The Psychosis Metabolic Risk Calculator (PsyMetRiC) - Validation, Recalibration and Model Revision Study

People with long term mental illness die earlier than people from the general population. This is mostly because of physical health problems like heart disease and diabetes. Preventing these physical health problems is an important way to improve the lives of people with long term mental illness.

A key step toward preventing these physical illnesses is by knowing who is at greatest chance of getting them. Steps could then be taken to adapt treatment early so that they don’t develop in the first place.

Using patient data from three parts of the UK, we developed a tool called the Psychosis Metabolic Risk Calculator (PsyMetRiC). PsyMetRiC was designed for young people with mental illness. It calculates the chance of developing physical illness up to six years later. To get the tool to the stage where it can be used by health care professionals, it needs testing in a large sample. Then, any necessary changes can be made to make sure it is accurate.

We aim to use the QResearch and CPRD databases to achieve this aim. QResearch and CPRD include up to 56 million UK patients and have been used to develop other tools that are used by healthcare professionals. We will test how accurate PsyMetRiC is these large databases. Then, we will make changes to the tool to improve accuracy and usefulness.

In future, we will test how well PsyMetRiC works in a real clinical trial. In this future trial, PsyMetRiC will be used to decide which physical health treatments are offered to young people with mental illness. We will look to see how good PsyMetRiC is at helping to improve the physical health of young people with mental illness. If it works, we hope it will be taken up and used by healthcare professionals in the NHS.

We aim to perform a cohort study within the QResearch and CPRD databases, including all primary care general practices (GP practices) available. Data from QResearch will be used as the model development sample. Non-overlapping data from CPRD will be used as the validation sample. We will limit the risk of overlapping data by only including practices in the CPRD dataset that have never used EMIS software.

Across both datasets, our study population will include people who were between 16-35 years old when they first received a recorded diagnosis of a psychotic disorder (ICD-10 code F2*), either in general practice data or hospital episode statistics, with the diagnosis recorded between Jan 1 2005 and December 31 2015, and who do not have complete missing data for the predictor and outcome variables (see below). Included participants will be followed up for a maximum of 15 years.

We will conduct three stages of primary analysis, described below. For each analysis, we will consider multiple imputation using chained equations to address the impact of missing data, depending on the structure and amount of missing data. If done, estimates will be combined using Rubin’s rules.

After completing all three stages of primary analysis, we will produce an online data-visualization tool as a more detailed and comprehensive update to http://psymetric.shinyapps.io/psymetric, which was developed alongside the original PsyMetRiC study.

Prof Julia Hippisley-Cox - University of Oxford

Professor Simon Griffin - University of Cambridge

Dr Emanuele Osimo - University of Cambridge

Prof Peter Jones - University of Cambridge

Prof Golam Khandaker - University of Bristol

Prof Rachel Upthegrove - University of Birmingham

Dr Jan Stochl - University of Cambridge

Dr Graham Murray - University of Cambridge