Comment on “Substance Use and the Associated Factors Among High School Students in Jimma Town,Jimma,Southwest Ethiopia”

Dear Editor,

We read with interest the article titled, “Substance Use and the Associated Factors Among High School Students in Jimma Town, Jimma, Southwest Ethiopia,” which reports substance use and associated factors among Ethiopian high school students using an adapted Brief Screener for Tobacco, Alcohol and other Drugs (BSTAD). ¹ Study reports a past year prevalence of any substance use of 27.2% and identifies living with parents, neutral parental opinions, lack of information, and peer khat use as key correlates.

One key concern relates to the internal consistency of the reported estimates and the interpretation of odds ratios. In the abstract, the prevalence of khat use is given as 48.5% with a 95% confidence interval (CI) of 1.75–1.96, and similar implausible CIs are presented for alcohol and tobacco; these appear to be mislabelled effect estimates rather than binomial CIs for proportions. The adjusted odds ratio (AOR) of 0.34 (95% CI = 0.15–0.77) attached to “lack of information” is then described as indicating increased odds of substance use, despite being <1, while the AOR of 2.76 (95% CI = 1.09–6.94) printed for “positive parental opinion” is interpreted as protective. Clarifying the reference categories and re-presenting the two-by-two tables would clarify whether these parental and information effects are real or artifactual.

A further concern is statistical precision and model complexity. Only 55 students reported substance use in the past year, yet multiple predictors were entered into multivariable models, producing wide CIs, for example, for living with parents (AOR = 9.25, 95% CI = 0.96–88.94). This counterintuitive association contrasts with larger African analyses, in which disrupted family structure and poorer supervision generally increase adolescent use, ² and with cohort data showing that higher parental monitoring halves the odds of recent substance use. ³ Cohort studies also show that adolescent polysubstance use predicts later mental and behavioral disorders and health service use, underscoring the need for stable, interpretable estimates. ⁴

Finally, the sampling frame (two intervention schools) and use of BSTAD beyond its validated age range, without local psychometric data, limit generalisability. Given evidence that one in four children in some settings live with a parent with substance use disorder, careful measurement and reporting of family variables are crucial. ⁵

We believe that revisiting the confidence intervals, clearly specifying reference categories, and simplifying the regression models would strengthen this important contribution and better inform school- and family-based prevention strategies.