Traditionally, public health has relied on averages, the so-called “average patient,” to design most programs. In South Africa’s HIV treatment programs, the average male is slightly more risky than the average adult female. But it’s very clear that all adult males are not the same and to assume so tends to lead to average programs that have average results. As the epidemic has matured and we now have large scale systems, we need more advanced approaches to targeting, tailoring and personalising services to different segments of the population.
What Palindrome's projects have shown is that we offer private sector data analytics, specifically AI-based segmentation and predictive risk scoring. So that helps our clients find more nuanced groups in their treatment programs that appear different and are having different treatment experiences to find out who is being underserved and who has been over-served. There are many potential efficiency gains that are waiting to be capitalised on.
For example, we showed some of our partners in South Africa that they have this large population of adult males and there's a portion of them that are actually the lowest risk group–the portion of them that go to the same clinic, at the same time, never late, and always take their medicines. They're actually one of the lowest risk groups in their entire portfolio. On the other hand, we helped them find a group of young professionals, male or female, that are employed, they look good on paper, they have disclosed their status to their family, and they've been on treatment for a while. However, if you give them appointments around payday weekend, they’re some of the highest risk groups in the entire portfolio. So maybe they need to tailor their services more towards professional groups by opening clinics at different times, keeping their appointments away from payday weekend, or other solutions like that. We believe that small changes in the program could lead to outsized impacts.
Palindrome's offerings have included two things. First of all, we provide data analytics to try and uncover what hidden segments are there and who we can serve better. Secondly, we work with user centred design to try and see how partners can adapt their programs, their tooling, and the services they offer to target the people who are in need and see if we can increase their impact per dollar.
Last year, we had two or three publications in journals on the data side or our business to show that these algorithms on existing data (not perfect data, but existing data) in Nigeria and South Africa can predict different outcomes in patients and effectively identify different groups. We're starting to feel that there is a critical mass of evidence and that it is possible to show that we can find these groups reliably. We've repeated those same results in different parts of South Africa and in Nigeria as well. So we have established credibility on the data analytics side of our business.
We are also implementing two studies in South Africa and Nigeria to see if the potential of the research evidence is actually feasible in practice, because that is quite challenging, especially in low-resource settings. Change management in any organisation is tough, so we’re trying to assess the feasibility of changing services to be proactive rather than reactive. It's nice that it works on paper in the data, but does it work in the real world where humans are experiencing lots of different pressures and incentives?
One challenge is operationalisation. There's great potential for it, but it has to be helpful to the health system users, district managers, health workers at points of care, as well as patients. They have to feel good about it and see those gains realised.
I think the other one is that there are a lot of misconceptions about AI and data privacy, which become barriers to trying new things and it leaves a lot of impact left on the table. We don't yet know how to advocate and communicate at that level. For policy openness, we often rely on bigger partners than ourselves to bring the more difficult topics up with large stakeholders, funders, and governments, because we're quite a small firm, and I think we probably haven't created the right communication materials for our partners yet. That's probably our next challenge to figure out. How do we communicate this to the market such that people have confidence in it and feel comfortable with what's possible and what's worth trying?
There is also the challenge of handling data and the fact that the public sector is less experienced and confident with it than the private sector is. Mostly in public health, we have missed the opportunity to engage with data to make things better.
There is also a misconception about AI replacing health workers, which is pretty ironic, for two reasons. Firstly, is the AI really so good that it can replace human touch? The second one is, in this sector, we've been talking about under-resourced African health workers for decades. This should be coming in to boost the health workers that we have, to optimise their time, to maximise their personal touch.
For people getting into an innovation, there really is no substitute to testing something with users. There will always be reasons to delay, but the refrain, “If you're not showing something you're embarrassed by, you probably aren't showing it early enough” is absolutely true. There is no substitute for spending time with users.
On the technology side, as a founder, it’s essential to know the details yourself and get your hands dirty. When your company is bigger you'll have better people than you, but in the beginning, there's no substitute for learning the details yourself. That will stand you in good stead, especially as the company evolves.
Villgro has impacted Palindrome, ironically, in just understanding who we are and what we're trying to offer to the world. I realise that's kind of basic, but running a small business, you get tied up in trying to keep the lights on and it becomes some strategic objective. Maybe you're neglecting the clarity of the mission as things evolve. The Villgro team has been incredibly patient with allowing us to fumble through the next version of what we're trying to envisage for our company, telling our story with sufficient clarity for larger markets. Right now, we have something really specific and technical, where if the person is already an expert in that area, they love it. But we've got broader ambitions and I think the Villgro team have been really patient in nurturing us to think bigger, providing customised support that directly fits where we are. I really appreciate it.