Psychometric properties and longitudinal validation of the self-reporting questionnaire (SRQ-20) in a Rwandan community setting: a validation study

This study took place to enable the measurement of the effects on mental health of a psychosocial intervention in Rwanda. It aimed to establish the capacities of the Self-Reporting Questionnaire (SRQ-20) to screen for mental disorder and to assess symptom change over time in a Rwandan community setting. The SRQ-20 was translated into Kinyarwanda in a process of forward and back-translation. SRQ-20 data were collected in a Rwandan setting on 418 respondents; a random subsample of 230 respondents was assessed a second time with a three month time interval. Internal reliability was tested using Cronbach's alpha. The optimal cut-off point was determined by calculating Receiver Operating Curves, using semi-structured clinical interviews as standard in a random subsample of 99 respondents. Subsequently, predictive value, likelihood ratio, and interrater agreement were calculated. The factor structure of the SRQ-20 was determined through exploratory factor analysis. Factorial invariance over time was tested in a multigroup confirmatory factor analysis. The reliability of the SRQ-20 in women (α = 0.85) and men (α = 0.81) could be considered good. The instrument performed moderately well in detecting common mental disorders, with an area under the curve (AUC) of 0.76 for women and 0.74 for men. Cut-off scores were different for women (10) and men (8). Factor analysis yielded five factors, explaining 38% of the total variance. The factor structure proved to be time invariant. The SRQ-20 can be used as a screener to detect mental disorder in a Rwandan community setting, but cut-off scores need to be adjusted for women and men separately. The instrument also shows longitudinal factorial invariance, which is an important prerequisite for assessing changes in symptom severity. This is a significant finding as in non-western post-conflict settings the relevance of diagnostic categories is questionable. The use of the SRQ-20 can be considered an alternative option for measuring the effect of a psychosocial intervention on mental health. Nederlands Trial Register NTR1120