Discharge Destination After Revision Total Joint Arthroplasty: An Analysis of Postdischarge Outcomes and Placement Risk Factors

Results

Comparison of Patient Characteristics and Comorbidities

A total of 9973 RJR patients from 2011 to 2013 were included for analysis. The most common discharge destination included home (66%), SNF (23%), and IRF (11%). Compared with those discharged home, SNF- and IRF-bound patients tended to be older, female, and nonsmokers (all P < .001; Table 1). Nonhome patients were also more likely to have body mass index (BMI) >40, diabetes, pulmonary disease, cardiac disease, hypertension, stroke, renal disease, bleeding-causing disorders, ASA class 3/4, greater days from admission to operation, and longer operative time (all P < .001). In terms of revision etiology, fracture (5.0% vs <1%), infection (12% vs 9.2%), and dislocation (10% vs 9.1%) were more common in the nonhome vs home cohort (P < .001).

Comparison of Adverse Events

Bivariate analysis of predischarge adverse events revealed greater incidence of SAEs (5.4% vs 2.5%), minor adverse events (2.3% vs 0.6%), and infectious complications (2.8% vs 1.6%) among nonhome discharged patients (all P < .001; Table 2). In particular, rate of unplanned return to the operating room (1.8% vs 0.4%), sepsis/septic shock (1.6% vs 0.6%), unplanned intubation (0.5% vs 0.1%), thrombolic event (0.4% vs 0.1%), and myocardial infarction (0.5% vs 0.2%) were drivers for the observed difference in predischarge SAEs between nonhome vs home cohorts. Average total length of stay (LOS; 5.5 vs 3.7 days) was higher for nonhome patients (P < .001). There was no difference in predischarge adverse events or average LOS for IRF vs SNF cohorts. Similar analysis of postdischarge adverse events revealed higher rate of SAEs (6.1% vs 4.1%, P < .001) and unplanned 30-day readmissions (9.3% vs 6.1%, P < .001) in nonhome vs home patients, driven by greater incidence of thrombolic events (1.3% vs 0.6%), unplanned return to operating room (3.6% vs 2.6%), and death (0.4% vs 0.2%) in nonhome patients (P ≤ .02 for all; Table 3).

Predictors of Discharge Destination

Multivariate analysis controlling for patient characteristics, comorbidities, and SAEs predischarge identified fracture (odds ratio [OR] = 2.91) and infection (OR = 1.36) etiologies relative to mechanical loosening, functional status (OR = 2.18), BMI >40 (OR = 1.35), renal disease (OR = 2.27), and ASA class 3/4 (OR = 1.54) as the strongest independent predictors of nonhome discharge destination (P ≤ .001; Table 4). Other significant risk factors included fracture etiology relative to mechanical loosening, 2-component revision (as opposed to 1), history of smoking, pulmonary disease, diabetes, hypertension, and bleeding-causing disorders (all OR ≥ 1.16, P ≤ .04).

Predictors of Postdischarge SAEs and Unplanned Readmission

In multivariate analysis, SNF and IRF patients were 1.51 and 1.30 times more likely to suffer an unplanned 30-day readmission, respectively, relative to home patients (P ≤ .05; Table 5). IRF patients were also 1.57 times more likely to suffer postdischarge SAEs as compared with those discharged home, whereas SNF destination was not a significant independent risk factor for postdischarge SAEs. The same analysis revealed that patients who suffered a predischarge SAE faced 1.82 times odds of postdischarge SAE and 1.80 times odds of unplanned readmission (P < .001). Infection etiology relative to mechanical complication, dependent functional status, BMI >40, and ASA class 3/4 were found to be significant risk factors for postdischarge SAEs (P ≤ .05). The same 5 factors and male gender were significant risk factors for unplanned readmission. Based on the area under the receiver operating characteristic curve (0.65 for predicting both unplanned readmission and SAE after discharge), the ability of this model to correctly classify those who will suffer unplanned readmission or SAE after discharge vs not is fair at best.

