Effectiveness of the Quick Medical Reference as a diagnostic tool

Prior research has explored the role of computer-based diagnostic systems in internal medicine. These systems aim to assist clinicians in identifying diagnoses from patient data. However, the effectiveness of such tools remains uncertain in real-world settings. Established knowledge includes the potential of decision support systems to improve diagnostic accuracy. That uncertainty drove this study, as no prior work had resolved how often these systems identify correct diagnoses among top results. The gap motivated an investigation into QMR's performance in a clinical context. This paper's contribution lies in its retrospective evaluation of a diagnostic program's accuracy. The study focused on a specific system's ability to rank correct diagnoses. It aimed to clarify the practical utility of such tools in clinical decision-making.

This study aimed to assess the effectiveness of the Quick Medical Reference (QMR) system in identifying correct diagnoses. The specific problem addressed was the frequency with which QMR ranks the correct diagnosis among top results. The motivation stemmed from the need to evaluate clinical decision support tools in real-world scenarios. The authors sought to determine if QMR could reliably assist physicians in diagnostic tasks. They focused on a teaching unit's patient records to test the system's performance. The study's goal was to provide empirical evidence of QMR's diagnostic accuracy. By analyzing retrospective data, the authors aimed to reflect typical clinical use. Their findings would help clarify the system's strengths and limitations in practice.

The study used a retrospective design to evaluate the Quick Medical Reference (QMR) system. Researchers reviewed charts of 1144 consecutive patients admitted to a teaching unit. They selected cases with undiagnosed illnesses and proven final diagnoses. Two physicians, familiar with but not experts in QMR, entered clinical data into the system. The data included patient information abstracted from medical charts. The correct diagnosis was identified if it appeared among the top five ranked diagnoses. Physicians independently used the system to test its performance. The study focused on the frequency of correct diagnosis rankings across cases.

The correct diagnosis appeared among the top five rankings in 62 of 154 cases for physician A. Physician B achieved success in 56 of 154 cases using the same system. These results suggest QMR's ability to identify correct diagnoses in a subset of cases. The success rate ranged from 36% to 40% across the two evaluators. No significant difference was reported between the two physicians' performance. The study highlighted cases where QMR failed to rank the correct diagnosis. Authors illustrated both strengths and limitations through example cases. The findings indicated that while QMR can assist, it is not consistently accurate.

The authors concluded that QMR can identify correct diagnoses in a portion of cases but is not consistently reliable. Their findings suggest that the system may assist physicians in some diagnostic scenarios. The study demonstrated that QMR's accuracy varies depending on the case. The authors proposed that QMR's performance is influenced by the quality of input data. They emphasized the need for further refinement of diagnostic tools like QMR. The system's limitations were illustrated through specific clinical examples. The authors did not claim QMR as essential for clinical decision-making. Their conclusion was based solely on the observed success rates in the study.