Pre-pandemic mental illness and risk of death from COVID-19

Although the mental health impact of a COVID-19 diagnosis and societal restrictions has been examined, whether pre-pandemic mental health problems predict susceptibility to COVID-19 has not been widely tested. Taquet and colleagues¹ recently found that a diagnosis of poor mental health was associated with an increased risk of COVID-19. We have previously explored the role of explanatory characteristics such as ethnicity,² lifestyle,³ and vascular risk indices⁴ in the relationship between mental health⁵ and COVID-19 using data from UK Biobank, a field-based prospective cohort study of around 0·5 million people. In those analyses, our outcome of interest was hospitalisation with COVID-19. We now use 2020 data on death from the disease.

In UK Biobank (2006–10), self-reported mental health was captured using two indices at baseline. Study members were asked whether they had ever been under the care of a psychiatrist for any mental health problem. Symptoms of psychological distress (a combination of anxiety and depression) were measured using the validated four item version of the Patient Health Questionnaire with a total score from 0 to 12 (higher scores denote greater distress). Socioeconomic status was quantified using self-reported educational qualifications (degree, other qualifications, no qualifications) and the Townsend index of neighbourhood deprivation (higher scores denote greater disadvantage). Ethnicity was categorised as White, Asian, Black, Chinese, Mixed, or other. Baseline vascular or heart problems, diabetes, chronic lung disease, and asthma, were based on a self-reported physician diagnosis. The presence of hypertension was defined as a blood pressure of 140/90 mmHg or higher or self-reported use of antihypertensive medication, or both. C-reactive protein, glycated haemoglobin, and high-density lipoprotein cholesterol concentrations were based on assays of non-fasting venous blood. Height, weight, and forced expiratory volume in 1 s were measured using standard protocols. Cigarette smoking, physical activity, and alcohol consumption were assessed using standard questions. Study members were linked to national mortality records and death from COVID-19, as denoted by the emergency ICD-10 code U07·1 (COVID-19, virus identified).

Of 447 463 individuals (241 883 women), 351 deaths were ascribed to COVID-19 between April 1, 2020, and Sept 23, 2020 (end of follow-up). As the comparator model, hazard ratios (HR) with accompanying 95% CIs were computed using Cox regression and adjusted for age, sex, ethnicity, and comorbidities (table ). Relative to the group reporting no symptoms of distress, rates of death from COVID-19 were most elevated in individuals within the high distress category (HR 1·76, 95% CI 1·34–2·32), and a dose–response effect was apparent (p<0·0001). The magnitude of this association was diminished by around a third after adjusting for socioeconomic status (1·51, 1·13–2·00; p=0·0030), and by around a half when lifestyle factors were taken into account (1·42, 1·07–1·90; p=0·012). Biological risk indices had little explanatory power, which was shown after we computed effect estimates in the most basic comparator model for the subsample with full biological data (appendix p 1). Although weaker relationships were seen for self-reported psychiatric consultation, there was a similar pattern of post-adjustment attenuation. Overall, the magnitude of the association distress and mortality from COVID-19 was higher than for hospitalisations from the disease.⁵

Table.

Pre-pandemic self-reported mental health and COVID-19 mortality

		Adjusted for age, sex, ethnicity, and comorbidities*	Adjusted for age, sex, ethnicity, and comorbidities plus socioeconomic status†	Adjusted for age, sex, ethnicity, and comorbidities plus lifestyle factors‡	Adjusted for age, sex, ethnicity, and comorbidities plus biomarkers§
Number of COVID-19 deaths		341	329	331	224
Number at risk of COVID-19		439 513	435 556	434 819	316 009
Psychological distress¶
	Group 1 (score 0)‖	1·0	1·0	1·0	1·0
	Group 2 (score 1–2)	1·38 (1·07–1·78)	1·37 (1·06–1·77)	1·29 (1·00–1·67)	1·54 (1·13–2·09)
	Group 3 (score 3–12)	1·76 (1·34–2·32)	1·51 (1·13–2·00)	1·42 (1·07–1·90)	1·76 (1·24–2·50)
	p value	<0·0001	0·0030	0·012	0·0007
	Per SD (2·11) increase in distress score	1·29 (1·17–1·44)	1·21 (1·10–1·34)	1·19 (1·09–1·31)	1·27 (1·13–1·43)
	Number of COVID-19 deaths	386	372	373	253
	Number at risk of COVID-19	486 887	481 578	480 810	349 078
Psychiatric consultation
	No‖	1·0	1·0	1·0	1·0
	Yes	1·35 (1·01–1·81)	1·26 (0·94–1·70)	1·22 (0·91–1·65)	1·44 (1·01–2·06)
	p value	0·040	0·125	0·183	0·042

Data analysed from UK Biobank, shown as hazard ratios (95% CI), unless stated otherwise.

^*Comorbidities were diagnoses of vascular or heart disease, diabetes, chronic bronchitis or emphysema, asthma, and hypertension.

^†Socioeconomic status was measured by educational attainment and Townsend deprivation index.

^‡Lifestyle factors were body-mass index, smoking status, alcohol intake frequency, and number of the types of physical activity in the past 4 weeks.

^§Biomarkers were forced expiratory volume in the first second, and blood concentrations of C-reactive protein, glycated haemoglobin, and high-density lipoprotein.

^¶Based on the Patient Health Questionnaire-4.

^‖Used as the referent group.