Feasibility study of a home-based artificial intelligence system to empower patients with type 2 diabetes

Jesús González-Lama, Pedro José Mompeán-Torrecillas, Matilde Romero-López, Nicomedes Rodríguez-Rodríguez, Macarena Espinilla-Estévez, José Luis López-Ruiz, David Pérez-Cruzado

Keywords: artificial intelligence; diabetes mellitus; e-health; self-care

Background:

The key to managing type 2 diabetes (T2DM) is a healthy lifestyle: to reach this goal patient involvement is essential. Artificial intelligence (AI) systems could help achieve this goal, if acceptable.

Research questions:

To evaluate the feasibility of an AI system (AI2EPD) in the monitoring of healthy behaviors in patients with T2DM.

Method:

A feasibility study was conducted in 10 patients with T2DM attending a primary care center. AI2EPD is an activity monitoring solution that reports the daily activities. Both monitoring and activity recognition are performed transparently and frees the user from direct interaction with the devices, using sensors installed at home (motion, opening/closing and environment), and an activity wristband. Its main objective is to provide feedback on compliance with the agreement signed prior to the study on treatment, meals, exercise, and other lifestyle recommendations. Data on satisfaction with the tool (System Usability Scale: 10 items with five response options, from ‘Strongly agree’ to ‘Strongly disagree’) were collected after 6 months of use.

Results:

Age ranged from 45 to 75 years (mean 64, SD 9), most were male (6/4), mean BMI 34.14 kg/m2 (SD 5.54), mean HbA1c 7.93% (SD 1.12). Users found the system easy to use (mean 4.2, CI95% 3.46 to 4.94), thought they would like to use it frequently (mean 3.9; CI95% 2.98 to 4.82), imagined that most people would learn to use it very quickly (mean 4.3; CI95% 3.71 to 4.89), and felt very confident using the system (mean 4.2; CI95% 3.75 to 4.65). The main problems reported were central node disconnected, sensor fixation failure, and Wi-Fi and/or electrical network failure.

Conclusions:

The AI2EPD connected health system, minimally invasive and without the need for patient intervention, was very well accepted by the participants despite some technical issues that need to be improved.

Points for discussion:

Is it feasible to use artificial intelligence to personalize lifestyle recommendations and adherence to therapeutic targets in individuals with T2DM?

Do patients accept the use of minimally invasive technology in their homes for medical supervision and self-monitoring?

#85