Discussion

Given Medicare's goal for >50% of payment to move through alternative payment models by the end of 2018, models such as bundled payments will need to target more complex and less clinically homogenous procedures/disease states to expand cost coverage and more effectively align incentives for physicians, hospitals, payers, and policy makers to improve quality and reduce costs [13]. Medicare's introduction of the CJR proposal that intentionally includes hip fracture etiology in the mandatory primary TJA bundle and the Oncology Care Model that rewards oncologists for managing cost of care over 6-month episodes (as opposed to 90 days) indicates a willingness on Medicare's part to make such moves 14 and 15. Given the rising prevalence and higher cost per case of revision joint arthroplasty (RJR) as compared with primary, the next application of bundles in orthopedics (excluding spine) will likely be RJR 16 and 17. Postacute care has been identified by several primary TJA studies and bundle demonstrations as a major area of opportunity to improve outcomes and reduce cost of care 7, 8 and 18. As such, the aim of this study was to compare postdischarge adverse events by discharge destination, identify risk factors for discharge placement, and stratify RJR patients based on these risk factors to identify the most appropriate discharge destination. A secondary objective was to compare how these results differed for primary vs revision TJA using the NSQIP database.

Understanding risk factors for nonhome discharge placement is critical for improving discharge planning and for addressing modifiable risk factors to increase the percent of patients discharged home after procedure. This is especially important in the case of RJR as it is well documented that 90-day complication rates and LOS in hospital, IRF, and SNF settings are often higher for revision vs primary TJA patients 19, 20, 21, 22, 23 and 24. Although several studies have evaluated drivers for longer LOS in nonhome discharge settings for RJR patients, none to our knowledge have evaluated factors influencing initial discharge disposition in this population 23 and 25. Multiple studies have assessed this for primary TJA and identified age, female gender, Medicare insurance, living alone, obesity, health failure, ASA class 3/4, longer LOS, and postoperative pain to be associated with increased risk of nonhome discharge 26, 27 and 28. Beyond these, our study identified infection, fracture, and dislocation etiology (relative to mechanical loosening); 2-component revision (relative to 1 component); functional status; history of smoking, diabetes, pulmonary disease, renal disease, and bleeding-causing disorders to be independent predictors of nonhome discharge destination. These findings are congruent with our expectations as patients with fracture, infectious, and dislocation etiologies of RJR have been shown to be more clinically complex, at higher risk for adverse events, and higher resource utilizers during and beyond the initial hospital stay as compared with those with aseptic loosening 20, 29, 30 and 31.

Among the factors we identified in this study, BMI >40, diabetes, hypertension, and history of smoking (all OR ≥ 1.21, P ≤ .003) are modifiable patient risk factors for nonhome discharge that should be optimized well before surgery. Several models, such as the perioperative surgical home, rely on a multidisciplinary team (including mid-level providers, anesthesiologists, and pharmacists) to activate the patient and optimize these factors before surgery. However, these models are being developed for patients undergoing primary TJA and require further testing before they can be adapted for RJR 32 and 33. Moreover, some etiologies for revision such as infection or fracture are often unpredictable and urgent, therefore limiting the applicability of the perioperative surgical home. This raises a potential opportunity for models such as the Integrated Practice Unit that seek to minimize risk of RJR by using demographic data, patient-reported outcomes, and clinical knowledge of the patient to develop a personalized risk score for complications after primary TJA (including need for revision) 34 and 35. These scores can help the care team determine which patients would benefit from targeted ancillary services (eg, case management, social work, nutrition/weight management) to address some of the preventable drivers of RJR (such as falls) and, in the case of elective TJA, can also help the team better determine which patients are appropriate primary TJA candidates in the first place 34, 35 and 36.

