AI-assisted Integrated Care to Promote Colonoscopy Uptake

Colonoscopy, Colorectal Neoplasms

colorectal cancer, colonoscopy, artificial intelligence, integrated care, trial

Colorectal cancer (CRC) ranks the second most common cancer and the fourth leading cause of cancer-related deaths in China. Early screening of CRC has been proven to reduce the incidence and mortality, with colonoscopy as the gold standard for CRC screening. This trial aims to evaluate the effectiveness of artificial intelligence-assistant integrated care for improving uptake rate of colonoscopy among high-risk individuals aged 40 to 64 in China. It's a two-arm, parallel cluster randomized controlled trial. The main question it aims to answer is whether the AI-assisted integrated care influence participants' screening-related knowledge, health beliefs, behavioral intention, and uptake of colonoscopy. Participants will: 1. Be recruited and allocated into one of two groups according to the assigned clusters. Participants in one group will be invited to receive usual specialty care. In addition to usual specialty care, participants in the other group will receive AI-assisted integrated care provided by specialist and general practitioners collaboratively. 2. Complete a questionnaire survey on their knowledge, health beliefs, behavioral intention on CRC screening. 3. Have their colonoscopy status checked at the middle and end of trial.

We will conduct a two-arm, cluster randomized controlled trial to evaluate the effectiveness of an AI-assisted integrated care (AICC) model in improving colonoscopy uptake rate among high-risk individuals aged 40-64. This will be followed by a pragmatic implementation science study to assess user engagement of AICC and identify the facilitators and barriers to its real-world implementation. Sample size calculation, based on detecting an increase in colonoscopy uptake from 15% to 30% with 80% power (α=0.05, two-sided), an ICC of 0.05, and 10 participants per cluster, indicates a need for 18 clusters per arm. Allowing for 10% attrition, the final sample size is determined to be 20 clusters per arm. Thus, a total sample size is 400 participants from 40 clusters. Participant recruitment will be conducted across 40 villages/communities in three representative counties/cities in China. An independent biostatistician will randomly allocate these villages/communities within each county/city to the study arms in a 1:1 ratio. The study procedure involves first identifying high-risk individuals for CRC through an initial risk assessment questionnaire and a fecal immunochemical test (FIT). Those who meet the criteria will then receive the intervention corresponding to their village's assigned study group. Participants in the intervention group will receive AICC. This includes a colonoscopy recommendation from a county specialist for both participants and their families, followed by an introduction to and guided registration for a CRC education chatbot with an initial 5-minute tutorial. Subsequently, general practitioners will conduct three monthly face-by-face follow-ups, each comprising a brief reminder of colonoscopy and a guided usage of CRC education chatbot. The control group will receive only a colonoscopy recommendation from a county specialist, with access to the chatbot granted only after the end of the 6-month study period. Post-intervention, all participants will complete a questionnaire assessing CRC screening knowledge, health beliefs, and behavioral intention. Colonoscopy uptake will be collected via the hospital information system at the 3- and 6-month follow-up. The primary analysis will follow the intention-to-treat (ITT) principle. The primary outcome is the uptake and timing of colonoscopy at 3 and 6 months after intervention. Secondary outcomes encompassed several domains: CRC screening knowledge, beliefs, and intention; chatbot usability and user engagement; and intervention costs. Between-group comparisons for continuous and categorical variables will utilize t-tests and chi-square tests. To account for potential confounders, the generalized estimating equation (GEE) will be employed to derive robust effect estimates. The timing of colonoscopy uptake will be analyzed using Kaplan-Meier survival curves and log-rank tests, and the intervention effects on the time-to-event will be quantified with a Cox proportional hazards model. Subgroup analyses will be conducted to elucidate the effect heterogeneity across populations stratified by baseline characteristics.