To the Editor: Mr Kohrt and colleagues1 assessed the mental health consequences of children who were conscripted as soldiers in Nepal. By comparing former recruits with a sample of controls, the authors implicitly assumed independence of exposure assignment (conscription) and potential outcomes (mental well-being). However, if exposure and potential outcomes were not independent, differences in outcomes attributed to child soldiering may actually have been due to preexposure characteristics that generated the selection process. The authors adjusted for observed covariates such as education and marital status, but their analytical strategy cannot account for unobserved covariates. In the absence of randomized exposure or an appropriate econometric strategy, their estimate of the effect of conscription on children's mental well-being may be biased (either upward or downward).
Tsai AC. Measuring Mental Health in Child Soldiers. JAMA. 2008;300(23):2729-2730. doi:10.1001/jama.2008.774