Assessing psychiatric disorder with a human interviewer or a computer

Abstract

Objective - To compare a self administered computerised assessment of neurotic psychiatric disorder (psychiatric morbidity) with an identical assessment administered by a human interviewer. In particular, to discover whether a computerised assessment overestimates or underestimates the prevalence of psychiatric morbidity in relation to a human interviewer. Setting - A health centre in south east London, UK. Subjects - A non-consecutive series of health centre attenders. Complete data were available on 92 subjects. Design - All subjects received both assessments on the same occasion but were randomised to receive either the computerised assessment first or the human interview first. Results - The mean total score on the assessment was the same for both methods of administration; computer 8.77 v human 8.69 (95% confidence interval for difference - 0.70, 0.87). The correlation between the human and interviewer assessments was 0.91. Conclusion - Self administered computerised assessments are valid, unbiassed measures of psychiatric morbidity. In addition to their use as a research tool, they have potential uses in primary care including screening for psychiatric morbidity and in forming the basis for clinical guidelines.