Physician utilization among the elderly: An analysis using structural equation modeling

Date of Award




Degree Name

Doctor of Philosophy (Ph.D.)



First Committee Member

Linda L. Belgrave - Committee Chair

Second Committee Member

Kenneth J. Smith - Committee Member


The purpose of this research was to examine physician utilization among the elderly within the framework of the Andersen behavioral model. The general research aim guiding this project was to determine which characteristics of individuals and their social environment predict physician use and to examine how these characteristics are interrelated.Using data from Version 3 of the Longitudinal Study on Aging, need, enabling, and predisposing factors were tested as predictors of physician use among a sample of community dwelling elders.Five latent variables were identified in three measurement models: social support network, health beliefs, perceived health, evaluated health, and chronic conditions in addition to remaining directly observed variables. The measurement models demonstrated a good fit between the three measurement models and the data. The clarification of these measurement issues helped to alleviate potential measurement problems in later specific aims.The analysis of the Andersen model based on the current literature revealed an R$\sp2$ of.20. The findings reported in the analysis for Specific Aim 4 support the idea that only analyzing the direct effects of predisposing, enabling, and need characteristics on physician use does not present a completely clear picture. The model developed had a reasonable fit and provided an improvement in the explained variance of physician use (R$\sp2$ =.25).The relative importance of major predictor variables in this model is consistent with the literature on health services use among the elderly. The inclusion of indirect effects of predisposing and enabling characteristics added clarification to understanding their roles in explaining health services use among the elderly. The full model incorporating structural equation and measurement models was tested and demonstrated that including the measurement component of predictor variables added to the understanding of the Andersen model. The full model revealed an R$\sp2$ of.25. Further clarification is still needed.


Gerontology; Health Sciences, Public Health; Sociology, General

Link to Full Text


Link to Full Text