Machine learning algorithms trained with pre-hospital acquired history-taking data can accurately differentiate diagnoses in patients with hip complaints

Background and purpose — Machine learning (ML) techniques are a form of artificial intelligence able to analyze big data. Analyzing the outcome of (digital) questionnaires, ML might recognize different patterns in answers that might relate to different types of pathology. With this study, we investigated the proof-of-principle of ML-based diagnosis in patients with hip complaints using a digital questionnaire and the Kellgren and Lawrence (KL) osteoarthritis score.Patients and methods — 548 patients (> 55 years old) scheduled for consultation of hip complaints were asked to participate in this study and fill in an online questionnaire. Our questionnaire consists of 27 questions related to general history-taking and validated patient-related outcome measures (Oxford Hip Score and a Numeric Rating Scale for pain). 336 fully completed questionnaires were related to their classified diagnosis (either hip osteoarthritis, bursitis or tendinitis, or other pathology). Different AI techniques were used to relate questionnaire outcome and hip diagnoses. Resulting area under the curve (AUC) and classification accuracy (CA) are reported to identify the best scoring AI model. The accuracy of different ML models was compared using questionnaire outcome with and without radiologic KL scores for degree of osteoarthritis.Results — The most accurate ML model for diagnosis of patients with hip complaints was the Random Forest model (AUC 82%, 95% CI 0.78–0.86; CA 69%, CI 0.64–0.74) and most accurate analysis with addition of KL scores was with a Support Vector Machine model (AUC 89%, CI 0.86–0.92; CA 83%, CI 0.79–0.87).Interpretation — Analysis of self-reported online questionnaires related to hip complaints can differentiate between basic hip pathologies. The addition of radiological scores for osteoarthritis further improves these outcomes.

Use of artificial intelligence (AI) techniques like data mining, machine learning (ML), and deep learning are now starting to erupt within healthcare, with first applications aimed at cancer diagnostics (Nguyen et al. 2018, Codari et al. 2019), cardiology (Nirschl et al. 2018) and image recognition in radiology (Wang et al. 2017, Fourcade and Khonsari 2019).AI is also emerging within the field of orthopedic surgery (Duffield et al. 2017). Earlier work using AI in orthopedic studies showed the ability of ML to classify knee osteoarthritis (OA) subjects versus healthy patients. Based on kinematic data Kotti et al. (2017) achieved an accuracy of 73%. In comparison with that study, which collected its data in a laboratory setting, Dolatabadi et al. (2017) used kinematic data from more unobtrusive sensors and were also able to distinguish OA subjects from healthy patients. Other ML-related publications in orthopedics report on spine pathology detection, fracture detection, and bone and cartilage image segmentation (Ashinsky et al. 2015).However, to our knowledge, no studies in orthopedics have developed ML algorithms for predicting a clinical diagnosis. In this paper we used information from digital intake forms, which were completed online by our patients before initial consultation with an orthopedic surgeon. We sought to determine (1) the accuracy of different ML algorithms to predict a pre-hospital diagnosis in patients suffering from hip complaints based on history-taking questions only, and (2) how much radiographic imaging results contribute to accurately predicting a diagnosis in these patients.