E, Walensky RP, Kessler CL, et al. Cost-effectiveness of total knee arthroplasty in the United States: patient risk and hospital volume. Arch Intern Med 2009;169:1113-21
- Esposito CI, Gladnick BP, Lee YY, et al. Cup position alone does not predict risk of dislocation after hip arthroplasty. J Arthroplasty 2015;30:109-13
- Mallory TH, Lombardi AV, Fada R, et al. Dislocation after total hip arthroplasty using the anterolateral abductor split approach. Clin Orthop Relat R 1999;358:166-72
- Matta JM, Shahrdar C, Ferguson T. Single-incision anterior approach for total hip arthroplasty on an orthopaedic table. Clin Orthop Relat Res 2005;441:115-24
- Zawadsky MW, Paulus MC, Murray PJ, et al. Early outcome comparison between the direct anterior approach and the mini-incision posterior approach for primary total hip arthroplasty: 150 consecutive cases. J Arthroplasty 2014;29:1256-60
- L’Hommedieu CE, Gera JJ, Rupp G, et al. Impact of anterior vs posterior approach for total hip arthroplasty on post-acute care service utilization. J Arthroplasty 2016;31(9 Suppl):73-7
- Christensen CP, Jacobs CA. Comparison of patient function during the first six weeks after direct anterior or posterior total hip arthroplasty (THA): a randomized study. J Arthroplasty 2015;30:94-7
- Rodriguez JA, Deshmukh AJ, Rathod PA, et al. Does the direct anterior approach in THA offer faster rehabilitation and comparable safety to the posterior approach? Clin Orthop Relat R 2014;472:455-63
- Barrett WP, Turner SE, Leopold JP. Prospective randomized study of direct anterior vs postero-lateral approach for total hip arthroplasty. J Arthroplasty 2013;28:1634-8
- Hamilton WG, Parks NL, Huynh C. Comparison of cup alignment, jump distance, and complications in consecutive series of anterior approach and posterior approach total hip arthroplasty. J Arthroplasty 2015;30:1959-62
- Taunton MJ, Mason JB, Odum SM, et al. Direct anterior total hip arthroplasty yields more rapid voluntary cessation of all walking aids: a prospective, randomized clinical trial. J Arthroplasty 2014;29(9 Supp):169-72
- Poehling-Monaghan KL, Kamath AF, Taunton MJ, et al. Direct anterior versus miniposterior THA with the same advanced perioperative protocols: surprising early clinical results. Clin Orthop Relat R 2014;473:623-31
- Rathod PA, Orishimo KF, Kremenic IJ, et al. Similar improvement in gait parameters following direct anterior & posterior approach total hip arthroplasty. J Arthroplasty 2014;29:1261-4
- Higgins BT, Barlow DR, Heagerty NE, et al. Anterior vs posterior approach for total hip arthroplasty, a systematic review and meta-analysis. J Arthroplasty 2015;30:419-34
- Quan H, Sundararajan V, Halfon P, et al. Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care 2005;43:1130-9
- Bannuru RR, Schmid CH, Kent DM, et al. Comparative effectiveness of pharmacologic interventions for knee osteoarthritis: a systematic review and network meta-analysis. Ann Intern Med 2015;162:46-54
- Parsons L, editor. Performing a 1:N case-control match on propensity score. 2004. Montréal, Canada: SUGI
- Graubard BI, Korn EL. Predictive margins with survey data. Biometrics 1999;55:652-9
- Austin PC. An introduction to propensity score methods for reducing the effects of confounding in observational studies. Multivariate Behav Res 2011;46:399-424
- Medicare program; comprehensive care for joint replacement payment model for acute care hospitals furnishing lower extremity joint replacement services [proposed rule]. Federal Register 2015;80(134)
- Maratt JD, Gagnier JJ, Butler PD, et al. No difference in dislocation seen in anterior vs posterior approach total hip arthroplasty. J Arthroplasty 2016;31(9 Suppl):127-30