Innovation in healthcare

When the tolerance for risk is, rightly, very low – how might we update a healthcare system that was not designed for the needs of patients with multiple, chronic conditions? We worked with UCL Partners to address this challenge, building a sandbox for decision makers to break-down the enormity of this task into tangible scenarios. 

UCL Partners, a health innovation partnership, asked us to model a radical change to healthcare service. This project stretched our definition of who an urban planner is, in the best possible way. Diagonal’s purpose is to improve the quality of life of those living in cities. While we typically work with clients who are responsible for infrastructure design and place planning, working with a health partner like UCL Partners allowed us to fold in population health into our model of the world. 

UCL Partners came to us with the question: How might we provide better care to a high-needs patient group, through a new poly-clinic care model? And will the demand be serviceable by the existing workforce and resources? ‘High-needs’ was defined by several parameters, focusing on long-term, chronic, comorbidities, including the following six comorbidities: diabetes, hypertension, COPD, depression, dementia, and cancer.

What we did

We built a system to explore the impact of these new poly-clinics on healthcare provision and workforce demand. We explored this problem at a neighbourhood level and across the whole population. To do this, we had to understand who was likely to have one or more chronic comorbidity and where they were likely to live. With this information, we could start to model new care scenarios. For example, what would happen if patients with comorbidities had a recommended number of routine appointments with specialist care providers, rather than many reactive appointments with GPs and Consultants? How many of these new appointments would be needed? Should the frequency of appointments be recommended by the number of comorbidities a person has? Or by their age? What communities could have the highest demand for these services? Where are there gaps in the existing health infrastructure? 

To explore these details, we built a ‘synthetic population’. We created a model to relate open data about demographics (age, sex, residential neighbourhood) to health information (prevalence of comorbidities). The result is a population that, in aggregate, reflects the real population, and allows us to zoom into neighbourhoods to understand the distribution of health conditions.

A privacy-first approach

Detailed population data, and population health data, is very sensitive information. We could have used real data. This would be obtained from patient records, GP surgeries and hospital records, though is a complex process. In fact, UCL Partners began the process of exploring how to collate and share some of this information, to help with benchmarking our analysis. But we did not follow this approach. 

Instead, we opted for an approach that would provide our client with a general understanding of its patients’ needs, without requiring sensitive, real patient data. We used open data from the UK Census, Health Survey for England (2019) and other sources (described in our GitHub repo.) Much of this data is provided at the national scale, so we had to link features together to get a local picture of population health. We did this by building a synthetic population. 

How did this approach help our collaborators?

A privacy-first approach provided two important benefits to UCL Partners. First, it allowed us to progress this strategic work, quickly. Gaining approval and managing access to sensitive patient information is timely and costly. By using synthetic data, we could focus on framing the questions and understanding the impacts of a new care model, without the up-front costs of accessing real data. 

Second, this approach allowed UCL Partners and their stakeholders to truly innovate. We could keep our analysis questions open, adapting our work as we understood more about the underlying trends. We did not need to set out a very restricted set of questions at the outset - with a requirement of strict adherence to gain access to patient data. For example, when we started this work, we didn’t know how the demand for a poly-clinic might change depending on the inclusion of more conditions. We started with three comorbidities, but expanded the exploration to include six comorbidities. This ability to change and expand our scope mid-project provided UCL Partners with greater insight into the complexity of multimorbidity care. If we were using real patient data, we would have had to re-start the approvals process to gain data authorisation. 

UCL Partners was able to take an innovative approach to envisioning new healthcare models, without many up-front privacy risks.

Our synthetic population ‘fit’ the national statistics of condition prevalence, while explorable at local census areas (MSOAs boundaries, which contain between 2,000 and 6,000 households). This meant we could estimate demand for chronic, comorbidity care by neighbourhood, and by GP. We modelled patient demand on the current system and compared this with future scenarios. We built an interactive tool for decision makers to explore the population model. They can sort, filter, and visualise the synthetic population to understand how condition prevalence changes by age, index of multiple deprivation, and neighbourhood. 

The synthetic population was also created more quickly than would have been possible if relying on the real data to be collated. This allows UCL Partners to work at pace in a fast changing complex healthcare landscape where data-informed decision making is imperative for improving services.

As UCL Partners develops its vision, and moves to more detailed implement plans, it is now equipped with a more specific understanding of what data it might need and why.  

What’s next

UCL Partners is using our work to operationalise this new health service model and design trials. Our tooling is supporting what is seen as a highly innovative program within the NHS and healthcare partner ecosystem. 

Not only did this project provide an important dimension to our understanding of neighbourhood service provision, it has inspired us to dive deeper into synthetic population data as a tool for place planning.

How does your work impact different communities? We can help you understand the consequences of your vision, without jeopardising personal or community privacy.

Get in touch today: claire@diagonal.works