Learning Health Systems: An Agile Team Science Approach
Can people develop feelings towards a robot therapist that are equivalent to those developed towards a human therapist? Should we worry more about sharing our mental
health data compared with our physical health data? Will digital technologies replace human clinicians? Should we embed Artificial Intelligence (AI) in clinical decision-making tools or would AI make bad, biased decisions about our healthcare? Is the importance of human contact lost from clinical encounters when new digital technologies are introduced (and does it matter if it is)? Such questions are often lumped together in a general panic about introducing digital innovation in healthcare.
These and other questions are explored in Reform’s recent report on data-driven mental healthcare. The report talks us through the importance of using data to improve healthcare and helps us understand how #datasaveslives. It also shows that successful transformation relies on building trust with patients and practitioners. This can only be achieved through rigorous research and a team science approach. We need to work with people to explore the risks, benefits and challenges of digital health technologies and data sharing.
What are the conditions that satisfy a social license for data sharing so that people feel safe to share their data for healthcare purposes? The Connected Health Cities programme ran Citizens’ Juries to explore how healthcare data could be used in a way that people found acceptable. Building trust and transparency around sharing health data is key. We need to use novel ways to explore with people how digital technologies can be usefully embedded in healthcare contexts. A large scale hospital simulation game, like enTRUSTed can be used to explore how digital tools and data analytics can be integrated into complex systems. This allows to brainstorm the digital possibilities across many areas, from process optimisation to patient-facing applications.
One of the best ways to embed digital technologies into routine care is to do so in ongoing cycles of continuous improvement. This involves creating a ‘learning health system’ that can adapt as it is embedded into different healthcare contexts. This parallels the ‘agile’ approach adopted in software development, where the software develops through ongoing collaboration and iteration. For example, in cystic fibrosis (CF) care we have developed a learning health system to support adherence to treatment in people with CF; we adapt and expand the platform as we learn from embedding it in new healthcare contexts and listen to feedback.
Sometimes we need to just keep iterating. There are complex ethical and practical dilemmas in the digital delivery of healthcare. There will be challenges and compromises. We can use AI in novel digital platforms, while also acknowledging that poorly-implemented AI can embed biases in unhelpful ways. AI can identify some of the biases of our existing human-driven systems. Maybe we can love our robot therapists but still value the human connection?
We can use AI in novel digital platforms, while also acknowledging that poorly-implemented AI can embed biases in unhelpful ways