A Viget Exploration
Patients who don’t take medication as prescribed are at risk of worsening health, increased hospitalization, and a greater chance of mortality.
Just in the U.S., non-adherence has been attributed to over $100 billion in avoidable healthcare costs.1 In developed countries, among patients with chronic illness, approximately 50% do not take medications as prescribed.2
"Increasing the effectiveness of adherence interventions might have a greater impact on the health of the population than any improvement in specific medical treatment."
“All of that medical jargon doesn’t make much sense to me but I don’t want to bother the doctor with my silly questions.”
“I feel just fine. I don’t know why that doctor thinks I’m sick. Why would I take medicine that I don’t need?”
“I took my prescription to the pharmacy but they said it would cost several hundred dollars! I just couldn’t afford it.”
“Taking 15 medications every day is tough. Especially since they all have different instructions. How am I supposed to keep track of everything?”
Access was our primary concern. Smartphone adoption is growing but 35% of the U.S. population still doesn’t own one.4 Reliance on personal computers is also problematic, especially among disadvantaged patients.
We narrowed our exploration to technology that was pervasive and familiar. Nearly 9 out of 10 American adults use a phone capable of texting.5 Furthermore, adherence studies using text reminders show promising results.6 Texting shows the potential to address difficult issues around the world and across verticals. In Libya, the development of the world’s first SMS voter registration system led to 1.5 million voter registrations.7
Could texts be used not only to send reminders but also to build an ongoing two-way dialogue with a trusted health partner? Would this conversation help improve medication adherence?
Our explorations led to the concept of Florence — a personal companion that would help patients understand and manage their medication regimen. A chatbot at heart, patients would converse with her through text and picture messages.
If developed, a simple, familiar interface would hide her powerful intelligence. Behind the scenes, machine learning and behavioral modeling would enable her to offer a tailored, human experience based on each patient’s needs.
Florence strikes up a conversation, learning details about your life, health, goals, and preferences in an engaging, human way.
This conversation, which unfolds over time, feels less like twenty questions and more like the start of a relationship.
Florence helps you remember what to take and when. Image recognition makes adding your medication as easy as taking a picture.
As you start taking new medications, Florence will follow up to see if you’re experiencing side effects.
She can offer additional information to help you understand if your experience is typical — or if you should talk with your doctor.
With complex regimens, days often blur together. Florence will show you what you’ve taken today — and what’s left to take.
Because Florence knows your regimen, she can find you ways to reduce the cost of care.
Florence can send you occasional notes of encouragement to keep you motivated.
For patients with asymptomatic conditions, this offers a sense of progress.
If Florence notices you’re not staying on top of your plan, she can message family members or others in your support network.
Side effects and drug interactions are currently available in services like Epocrates and Medscape. Though designed for long-form reading, it could be pruned and adapted to fit messaging constraints.
Generating believable messages that don’t feel artificial or templated is key to maintaining the illusion. Tools like Automated Insights Wordsmith hint to possibilities where content can be produced on the fly and personalized to each user.
Florence’s conversational style will encourage users to speak to her like a friend. Natural language processing, like those offered by IBM’s Watson, will be needed to extract meaning and respond intelligently.
Florence can generate many different messages which, unchecked, could make her a little too friendly. A system that learns the preferences of users will be helpful in tailoring which kinds of messages are sent, when, and in what frequency.
The emergence of connected adherence devices like Proteus pills and Smrxt bottles, combined with the popularity of health devices like Fitbits and Qardio, can provide a more rounded view of patients. Aggregating and synthesizing that information, perhaps using services like Validic, would help Florence interact more appropriately.
Florence should keep patient data secure. Managing Personal Health Information (PHI) exposure is a real challenge and SMS is an insecure medium. While HIPAA doesn't preclude sending PHI via SMS, it's unclear if remediation tactics like end-user consent forms satisfy the Security Rule. 8
If created, Florence will use technology to help address significant causes of medication non-adherence. The broader challenges are complex, however, and work must be done on a variety of factors such as socio-economic, education, healthcare access, and treatment models. Technology is just one means of attack, not the only one.
If you’re interested in learning more about this problem, we encourage you to read the World Health Organization’s report Evidence for Action. It provides a digestible overview and deep analysis into the problem.