A web-based expert system developed by M Zaki Putra and Yohanni Syahra from the Information Systems Study Program, Faculty of Computer Science and Information Technology, Universitas Muhammadiyah Sumatera Utara, Medan, Indonesia, achieved 92% accuracy in identifying rheumatoid arthritis (RA) among older adults. Published in September 2026, the study demonstrates how digital technology can support early screening for a disease that is often difficult to distinguish from other joint disorders in elderly patients.
The findings are important for primary healthcare facilities, including community health centers (Puskesmas) and integrated service posts for older adults (Posyandu Lansia), where access to medical specialists may be limited. By translating clinical assessment criteria into a digital decision-support system, the researchers offer a potential tool to help healthcare workers recognize suspected cases and determine when further medical evaluation is needed.
Why Rheumatoid Arthritis Can Be Difficult to Identify in Older Adults
Rheumatoid arthritis is an inflammatory disease that affects the joints and can cause swelling, stiffness, and pain. When the condition first appears in older adults, commonly referred to as elderly-onset rheumatoid arthritis (EORA), its symptoms may overlap with those of osteoarthritis and other age-related joint conditions.
The distinction matters because delayed identification of rheumatoid arthritis can postpone appropriate treatment and increase the risk of lasting joint damage. However, primary healthcare facilities may face challenges in assessing complex symptoms, particularly when patients describe their complaints inconsistently or when specialist services are not readily available.
To address this problem, M Zaki Putra and Yohanni Syahra developed a digital expert system that combines established medical classification criteria with a method for calculating diagnostic certainty. The system is designed to process reported symptoms and provide an initial assessment to support healthcare decision-making.
How the Researchers Developed the Digital System
The researchers used a quantitative approach and a System Development Life Cycle (SDLC) framework to design, implement, and evaluate the application.
The development process included a literature review, interviews with a rheumatology specialist, the preparation of diagnostic rules, implementation of the calculation algorithm, and system performance testing.
The system applies the 2010 American College of Rheumatology and European Alliance of Associations for Rheumatology (ACR/EULAR) classification criteria. These criteria consider several clinical factors, including joint involvement, blood-test results, inflammatory markers, and the duration of symptoms.
The researchers incorporated these criteria into a set of IF-THEN rules, allowing the system to connect specific symptom combinations with possible diagnostic conclusions.
The application also uses the Certainty Factor method, which calculates the degree of confidence associated with the available evidence. This approach helps the system process uncertain or subjective symptom reports rather than treating every response as an absolute yes-or-no answer.
To evaluate performance, the researchers tested the application against 50 historical medical records of older patients from an autoimmune referral center. The records had been validated by a rheumatology specialist and served as the reference for comparison.
System Achieves 92% Accuracy in Testing
The evaluation showed that the expert system performed well when its predictions were compared with the specialist's diagnoses.
The main findings were:
- 92.00% accuracy: The system correctly classified 46 of the 50 medical records.
- 93.33% precision: Of the 30 cases classified as positive by the system, 28 were confirmed as positive by the specialist.
- 93.33% recall: The system identified 28 of the 30 patients classified as positive by the specialist.
- 93.33% F1-score: The combined measure of precision and recall indicated balanced performance in the evaluation.
The results included four classification errors. The system incorrectly identified two patients as positive and failed to identify two patients who had been classified as positive by the specialist.
According to the researchers, these errors were associated with overlapping symptoms between early-stage elderly-onset rheumatoid arthritis and severe osteoarthritis accompanied by secondary inflammation. Some early rheumatoid arthritis cases also presented without the expected positive serological profile.
The study further demonstrated how the Certainty Factor method calculates confidence from multiple symptoms. In one simulated case involving a 65-year-old patient, four symptom inputs produced a diagnostic certainty score of 96.65%.
Across the confirmed positive samples, the average diagnostic certainty reached 88.50%.
These findings suggest that the combination of clinical classification criteria and certainty-based reasoning can support preliminary identification of suspected rheumatoid arthritis cases. However, the results reflect a limited evaluation and do not establish that the system can replace a medical professional's diagnosis.
Potential Benefits for Primary Healthcare Services
The system could offer practical benefits for healthcare facilities that need structured tools to support the initial assessment of older patients.
First, it can help organize symptom information into a consistent digital format. This may reduce reliance on informal symptom descriptions and make preliminary assessments easier to document.
Second, the application could support referral decisions by helping healthcare workers identify patients who may require further examination by a specialist. This is particularly relevant in areas where rheumatology services are not easily accessible.
Third, the transparent calculation process may provide an educational resource for nurses and health cadres working in Posyandu Lansia. By showing how individual symptoms contribute to the final certainty score, the system may help users better understand the reasoning behind its recommendations.
The researchers also proposed expanding the knowledge base to cover other joint disorders, including osteoarthritis, gout arthritis, and polymyalgia rheumatica. They recommended developing a mobile version that could operate offline, potentially making the tool more accessible to healthcare workers in remote areas.
Nevertheless, broader testing remains necessary. The evaluation involved only 50 medical records from a single referral center, so the reported performance should be validated using larger and more diverse patient populations before wider implementation.
Digital Decision Support, Not a Replacement for Medical Diagnosis
The study highlights the potential of artificial intelligence-based expert systems to support healthcare services by converting specialist knowledge into structured digital rules.
For older adults experiencing joint pain, swelling, or prolonged morning stiffness, such technology may help healthcare workers recognize patterns that warrant further assessment. Earlier recognition can support timely referral and clinical evaluation, although the expert system itself does not confirm a medical diagnosis.
The 92% accuracy reported by Putra and Syahra provides encouraging initial evidence for the proposed approach. Future studies involving multiple healthcare facilities, larger datasets, and additional diagnostic comparisons will be important to establish how reliably the application performs in everyday clinical settings.
Author Profiles
- M Zaki Putra -Universitas Muhammadiyah Sumatera Utara, Medan, Indonesia.
- Yohanni Syahra - Universitas Muhammadiyah Sumatera Utara, Medan, Indonesia.
Research Source
Journal: International Journal of Educational Technology Research (IJETR)
Publication: Volume 4, Number 3, September 2026, pages 217–226
DOI: https://doi.org/10.59890/ijetr.v4i3.14
Official Journal URL: https://journalijetr.my.id/index.php/ijetr
0 Komentar