Unjani Clinics is a Non-Profit Company, on a mission to empower black women by building a sustainable network of nurse-led primary care clinics in South Africa. The group is on a growth path, with a target of having more than 600 care settings by 2031.
Over the past 18 months we have set up a data analytics platform, a data governance framework, and dozens of data dashboards for Unjani Clinics. More recently, we have begun to train machine learning models on Unjani Clinics’ in-house data, and we are working on integrating these models into their privately owned electronic health record system. All of these activities serve the overarching goal of helping the organisation learn as much and as fast as possible from their own data, to optimise efficiency and maximise impact.
From the start of our collaboration in late 2023, I have noticed a striking contrast between Unjani Clinics’ bold can-do attitude towards the adoption and integration of health analytics, machine learning and other AI, and the apparent fear and hesitation that is widely spread among most other healthcare organisations.
A few weeks ago I sat down with their CEO, Lynda Toussaint , to check in on our various work streams, and I used the opportunity to ask what makes Unjani Clinics different. Specifically, I wanted to understand why they were not paralysed by fear or complexity. Here's what I learned:
Ownership
Not only does Unjani Clinics own the source system that is used to capture all of its electronic health records, they also have full ownership over the analytics platform, the data governance framework, the data dashboards, and every data pipeline and machine learning model that we have built. This gives them the assurance that they are in control. What data gets used, for what purpose, and who gets access to inputs, source code and model outputs. There is clarity and transparency, and changes can be made and implemented instantly.
Strategy
Back in 2023 when we first conceived our partnership, we didn’t go straight for the new and shiny objects of generative AI (ChatGPT was introduced in November 2022). Instead, we adopted a deliberate strategy of foundational priorities and incremental complexity. The goal of the very first work stream was to set up a scalable data platform with automated data pipelines from the source system, via curated fit-for-purpose data layers, to easy-to-use dashboards.
The focus was on building the foundation on which all of the future health analytics, machine learning and language models could be built, efficiently and with confidence. Hence, for the first year, we did not worry about the accuracy of predictive models or the validity, ethics and compliance of language models. The analytical workflows were simple yet crucially important. They presented the data back to the leadership of Unjani Clinics via self-service dashboards. In this way, they facilitated an iterative process of getting answers to simple questions, which in turn triggered more nuanced questions. Gradually, the insight and understanding into their own data deepened. Where gaps in data quality were identified, remedial actions were taken to improve the source systems and their adoption. Only once we had designed and implemented a modular data system architecture, applied best practices in DataOps, and gained trust and understanding of the raw data, did we allow ourselves to ideate about use cases for supervised and unsupervised machine learning models for forecasting and anomaly detection.
Trust
Much of the fear and apprehension that many organisations feel originates from a lack of trust. To adopt and integrate AI decisively and effectively, one needs to trust in the accuracy and validity of the source data, as well as in the team that builds and integrates the technology. The mutual trust between Unjani Clinics and Wimmy is explained by two key factors: alignment of our value systems and the depth and breadth of our interdisciplinary expertise. Both our organisations exist to do purpose-driven work at scale. Also, both organisations put a premium on quality and invest in continuous professional development of every member of the team. In addition, Wimmy's interdisciplinary team includes medical doctors, statisticians, machine learning engineers, software developers and UX/UI designers, and many of our team members are skilled in at least two of these domains. This means that we don’t need a lot of time to understand the challenge, conceive and build a solution, test it in a sandbox environment, and explain the results back to Unjani Clinics’ leadership, before moving the solution to production.
Intrinsic Innovative Identity
Lastly, it is in Unjani Clinics’ very nature to be pioneers and disruptors. Their business model – improving access to high quality, affordable healthcare through nurse ownership of clinics – did not exist, and they were told it couldn’t be done. Yet they ignored the critics and forged ahead anyway, finding a solution for the patient. Another defining identity trait is their leanness. For the Non-Profit Company to succeed in its very bold growth targets, it needs to be a lean, ‘digital-first’ organisation. And that means it favours automated workflows, data-driven decision-support tools and AI over human resources where that is possible and safe.
In conclusion, Unjani Clinics' boldness in adopting machine learning and other AI to guide patients, clinics and the organisation's overall strategic direction, stems from their imperative to take ownership of data and technology, their strategy of gradually building solutions within a solid data platform, trust in the data and the humans behind the AI, and their intrinsic identity of innovators and disruptors.
