The use of Bayesian network models to identify factors related to frailty phenotype and health outcomes in middle-aged and older persons.

BACKGROUND

Studies on frailty frequently only include older persons. The mapping of factors related to this syndrome and negative outcomes associated with it, also in middle age, may assist in health strategies to each age group.

OBJECTIVES

To investigate social and health factors related to the frailty phenotype and to analyze the probabilistic relationships between frailty, falls and hospitalization in middle-aged and older persons.

METHODS

This is a cross-sectional study using data for 4442 middle-aged (50-59 years) and older participants (60 years or older) from the Brazilian Longitudinal Study of Aging (ELSI-Brazil) 2015 and 2016. Bayesian network models were estimated with the score-based hill-climbing algorithm to identify factors associated with frailty, falls and hospitalization.

RESULTS

Mean age was 63.7 years, and prevalence of frailty was 8.5 % and 11.9 % among middle-aged and older participants, respectively. In the former, the probability of frailty increased when "poor" self-rated memory was considered in the model; and in the latter, the probability of frailty was greatest among individuals who did not participate socially and had the lowest level of education. In both age groups, frailty was an important factor that influenced the probability of negative health outcomes such as falls and hospitalization. However, this result depended on combinations of health factors in each sample.

CONCLUSIONS

This study has identified potential vulnerabilities that should be considered when undertaking a comprehensive assessment of middle-aged and older persons and developing suitable health strategies for each of these phases of life.