Dr Sperrin researches new statistical methodology to make inference with observational health data, collaborating closely with clinicians, epidemiologists, health informaticians, software engineers and statisticians. His research can be categorised in three areas:
1. Understanding the observation process. When data are observational it is crucial to understand why data are present (and why not present). For example, a blood pressure measurement in a medical record implies both that the patient has made contact with a healthcare professional, and the professional has deemed it appropriate to measure blood pressure. In other words, the presence of the measurement (or the absence) can be just as important as the measurement itself. This is often ignored in statistical analysis, which can lead to biased results.
2. Inferring trends and patterns from data. This can range from simple applied questions (is X associated with Y?) through to advanced statistical techniques to uncover hidden structure in a population. For example, a disease such as asthma may actually consist of a number of subdiseases (endotypes) that may have different outcomes and require different treatment.
3. Making predictions and decisions. Given what we know about a patient now, what do we think will happen to them in the future, and (therefore) what should happen next? For example, this involves developing prediction models for disease incidence and mortality, and complex simulation models to understand how disease may progress under different scenarios, both at an individual and population level.
A review of statistical updating methods for clinical prediction models
Statistical Methods in Medical Research (2018) 27(1)