Uso do informante secundário em inquéritos de saúde

Abstract

Introduction: To investigate the possibility and direction of bias introduced by proxy respondents in health surveys is of major interest, as proxies are often used as the primary source of information in general population and for patients with limiting health conditions. Objectives: The main objectives are two: 1) to estimate the reliability between pairs of primary-secondary informants in relation to some health indicators and 2) to compare two explanotory models for bad and very bad health evaluation, one based on self-report and the other on proxy report. Materials and methods: Data obtained from household cross sectional study of 710 pairs of primary-secondary informants: 239 elderly-adult, 239 adult-elderly e 232 adult-adult. The three pairs of informants were compared and reliability estimated by prevalence adjusted kappa and proportional bias estimated. The influence of proxy characteristics in the disagreement between pairs of informants were assessed by Poisson regression with robust variance of error. The explanotory models of overall evaluation of elderly health were obtained from logistic regression analysis of 230 pairs elderly-proxy. Results: in general, there was a good agreement and reliability of information obtained from proxy in relation to tobacco use, overall health evaluation, reported morbidity and medical visits. The poorest agreement and reliability was found among elderly-adult pairs. The results also show that the use of proxy can introduce bias in mensuration of exposures and health events depending on question type (presence of bias was found for hypertension and medical visits) and proxy characteristics, such as schooling and age. The comparison of final models based on elderly self rated health and on the proxy answers show important differences. While elderly model identified age and disabilities in activities of daily living and/or to mobility as the variables independently associated with bad health evaluation, the model based on proxy response replicate the biomedical model which gives more value to the formal diagnostic of a chronic disease than its consequences to elderly life and independence. Besides, the proxies with a bad self-rated health had almost three times the probability of rating the elderly health as also bad/very bad. Conclusions: the results support the recommendation of caution in the use of proxy information, mainly in relation to subjective factors, which must be obtained from directly from individuals involved. They reinforce the comparability of self and proxy information in relation to many objective measures, but advert to need to avoid bias inferences in analysis of subjective health evaluation obtained from secondary informants.