MORTALITY PREDICTION MODELS IN NURSING HOME RESIDENTS

Emory University, Atlanta GA

Investigators

Linked publications & trials

Abstract

A readily accessible, clinically useful method for predicting mortality in nursing home residents is highly desirable. Very few models for predicting one-year mortality in nursing home residents have been proposed, and these have generally been unsuitable for use in clinical practice due to the complexities of obtaining and utilizing the required data. Cumbersome, time or personnel intensive approaches are unlikely to be widely used by clinicians. The Minimum Data Set (MDS) is a validated instrument which includes assessments of over 350 individual items related to resident functioning in domains of cognition, mood and behaviors, physical functioning, sensory function, continence, disease diagnoses, health conditions, nutritional status, dental status, skin condition, activity patterns, medications, and special treatments. The MDS is readily available and administered to all Medicare and Medicaid certified nursing home residents in the United States, and has been previously used to develop models to predict adverse events such as falls and functional decline. The MDS however contains no information on the date of the resident's death. This information regarding death however is readily obtainable from the National Death Index, a central, computerized index of death records compiled from information submitted under contractual arrangements to the National Center for Health Statistics by the state vital statistics offices. Further, this database will allow identifications of factors associated with one-year mortality in newly-admitted and long-stay residents, as well as gender and ethnic specific analyses. The population will include all New York State nursing home residents greater than 65 years of age in the time period of 1994-1996. Bivariate proportional hazards regression analysis will identify candidate factors associated with mortality, and multivariate proportional hazards regression will be used to develop the final all-inclusive models for newly-admitted and long-stay residents. The same analysis will be used to develop gender and ethnic specific mortality prediction models. We anticipate that the outcome of this proposal will be mortality prediction models for nursing home residents that can be readily accessed by any clinician in a nursing home in the United States. This mortality information will then be of great for primary care teams, residents, and families when developing and modifying care plans for nursing home residents.