The methodology of this ScR was based on the method described by Arksey and O’Malley [27], and the report followed the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses Extension for Scoping Reviews) guidelines. The protocol was developed and registered in the International Platform of Registered Systematic Review and Meta-analysis Protocols (INPLASY; #202290061).

This ScR aimed to answer 3 research questions (RQs):

This study aimed to include all relevant literature and conference abstracts in English or Chinese. To identify relevant studies, an extensive search was conducted across 7 electronic databases: PubMed, Embase, the Cochrane Library, the Web of Science, the China National Knowledge Infrastructure (CNKI), Wanfang, and VIP. All searches were performed from the date of database establishment up to June 30, 2022. In addition, searches of additional sources, such as Google, Baidu, and official conference websites, were conducted for academic conference abstracts from the period of 2021-2022. Details are listed in Multimedia Appendix 1.

Studies were included if they reported the clinical application of CDSSs in diabetes care. CDSSs in this ScR referred to any technology-based systems (ie, mobile/tablet, web-based, or computer-based app) that can provide support for clinical decision-making and be applied across the whole spectrum of diabetes care, such as CDSSs used for treatment recommendation, complication risk assessment, and blood glucose monitoring.

The exclusion criteria were as follows:

Two researchers (authors XL and YL) independently evaluated the titles and abstracts of the identified studies based on the eligibility criteria. The full texts of potentially eligible studies were retrieved and then independently screened by the same 2 researchers (XL and YL). The 2 researchers also recorded the reasons for exclusion, and disagreements were resolved by a third senior researcher (author SH).

Two researchers (authors SH and JL) independently collected data using a standardized data sheet. Disagreements were resolved by a third researcher (XL). We collected the following variables that were pertinent to the aims of this research: (1) study characteristics (eg, publication year, number of subjects, follow-up period), (2) CDSS characteristics (eg, decision-making mechanism, functions) and clinical applications, and (3) evaluation of CDSSs in diabetes care (eg, user experience, user adherence, effectiveness and safety).

Continuous variables were summarized into categories to allow for a more meaningful summary. Categorical variables were summarized using frequency counts and percentages. The number of papers reporting the corresponding outcome was used as the denominator for each variable.

A total of 11,569 studies were identified from included sources. After removing duplicated publications, 9607 (83%) studies were available for title and abstract screening. Finally, 237 (2.5%) studies were evaluated based on their full text, and 85 (35.9%) studies [23,28-111] (including 13/18, 15%, conference abstracts) were selected for analysis. The PRISMA-ScR flow diagram is shown in Figure 1 and the PRISMA-ScR checklist in Multimedia Appendix 2.

In the 85 studies included in this ScR, a total of 159,475 subjects were enrolled. The number of publications was undergoing rapid growth during 2021-2022, and 45 (53%) of the studies were published in the past 5 years (2018-2022), as shown in Multimedia Appendix 3). Tables 1-7 summarize the characteristics of the included studies. The most common follow-up period was less than 12 months (<6 months: 20/75, 27%; 6-12 months: 38/75, 51%). The majority of studies were conducted in North American or European countries, with 34% (22/64) being conducted in the United States and 9% (6/64) in Austria and Germany. About half of the studies (34/65, 52%) were multicenter and 48% (31/65) were single-center studies. Most studies were RCTs (36/85, 42%), followed by single-arm trials (22/85, 26%), observational studies (18/85, 21%), and pragmatic clinical trials (PCTs; 5/85, 6%).

The characteristics of CDSSs are summarized in Table 8. Most CDSSs included in this ScR were knowledge based (44/58, 76%), although non-knowledge-based CDSSs have been increasing in recent years. In 59% (48/82) of the studies, physicians were the users of CDSSs. In 51% (42/82) of the studies, patients were the users of CDSSs. Both physicians and patients were the users of CDSSs in 10% (8/82) studies. In 30% (25/82) of the studies, nurses, medical assistants, and pharmacists supported physicians in using CDSSs. The types of outputs provided by CDSSs were vast, and we classified them based on the 6 categories of intervention types reported in the Clinical Decision Support Implementers’ Workbook [112]. Most CDSSs facilitated users by providing proactive order suggestions and order sets (65/83, 78%) and supporting guidelines, complex protocols, algorithms, and clinical pathways (25/83, 30%).

