The Impact of Biological and Social Factors on Mortality in Older Adults Living in Rural Communities

Abstract

Background: Information on factors that increase mortality in remote settings is limited. This study aims to estimate the independent and joint role of several factors on mortality risk among older adults living in rural Ecuador.

Methods: Participants were selected from community-dwelling older adults who were included in previous studies targeting mortality risk factors in the study population. Generalized structural equation modeling (GSEM) was utilized to evaluate prior causal assumptions, to redraw causal links, and to introduce latent variables that may help to explain how the independently significant variables are associated with mortality.

Results: The study included 590 individuals (mean age: 67.9 ± 7.3 years; 57% women), followed for a median of 8.2 years. Mortality rate was 3.4 per 100 person-years. Prior work on separate multivariate Poisson and Cox models was used to build a tentative causal construct. A GSEM containing all variables showed that age, symptoms of depression, high social risk, high fasting glucose, a history of overt stroke, and neck circumference were directly associated with mortality. Two latent variables were introduced, 1 representing the impact of biological factors and another, the impact of social factors on mortality. The social variable significantly influenced the biological variable which carried most of the direct effect on mortality.

Conclusions: Several factors contributed to mortality risk in the study population, the most significant being biological factors which are highly influenced by social factors. High social risk interact with biological variables and play an important role in mortality risk.

Keywords: generalized structural equation modeling approach; mortality risk; population-based longitudinal cohort; social factors influencing mortality.

Figure 1.

Causal generalized structural equation modeling with arrows pointing to the hypothesized direction of causality. The Odds ratios or Incidence Rate Ratios were added in significant associations to understand multivariate structure and corresponded to separate Poisson or logistic regressions that were fitted for each variable that accumulates arrows toward it. Data below the boxes represented associations not initially hypothesized within this construct that resulted to be significant.

Figure 2.

Measurement generalized structural equation modeling showing the estimation of the associations between risk factors and confounders (the diagram shows only significant associations).

Figure 3.

Specification of a generalized structural equation modeling for predicting mortality from known risk factors and 2 hypothesized latent variables (biological and social). Increasing age was associated with mortality independently of latent variables.