Does information technology improve medical diagnoses?

The research team conducted a clinical study in Swiss hospital emergency departments to evaluate the effect of computer-aided diagnosis systems (CDDS) on diagnostic processes and patient outcomes. They focused on three common and often misdiagnosed symptoms, i.e. fever, abdominal pain and fainting. In this study, patients were diagnosed either conventionally, i.e. according to common local practice, or with the help of CDDS. Patients were monitored for 14 days to evaluate their recovery.

The health care system is becoming increasingly digitalised. Computer-aided diagnosis systems, which collect data regarding patients' complaints and subsequently suggest possible causes, are a prominent example of this development. Initially, CDDSs were designed to support physicians. In the meantime, the term also includes so-called “symptom checkers”, which provide an interface for patients and directly recommend measures, such as a medical consultation.

It is not known to what extent the use of CDDSs actually improves the quality of medical diagnoses and thus the health of individual patients. In particular, it is not known whether and how effectively CDDSs work in less structured tasks such as the assessment of common symptoms, e.g. fever, abdominal pain or fainting.

The aim of this research project was to understand, what (if any) effects CDDS have on the diagnostic process, diagnostic quality and patient outcome in emergency medicine.

Specific goals:

• Comparison of the accuracy of diagnoses established conventionally or with the support of a CDDS.
• Comparison of the diagnostic workflow in the above-mentioned groups.
• Assessment of the impact of CDDS use on economic benefits and the use of resources, and identification of the pedagogical implications for medical education.

The research project contributed to increased diagnostic certainty. The team determined the implications of the use of CDDS for patients and the processes in the clinic, and derived consequences of relevance to patients, the medical profession, emergency rooms, political decision makers, health insurance companies, educators as well as CDDS manufacturers and researchers.

Three main messages:

1. Diagnostic errors are frequent and consequential problems in emergency medicine.
2. Computerised diagnostic decision support does not improve diagnoses made in emergency medicine.
3. Computerised diagnostic decision support impacts the diagnostic process in unexpected ways. It makes trainees less likely to seek advice, and its reliability and application vary significantly depending on the physician and the context.

The digital diagnostician: how information technology affects medical diagnoses

Prof. Dr. Wolf Hautz, Universitäres Notfallzentrum, Inselspital, Universität Bern