Users could leverage CDSSs in the clinical management of diabetes in various ways (Table 9). The most common application scenario of CDSSs was to provide treatment recommendations (63/81, 78%), not only for physicians (36/47, 77%), but also for patients (29/42, 69%). Of the 36 (77%) studies, CDSSs were commonly used by physicians for drug recommendations (n=22, 61%) and insulin dose adjustment (n=14, 39%). Of the 29 (69%) studies, CDSSs were used by patients not only for insulin dose adjustment (n=16, 55%) and drug recommendations (n=6, 21%) but also for suggestions for diet and exercise (n=11, 38%). Other application scenarios included medical education (13/81, 16%), complication risk assessment (12/81, 15%), blood glucose monitoring (12/81, 15%), diabetes screening (4/81, 5%), and appointments for examinations (3/81, 4%). When categorizing physician users according to their medical disciplines, primary care physicians (20/39, 51%) were the most common, followed by endocrinologists (15/39, 38%) and nonendocrinologists (8/39, 21%). For primary care physicians as users in 20 (51%) studies, CDSSs were mainly used for treatment recommendations (n=16, 80%) and no application scenario for blood glucose monitoring was found. For endocrinologists as users in 15 (38%) studies, CDSSs were mainly used for treatment recommendations (n=13, 87%) and no application scenarios for medical education, appointments for examinations, and diabetes screening were found. For nonendocrinologists as users in 8 (21%) studies, CDSSs were only used for treatment recommendations (n=4, 50%), complication risk assessment (n=3, 38%), and diabetes screening (n=1, 13%).

CDSSs in diabetes care have been evaluated using various dimensions, including the effectiveness, safety, consistency, and diagnostic accuracy of CDSSs; user behavior, user adherence, and user experience; and cost-effectiveness. Studies that evaluated the effectiveness and safety of CDSSs regarding biomarkers were the most prevalent, and the results of the effectiveness of CDSSs for biomarkers are summarized in Table 10. Regarding the safety of using CDSSs, the risk of hypoglycemia significantly decreased in 34% (12/35) studies [36,49,60,61,67,69,73,82,86,87,98,110].

Of the 85 studies included, 64 (75%) assessed the effectiveness of CDSSs in improving patients’ blood glucose (n=63, 98%), blood pressure (n=18, 28%), and blood lipid levels (n=21, 33%). Significant improvements in biomarkers were based on the reported results of the included studies. A significant improvement in any 1 biomarker was considered significant. CDSSs significantly improved patients’ blood glucose (45/63, 71%), blood pressure (12/18, 67%), and blood lipid (8/21, 38%) levels. Specifically, CDSSs significantly improved glycated hemoglobin (HbA1c; 30/43, 70%), glucose variability (GV; 5/7, 71%), diastolic blood pressure (DBP; 10/15, 67%), time in range (TIR; 12/18, 67%), systolic blood pressure (SBP; 11/18, 61%), and mean blood glucose (MBG; 8/14, 57%).

In addition, 35 (41%) studies evaluated the safety of CDSS use in diabetes care, indicating that CDSSs would not increase the risk of hypoglycemia. Meanwhile, CDSSs significantly decreased the risk of hypoglycemia in 34% (12/35) of the studies.

Furthermore, 3 (4%) studies analyzed the consistency of insulin dose adjustments determined between CDSS algorithms and physicians, suggesting that the recommendations made by CDSSs are like those made by physicians and have an acceptance rate of more than 90% among HCPs. Additionally, 2 (2%) studies demonstrated the diagnostic accuracy of CDSSs in predicting the risk of diabetic retinopathy.

