The accuracy of substitution rules for health risk appraisals

Abstract Health Risk Appraisals (HRAs) have been in increasingly wide use over the past 20 years as health education tools designed to encourage healthy behaviors. A major criticism of HRAs has been the error introduced by missing or incorrectly-specified responses by clients, especially regarding physiological measures such as blood pressure and cholesterol. Four HRAs were randomly assigned to members of a random sample of Massachusetts residents as part of a trial to assess the reliability and validity of HRA instruments. Two rules for substituting unknown physiological measures were evaluated: the existing rule for an HRA and a new rule, which simply substitutes the age- and gender-specific mean physiological values. Ordinary least squares regression was used to compare risk scores generated from known physiological measures, taken in respondents' homes, with risk scores generated by substitution. Regression slopes were mostly close to 1.0, most intercepts were relatively close to 0.0, and correlations were all extremely high, suggesting good performance of substitution rules on a group level. However, outliers from the regressions tended to be individuals at high risk whose risk was underestimated by the substitution rules. These results raise ethical questions about the use of substitution rules and reinforce the need for taking actual physiological measurements.