A decision support system for diagnostic consultation in laboratory tests

Correct interpretation of laboratory test results is essential in primary care for accurate diagnosis and cost-effective management. The Japan Society of Clinical Pathology has outlined recommended screening tests to guide practitioners. However, gaps remain in how these tests are applied in real-world settings. Prior research has shown that diagnostic accuracy can vary depending on the disease category and the availability of supporting information. No prior work had resolved how to integrate these screening items into a structured decision-making tool. The challenge lies in translating standardized test recommendations into actionable diagnostic support. This uncertainty drove the development of a system that could assist general practitioners in interpreting test results. The need for a streamlined approach to diagnosis using basic laboratory tests is clear. A tool that reduces unnecessary testing while improving diagnostic confidence is a key innovation in this field.

This study aimed to create a decision support system for primary care physicians using standardized laboratory tests. The goal was to improve diagnostic accuracy through a structured approach to test interpretation. The system was designed to guide users toward correct diagnoses using a set of 26 essential test items. The motivation was to reduce diagnostic errors and unnecessary testing in primary care settings. The system was intended to function as a practical tool in real-world clinical environments. It was also designed to provide recommendations for follow-up tests when needed. The study sought to evaluate the system's effectiveness in identifying 11 diagnostic categories. The ultimate aim was to support general practitioners in making informed diagnostic decisions.

The system was built as a microcomputer-based tool running on a PC-9821Bp under Windows 3.0. It incorporated 26 laboratory test items, including urinalysis and blood tests, as input parameters. The system used a decision-tree logic to categorize possible diagnoses. It was structured into two main processes: diagnostic classification and follow-up recommendations. The first process assigned patients to one of 11 diagnostic categories. The second process suggested additional tests or maneuvers for confirmation. The system was tested on 219 patients with confirmed diagnoses at Nagoya University Hospital. The test cases were selected to represent a range of disease categories. The system’s diagnostic accuracy was measured against the confirmed diagnoses.

The system correctly identified 14 of 14 anemia cases, 42 of 55 reno-urinary disease cases, and 8 of 14 diabetes mellitus cases. It also correctly diagnosed 14 of 18 hyperlipidemia cases. Lower accuracy was observed in infectious disease, with 10–20 cases correctly identified. Malignant tumor cases had 34 of 74 correct diagnoses. Muscular disease accuracy was low, with only 2 of 6 cases correctly identified. Bone disease had 3 of 26 correct diagnoses. The system provided follow-up test recommendations for confirmation in all categories. It demonstrated high accuracy in four key diagnostic areas: anemia, reno-urinary disease, diabetes, and hyperlipidemia.

The system showed strong performance in identifying anemia, reno-urinary disease, diabetes mellitus, and hyperlipidemia. It provided minimum and correct test items for diagnosis in primary care settings. The authors suggest that this system can reduce unnecessary testing when used alongside patient history and physical examination. It may also improve diagnostic confidence for general practitioners. The system’s accuracy was lower for infectious disease, malignant tumors, muscular disease, and bone disease. The authors propose that additional test items are needed for these categories. The system supports the efficient use of basic laboratory tests in primary care. It may serve as a model for integrating standardized tests into clinical decision-making tools.