Medical expert systems for developing countries: Evaluation in practice

Abstract

In this article the authors describe the evaluation of ESTROPID (an Expert System for TROPIcal Diseases) in a developing country context. The system was designed to be used by paramedical staff in an outpatient clinic setting, with the aim of providing diagnostic support. The evaluation showed that use of the system increased the time the clinical officer (CO) spent with patients and increased the number of symptoms elicited from the patient at the beginning of a consultation (demonstrating a check-list effect). The scale of disagreement per patient was 24% between CO and doctor when the computer was used and 14% under normal practice. This was a small study, and the limited statistical data collected on concordance levels between the different types of consultations are not necessarily generalisable. What was more interesting were the additional observations of normal clinical practice and the CO's assessment of the system's worth. This enabled the integration of quantitative and qualitative measures as complements rather than alternatives. The tentative conclusion of this trial is that preliminary baseline studies need to be conducted before the effect of a decision support tool can be investigated. Health personnel may need help, but their need of computerized decision-aid technology is questionable—perhaps manual paper-based means are adequate.