A Tale of Two Technologies: Tele-Medicine And Chatbots (Part 2)
Why Healthcare Chatbots Show Potential but Still Struggle to Deliver Meaningful Outcomes
When it comes to AI-driven health interventions, chatbots have been the flavor of season for a while now. This is driven largely by the fact that Natural Language Processing and the Generative AI capabilities of Large Language Models – technologies on which chatbots rely – have dramatically improved over the past decade or so.
This technological breakthrough, combined with the exponential growth in data and computational-capacity, has resulted in a rapid growth with chatbots. And this growth of chatbots appears to be especially florid in healthcare. Estimates are, between 2022 and 2032, the chatbot market in health is predicted to grow 10-fold to USD 1.2 billion from $196 million.
A few factors make chatbots inherently attractive in healthcare. Artificial Intelligence enabled chatbots allow 24 x 7 access to usefully curated and parsed information via a user-led, bi-directional, simulated conversation in a human language. This is very useful in healthcare, an information-heavy industry where large volumes of information of variable complexity have to flow seamlessly in several directions and channels.
Chatbots are also attractive in healthcare because health systems around the world are just perennially short of staff. Chatbots can offer a workaround of sorts around this crucial bottleneck, especially when the role involves the information-distribution of predictable nature. On a related note, because chatbots may save man-hours, there is also the expectation that chatbots may save health systems money. This has made chatbots especially attractive in resource-limited settings across the world, where chatbots have been deployed as hopeful quick-fix solutions to the perennial problem of inadequate access to health services.
Several reviews of chatbot implementation across healthcare have found that users and patients are often drawn to their usability and accessibility, especially compared to traditional web-portals or telehealth platforms. Users and patients also have found them particularly useful for health issues where access to information is central to service provision. For example, behaviour change communication, health promotion, and even mental health services. Then there is also the fact that some patients and users of health services appear to prefer a private chatbot over a person; users feel human providers could be judgemental and may not honor their privacy.
However, concerns remain. Data safety and privacy, implementation bottlenecks, and the holy grail–whether chatbots actually achieve outcomes that matter i.e. improved access to health services, cost-effectiveness of services, and improved health outcomes.
A recent review of chatbots evaluated their implementation across four countries in resource-limited settings. These chatbots, implemented across Africa and South America, provided services ranging from sexual health to mental health found that while chatbots could help improve access to services, a few common bottlenecks were recurrent–chatbots implementations failed to acknowledge and plan for the realities of health systems in resource-limited contexts, including infrastructural and technical bottlenecks. Chatbots were thought of as a standalone technical intervention with little thought on how they fit into the overall health system and how they might advance broader systemic goals. Chatbots also did not adequately consider user-perspectives. This resulted in the lack of crucial buy-in from people who are the most important stakeholders in any health system–paitents and health service providers.
As a result, while chatbots still hold promise, their ability to improve health outcomes that matter have been conflicting at best. A meta-analysis of mental health chatbot found no conclusive evidence of improved mental health outcomes, to improve mental health found inconclusive, although another meta-analysis of a chatbot intervention to improve lifestyle behaviour did find benefit.
Chatbots, it appears, could indeed help improve health services delivery and improve health outcomes, but it will be a while before we get there reliably and consistently.
Coda
The Guardian recently published a long-form audio podcast of the story of a Chinese woman who’d grown to like and rely on a DeepSeek chatbot over her transplant nephrologist. This nephrologist was a doctor she’d been relying on for the medical care of her transplanted kidney for a long time; one she’d travel for two days to see no less. The woman is reported to have remarked, “Deepseek is humane, doctors are more like machines.”
A fascinating story, and an early insight into the brave new world healthcare is venturing into. This feature, written by Viola Zhou, was first published in Rest of World with the title My mom and Dr DeepSeek.
Kiran Raj Pandey is a physician and a health services & systems researcher. Learn more about him and his work at kiranrajpandey.com.
 | A guest post by
|