Currently, the most prevalent tool for predicting discharge destination after joint arthroplasty is the Risk Assessment and Prediction Tool (RAPT); however, 2 studies analyzing the accuracy of RAPT in 3213 primary TJA patients have shown that the tool is only 62%-65% accurate for the patients (>50% of the total cohort) who received intermediate scores between 6 and 10 (RAPT is on a 12-point scale) 37 and 38. The predictive accuracy of RAPT could be equally poor for intermediate scoring RJR patients, as such building a modified RAPT tailored to RJR patients that includes some of the strong nonhome discharge predictors we identified (including revision etiology, 2-component revision, ASA class 3/4, diabetes, smoking, and BMI >40) could improve RJR discharge planning.

Although understanding drivers of nonhome discharge optimizes around today's discharge patterns, analysis of outcomes in those settings is required to assess the value of nonhome discharge destination for RJR patients. To our knowledge, no studies have assessed differences in outcomes by setting for RJR patients although a study by Cram et al [39] evaluated 348,596 RJR patients over 18 years and found that the percent of patients discharged to nonhome discharge destination increased ∼16% while the rate of 30-day and 90-day unplanned readmission increased by 5.9% and 8.0%, respectively. Multiple studies have evaluated the influence of discharge disposition on postoperative outcomes for primary TJA including one by Bini et al [40] which found that of 9150 primary TJA patients, those discharged to a SNF had significantly higher readmission rates (5.2% vs 2.4%, P = .001) as compared with those discharged home. Our prior analysis of 109,153 primary TJA cases using the NSQIP database also found similar results, with SNF/IRF patients experiencing higher rates of 30-day SAEs (2.3% vs 1.2%, P < .001) and unplanned readmissions (5.2% vs 2.9%, P < .001) as compared with patients discharged home. Similar to these studies, we observed that nonhome discharge RJR patients had significantly higher rates of 30-day SAEs (6.1% vs 4.1%, P = .002) and unplanned 30-day readmission (9.3% vs 6.1%, P < .001). This is likely due to the fact that SNF and IRF RJR patients are more clinically complex than those discharged home as is shown in our Table 1 and in the other studies mentioned previously 28 and 40.

We found SNF and IRF discharge destination to be independent predictors of 30-day unplanned readmission (P < .001 for both) which is consistent with our previous findings and those of Bini et al [40]. However, in contrast to those studies, IRF was found to be an independent risk factor for 30-day postdischarge SAEs (P = .001), while SNF was not. One likely possibility for this is that the actual medical (eg, pancreatitis, anemia, acute joint pain/stiffness) causes of unplanned 30-day readmission in RJR patients are somewhat different from the general set of adverse events captured by NSQIP and that these causes are independently more likely to occur in the SNF setting 30 and 31. As with primary TJA, our results suggest that the greater surveillance and care resource intensity offered by IRF and SNF do not outweigh the additional risks RJR patients face by virtue of being in these settings. As noted by Carter et al [41], this is not surprising given that (1) the “nurse” component of Medicare's SNF/IRF reimbursement system does not correlate to actual nurse time spent with the patient, (2) costs of nonrehab ancillary service payments (eg, for expensive medications, ventilator) are not reflected in reimbursement incentivizing underprovision of care, and (3) no component of SNF reimbursement is tied to risk-adjusted 30/90-day clinical outcomes such as unplanned readmission. Moreover, a recent study by Unroe et al [42] showed that Medicare SNF quality ratings are only moderately associated with 90-day readmission and mortality rates for heart failure patients. Unfortunately, the current SNF reimbursement and rating systems provide very little incentive for SNF staff to work with the patient's initial care team (including the orthopedic surgeon) when complications arise and to manage such instances in-house as opposed to sending the patient back to hospital [43]. Although Medicare' CJR legislation is one step in the right direction for incentivizing SNFs to provide more coordinated and higher quality, lower cost care, true alignment with value can only be achieved by tying SNF reimbursement to 90-day clinical outcomes and patient-reported outcomes [11].

