Relative validity of a brief dietary survey to assess food intake and adherence to national dietary guidelines among Sri Lankan adults.
Renzella J., Fernando S., Kalupahana B., Townsend N., Rayner M., Wickramasinghe K., Katulanda P., Scarborough P.
BACKGROUND: Suboptimal diet is the leading cause of global morbidity and mortality. Addressing this problem requires context-specific solutions informed by context-specific data collected by context-specific tools. This study aimed to assess the relative validity of a newly developed brief dietary survey to estimate food intake and adherence to the Food Based Dietary Guidelines for Sri Lankans. METHODS: Between December 2018 and February 2019, we interviewed 94 Sri Lankan adults living in Colombo (Western Province), Kalutara (Western Province), and Trincomalee (Eastern Province). We assessed the relative validity of the Sri Lankan Brief Dietary Survey (SLBDS) with Wilcoxon rank-sum tests, Spearman's Rho correlation coefficients, Bland-Altman plots, and Cohen's kappa tests using a 24-h Dietary Recall (24DR) as reference. RESULTS: Ninety-four adults (40.7 years ±12.6; 66% female) completed both surveys during the same interview. With the exception of 'Fish, pulses, meat and eggs' food group median intake, which was underestimated by the SLBDS compared to the 24DR, there was no strong evidence of difference between median intakes reported by the two methods. Correlation coefficients were highest for 'Milk and dairy products' (0.84) at the food group level and for 'dosa', 'hoppers', 'milk rice', and 'dried fish' (1.00) among individual food and beverages. Visual exploration of Bland-Altman plots showed acceptable agreement between the SLBDS and 24DR, with the SLBDS tending to overestimate consumption as the number of servings of 'Rice, bread, other cereals and yams' and 'Vegetables' consumed increased and slightly underestimate consumption as the number of servings of 'Fish, pulses, meat and eggs', 'Milk and dairy products', and 'Nuts' increased. Kappa values ranged from from 0.59 (95% CI: 0.32-0.86) for 'Vegetables' to 0.81 (95% CI: 0.66-0.96) for 'Fruit' indicating a moderate to strong level of agreement. CONCLUSIONS: Having been developed for and relatively validated with the study population in question, our study shows that the SLBDS can be used as a fit for purpose research tool. Additional research is needed to assess SLBDS test-retest reliability and to validate further the reporting of salt, oil, and coconut intake.