Table 3.

Type, features, and results of the CDSSs^a for diagnostic support.

CDSS name and study	CDSS type	Features of the CDSS	Results
Benditz et al [38], 2019	Knowledge based	Questionnaire-based CDSS A computerized tool with disease-specific algorithms cascading the next best questions leading to the most probable diagnosis and actions	Diagnosis of CDSS compared with spinal surgeons: Cramer V=0.424b; P<.001
Benditz et al [39], 2021	Knowledge based	Decision tree algorithm and app-based questionnaire Questionnaire will ask the patient to identify the location of their pain and present dichotomous questions to suggest a diagnosis. If the patient scores >65% on these questions, the diagnosis is confirmed. If not, the questionnaire will ask additional questions about the second most likely diagnosis. If the patient still scores <65% after these questions, they are advised to consult with a physician.	Diagnosis of CDSS compared with spinal surgeons: Cramer V=0.711b; P<.001 Concordance: 67.4% A total of 15.1% overestimated A total of 7% underestimated
Lin et al [40], 2006	Knowledge based	Knowledge from 2 highly experienced physical therapists and web-based questionnaire Patients can start a self-diagnosis session with or without clinician’s assistance. Questions regarding specific pain symptom or assessment will be presented through typically 13 to 15 web pages, depending on the number of follow-up questions triggered. After completion of the questions, a diagnosis that may consist of ≥1 parts, based on patient information, clinical evidence provided by the user, and system’s rule activation, will be generated and the clinician can override any parts of the diagnosis. The explanatory panel can be activated by the clinician to review the system’s reasoning process. The clinician can add, remove, or modify an existing rule in reference to its observation, decision outcome, or certainty level.	Not reported
Peiris et al [41], 2014	Knowledge based	Recommendations from 15 guidelines for back pain management After excluding serious pathology, the CDSS will continue to assess for the most probable diagnosis and treatment through a series of questions. A personalized information sheet will be printed.	Not reported
Kim et al [42], 2022	Nonknowledge based	Computer vision–based posture analysis system The CDSS uses a Kinect sensor and specialized software to analyze a person’s skeletal structure and gait. The CDSS captures an image and records a moving video of the participant. The software then identifies the participant’s joints and uses them to determine the skeletal structure and gait. Furthermore, it uses a set of algorithms to judge the probability of scoliosis by analyzing the curvature of the participant’s central coronal axis, which is determined by a line connecting the eyes, shoulders, and pelvis. The CDSS classifies scoliosis as normal (≤3 mm curvature), 20% scoliosis (3 mm to 10 mm curvature), or 50% scoliosis (>10 mm curvature).	Postural deformations: assessed with 94% accuracy (comparable with radiographic assessments) Normal or mild scoliosis: conformity assessment accuracy of 98.57% CDSS’s diagnostic accuracy for scoliosis was 0.94, with the most influential factors being spinal curvature and pelvis height, which accounted for 79.97% and 19.86% of the variance in the data, respectively
Vertebral Compression Fracture tool, Wang et al [43], 2011	Knowledge based	Logistic regression and web-based checklist Uses checklists for dichotomous and nondichotomous discrete variables based on MRIc features to generate a probability of malignancy and a text report. The model captures inputs from these variables to make its assessment.	Not reported

^aCDSS: clinical decision support system.

^bInterpretation of Cramer V effect size measurement of association: effect size ≤0.2: weak association, <0.2 effect size ≤6: moderate association, and effect size >0.6: strong association.

^cMRI: magnetic resonance imaging.