Evaluating the quality of self-reports of hypertension and diabetes

Abstract

Increasingly, researchers and health specialists are obtaining information on chronic illnesses from self-reports. This study validates self-reports of two major health conditions, hypertension and diabetes, based on a recent survey in Taiwan (SEBAS 2000). These data, based on a large, nationally representative sample of respondents aged 54 and older, include both self-reported health information and a physical examination. Average blood pressure readings, laboratory measures of glycosylated hemoglobin, and information on whether the respondent was taking medication for hypertension or diabetes are used to validate respondents' reports of high blood pressure and diabetes. The resulting comparisons reveal that self-reports vastly underestimate the prevalence of hypertension, but yield a reasonably accurate estimate of the prevalence of diabetes. Significant correlates of the accuracy of the self-reports include age, education, time of the most recent health exam, and cognitive function.

Introduction

Increasingly, researchers and health specialists obtain information about chronic illnesses, conditions, and risk factors for disease from self-reports of the target population. The data, often collected via face-to-face interviews, mail-back questionnaires, or telephone interviews, have an obvious advantage over clinical records, namely that they can be obtained readily for a large and representative sample of the population without great expense. Unfortunately, they have a potential drawback as well: the accuracy of self-reports depends on the respondents' knowledge of the relevant information, ability to recall it, and willingness to report it.

In an effort to assess the accuracy of such data, researchers have compared them with “gold standards” such as medical or administrative records, medical provider surveys, or measurements obtained from physical examinations. Several notable problems characterize the majority of these comparisons. First, as Vargas et al. [1] lament, few validation studies have been based on a nationally representative sample. (One notable exception is the NHANES [1].) Most studies have been restricted geographically 2, 3, 4, 5, 6, 7; relied on volunteers 4, 8; or included only persons in good health [8], participants in a particular organization (such as an HMO) or screening program 2, 9, or hospital patients [6]. Second, many validation studies have been based on small samples 4, 5, 8 that limit the ability of the analyst to identify characteristics of the respondents that are associated with inaccurate reporting. Finally, almost all studies have been carried out in wealthy Western populations, even though health interview surveys have become critical sources of data in developing and newly industrialized countries.

This study uses a recent survey in Taiwan to validate respondents' reports of two major health conditions: hypertension and diabetes. Taiwan provides an interesting case study for comparison with Western European and North American countries in that rapid industrialization and economic growth have produced a society with a modern health care system and high life expectancy, yet a substantial fraction of the older population is illiterate or lacks formal education. The survey, which includes both self-reported health information and a physical examination, is based on a large nationally representative sample. The resulting data permit not only unbiased estimation of the accuracy of self-reports, but also an examination of how demographic and social characteristics of participants affect the accuracy. The analysis also includes an assessment of the extent to which participants in the physical examination represent a select subsample of those interviewed.