[Measurement reliability in a study on family violence and severe acute malnutrition]

This study is nested within a research program related to family violence and severe childhood malnutrition. Its aim is to evaluate the reliability of the data collection process in a case-control study. Four components of the main instrument are addressed: (a) CTS (Conflict Tactics Scales) used to measure violence at the family level; (b) CAGE (Cut-down; Annoyed; Guilty & Eye-opener) questionnaire used to gauge suspicion of drinking problems; NSDUQ (Non-student Drugs Use Questionnaire) used to indicate illicit drug consumption; and (d) height/length measurements.

Stability (intra-observer or test-retest reliability) and equivalence (inter-observer reliability) were evaluated for the cited components (a), (b) and (c). Information was replicated among the first 50 subjects selected for the underlying case-control study. The Kappa index (k) was used in the analysis. A pseudo-Bayes adjustment was carried out in order to handle estimation problems. Regarding (d), only equivalence was evaluated (n = 73), using the Intraclass Correlation Coefficient as the estimator.

By and large, all components showed acceptable stability and equivalence. Regarding stability, the estimates of k were around 0.70, 0.78 and 0.85, for CTS, CAGE e NSDUQ, respectively. With respect to equivalence, k was 1.0 for CTS and NSDUQ and 0.75 for CAGE. Equivalence for height/length estimated through the ICC was 0.99. Nevertheless, some deviant situations were detected and are further discussed. The results point to an adequate standardization of observers and reflect the good quality of the data collection procedure concerning the main study, encouraging the research team to press forward with greater assurance.