Of the 85 studies included, 16 (19%) evaluated users’ adherence and the results are summarized in Multimedia Appendix 4. In 7 (44%) studies, users’ adherence to insulin dose suggestions was over 90%. In 7 (44%) studies, CDSSs improved users’ adherence to follow-up and examination appointments, diabetes care guidelines, and drug usage. In the remaining 2 (13%) studies, users’ adherence to suggestions for lifestyle changes ranged from 50% to 80%.

In addition, 25 (29%) studies evaluated the user experience of CDSSs. All studies reported positive comments, with 9 (36%) also reporting negative comments. Users provided favorable comments for CDSSs, such as “It was easy to use” [81] and “It offered useful information” [62]. Meanwhile, challenges and limitations associated with using CDSSs were exposed by negative comments, for example, “Software glitches influenced physicians’ acceptance of CDSSs” [23], “A lack of integration with the electronic health record (EHR) system would result in a more complicated workflow” [23], “Some recommendations provided by CDSSs did not consider comorbidities or patient adherence” [23], and “CDSSs were not up to date on the most recent guidelines” [72].

To the best of our knowledge, this is the first ScR to provide a comprehensive analysis of the use of CDSSs in diabetes care. Our findings suggest a significant increase in the number of studies and relevant study participants in recent years, reflecting a growing interest in using CDSSs in diabetes care. Most CDSSs are knowledge based, while the number of non-knowledge-based CDSSs has been increasing in recent years. CDSSs can be used by diverse users (even nonendocrinologists, nurses, medical assistants, and pharmacists) in various application scenarios, including treatment recommendations, medical education, complication risk assessment, blood glucose monitoring, appointments for examinations, and diabetes screening. The included studies demonstrated that CDSSs are effective and safe for diabetes care.

Studies assessing the effectiveness of CDSSs primarily used biomarkers (eg, HbA1c, TIR, or low-density lipoprotein [LDL]) as endpoints. The most common follow-up period in the included studies was less than 12 months. Our review found that CDSSs significantly improve blood glucose, blood pressure, and lipid profile (71%, 67%, and 38% of the studies, respectively) and that the risk of hypoglycemia does not increase correspondingly. This aligns with the results of previous reviews [15,16,18,113-115]. In recent years, there has been an increasing focus on long-term outcomes in diabetes care [116-122]. The long-term outcomes of implementing CDSSs are still unknown; thus, further research with long-term outcomes is needed.

An evident disparity was observed between the care recommended by clinical guidelines and the actual care provided to patients, ultimately leading to suboptimal glycemic control outcomes [123]. CDSSs might play a vital role in improving the quality of diabetes care in the following ways:

CDSSs might be most useful for HCPs with limited formal education and practical experience in diabetes care or for patients with limited access to medical resources [84,128]. As the incidence of diabetes continues to rise and the number of qualified endocrinologists remains inadequate [11], primary care physicians might find themselves increasingly responsible for managing patients with diabetes [84,129]. Primary care physicians face challenges in diabetes management as they usually deal with multiple health issues and have little experience with the standard of care for diabetes [130]. Our review revealed that primary care physicians are the main users of CDSSs in diabetes care, especially in the scenario of drug recommendations, likely due to their lack of knowledge of the latest guidelines compared to specialized endocrinologists. Regarding glucose management, nonphysician HCPs (eg, nurses, medical assistants, and pharmacists) and nonendocrinologists face comparable situations to those of primary care physicians, suggesting that CDSSs have great potential for application in these situations. Our review found that some nurses, medical assistants, pharmacists, and nonendocrinologists have initiated the use of CDSSs in diabetes care.

