Table 1.

Summary of key findings from reviewed articles. Characterizes each article in the review by their focus on digital health (DH) or artificial intelligence (AI) as a discipline, the aspect of inflammatory bowel disease (IBD) care that it addresses (diagnosis, treatment, monitoring, prognosis), and the key finding(s). Underlined and italicized are the categories of each approach to IBD care. In bold are the specific approaches utilized by the investigators. Abbreviations: Crohn’s disease (CD), area under the curve (AUC), random forest (RF), ulcerative colitis (UC), artificial intelligence (AI), inflammatory bowel disease (IBD), area under the receiver operator curve (AuROC), tumor necrosis factor (TNF), machine learning (ML), Telemonitoring of Crohn’s Disease and Ulcerative Colitis (TECCU), Simple Clinical Colitis Activity Index (SCCAI), TELEmedecine for Patients with Inflammatory Bowel Disease (TELE-IBD), quality of life (QoL), fecal calprotectin (FC), and health-related quality of life (HRQoL).

	Digital Health
Diagnosis
Treatment	Treatment Adherence and Maintenance “Constant-care” web service Significantly improved adherence to 5-aminosalicylate treatment, knowledge of IBD, and QoL compared to patients receiving standard care [75]. Helped UC patients optimize their maintenance treatment using mesalazine and improve treatment adherence, disease activity, and QoL [76].
	Treatment Management Virtual clinic for anti-TNF therapy management Significantly shortened time until treatment success, provided suitable dose intensification, improved disease control, and improved treatment de-escalation compared to standard CD care [77].
Monitoring	Telemedicine and Telemanagement Approaches	Mobile Applications
	myIBDcoach Significantly reduced the number of outpatient visits compared to IBD patients using standard care while maintaining QoC and disease monitoring (p < 0.0001) [49]. TECCU Reduced outpatient clinic visits among IBD patients. TECCU users experienced improvements in disease activity and 81% of these patients were in clinical remission by the end of the study, compared to 71.4% of patients receiving standard care [48]. CRONICA The self-administered SCCAI via the CRONICA web platform was a trustworthy self-assessment tool for UC patients to monitor their. Online SCCAI scores were 85% in agreement with physician’s assessments of remission or UC disease activity [78]. IBD telemedicine clinic Appointments were evaluated to assess the quality of care provided at a low cost in comparison to standard care. Telemedicine patients saved a mean of $62 in travel costs and at least half a day of time without negative impacts on quality of care [79].	HealthPROMISE Led to a significant reduction in hospitalizations and emergency room visits within one year among IBD patients compared to those who received standard care [80]. TELE-IBD TELE-IBD groups experienced a decline in IBD-related hospitalizations, with a significant decrease when receiving TELE-IBD messages weekly compared to standard care. TELE-IBD educational messages did not significantly improve disease activity and QoL in comparison to standard care, potentially due to the patients having more severe CD and UC [81]. Interviews with patients using TELE-IBD revealed that they considered the service a beneficial supplement to traditional follow-ups and a useful component in IBD self-management to stay educated on IBD, monitor their symptoms, and connect with their physician [82]. IBD-Home 29% of patients were compliant to the IBD-Home application and FC test kit after one year. Patients who were compliant experienced a rise in medical treatment, providing the value to remote disease monitoring [83]. Self-monitoring applications (IBDsmart and IBDoc) Led to significantly fewer outpatient appointments than standard care patients (mean of 0.6 vs. 1.7) without affecting health outcomes or HRQoL (p < 0.001) [84].
Prognosis	IBD-Related Predictions Web-based symptom diary for CD Patient-reported IBD-related symptoms were associated with significant increases in hospitalizations, unscheduled visits, and bowel resection surgeries among CD patients with more severe disease [85].
	Artificial Intelligence
Diagnosis	IBD Detection Tri-matrix factorization model used a combination of exome sequencing data and biological knowledge to differentiate healthy individuals from CD patients (AUC = 0.816) [86]. RF model differentially diagnosed CD and UC using descriptions of colonoscopy images (AUC = 0.936) [87]. AI system built using a probabilistic neural network assessed intestinal crypt architecture distortion and mucosal damage from patient biopsies and diagnosed IBD with 98.31% precision and recall [88]. Deep neural network for evaluation of UC predicted endoscopic remission with 90.1% accuracy and histologic remission with 92.9% accuracy using endoscopic images and biopsies from UC patients [89].
Treatment	Treatment Response Predictions RF algorithms predicted clinical responders and non-responders (AuROC = 0.856) and non-adherence to thiopurine therapy (AuROC = 0.813). Can be used to personalize thiopurine dosages [90]. RF model predicted corticosteroid-free endoscopic remission at 52 weeks of vedolizumab treatment using data acquired during week 6 of therapy (AuROC = 0.73) [91].
Monitoring	Inflammation and Disease Activity Monitoring Deep neural network for evaluation of UC predicted endoscopic remission with 90.1% accuracy and histologic remission with 92.9% accuracy using endoscopic images and biopsies from UC patients [89]. Proprietary ML algorithm was 91% accurate at detecting histologic inflammation from endocytoscopic images and therefore assessing disease activity and risk of clinical exacerbation [92].
Prognosis	IBD Assessment and Predictions Proprietary ML algorithm was 91% accurate at detecting histologic inflammation from endocytoscopic images and therefore assessing disease activity and risk of clinical exacerbation [92]. RF model constructed from medical records of IBD patients predicted IBD-related hospitalizations and outpatient steroid use (AuROC = 0.85) [93].

The underlined and italicized terms should ideally be grouped with the text underneath it.