EFFICACY OF AN AI ENABLED PATIENT-FACING CLINICAL DECISION SUPPORT SYSTEM IN THE DIAGNOSES OF LUMBAR RADICULOPATHY

S. Thomas¹, F. Rahman Khan², M. Faisal Chevidikunnan³, F. Hassan Badahman³, A. Khalid Alzahrani³, V. Mary Vergis¹

¹THERAPHA, Inc, New City, New York, United States, ²King Abdulaziz University Hospital, Deaprtment of Physical Therapy, Jeddah, Saudi Arabia, ³King Abdulaziz University, Department of Physical Therapy, Jeddah, Saudi Arabia

Background: Lower back with or without radicular symptoms are of very high prevalence worldwide and is one of the leading causes of disabilities. Irrespective of numerous guidelines and campaigns, imaging studies have been routinely performed in the management of these pathologies leading to over diagnoses, over treatment and drug dependencies. In the absence of any red flag symptoms, early imaging lead to low value care and confusion in both treating clinicians and patients alike. In the conservative care model, both clinicians and patients require some form of validation that symptoms are not of any sinister pathologies but can be managed with few corrective exercises, life style modifications, ergonomic changes and staying active for recovery and prevention. Though 75-83% of out patient clinical diagnoses can be obtained from a skilled history taking alone, modern medicine has deliberately ignored it's significance and rely on imaging studies in establishing a musculoskeletal diagnosis. With the help of a patient-facing clinical decision support system, this study is bringing back the tremendous power of history taking by harnessing advancements in informational technology and artificial intelligence. In short, history taking is used as a diagnostic technology by replacing the traditional intake form with a chatbot. The study highlights the significance of patient interviews just as physical examinations in clinical diagnoses and underlines the irrelevance of over-reliance on imaging studies in conservative care delivery.

Purpose: To compare the efficacy of an AI enabled decision support versus MRI or CT studies in predicting diagnoses for LBP with radicular symptoms.

Methods: A comparative study between current Gold standards (MRI/CT scans) and an AI enabled decision support tool in the diagnosis of lumbar radiculopathy. Inclusion criteria were ages above 18, LBP with leg symptoms and indications for MRI/CT studies. Exclusion criteria were known lumbar disc pathologies, other known pathologies causing similar symptoms and presence of any red flag including but not limited to cancer, infection etc.

Results: With the mean age of 47.2 + 14.4 and BMI of 28.5 + 4.5 in the subjects, Therapha demonstrated a very high sensitivity of 97% and specificity or 21.4%. These values were determined using the Receiver Operating Curve (ROC) and the area under the curve showed a value of 0.775 evidencing the strength of the study. Other highlights are the high Positive and Negative Predictive Values, which were 80% and 75% respectively. The lower specificity may be due to the selection bias and the age group of the selected subjects.

Conclusions: The study unequivocally revealed that, Therapha should be an adjunct in routine clinical evaluations for patients of LBP with radicular symptoms. The tool has demonstrated very high performance in predicting lumbar disc herniations even with probable nerve root compromises.

Implications: The AI enabled chatbot has tremendous value as a low cost decision support tool in the musculoskeletal care especially spine care. This easy to adopt technology should be part of the curriculum to train medical students in developing clinical reasoning skills and look beyond one's inert bias to establish a better clinical picture and diagnoses.

Funding acknowledgements: The authors declare no external funding in this study

Keywords:
Lumbar radiculopathy
Artificial Intelligence
Clinical Decision Support System

Topics:
Innovative technology: information management, big data and artificial intelligence
Musculoskeletal: spine
Musculoskeletal

Did this work require ethics approval? Yes

Institution: King Abdulaziz University Hospital (KAUH), Jeddah

Committee: Institutional Review Board (IRB) || Ministry of Health

Ethics number: HA-02-J-008 || HAP-02-T-067