To identify the most appropriate discharge destination for patients of varying clinical complexity, we used the 6 significant risk factors for unplanned 30-day readmission and stratified patients based on number of risk factors and discharge destination (Table 6). Home discharge destination yielded significantly lower rates of unplanned readmission as compared with IRF patients at all risk levels (3 of 4 risk levels, P ≤ .05) except those with ≥3 risk factors (P = .11) and SNF patients with only 1 risk factor (P = .001). These results suggest that home discharge destination (when feasible) is the optimal strategy compared with IRF; however, the evidence is not strong enough to make the same statement for home vs SNF  Although we recognize the reality that certain types of RJR patients may need the resources provided by IRF/SNF setting and that our exploratory analysis does not perfectly risk adjust (ie, it only used the 6 independent predictors of unplanned readmission), we do believe that it can be used as one input into the shared decision-making dialogue between patient and provider regarding appropriate discharge destination after RJR.

There are several limitations to our study. First, within the home cohort, we were unable to distinguish between home self-managed patients vs those discharged home with home health services as these data are not collected by NSQIP. It is likely, however, that most patients who go home after RJR will receive home health services. Second, approximately 26% (2634 patients) of our revision TJA cohort had a revision etiology of “other/unknown” because of absent or incorrectly coded primary ICD-9 diagnosis codes. Third, because NSQIP does not collect outcomes beyond 30 days after principal procedure, we were unable to analyze the 90-day postprocedure adverse event risk. Finally, adverse events and readmissions occurring outside of the index hospital or health system are not captured in the NSQIP data; thus, the actual rate of these events is likely underestimated in our analysis.

Conclusion

RJR patients who are discharged to SNF or IRF have significantly increased risk for unplanned readmissions as compared with patients discharged home, whereas only IRF patients have a higher risk of suffering 30-day postdischarge SAEs. Across risk levels, home discharge destination (when feasible) is the optimal strategy compared with IRF, although the distinction between SNF and home is less clear. Modifiable risk factors for nonhome discharge and postdischarge adverse events should be addressed preoperatively to reduce the need for higher acuity discharge disposition and improve patient outcomes across discharge settings.

References

• • 8
  • K.J. Bozic, L. Ward, T.P. Vail, et al.
  • Bundled payments in total joint arthroplasty: targeting opportunities for quality improvement and cost reduction
  • Clin Orthop Relat Res, 472 (1) (2014), p. 188
• • 9
  • R. Snow, J. Granata, A.V. Ruhil, et al.
  • Associations between preoperative physical therapy and Post-acute care utilization patterns and cost in total joint replacement
  • J Bone Joint Surg Am, 96 (19) (2014), p. e165
• • 10
  • S.S. Liu, A.G. Della Valle, M.C. Besculides, et al.
  • Trends in mortality, complications, and demographics for primary hip arthroplasty in the United States
  • Int Orthop, 33 (3) (2009), p. 643
• • 11
  • M.E. Porter
  • What is value in health care?
  • N Engl J Med, 363 (26) (2010), p. 2477
• • 12
  • A.J. Pugely, J.J. Callaghan, C.T. Martin, et al.
  • Incidence of and risk factors for 30-day readmission following elective primary total joint arthroplasty: analysis from the ACS-NSQIP
  • J Arthroplasty, 28 (9) (2013), p. 1499

• • 25
  • S.M. Yeung, A.M. Davis, R. Soric
  • Factors influencing inpatient rehabilitation length of stay following revision hip replacements: a retrospective study
  • BMC Musculoskelet Disord, 11 (2010), p. 252
• • 26
  • G. Dejong, S.D. Horn, R.J. Smout, et al.
  • Joint replacement rehabilitation outcomes on discharge from skilled nursing facilities and inpatient rehabilitation facilities
  • Arch Phys Med Rehabil, 90 (8) (2009), p. 1284