Here we talk with Dr Konstantin Georgiev, shortly after completing his PhD research in January 2026. His thesis can be found in the Edinburgh Research Archive: Konstantin’s thesis.
1. What was your PhD research about?
My PhD was focused on a central question: Can we use the routine data that NHS systems collect every day to better predict which older patients are most at risk of worsening health/quality of life and what level of care they will need?
The NHS sees enormous numbers of older people in emergency departments, and clinicians have very little time to make complex decisions. My research aimed to show how data-driven methods, such as machine learning and process mining, could help analyse health records to highlight:
a) which patients arriving at A&E might be in need of specialist geriatric care,
b) which patients might need extensive nursing and rehabilitation after presenting with COVID-19, or even
c) which individuals could develop dementia years before symptoms appear.
2. Why were electronic health records necessary for your research?
The insights provided by routine data may not be clear to a clinician in a single patient consultation, but they become obvious when you can look at patterns across hundreds of thousands of records.
Furthermore, to answer questions about risks in older populations, you need data that reflects how people move through the healthcare system. Electronic health records are well-placed to provide this: they capture primary care contacts, secondary care episodes, prescribing histories, diagnostic codes, and care-provider interactions, all linked at the patient-level. For something like dementia prediction, the most clinically useful time is years before formal diagnosis, when subtle signals (GP contacts, medication changes, signs of frailty) are already present in the health record.
3. Why did you choose to work with DataLoch?
DataLoch was simply the right fit. I needed secure, reliable access to health records across multiple hospital sites to develop my algorithms. DataLoch provides extracts of linked primary and secondary care data from NHS Lothian, including long-term datasets that show how older people move between community and hospital settings.
Also, before my PhD, I worked at RedStar AI, and during this time I worked on collaborations with DataLoch through projects related to hip fractures and management of brain health. (It was in this industry role, where I first acquired the skills to work with large health data.)
4. How was your experience of working with DataLoch?
Overall, it was very positive, particularly once I had become familiar with the Trusted Research Environment through which I could access my research data. The team itself was supportive throughout. Having access to my supervisor’s (Dr Atul Anand) knowledge of DataLoch was enormously helpful for navigating the governance steps, data access applications, and understanding what was available in the linked datasets.
5. If you had to start over, would you do anything differently?
I would invest more time at the very beginning in understanding the data before diving into analysis!
In practice, I spent a significant portion of the early PhD period discovering data quality issues, coding inconsistencies, and structural quirks of the health service through trial and error. That said, I think this is almost inevitable when working with routine data. These data are not collected for research purposes and its caveats only become apparent as you develop and test a pilot model and then see its systematic biases, figuring out how to address them in the next version.
6. After your PhD – what's next?
I am happy to say that I am continuing my collaboration with DataLoch, now as a Postdoctoral Fellow in Health Data Science at the University of Edinburgh, where I am harnessing data in a new research direction.
My project, funded by the British Heart Foundation, focuses on developing AI algorithms to predict acute cardiac conditions in the Emergency Department, integrating electronic health records, electrocardiogram (ECG) data, and chest X-ray imaging.
This research builds on what I learned during my PhD about working with routine NHS Lothian data. However, I am extending this by also using imaging data and other medical measurements to improve the risk assessment of patients in cardiovascular care. My research could become the foundation for new decision-support tools that would help clinicians understand their patients’ needs more clearly, thereby ensuring these patients get the right treatment at the right time.
7. Is there anything else you would like to say?
I think what struck me most across nearly six years of working with DataLoch was how much useful information is already sitting in routine healthcare data waiting to be analysed with the right methods. We often talk about needing new data sources or new technology, but a significant amount of what clinicians and services need is achievable right now using data the NHS is already collecting. Platforms like DataLoch are a crucial part of that infrastructure, and I hope to see more researchers who can harness this data in exciting and innovative ways that truly make an impact on our population's health.
Finally, there is one additional publication from my PhD which I didn’t manage to touch on before about hospital activity and people with multiple long-term conditions!
Case Studies
