Understanding the relationship between depression and trajectories of physical multimorbidity accrual: longitudinal analysis of UK Biobank data

Lead Research Organisation: University of Edinburgh
Department Name: Centre of Population Health Sciences


There are increasing numbers of people who have multimorbidity which means living with multiple physical and/or mental health conditions. Increasing multimorbidity has a number of causes. People are living longer and are more likely to survive life-threatening illness than in the past (for example, anyone who survives a heart attack will then have to live with chronic heart disease). Living longer and surviving life-threatening illness are of course good things, but multimorbidity poses challenges to health services and research which often focuses on single conditions. We need to understand how people develop multiple conditions better, and this is the focus of this research.

The combination of physical conditions with depression is very common. About one in five people with a physical condition have depression, and this combination is associated with worse physical and mental health outcomes. It is therefore important to understand how physical conditions and depression are related. Existing research has often just looked at people at one point in time. This means that it isn't possible to know if depression causes physical disease, or if physical disease causes depression, or both. Existing research also does not always account for things which might be related to both physical disease and depression such as heavy drinking or lack of exercise.
The aim of this research is to examine the relationship between physical disease and depression using data for half a million middle-aged people followed up for ~10 years as part of the UK Biobank study. There are three elements to the planned work:

First, we will develop a number of different ways of measuring the patterns of physical conditions that people get in middle and older age. There are three different ways of doing this: (1) Simply counting how many physical conditions people develop over 10 years; (2) Identifying the 'shape' or trajectory of how they develop new physical conditions (eg developing none vs slowly developing new conditions vs rapidly developing them); and (3) Identifying if there are groups of physical conditions that develop together (for example, different conditions caused by smoking, but there may be unexpected groupings as well). We will also explore different ways of measuring depression in the data, because we don't know whether the severity of depression affects the relationship with physical conditions.

Second, we will examine whether depression before the start of the study predicts how people develop physical conditions over the next 10 years. We will do this using a number of methods, based on the different ways of measuring patterns of physical conditions described above. We will carefully account for the effect of other individual characteristics which might be linked to both depression and patterns of physical condition such as gender, smoking, alcohol, obesity, and how rich or poor a person is.

Third, we will examine whether different patterns of physical condition predict whether or not people develop new or recurrent depression. Again, we will carefully account for the effect of other individual characteristics which might be linked to both depression and patterns of physical conditions.

The study will add significantly to our understanding of the patterns of physical conditions which people develop in middle age, and the links between physical conditions and depression. The same approach can be used in the future to examine how physical conditions are linked to other mental health problems like anxiety or schizophrenia. It will also make possible additional studies that examine if there are common genetic (inherited) causes of both depression and physical conditions, or if there are common mechanisms affecting different diseases like inflammation. Finally, better understanding of patterns of disease will help identify people who may need more help from health services to improve their quality of life and future health.

Technical Summary

Importance and aim. Multimorbidity poses major challenges to researchers and to health systems. Physical condition-depression multimorbidity is particularly interesting because it is common, appears to have a bidirectional relationship, and is associated with worse outcomes. The aim of this study is to use UK Biobank data to identify trajectories of physical condition accrual in middle and older age, and examine how associations with depression at both baseline and follow-up.

In objective 1, we will build on our previous work to define which conditions we are going to count, and define trajectories of physical condition accrual in three ways: (1) Unweighted counts of number of conditions incident during follow-up; (2) Latent class trajectory modelling to classify patients in terms of the count trajectory, and; (3) Agglomerative hierarchical and k-means clustering to classify patients in terms of the types of conditions they accrue during follow-up. We will additionally define a number of measures of depression at both baseline and follow-up.

In objective 2, we will examine how baseline depression (and sociodemographic, lifestyle and risk factor characteristics) is associated with subsequent physical condition accrual using Poisson regression and multiple event survival analysis (outcome = condition count) and multinomial regression (outcomes = count trajectory and trajectory-by-condition-cluster cross-classification).

In objective 3, we will examine how physical condition accrual during follow-up is associated with subsequent incident and recurrent depression using logistic regression for multiple depression phenotypes.

Outputs. The substantive findings will be of widespread interest, but the derived variables (individual morbidities, trajectories and condition clusters) and methods will also support future studies in Biobank using blood assay and genotype data, and implementation in other datasets.

Planned Impact

The substantive findings and methods will be of interest and value to academic researchers; NHS and healthcare policy analysts, managers and clinicians; and patients, the public and third sector organisations.

Beneficiary group 1 are academic researchers in a range of disciplines. Physical health conditions and depression are one of the commonest patterns of multimorbidity seen, but most previous analysis has been relatively narrow, often focusing on cross-sectional relationships between single physical conditions and depression (ie taking a comorbidity approach, where the researchers are typically interested in depression in people with physical condition X, or physical condition Y in people with depression). Depending on the results, the findings of this study have the potential to reframe understanding of the relationship. We expect the findings to identify new avenues of productive research, including directly supporting future analysis using linked genomic and biomarker data in UK Biobank, but will additionally be hypothesis generating for analyses in other datasets.

