Development and comparative validation of a dietary fat screener for grade six children

Abstract

Background Risk factors for chronic non-communicable diseases have been shown to track from childhood into adulthood. Cost-effective intervention starts with valid screening. The aim of this research was development and comparative validation of a dietary fat screener in grade six learners. Methods A pictorial, quantitative food frequency questionnaire type, scored dietary fat screener (test method), consisting of ten food categories associated with high fat intakes, was developed and subjected to developmental evaluations in the target group. Subsequently the test method was administered to learners of an urban middle-class school (Pretoria, South Africa). Test-retest reproducibility was checked in a random sub-sample. Two reference methods were used for comparison: Parental completion of the screener and a three-day food record by the children. Reliability testing of the test method involved measuring internal consistency and test-retest reproducibility. Credibility of energy intakes in the food record was checked. Mean cholesterol intake and percentage fat and saturated fat energy were determined. Comparative validation was based on correlations, mean differences and the Bland Altman method for continuous variables. Percentage agreement, kappa statistics and the McNemar tests were determined for categorical data, as were sensitivity, specificity and predictive values. Receiver operating characteristic curves were plotted. Results Sample: Out of 108 children, 39 (100%) were re-tested, 93 (86%) provided usable food records and 78 (72%) parents responded. Mean age was 148±4.4 months. Reliability: The test method was internally consistent. Test-retest reproducibility of portion size and frequency of intake estimates depended on the food category. No systematic error between administrations was noted as mean category and final score differences between the two administrations did not differ significantly from zero. A significant (r=0.36, P=0.02) correlation existed between administrations, but boys were characterised by random error and a lack of reproducibility (r=0.26, P=0.29), whilst for girls reproducibility could be established (r=0.58, P=0.01). Comparison to screener by parents: Children and parents did not agree in respect of reported portion size and frequency of intake. Parents had lower values for all scores. Correlation between children’s and parents’ final scores was 0.23 (P=0.04) (boys: r=0.13, P=0.46; girls: r=0.33, P=0.04), but the mean difference in final scores differed significantly from zero (P=0.0001). Classification was identical in 74% of cases, but when corrected for chance this agreement was also poor. Comparison to food record: The food record appeared to be a plausible reflection of energy intakes during the recording period. For girls a significant (P<0.05) correlation between test method final score versus cholesterol intake and energy from total and saturated fat was found. The sensitivity of the test method was very high (over 90%). Chance corrected agreement between test method classification and measures of fat intake from the food record was poor. Changing the cut-off of the test method final score could not achieve high sensitivity and high specificity simultaneously. Conclusion The dietary fat screener cannot yet be used as sole indicator of high fat intake in grade six learners. Until the discriminatory abilities have been improved, its value lies in creating awareness of high fat intakes and providing a food-based starting point for anticipatory guidance.