Using predictive algorithms to recognise risk factors and prompt testing

My Topol fellowship problem / project:

The problem I would like to explore is missed diagnosis of HIV and Hepatitis B and C.

People with these viruses may attend healthcare many times without being diagnosed because the risk factors for those conditions are subtle and not recognised by the clinicians looking after them. These represent missed opportunities for diagnosis of HIV and hepatitis B and C in hospitals and in the community. The diagnoses are a low frequency but high impact event, the low frequency means many clinicians are unaware of the many triggers to test and the high impact means every diagnosis makes a significant difference to patients, their families and associates through treatment and prevention of complications and onward transmission. The triggers to diagnose are present in the patient records but current systems do not identify and quantify these risks and highlight them to clinicians.

It is estimated that over 100,000 people in the UK with hepatitis B, C and HIV are undiagnosed and the majority will see a clinician 4 to 5 times before a diagnosis is made. However, the problem lies with healthcare workers who do not recognise the risk of infection in their patients. All clinicians can experience the problems of under investigation when there is no prompt to investigate. The consequence of which is late diagnosis for the patient.

The consequences of a missed diagnosis occur on several different levels. For the patient, this can manifest as a cumulative increased risk of morbidity and increases the risk of early mortality 8-fold.

Diagnosis is then made when presenting later on extremely unwell and after avoidable potential onward transmission. For family members and associates, this can result in illness and distress that was potentially avoidable. For clinicians missing a diagnosis and finding out in hindsight that you could have avoided subsequent morbidity can be a distressing experience  and can then lead to behaviour changes that over emphasise certain risk factors leading to defensive medicine. At a system level it can lead to inefficiencies of resources within the healthcare system with over investigation and changes to improve system security to avoid missed diagnoses in the future, however, without the appropriate knowledge and skills this can be highly inefficient.

To solve this problem my project will develop a predictive risk algorithm that prompts testing of patients at increased risk based on their individual clinical and epidemiological history using nationally recognised standards.

My medical career has taken me from Scotland to London, Liverpool to Latin America and Zambia to the North East with many a stop-over in between. Having completed my training last year in 2021 I launched myself into my role as an infectious diseases consultant at the tip of the North York Moors at South Tees Hospitals where I have the freedom and support to pursue my various research interests.

A benefit of a varied source of training is the exposure to different ways of working and different challenges for staff and patients.

Even across the UK expectations and facilities vary hugely. This variation illustrated to me how certain communities and conditions are susceptible to differing management and diagnosis.

My research background has focused on these groups with limited access to critical services in the UK and has culminated in me joining a research group looking into late diagnosis of HIV, which is the foundation for my Topol project.