Laurie Tomlinson: From patient to data and back again—how anonymised patient records can improve prescribing guidance

We often hear about the importance of “bench to bedside” medicine, how basic science research is translated into novel treatments. But this paper demonstrates an equally important concept: how anonymised health records provide a rich data source to address clinical questions, which in turn improves patient care.

I work as a clinical academic, undertaking outpatient work in nephrology and in my research time, using electronic health records to examine drug effects—often as they relate to kidneys. Over recent years NHS England ran a programme called “Think Kidneys“, aiming to develop resources to improve the detection and treatment of patients with acute kidney injury (AKI). I was lucky enough to be part of the Intervention Workstream, which aimed to develop guidance for clinicians about ways to reduce harm to patients with AKI, particularly related to prescribed medication. Unfortunately, this led to much time spent discussing the inadequacies of the evidence base on which to base suggested advice.

There is often very limited research related to drug effects for people with kidney disease as they are frequently excluded from clinical trials due to concerns about impaired drug excretion or metabolism, and therefore increased risk of side effects. In addition, the evolving definitions of acute renal failure and subsequently acute kidney injury have meant that many older trials recorded limited or inconsistent data about sudden drops in kidney function. Therefore, much prescribing guidance is based on low quality or highly confounded observational studies, often in select patient groups.

The situation when we came to discuss possible guidance about trimethoprim prescribing was slightly different. Here, there were basic science studies of the biochemical effects of trimethoprim, and some high quality observational research using population cohorts. However, these related only to patients taking renin-angiotensin system blockers or spironolactone, and only to use of co-trimoxazole rather than trimethoprim alone (which we mainly use in the United Kingdom). With these uncertainties, the workstream wondered what advice should we give for the general population? Were the risks for trimethoprim the same as co-trimoxazole? How common was hyperkalaemia among people taking trimethoprim without other risk factors for high potassium levels? It struck me that this was something our research group could look at using UK primary care data, and we set about planning the study, completing the loop of clinical question to results suitable for informing guidance.

The availability of large scale anonymised healthcare records and growing skills in conducting such research have enabled increasingly rapid research into drug effects after issues of clinical concern have arisen. This is particularly true for the UK, which holds a number of excellent resources, such as Hospital Episodes Statistics and databases of anonymised primary care data including the Clinical Practice Research Datalink. However, public confidence in the use of such data may be falling, particularly following the saga. Increasing rates of individuals refusing consent for use of their anonymised data would greatly reduce the generalisability and validity of this type of research. Hopefully, examples of research where the careful use of these vitally important resources led directly to guidance to improve healthcare will maintain the trust of our patients.

Laurie Tomlinson is an associate professor at the London School of Hygiene and Tropical Medicine, and an honorary consultant nephrologist at Brighton and Sussex University Hospitals NHS Trust. Her research focuses on the epidemiology of kidney disease, particularly kidney-related adverse drug reactions.

Competing interests: None declared.