Additionally, effective diabetes care is a multifaceted process that relies not only on the expertise of HCPs but also on the active participation of patients in managing their diet, exercise routine, medication management, and other important health factors [131]. Benefiting from the development of mobile internet technology, our review found an increasing trend of CDSSs being developed for patient-oriented care. These CDSSs could facilitate patient self-management in diverse application scenarios, such as providing recommendations for insulin dose adjustment, providing suggestions for diet and exercise, providing medical education, and monitoring blood glucose. This might be especially appealing to patients who live in rural areas or have limited access to in-person physician visits [84].

Sirajuddin’ et al [132] stated that for modern CDSSs to be effective, they should follow the Five Rights model. This model emphasizes that delivering “the right information to the right person, in the right format, through the right channel, and at the right time” is crucial for achieving lasting improvements. This ScR found that not all CDSSs could fit the model.

One of the challenges identified in this study was the suboptimal format and channel, such as the lack of integration with hospital electronic systems and the unfavorable design of the CDSS’s human-computer interaction [23]. Integration of CDSSs into hospital systems to reduce physicians’ workload [133], expediting software iteration, and developing CDSSs in collaboration with physicians [134] could help resolve these challenges.

Some CDSSs in the studies included reported challenges in providing the “right information.” User acceptability of CDSS recommendations has decreased due to incomplete data collection [23] and delayed updates to the CDSS knowledge base [72]. Smart wearable devices could be leveraged to improve the efficiency and accuracy of data collection [135] and assist in making specific recommendations as opposed to a variety of suggestions. It is challenging to timely manage and maintain the rules (created based on expert knowledge, guidelines, etc) of CDSSs. Outdated rules could lead to inaccurate suggestions for treatments or preventive services. The extensive range of available data sets has led to the application of new methods (eg, association rule mining and machine learning algorithms) to explore novel modes of knowledge, which might reduce the cost of updating and maintaining the knowledge base [128].

In addition, our review found that although knowledge-based CDSSs remain the most commonly used type, the rise of AI and big data has led to an increase in non-knowledge-based CDSSs, which are primarily used to provide treatment recommendations for insulin dose adjustment and predict the patient’s risk of complications. It is possible that we are currently undergoing a transformation from the rule-based approach to new methods, such as machine learning combined with voluminous clinical databases, offering more precise and personalized approaches to health care [128,136].

Furthermore, CDSSs might be effective and safe in improving diabetes care, but the cost of design, local implementation, ongoing maintenance, and user support for CDSSs could be high [137,138], which might be a significant barrier to fully implementing CDSSs. In the view of service payers (eg, health care facilities, insurers, and policy makers) promoting the use of CDSSs, it is important to find evidence about whether CDSSs are cost-effective in improving diabetes care. However, few studies have reported the cost and economic benefits of CDSS implementation.

As discussed before, there are several research gaps. Future research should:

This ScR has several limitations. First, publication bias could exist in the studies included as negative results may not always be published. Second, this ScR could be subject to information bias due to certain data being collected based on subjective judgment. However, senior researchers and experts participated in data validation and verification to minimize potential bias. For instance, it is difficult to distinguish between knowledge-based and non-knowledge-based CDSSs. To address this issue, we enlisted the assistance of industry professionals to identify the decision-making mechanisms of CDSSs, but misclassification might still exist. Lastly, the great heterogeneity in CDSSs’ design, purpose, and targets for evaluation prevented us from conducting a quality assessment and a meta-analysis, which according to ScR guidelines is usually not required.

This ScR found that CDSSs are being increasingly used in diabetes care and have been widely implemented by diverse users across various scenarios. They have been shown to be effective and safe in improving diabetes care, implying that CDSSs can be a reliable assistant for physicians and might be particularly helpful for physicians with limited experience and patients with limited access to medical resources. CDSSs also face some challenges and necessitate ongoing optimization iterations. Future studies should focus on further improving CDSS performance, evaluating their long-term effects and cost-effectiveness, and promoting their usage among HCPs and patients beyond endocrinology.