Consistent Patient Signal for Improving HIV/AIDS Treatment Outcomes

Consistent Patient Signal for Improving HIV/AIDS Treatment Outcomes

This article was originally published on the ICTworks site.

Medication adherence still a major issue in the world of free ARTs and services

Over 50% of HIV-positive adolescents in Western Kenya fail to adhere to antiretroviral therapy (ART), despite the availability of free services and medications.[1] Non-adherence puts these patients at risk of life-threatening complications and contributes to the transmission of HIV. Common adherence blockers are stigma, lack of family/peer support, lack of resources to travel to the clinic, enrolling in boarding school, and side effects. Adolescents worldwide typically achieve medication adherence rates that are lower than the rest of the population.

While common reasons for non-adherence are understood, individual reasons are rarely identified by the health system, much less communicated to clinicians. This makes it hard for the health system to respond. Clinics are usually not aware of forming non-adherence habits, since they have no information about the patients between clinic visits which normally have a frequency of 1–3 months.[2] While clinics increasingly employ mechanisms for community outreach and follow-up, the communication is initiated by health system employees, which brings large inefficiencies.

The DREAMS Innovation Challenge, funded by the U.S. President’s Emergency Plan for AIDS Relief (PEPFAR), aims to reduce HIV infections among adolescent girls and young women across 10 sub-Saharan African countries. Premise Data, a San Francisco-based startup, has been working with DREAMS in Kenya to address the challenges of consistent patient data signal and reasons for non-adherence among adolescents. This post presents a summary of strategies involved.

Filling information gap and patient engagement

Premise patient-centric Patient 360 Signal addresses the information gap between clinic visits with persistent data stream directly from users via an Android mobile device (the most commonly used phone at the time of writing this article is a $45 TECNO S device). The platform is configured to run well in low-connectivity environments and accommodates cases of non-persistent connectivity. Premise encourages patients to use the app daily to report on the key information points related to their adherence journey. The daily interaction with the app (it takes about one minute) centers around learning whether the patient took medication (and if not, why). The more rare interactions (such as weekly or monthly tasks) address more longitudinal issues such as stigma, clinic visits and medication refills.

Premise then uses data science to learn from frequency, quality, and content of patient reporting and adapts the level of support to the individual’s needs. Key metrics of Patient 360 Signal in Kenya are app engagement, self-reported adherence and clinic retention. In the long run, self-reported adherence data can be correlated with patient viral loads, the standard high-reliability metric of medication adherence.

Several strategies are employed to ensure regular engagement. First is a monetary rewards system, which is designed to compensate users for daily engagement — mainly for the cost of mobile internet bundles and electricity (in many low-electricity areas, phone charging is transactional). The rewards are typically small sums and only apply to patients who engage every day without interruptions (minimum is 7 days streak for a reward). In Kenya, a monthly Safaricom 1GB data bundle costs 500 KES (about $5) and Premise users can earn that if they engage every day of the month. This ensures data collection includes economically weak patients, who are disproportionally affected by the pandemics.

The Premise monetary rewards system has some similarities with conditional cash transfers (CCT), but there are some key distinctions. Most importantly, users are rewarded for engagement, and not for adherence (even if they consistently report being non-adherent). The value lies in the granularity of data, which is then used to ultimately improve health outcomes and to target care better.

Second set of features designed to engage users is individualized dashboard; it shows each patient their progress in adherence, clinic retention, and other metrics. After longer engagement, the platform has the capability to communicate personalized progress compared with peers (social benchmarking). Users are also encouraged to challenge themselves with “adherence streaks” in order to improve patient’s health outcomes.

Closing the loop — data analytics and the health system

Most resources should be invested within the first year after initiation of ARV treatment, as this is when adherence habits are shaped. Premise platform gives patients a sense of ownership and responsibility for their adherence, all within a supportive and encouraging app environment. The collected data feeds directly into patient in-app features and communications, making the experience extremely personalized and relevant. The data generated by patients also give clinics a deeper understanding of individual and aggregate patient behavior which is acted upon.

The set of in-app treatment-related advice, called “Digital Coach”, advises patients on their treatment journey. While it is not designed to substitute for any medical advice provided by the clinic, it provides things such as clinic visit reminders, general advice on side-effects and other information relevant to the specific user.

With granular patient-level information, the HIV clinics significantly strengthen their mechanisms to respond to emerging trends of non-adherence. In the past, self-reported information on patient adherence has been collected by the clinics and acted-upon by dedicated field workers. However, it has been a high-effort with low yields, typically based on in-person visits and phone calls. Premise clinic dashboards surface risk cases in real time and are used by dedicated clinic staff to respond. Creating risk profiles based on past behaviors of patients allows to predict likelihood of non-adherence patterns in patients ahead of time, further allocating resources to where it matters most.

Matej Novak, Head of Programs with support from Vikram Chatterji



This article was funded by a grant from the United States Department of State as part of the DREAMS Innovation Challenge, managed by JSI Research & Training Institute, Inc. (JSI). The opinions, findings, and conclusions stated herein are those of the author and do not necessarily reflect those of the United States Department of State or JSI.