Beneficiary group 2 is anyone who wishes to examine individual or multiple morbidities in routine clinical data, which includes academic researchers but also analysts working for industry and in policy and the NHS. The benefit will be realised by our provision of derived variables (individual morbidities, counts, trajectories and condition clusters) back to UK Biobank, and publication of codesets and methods to facilitate replication in other datasets. Our previous multimorbidity work where we defined 40 individual morbidities to measure multimorbidity demonstrates the potential value, since that set of conditions has proved the starting point for multiple other studies implemented in different datasets internationally.

Beneficiary group 3 is NHS and healthcare policy professionals. Better understanding trajectories of condition accrual and condition clustering in middle and older age, and how physical morbidity is associated with both prior and subsequent depression will be of interest to NHS clinicians and managers and those working in healthcare policy. Initial impact is likely to be to reframe taken-for-granted assumptions about how healthcare should be organised (and in particular, the near complete division between physical and mental health services in the UK and many other healthcare systems). Longer term impact will need further research that develops and evaluates interventions in this population, but such developing such interventions requires the kind of good epidemiological data that this project will provide, for example in terms of identifying groups of patients and particular points in trajectories to intervene in.

Beneficiary group 4 are patients, the public and third sector organisations. Living with one or more physical conditions and depression is one of the commonest patterns of multimorbidity, particularly in middle and early older age. In the short term, the findings will be of value to patients, the public and third sector organisations by challenging current assumptions about the separation between physical and mental health (including for example, in how charities and charitable research funding is organised). In the medium term, the findings will support the development, evaluation and implementation of interventions to improve care and health for people with both, and in the medium to longer term support new basic science discovery that builds on the epidemiological data and derived variables.

Capacity building and training. The two employed researchers will gain valuable experience and expertise in managing large clinical datasets, and in the analysis of longitudinal data, both of which are shortage skills.


10 25 50
Description Artificial Intelligence and Multimorbidity: Clustering in Individuals, Space and Clinical Context (AIM-CISC)
Amount £3,926,146 (GBP)
Funding ID 202639 
Organisation National Institute for Health Research 
Sector Public
Country United Kingdom
Start 07/2021 
End 05/2025
Title Multiple derived variables returned to UK Biobank for other researchers to use 
Description (1) Variables defining 80 chronic conditions derived from UK Biobank baseline data, linked hospital data, and linked GP electronic health record data (presence of condition, date of first diagnosis); (2) Variables defining observability in GP data (ie dates of continuous GP registration before and after UK Biobank baseline). UKB can make available to other researchers (not under our direct control). 
Type Of Material Improvements to research infrastructure 
Year Produced 2024 
Provided To Others? Yes  
Impact Will support other multimorbidity research 
Title Clinical Terms Version 3 (CTV3) codesets for common chronic morbidities 
Description We are using UK Biobank data and defining morbidities using Read v2 and ICD-10 codesets developed in UCL and available through the HDR-UK phenotype library. However, most of the primary care data in UK Biobank used Clinical Terms Version 3 (CTV3) for which there are no codesets available. We have mapped core morbidities to CTV3. 
Type Of Material Data handling & control 
Year Produced 2021 
Provided To Others? No  
Impact We have shared these with other researchers on a one-off basis so far, but will publish in the HDR-UK Phenotype Library later in 2022, and put back into UK Biobank for others to use. 
Description NIHR AI and Multiple Long-Term Conditions Programme - supported data management in two of the funded consortia AIM-CISC and AI-Multiply. 
Organisation Newcastle University
Country United Kingdom 
Sector Academic/University 
PI Contribution Both AIM consortia have drawn on condition definitions in UKB to reduce their data management time, and to be consistent/replicable.
Collaborator Contribution Further extends understanding of trajectories of morbidity accrual
Impact Analysis still in progress
Start Year 2023
Description Multimorbidity methods workshop 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach National
Primary Audience Professional Practitioners
Results and Impact Masterclass presentation about measuring multimorbidity to multiple research consortia including professional collaborators, public contributors and students
Year(s) Of Engagement Activity 2022
Description Online conference organised by us - Methodological challenges in research on physical-mental health multimorbidity 
Form Of Engagement Activity Participation in an activity, workshop or similar
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Other audiences
Results and Impact We organised an online conference involving two other groups funded in the same MRC/NIHR SPA as us, focused on methodological problems in researching physical-mental health multimorbidity (although the problems are generally applicable to all multimorbidity research). Over 150 people attended from a wide range of backgrounds, with others watching the recording subsequently.
Year(s) Of Engagement Activity 2022
Description Public engagement event - Mind and Body in (dis)Harmony 
Form Of Engagement Activity Participation in an activity, workshop or similar
Part Of Official Scheme? No
Geographic Reach Regional
Primary Audience Public/other audiences
Results and Impact Public engagement event involving presentation and discussion of research findings, personal testimony from our public contributors, and music
Year(s) Of Engagement Activity 2022
URL https://www.ed.ac.uk/usher/advanced-care-research-centre/news/thought-provoking-engagement-from-mind...