Risk of Death Among Nursing Home Residents: A Cross-National Perspective.

OBJECTIVES: Describe the rate of death over 4 consecutive quarters and determine optimal categorization of residents into risk-of-death categories, expanding the Changes in Health, Endstage Disease, Signs and Symptoms (CHESS) scale. DESIGN: Using secondary analysis design with Minimum Data Set (MDS) data, the CHESS scale provided the base upon which the DeathRisk-NH scale was developed. SETTING AND PARTICIPANTS: Baseline and 4 quarterly follow-up analyses of Canadian (n = 109,145) and US (n = 1,075,611) nursing home resident data were completed. METHODS: Logistic regression analyses identified predictors of death, additive to CHESS, to form the DeathRisk-NH scale. The independent variable set used MDS items, focusing on clinical complexity indicators, diagnostic conditions, and measures of severe clinical distress. RESULTS: Country cohorts had similar percentages of residents with mean activities of daily living hierarchy scores, dependence in mobility, continence, memory, and overall CHESS scores. The percentage of individuals who died increased from 10.5% (3 months) to 30.7% (12 months). The average annual death rate for this cohort was 5.5 times higher than the national annual death rate of approximately 5.6%. CONCLUSIONS AND IMPLICATIONS: The DeathRisk-NH is an effective prediction model to identify residents at risk of death within the first 12 months after admission to the nursing home. The tool may be helpful in patient care planning, resource allocation, and excess death monitoring.

Cognitive Change Among Nursing Home Residents: CogRisk-NH Scale Development to Predict Decline.

OBJECTIVES: Examine cognitive changes over time among nursing home residents and develop a risk model for identifying predictors of cognitive decline. DESIGN: Using secondary analysis design with Minimum Data Set data, cognitive status was based on the Cognitive Performance Scale (CPS). SETTING AND PARTICIPANTS: Baseline and 7 quarterly follow-up analyses of US and Canadian interRAI data (N = 1,257,832) were completed. METHODS: Logistic regression analyses identified predictors of decline to form the CogRisk-NH scale. RESULTS: At baseline, about 15% of residents were cognitively intact (CPS = 0), and 11.2% borderline intact (CPS = 1). The remaining more intact, with mild impairment (CPS = 2), included 15.0%. Approximately 59% residents fell into CPS categories 3 to 6 (moderate to severe impairment). Over time, increasing proportions of residents declined: 17.1% at 6 months, 21.6% at 9 months, and 34.0% at 21 months. Baseline CPS score was a strong predictor of decline. Categories 0 to 2 had 3-month decline rates in midteens, and categories 3 to 5 had an average decline rate about 9%. Consequently, a 2-submodel construction was employed-one for CPS categories 0 to 2 and the other for categories 3 to 5. Both models were integrated into a 6-category risk scale (CogRisk-NH). CogRisk-NH scale score distribution had 15.9% in category 1, 26.84% in category 2, and 36.7% in category 3. Three higher-risk categories (ie, 4-6) represented 20.6% of residents. Mean decline rates at the 3-month assessment ranged from 4.4% to 28.3%. Over time, differentiation among risk categories continued: 6.9% to 38.4.% at 6 months, 11.0% to 51.0% at 1 year, and 16.2% to 61.4% at 21 months, providing internal validation of the prediction model. CONCLUSIONS AND IMPLICATIONS: Cognitive decline rates were higher among residents in less-impaired CPS categories. CogRisk-NH scale differentiates those with low likelihood of decline from those with moderate likelihood and, finally, much higher likelihood of decline. Knowledge of resident risk for cognitive decline enables allocation of resources targeting amenable factors and potential interventions to mitigate continuing decline.

Machine-Learning Modeling to Predict Hospital Readmission Following Discharge to Post-Acute Care.

OBJECTIVES: Primary purpose was to generate a model to identify key factors relevant to acute care hospital readmission within 90 days from 3 types of post-acute care (PAC) sites: home with home care services (HC), skilled nursing facility (SNF), and inpatient rehabilitation facility (IRF). Specific aims were to (1) examine demographic characteristics of adults discharged to 3 types of PAC sites and (2) compare 90-day acute hospital readmission rate across PAC sites and risk levels. DESIGN: Retrospective, secondary analysis design was used to examine hospital readmissions within 90 days for persons discharged from hospital to SNF, IRF, or HC. SETTINGS AND PARTICIPANTS: Cohort sample was composed of 2015 assessment data from 3,592,995 Medicare beneficiaries, including 1,536,908 from SNFs, 306,878 from IRFs, and 1,749,209 patients receiving HC services. MEASURES: Initial level of analysis created multiple patient profiles based on predictive patient characteristics. Second level of analysis consisted of multiple logistic regressions within each profile to create predictive algorithms for likelihood of readmission within 90 days, based on risk profile and PAC site. RESULTS: Total sample 90-day hospital readmission rate was 27.48%. Patients discharged to IRF had the lowest readmission rate (23.34%); those receiving HC services had the highest rate (31.33%). Creation of model risk subgroups, however, revealed alternative outcomes. Patients seem to do best (i.e., lowest readmission rates) when discharged to SNF with one exception, those in the very high risk group. Among all patients in the low-, intermediate-, and high-risk groups, the lowest readmission rates occurred among SNF patients. CONCLUSIONS AND IMPLICATIONS: The proposed model has potential use to stratify patients' potential risk for readmission as well as optimal PAC destination. Machine-learning modeling with large data sets is a useful strategy to increase the precision accuracy in predicting outcomes among patients who have nonhome discharges from the hospital.

Developing quality indicators for in-patient post-acute care.

BACKGROUND: This paper describes an integrated series of functional, clinical, and discharge post-acute care (PAC) quality indicators (QIs) and an examination of the distribution of the QIs in skilled nursing facilities (SNF) across the US. The indicators use items available in interRAI based assessments including the MDS 3.0 and are designed for use in in-patient post-acute environments that use the assessments. METHODS: Data Source: MDS 3.0 computerized assessments mandated for all patients admitted to US skilled nursing facilities (SNF) in 2012. In total, 2,380,213 patients were admitted to SNFs for post-acute care. Definition of the QI numerator, denominator and covariate structures were based on MDS assessment items. A regression strategy modeling the "discharge to the community" PAC QI as the dependent variable was used to identify how to bring together a subset of seven candidate PAC QIs for inclusion in a summary scale. Finally, the distributional property of the summary scale (the PAC QI Summary Scale) across all facilities was explored. RESULTS: The risk-adjusted PAC QIs include indicators of improved status, including measures of early, middle, and late-loss functional performance, as well as measures of walking and changed clinical status and an overall summary functional scale. Many but not all patients demonstrated improvement from baseline to follow-up. However, there was substantial inter-state variation in the summary QI scores across the SNFs. CONCLUSIONS: The set of PAC QIs consist of five functional, two discharge and eight clinical measures, and one summary scale. All QIs can be derived from multiple interRAI assessment tools, including the MDS 2.0, interRAI-LTCF, MDS 3.0, and the interRAI-PAC-Rehab. These measures are appropriate for wide distribution in and out of the United States, allowing comparison and discussion of practices associated with better outcomes.