A team of researchers from Dementias Platform UK (DPUK) used machine learning techniques to analyse 1,175 participants aged 40-70 in the UK Biobank cohort. They found that mental health had a more important predictive effect on future thinking skills than other tested variables, which included a combination of demographic, socioeconomic and lifestyle factors such as BMI, education level and smoking.
Psychiatric conditions, say the researchers, may therefore ‘have a deleterious effect on long-term cognition’.
With recent research estimating that 40% of dementia cases could be delayed or prevented by addressing modifiable risk factors, these results suggest mental health has an important role to play in strategies for arresting cognitive decline and identifying those at risk of developing dementia.
The research was carried out in the DPUK Data Portal and is published in Evidence-Based Mental Health, part of the BMJ group of journals.
Senior author Dr Sarah Bauermeister, senior data and science manager at DPUK, said: ‘Our previous work has shown that poor mental health is associated with accelerated decline in cognitive ability as people get older. What we did here, for the first time, was to look at some of the other “comorbidities” of cognitive decline side by side to gain greater insight into the relative importance of the factors that affect our brain health trajectories.’
Comorbidity is defined as the presence of two or more chronic conditions in the same person. The comorbidities of impaired cognition examined as part of this study were anxiety, depression, cardiovascular disease and diabetes. The machine learning model used to carry out the analysis was also given information on age, BMI, lifestyle factors (including smoking and alcohol consumption), and socioeconomic status (for example, income and level of education).
The researchers found that both anxiety and depression were better at predicting future cognitive decline (measured in the UK Biobank cohort through a reaction time test) than cardiovascular disease, diabetes, or the reference covariant model of demographic, socioeconomic and lifestyle information. Using a machine learning measurement known as ‘area under the curve’, anxiety returned a result of 0.68 and depression 0.63, compared with 0.60 for cardiovascular disease and diabetes, and 0.56 for the covariant model.
Dr Bauermeister, who leads a research programme on the emerging area of childhood adversity and brain health, added: ‘The machine learning approach allowed us to interrogate a lot of data at once and compare these comorbidities alongside one another, giving robust and accurate predictions based on the longitudinal data in the UK Biobank study.
‘We have shown the importance of mental health as a predictor of cognitive decline, but also the importance of making sure we don’t look at mental health – or other factors – in isolation. Looking after our mental health works in tandem with keeping healthy in other ways to reduce our risk of cognitive decline or dementia in later life.’
The study made use of UK Biobank data across three time points, allowing the researchers to identify the predictive effect of each comorbidity. Mental health status was self-reported via questionnaires, with responses used as proxies for diagnoses of anxiety or depression disorders.
Future research by Dr Bauermeister and colleagues will explore the same question in other population cohorts. The researchers say further work would benefit from a broader range of cognitive measurements, and by looking at individuals with pre-diagnosed health conditions (UK Biobank recruited healthy individuals for initial testing), including formal mental health diagnoses.
Read the full study, ‘Psychiatric comorbid disorders of cognition: a machine learning approach using 1175 UK Biobank participants’, in Evidence-Based Mental Health.