Introduction Composite health scores are useful in public health research, but components must be operationalised to be feasible in community settings. This paper compared multiple methods to calculate the World Cancer Research Fund and American Institute for Cancer Research (WCRF/AICR) 2018 Cancer Prevention Recommendations Adherence Score, which scores recommendations to: (1) be a healthy weight; (2) be physically active (PA); (3) eat whole grains, fruit and vegetables (FV) and beans; (4) limit fast/processed foods; (5) limit red/processed meat; (6) limit sugar-sweetened beverages (SSBs) and (7) limit alcohol. Methods We conducted cross-sectional comparisons of validated surveys and alternate measurements (pedometer readings, 24-hour diet recall) for 181 US adults. Across measurement approaches, we calculated percentage agreement and linear weighted kappa statistics for component classifications, and compared composite scores using Spearman’s rank-order correlations (r s ), paired t-tests and intraclass correlation coefficients (ICC). Results Survey and alternate measures had slight agreement for adherence scoring of FV and alcohol (p<0.001), red/processed meat and fast/processed foods (p<0.01) and fibre and PA (p<0.05), and no agreement for SSBs. WCRF/AICR composite scores using alternate measures were moderately or strongly correlated (r s =0.44–0.87) to a survey-only composite. Mean WCRF/AICR composite score was 3.34 based on survey data, but lower when substituting pedometer readings (−0.36, p<0.001) and higher with the 24-hour diet recall (+0.54, p<0.001). Reliability of composite scores was moderate to excellent (ICC=0.63–0.94), with no evidence of systematic bias. Conclusions In this sample, measurement approach influenced WCRF/AICR component adherence classifications substantially and composite scores somewhat. Self-reported data are a cornerstone of community-based research, but unequal units and reference periods and biases in self-reported data all likely contributed to observed discordance in component adherence and composite scores. Future research is needed to identify feasible and accurate approaches to collecting data for composite health metrics.
Hanson et al. (Wed,) studied this question.