Predicting mental illness risk using health records

Serious mental illnesses like bipolar disorder, suicidal thoughts, and post-traumatic stress disorder (PTSD) can significantly shorten life expectancy and negatively affect people’s lives. Often, these conditions are not diagnosed or treated until much later, which can lead to worse outcomes.  

Since most people first talk to their GP about mental health concerns, it’s important to find ways to identify those at risk early in this setting. Electronic health records (EHRs) – which include information like GP visits and prescribed medications – might help predict who is at higher risk. However, these records often have missing information, like a patient’s ethnicity, which can make prediction more difficult. 

Project aims

This project has two main goals. We want to: 

• Find out whether information from EHRs can accurately predict who is at risk of developing serious mental illnesses 

• Test different methods for handling missing data to see which ones work best when building these prediction models. 

What we hope to achieve

We aim to create a risk prediction model that GPs could use to help identify people who may be at higher risk of serious mental health problems before a formal diagnosis is made. This could lead to earlier support or treatment.  

We also hope to better understand how different ways of dealing with missing data affect the accuracy and usefulness of these models.  

If our model works well, the next step would be to test it in other settings to see if it holds up in different groups of patients. Ultimately, this research could help improve early mental health care and outcomes for patients.