Comparison of Longitudinal and Cross-Sectional Approaches in Studies About Knowledge, Attitude, and Practices Related...

Cross-sectional studies on the association between knowledge, attitudes, and practices of antibiotic misuse are abundant. Conversely, longitudinal studies are unavailable.

The objective of this study was to compare findings from cross-sectional and longitudinal data analyses of the same study on knowledge and attitudes related to practices of antibiotic misuse in the general adult population.

A validated knowledge, attitudes, and practices questionnaire was self-administered at baseline and participants were followed up bimonthly. Logistic regression was used on baseline data to compute adjusted odds ratios in the cross-sectional data analysis. In the longitudinal approach, Poisson regression and generalized linear mixed models were used to estimate adjusted incidence rate ratios and adjusted odds ratios, respectively.

Seven hundred and forty-seven adults completed the baseline and the follow-up questionnaires. In the cross-sectional analysis, 78 participants misused antibiotics and strong associations were observed between agreement with various knowledge and attitudes statements and antibiotic misuse practices, with odds ratios between 1.94 and 5.66. In the longitudinal data analysis using Poisson regression, 46 events of antibiotic misuse were detected. No associations were observed between knowledge, attitudes, and misuse practices, except for two statements about the preference of storing antibiotics [adjusted incidence rate ratio: 2.33 (95% confidence interval 1.20-4.54)] and taking antibiotics when getting a cold [adjusted incidence rate ratio: 2.28 (95% confidence interval 1.24-4.21)]. The magnitude of these associations was lower than that observed in the cross-sectional analysis. Findings of the generalized linear mixed-model analysis were similar to those yielded by Poisson regression.

The cross-sectional and the longitudinal approaches yield divergent results in our setting. Risk of bias in both analyses should be scrutinized.