Gregg Barrett is the CEO of Cirrus, Africa’s AI initiative. Gregg is a technology executive who builds and scales innovative operations at large private and public sector organisations, non-profits, and start-ups globally. He co-founded an organisation that became the market leader in contract management systems in Africa and was selected to provide the E-commerce platform for the Southern African Development Community (SADC) common market. He also led the International Association for Contract and Commercial Management (IACCM) in Southern Africa to develop and advance contract and commercial management in the region. He is a seasoned executive with extensive and diverse experience in strategy, building and managing relationships, deal-making, communication, developing high-performance teams, organisational leadership, and problem-solving across a range of areas. Over the last decade, Gregg has led work in data science, machine learning, corporate research, and corporate venture capital.
Healthcare model
If the world of healthcare was to be reimagined and Artificial Intelligence introduced, what would be necessary and what would such a world look like? The key to an AI-enabled healthcare world is data. Healthcare data is unfortunately siloed across different institutions, devices, and systems, with no unified, high-dimensional health data available for individuals, samples, and populations. “High dimensional” data is to refer to all the relevant input variables (i.e. radiology, musculoskeletal imaging, digital pathology, environmental, genetic, clinical trials, wearables) needed for the implementation of a healthcare model to provide predictive, preventative, and personalised care. The existence of such a model would help to reduce demand, drive down costs, increase access, and radically improve outcomes.
Book of Life
Since large amounts of data generated at various points in the healthcare process are siloed it cannot readily be exchanged across time and institutions to create a single unified dataset for an individual. Data fragmentation is a long-standing challenge in healthcare impeding improvements in efficiency, delivery, and research. To move beyond data fragmentation is to move towards the creation of a “book of life” for an individual. This is to create an all-encompassing single unified dataset of all health information pertaining to an individual that spans the individual’s entire life history.
Electronic Health Records
The implementation of Electronic Health Records (EHR) systems has not solved the problem as the dimensionality of health data is too vast. Aside from EHR systems not capturing the full dimension of an individual’s health information, EHR systems have created their own set of problems for medical professionals. According to a study by the Mayo Clinic: “Physicians now spend one to two hours on EHRs and desk work for every hour spent in direct face-to-face contact with patients, as well as an additional one to two hours of personal time on EHR-related activities outside of office hours.” Further, the study reports: “The usability of current EHR systems received a grade of F by physician users when evaluated using a standardized metric of technology usability. A strong dose-response relationship between EHR usability and the odds of burnout was observed.” In response, The American Medical Association stated: “It is a national imperative to overhaul the design and use of EHRs and reframe the technology to focus primarily on its most critical function — helping physicians care for their patients.”
Data management platform
In the world of AI, data management platforms play a critical role in enabling the data foundation that supports the training of AI models and the deployment of those models (inference). At a high level, a data management platform provides the capability needed to store, manage, share, find, and use data for AI. A database is not a data management platform. Rather a data management platform is required for a single point of data ingestion. As the name implies this is to ingest data in all its forms (structured, unstructured, and semi-structured), and with all key data transfer approaches (batch, micro-batch, and streaming). To do this the data management platform provides a highly configurable set of data integration tools that extend far beyond typical extract-transform-load (ETL) or extract-load-transform (ELT) solutions. Within a highly federated healthcare system with heterogenous databases and schemas, diverse sources, forms, and transfer approaches, a data management platform serves as a single interface supporting the many-to-one and one-to-one relationship that would constitute an individual’s personal book of life.
As much as 30% of the entire world’s stored data is generated in the healthcare industry. A single patient typically generates close to 80 megabytes each year of imaging and electronic medical record data. (Marco D. Huesch, Timothy J. Mosher, 2017)
The data management platform provides the access control framework to restrict access to sensitive information at a granular level, ensuring that healthcare professionals see only the specific data points that are necessary to complete their work. The deployment of a data management platform at scale effectively serves as the data operating system upon which the deployment of healthcare-related AI services is then provided.
Economics
With the technology now in existence, efforts are underway in South Africa as a once-in-a-generation attempt to revolutionise healthcare to serve as a model for the rest of the world. The undertaking is to provide a state-of-the-art data management platform for healthcare as a public good. The intention is for the platform to be provided freely to all stakeholders to ensure rapid and widespread adoption under a public benefit organisation (PBO). While the implementation and maintenance of the platform constitute a substantial investment, the opportunity is significant. In 2019, Ernst & Young estimated that unlocking NHS health data could be worth up to £10 billion per annum through operational efficiencies, improved patient outcomes, and wider economic benefits. Using the EY numbers (conservatively) as a guide, a patient record valued at a quarter of that used in the report, and assuming a penetration of 50% of the patient records in South Africa[i] results in a revenue generation figure of more than 600 million pounds annually. An amount sufficient to cover the implementation and maintenance of the platform ensuring its service can be provided at no charge. The economic and societal impact of the platform will be far more substantial in creating new opportunities for South Africa in collaborating with life sciences firms to access health data, where appropriate, which would generate funding for greater investment in public research and development. Importantly, a governance framework would need to be instituted to ensure that additional proceeds be remitted to stakeholders for the provision of healthcare services to those in need – ensuring the PBO is always patient-centric.
With the foundation of a data management platform in place, the application of AI to advance research and healthcare services to individuals can be implemented. The economics of the model ensure that AI-enabled healthcare services to individuals can be provided at no cost. As with the data management platform, the technology exists today to enable such services. Frameworks like the Generally Nuanced Deep Learning Framework (GaNDLF) have been developed as a general-purpose framework for AI in healthcare to define end-to-end clinical workflows for various healthcare use cases such as tumor segmentation and molecular classification. Frameworks like GaNDLF follow zero/low code principles and hence alleviate programming requirements for medical researchers to easily take advantage of the latest AI developments and analyse healthcare data. Further, the framework handles multiple types of data and AI workloads, and it operates in a privacy-preserving manner, without sharing any patient data.
Summary
For too long there have been discussions about democratizing healthcare and AI, with initiatives and efforts spanning the full gamut of the imagination. The key, the aggregation of data, must now be the focal point not only in the discourse but in the implementation. Individuals, medical professionals, and researchers must be empowered with the data needed to implement the healthcare model providing predictive, preventative, personalised care, and advanced research. The time is now to discard the approach of relying heavily on anecdotal accounts to implement a holistic view of the patient’s wellness by providing a detailed and data-oriented book of life. The more connected medical professionals are with their patients, the more capable they will be in addressing underlying health issues rather than merely treating symptoms. In low-resource environments pervasive in the developing world trained healthcare professionals and physical clinics are a limited and expensive resource. Through the application of AI location can be removed from the equation, shifting care to the point of the patient, increasing access, and lowering the cost of delivery. The negligible incremental cost to provide AI services per person makes the approach infinitely scalable and accessible.
Moving to electronic health records has not solved the problem. With data sources proliferating, only the aggregation of data onto a patient-centric privacy-preserving platform will enable everyone to their book of life. With the technology here and now, the delta between the world of today and the world that can be, is not one of technology, but of leadership. Persistent and knowledgeable leaders with the courage to think differently will have direct implications on our lives and the lives of loved ones. Leaders who will pilot a new course to institute the technology foundations to realise the world of tomorrow.