Date Presented 4/19/2018
Self-care and mobility are two main functional areas in occupational therapy. This study determined the optimal method to analyze functional scores (self-care and mobility) to predict successful community discharge for Medicare beneficiaries across postacute settings.
Primary Author and Speaker: Chih-Ying (Cynthia) Li
Additional Authors and Speakers: Amol Karmarkar, Kenneth Ottenbacher, and Hemalkumar Mehta
PURPOSE: Currently, three functional assessments are practically used across postacute care (PAC) settings: Inpatient Rehabilitation Facility Patient Assessment Instrument (IRF–PAI) in inpatient rehabilitation facilities (IRFs), Minimum Data Set (MDS) in skilled nursing facilities (SNFs), and Outcome and Assessment Information Set (OASIS) in home health agencies (HHAs; Centers for Medicare & Medicaid Services, 2014). Determining the optimal self-care and mobility functional categories across the three assessments to predict successful community discharge is crucial but has not been established.
This study determined the optimal method to analyze functional score categories (self-care and mobility) using Medicare claims data to predict successful community discharge. Accurately categorizing functional ability can reflect reliable rehabilitation outcomes and further improve long-term care planning for patients, which is important for all stakeholders, including patients, practitioners, researchers, and health care policy makers.
METHOD: In this retrospective cohort study, we included only patients who were alive at index acute discharge and excluded those discharged with same-day readmission. We included 125,767 people with stroke (ischemic or hemorrhagic) across IRFs, SNFs, and HHAs. The primary outcome was successful community discharge, defined as staying in the community >30 days after discharge from the hospital. We compared outcome predictions using the three methods for generating functional score categories: (1) admission summed-score distribution proportion (quartile), (2) change score (between admission and discharge) distribution proportion (quintile), and (3) person strata based on Rasch modeling (Andrich, 1982). Rasch person strata generated four categories for self-care and three for mobility empirically. We used six indices to determine the best predictive model: area under the curve (AUC), Somer’s delta (Somer’s D), Akaike information criterion (AIC), Bayesian information criterion (BIC), integrated discrimination improvement (IDI), and net reclassification improvement (NRI).
RESULTS: A total of 50,424 cases (40.1%) were assessed with IRF–PAI in IRFs, 47,226 cases (37.6%) with MDS in SNFs, and 28,117 cases (22.4%) with OASIS in HHAs. All six indices (Pencina, & D’Agostino, 2004; Pencina et al., 2008) reported consistent results, with Method 3, Rasch person strata, the optimal method to predict outcomes (successful community discharge; AUC = 85.31%, Somer’s D = 0.706, AIC = 112072.48, BIC = 112285.72), followed by Method 2, quintile change score distribution proportion (AUC = 84.92%, Somer’s D = 0.6983, AIC = 113343.27, BIC = 113585.59), then Method 1, quartile admission summed-score distribution proportion (AUC = 84.2%; Somer’s D = 0.6838; AIC = 115601.73; BIC = 115824.67). For Method 1 vs. Method 2, IDI = 0.015 and NRI = 0.158; for Method 2 vs. Method 3, IDI = 0.00078 and NRI = 0.190; and for Method 1 vs. Method 3, IDI = 0.023 and NRI = 0.448.
CONCLUSION: For decades, score distribution proportion (e.g., tertile, quartile, or quintile) has been the most widely used method to categorize patient function, especially in administrative large data research. However, our results demonstrate that Rasch person strata generated the most optimal functional categories to predict successful community discharge for Medicare beneficiaries compared with the traditional distribution proportion method. Using inaccurate functional categories may underestimate rehabilitation outcomes and skew outcome predictions. Occupational therapy can pioneer large data research using optimal functional categories, thereby providing the best predictive treatment model in rehabilitation.
References
Andrich, D. (1982). An index of person separation in latent trait theory, the traditional KR20 index, and the Guttman scale response pattern. Educational Psychology Research, 9, 95–104.
Centers for Medicare & Medicaid Services. (2014). IMPACT Act of 2014 and cross setting measures. Retrieved from https://www.cms.gov/Medicare/Quality-Initiatives-Patient-Assessment-Instruments/Post-Acute-Care-Quality-Initiatives/IMPACT-Act-of-2014/IMPACT-Act-of-2014-Data-Standardization-and-Cross-Setting-Measures.html
Pencina, M. J., & D’Agostino, R. B. (2004). Overall C as a measure of discrimination in survival analysis: Model specific population value and confidence interval estimation. Statistics in Medicine, 23, 2109–2123. https://doi.org/10.1002/sim.1802
Pencina, M. J., D’Agostino, R. B., Sr., & D’Agostino, R. B., Jr. (2008). Evaluating the added predictive ability of a new marker: From area under the ROC curve to reclassification and beyond. Statistics in Medicine, 27, 157–172. https://doi.org/10.1002/sim.2929
