Background/Purpose: Diagnosis in rare diseases cases is often delayed by several years. Main factors for delayed diagnosis are believed to be lack of awareness and knowledge about rare diseases among health care professionals. Diagnostic decision support systems (DDSS) have the potential to enhance clinical diagnosis by assessing case data based on incorporated medical knowledge and by suggesting relevant differential diagnoses. DDSS can contribute to professional support and education by visualising medical information and reasoning. We report about the use of Ada/DX, a DDSS in development, in an outpatient clinic for rare inflammatory systemic diseases. Presenting preliminary results, we evaluate the system’s diagnostic accuracy and assess the potential impact of this diagnostic and educational tool on the time to diagnosis.

Methods: This retrospective study is being conducted at the outpatient clinic for rare inflammatory systemic diseases at the Hannover Medical School, Germany. Ethical approval was obtained from the local ethics committee. To date, 82 (of a total 120) patient cases with confirmed diagnosis were included. The time of the visit of first documented symptoms and the time of diagnosis were identified. Time to diagnosis (TD) was calculated. Documented clinical evidence from the medical record was transferred to the DDSS and the disease suggestions in the DDSS were evaluated. Primary endpoint was the correctness of top disease suggestions for the visit of diagnosis. In these cases, secondary endpoints were the time to first correct top rare disease suggestion (T1R) and the time to first correct top 5 rare disease suggestion (T5R). The difference between TD and T1R and the difference between TD and T5R was calculated. Wilcoxon signed-rank test was conducted.

Results: On preliminary evaluation, primary accuracy of top suggestions of the DDSS at the time of diagnosis was 80.5% (71.9% to 89.1%, 95% CI). The table shows a comparison of the original time to diagnosis without the use of the DDSS and the time to correct disease suggestions with the use of the DDSS. (All times are expressed in months.)

The Wilcoxon signed-rank test shows a significant difference for TD – T1R (z-score -5.37, α=0.05, p<0.001) and TD – T5R (z-score -6.03, α=0.05, p<0.001). Main reasons for incorrect DDSS disease suggestions were multi-morbidity (cases with multiple relevant diagnoses), atypical disease presentation and high level of case complexity.

Conclusion: The DDSS suggested the correct diseases based on information from the medical record in most of the analysed rare disease cases. The DDSS often suggested the correct diseases at times prior to the visit of diagnosis. DDSS could be used as educational tools that suggest relevant differential diagnoses in rare disease cases. They might help to reduce time to diagnosis and improve patient outcomes. Prospective research is needed to verify the results.

ACR Meeting Abstracts - https://acrabstracts.org/abstract/the-accuracy-and-potential-impact-of-a-diagnostic-decision-support-system-in-rare-disease-cases/