Factors associated with long-stay nursing home admissions among the U.S. elderly population: comparison of logistic regression and the Cox proportional hazards model with policy implications for social work.

Two statistical methods were compared to identify key factors associated with long-stay nursing home (LSNH) admission among the U.S. elderly population. Social Work's interest in services to the elderly makes this research critical to the profession. Effectively transitioning the "baby boomer" population into appropriate long-term care will be a great societal challenge. It remains a challenge paramount to the practice of social work. Secondary data analyses using four waves (1995, 1998, 2000, and 2002) of the Health Retirement Study (HRS) coupled with the Assets and Health Dynamics among the Oldest Old (AHEAD) surveys were conducted. Multivariable logistic regression and Cox proportional hazards model were performed and compared. Older age, lower self-perceived health, worse instrumental activities of daily living (IADL), psychiatric problems, and living alone were found significantly associated with increased risk of LSNH admission. In contrast, being female, African American, or Hispanic; owning a home; and having lower level of cognitive impairment reduced the admission risk. Home ownership showed a significant effect in logistic regression, but a marginal effect in the Cox model. The Cox model generally provided more precise parameter estimates than logistic regression. Logistic regression, used frequently in analyses, can provide a good approximation to the Cox model in identifying factors of LSNH admission. However, the Cox model gives more information on how soon the LSNH admission may happen. Our analyses, based on two models, dually identified the factors associated with LSNH admission; therefore, results discussed confidently provide implications for both public and private long-term care policies, as well as improving the assessment capabilities of social work practitioners for development of screening programs among at-risk elderly. Given the predicted surge in this population, significant factors found from this study can be utilized in a strengths-based empowerment approach by social workers to aid in avoiding LSNH utilization.