Association between Mental Disorders and Subsequent Medical Conditions

A study by Natalie C. Momen et. al.

Individuals with mental disorders (MDs) are at higher risk of subsequently developing a range of GMCs. Our study produced clinically relevant estimates of risk (e.g., absolute risk) for over 400 different MD-GMC pairs, the majority of which observed a statistically significant risk increase. Moreover, the magnitude of risk varies greatly across MDs, GMCs, sex, time since diagnosis, and more. To aid in the understanding of how these various factors influence risk, we've carefully designed a suite of interactive visualizations for exploring these risk estimates, and also a handful of video demonstrations of drawing insight from them.

Interactive demos

Relative risk

The first visualization shows hazard ratios (HRs), a measure of relative risk, where HR > 1 implies those diagnosed with the given MD are at a higher risk than those without that diagnosis. Each dot represents a MD-GMC pair, and displays the relevant HR on hover. This video shows a few extreme HRs. For example, given substance use, liver disease is 12.72 times more likely). Liver disease is a specific GMC (one of 31) which fits into the more general Gastrointestinal GMC category (one of 9).

Highlighting and conditioning

Clicking on a dot highlights the relevant MD and adds 95% confidence intervals, making it easier to assess risk (and uncertainty) across MDs and GMCs. For example, Substance Use ranks higher than most other MDs in many GMC categories. In some cases, these estimates can change dramatically conditional on the sex of the patient.

Multiple comparisons

Compare multiple MDs at a time via button click and hover. This feature makes it easy to compare overall risk across numerous MDs; for example, a Mood, Neurotic, and Personality MD diagnosis have very similar comorbidity patterns. To remove highlighting, click the 'clear selection' button and/or click on an already highlighted MD.

Linked views

Highlighting is reflected throughout all the visualizations which, among other things, is useful for seeing how the time since diagnosis impacts the relative risks (i.e., lagged HRs) of interest. In these small multiple displays, the y-axis range varies depending on the shown GMC; however, they can also be fixed to a global range (with log scaling), which is useful for tracking changes across model or sex.

Absolute risk

In addition to relative risks (HRs), we’ve also estimated time-specific absolute risks as a cumulative incidence proportion (CIP). To obtain these estimates, each person diagnosed with specific MD was matched to up to five ‘reference individuals’ of the same sex and born within 30 days of them. For each MD-GMC pair, CIPs are estimated as a function of time of diagnosis, for both the reference group (dashed line) and those who were diagnosed (bold, solid line).

Model types:
Mental disorders:
Hazard ratio overview
Click to open