JMIR Publications

ABSTRACT

Background:

In South Africa, diabetes is a leading cause of morbidity and mortality, which was exacerbated during the coronavirus pandemic. Most education and counselling was stopped during lockdown and the Great4Diabetes WhatsApp Chatbot was innovated to fill this gap

Objective:

To evaluate the implementation of the Chatbot in Cape Town, South Africa, between May and October 2021.

Methods:

Convergent mixed methods evaluated implementation outcomes: acceptability, adoption, appropriateness, feasibility, fidelity, cost, coverage, effects and sustainability. Quantitative data was derived from the Chatbot and analysed with the Statistical Package for Social Sciences. Qualitative data was collected from key informants in the health services, Aviro Health and Stellenbosch University and analysed using the framework method, assisted by Atlas-ti. The Chatbot provided users with 16 voice messages and graphics, in English, Afrikaans or Xhosa. Messages focused on coronavirus and self-management of type-2 diabetes. Users had to reply to a question after each message to receive the next message and give brief feedback at the end of the programme.

Results:

The Chatbot was adopted by the Metro Health Services to assist people with diabetes who had restricted health care during lockdown and yet were more at risk of hospitalisation and death from coronavirus. The Chatbot was disseminated via healthcare workers in primary care facilities and local non-profit organisation as well as via local media and television. Two technical glitches interrupted the dissemination, but did not substantially affect user behaviour. Minor changes were made to the Chatbot to improve its utility for users. Many patients had access to a smartphone and were able to use the Chatbot via WhatsApp. Overall 8158 people connected with the Chatbot and 4577 (56.1%) proceeded to listen to the messages, with 12.6% of them listening to all 16 messages, mostly within 32 days. Incremental set-up costs were $5295 and operational costs over 6-months were $17304. More than 90% of users that listened to each message found them useful. Of the 533 that completed the whole programme 71.1% said they changed their self-management “a lot” and 87.6% were more confident. Most users changed their lifestyle in terms of diet (76.1%) and physical activity (53.6%). Healthcare workers also saw the benefits to patients and recommended the service continue. Sustainability of the Chatbot will depend on the future policy of the provincial Department of Health towards mHealth and willingness to contract with Aviro Health. There is potential to go to scale and include other languages and chronic conditions.

Conclusions:

The Chatbot shows great potential to complement traditional health care approaches for people with diabetes and assist with more comprehensive patient education. Further research is needed to fully explore the patient’s experience of the Chatbot and to evaluate the effectiveness in our context. Clinical Trial: NA