Spatio-temporal trends in antidepressant prescribing in England

I was the lead supervisor of an MSc in Health Data Analytics and Machine Learning student summer dissertation in 2022/23 academic year with Prof. Marta Blangiardo as a co-supervisor. The dissertation was titled Spatio-temporal trends in antidepressant prescribing in England. The student received a distinction for their dissertation. The dissertation proposal is below.

Dissertation Proposal

Background

In English primary care, overall antidepressant prescriptions have more than tripled between 1998 and 2018 from 377 items per 1000 to 1266 per 1000. This rapid growth is well known, as is the substantial variation in prescribing behavior between practices. Whilst the overall growth can be attributed to factors such as population growth, use of antidepressants for non-depressive indications, and longer-term prescribing, not much is known of the cause of variation between practices.

Many mental ill-health disorders, including depression, have a strong association with social inequalities. An important metric to capture social inequalities is the Index Multiple Deprivation (IMD), which quantifies relative deprivation of small areas. Given the location an individual lives reflects their socioeconomic status, which is associated with mental ill-health disorders, the variation in the number of antidepressants prescribed for a given practice may be due to its location.

Aims

The goal of this project is to use practice-level summaries of prescribing activity that is available on NHS digital to examine spatial and temporal trends in the overall antidepressant prescriptions in English primary care as well as the antidepressant-specific trends. After adjusting for spatially varying coefficients, such as England’s IMD, the student will ascertain if the variation between different practices is due to the location of the practice itself.

In addition to the overall number of antidepressants dispensed per practice, NHS digital has information on the type of antidepressants being dispensed per practice. The student will also explore the spatial and temporal trends in the different types of antidepressants being dispensed in England.

Finally, the student will use their results to predict future trends in both the overall and individual types of antidepressants being dispensed both nationally and locally.

Proposed methods

The student will use Bayesian spatio-temporal models (as introduced in the Bayesian modelling for spatial and spatio-temporal data module) to examine the spatial and temporal trends in the overall and specific antidepressant prescriptions in English primary care.

Practice-level data on the overall number and specific antidepressants being dispensed will come from NHS digital. This will be combined with the additional data sets, such as England’s IMD from gov.uk, to account for spatially varying coefficients.

A cross validation procedure will be used to evaluate the model’s predictive capabilities. The results will be displayed in a variety of different temporal and spatial scales.