Automated Structured Education Based on an App and AI in Chinese Patients With Type 1 Diabetes

Automated Structured Education Intervention Based on an App and Artificial Intelligence in Chinese Patients With Type 1 Diabetes

In recent years, more and more attention has been paid to diabetes self-management. Glycemic control and self-management skills of patients with type 1 diabetes (T1DM) in China are poor. Artificial intelligence (AI) and the Internet offer a new way to improve the self-management skills of patients with chronic diseases. Few studies have combined AI technology with structured education intervention of type 1 diabetes. This study is innovative in that it compares the effectiveness of smartphone app between usual care, as well as automatic and individualized app education and standardized app education to explore whether the individualized treatment advocated by the latest guideline will bring any additional benefit to T1DM patients. The ultimate goal is to provide an effective and convenient approach for glycemic control of type 1 diabetes and reduce related disease burden in China.

This is a single-blinded, 1:1 paralleled group cluster randomized controlled trial (RCT). The intervention will last for 24 weeks. The laboratory staff who test the HbA1c level, the outcome assessor who collects the blood glucose data, and the statisticians will be blinded to the treatment allocation. Sample size estimation: We propose to enroll 138 patients with type 1 diabetes (T1DM) by considering withdrawals, 69 in the smartphone app groups and 69 in the routine care group. Sample size estimation is based on hypothesized changes in the primary outcome HbA1c. In order to ensure high quality data, two staff are responsible for the input of original data into the database to check and confirm the accuracy. When the data entered by two staff independently, the auxiliary staff decides which data to use. Data analysis will be conducted under the intention-to-treat principle by including all the randomized patients in the data analysis. Missing data will be filled in with multiple imputation method. Any substantial difference in baseline characteristics will be adjusted with mixed-model regression analysis. Sensitivity analysis will be conducted by using per-protocol data by excluding those patients who drop out of the RCT.

Subjects will be given instructions to install the patient-end App, which includes the following functions: diabetes education, patient-doctor communication, diabetes diary, peer support, reminder for blood sugar test and related abnormal results. They receive push notifications that provides recommended education materials which meet the needs of the patient by considering his/her baseline diabetes-related knowledge.

Subjects only receive the education provided by health-care professionals in the outpatient department

In the 24-week intervention period, the experimental group receives automated push notifications supported by artificial intelligent algorithm.

Diabetes Quality of Life scale (DQOL) is a wildly used 46-item tool for assessing the quality of life related to diabetes in three aspects: diabetes satisfaction (15 items), impact (20 items), and worry (11 items). Each item is responded to on a 5-point Likert scale, with score of 1 represents "always affected", "always worried", or "never satisfied" and a score of 5 indicates "no impact", "no worries", or "always satisfied". Higher total score reflects better quality of life. A Chinese version of DQOL has been translated and validated in the diabetic population from Mainland China and will be adopted in this trial.

Diabetes Self-Management Scale is used to assess diabetes self-management behaviors. This scale contains six aspects with a total of 14 items: dietary management, physical activity, self-monitoring of blood sugar, medical treatment, foot care and smoking. Except for smoking, the other five aspects with 11 items ask the number of days during the last week (ie. how many days did you test your blood sugar during the last 7 days?...). One of the dietary questions (ie. days of high-fat diet consumption) is reversely scored (the more days the fewer score), and the rest are positively scored in 0-7 points. The overall score uses the above five aspects of 11 questions, with a minimum score of 0 and maximum score of 77. Higher score reflects better the self-management behaviors.

Diabetes Self-Care Activities (SDSCA) is used to assess diabetes self-care behavior. This scale contains six behavior related scales: general dietary, specific dietary, glucose monitoring, physical activity, foot care, and smoking. Absolute weekly frequency or consistency of diabetes self-care activities are scored with a 0-7 ranged scale, with higher scores reflecting better performance in self-care behaviors. The internal consistency reliability and construct validity of SDSCA was supported by its psychometric test based on an adult diabetes population. A validated Chinese version of the SDSCA (C-SDSCA) is available for this trial.

Patients' diabetes-related psychosocial self-efficacy will be evaluated with the Diabetes Empowerment Scale-Short Form, which was a short form of Diabetes Empowerment Scale developed from the America population with type 1 or type 2 diabetes. A revised Chinese version is available for the Mainland China population. The Chinese version DES-SF includes 8 domains with 1 item for each (i.e., assessing the need for change, developing a plan, overcoming barriers, asking for support, supporting oneself, coping with emotion, motivating oneself, and making diabetes care choices appropriate for one's priorities and circumstances). Each item is responded on a 5-point Likert scale, with 1 indicating strongly disagree and 5 indicating strongly agree. Total score ranges from 8 to 40, with higher scores reflect a better psychosocial self-efficacy.

State-Trait Anxiety Inventory (STAI) is used for assessing patients psychological status. The Chinese version STAI consists of two sub-scales to measure both state and trait anxiety states. Each of the two anxiety states will be measured with a 20-item sub-scale, and each item will be scored from 1 to 4. The total score for both state and trait anxiety range from 20 to 80, with high scores indicating more serious anxiety.

Beck's Depression Inventory (BDI) is used for assessing patients psychological status. The Chinese version BDI (CBDI) consists of 21 self-rated items. Each item will be scored from 0 to 3, with the total score ranges from 0 to 63, and a higher score indicates more serious depression.

Patients' engagement with the app will be measured in terms of communications with the clinician and the utilization of the smartphone app. Specifically, the number of messages sent to patients, the number of message responses, the number of video calls/phone calls with patients, the number of logs entered by patients, and time spent in the health education module will be collected.

Safety-related outcomes including hypoglycemic events, hospitalization, and emergency room visits will be collected at each follow-up time point including the monthly telephone interview.

Inclusion Criteria: Individuals diagnosed with Type 1 Diabetes according to the 1999 World Health Organization report Insulin dependence from disease onset Aged 18-50 years With a disease duration over 6 months With a HbA1c level over 7% Treated T1DM with multiple daily injections or insulin pump Individuals who own smartphone and are capable of using wechat or apps Exclusion Criteria: Age below 18 years or above 50 years Being pregnant With mental disorders Have any other condition or disease that may hamper from compliance with the protocol or complication of the trial Already using a smartphone app for managing diabetes Having chronic complications including diabetic retinopathy, diabetic nephropathy or diabetic foot, diabetic neuropathy

Huang F, Wu X, Xie Y, Liu F, Li J, Li X, Zhou Z. An automated structured education intervention based on a smartphone app in Chinese patients with type 1 diabetes: a protocol for a single-blinded randomized controlled trial. Trials. 2020 Nov 23;21(1):944. doi: 10.1186/s13063-020-04835-9.