Examining the association between family status and depression in the UK Biobank

Background: We examined associations between family status (living with a spouse or partner, number of children) and lifetime depression. Methods: We used data from the UK Biobank, a large prospective study of middle-aged and older adults. Lifetime depression was assessed as part of a follow-up mental health questionnaire. Logistic regression was used to estimate associations between family status and depression. We included extensive adjustment for social, demographic and other potential confounders, including depression polygenic risk scores. Results: 52,078 participants (mean age = 63.6, SD = 7.6; 52% female) were included in our analyses. Living with a spouse or partner was associated with substantially lower odds of lifetime depression (OR = 0.67, 95% CI 0.62-0.74). Compared to individuals without children, we found higher odds of lifetime depression for parents of one child (OR = 1.17, 95% CI 1.07-1.27), and parents of three (OR = 1.11, 95% CI 1.03-1.20) or four or more children (OR = 1.27, 95% CI 1.14-1.42). Amongst those not cohabiting, having any number of children was associated with higher odds of lifetime depression. Our results were consistent across age groups, the sexes, neighbourhood deprivation and genetic risk for depression. Exploratory Mendelian randomisation analyses suggested a causal effect of number of children on lifetime depression. Limitations: Our data did not allow distinguishing between non-marital and marital cohabitation. Results may not generalise to all ages or populations. Conclusions: Living with a spouse or partner was strongly associated with reduced odds of depression. Having one or three or more children was associated with increased odds of depression, especially in individuals not living with a spouse or partner.


44
Major depressive disorder currently affects up to 322 million people worldwide, making it the most 45 prevalent mood disorder (1). At least half of the global population report experiencing depressive 46 symptoms at some point in their lives (2). Up to a third of older adults suffer from depression, many 47 of whom experience functional decline and cognitive impairment (3). Depression is the fifth leading 48 cause of long-standing disability amongst all diseases, and the first amongst mental health disorders 49 (4). It has been associated with lower productivity (5,6), lower earnings (7,8), increased risk of suicide 50 (9,10) as well as higher rates of cardiovascular disease (11), cancer (12) and all-cause mortality (13). 51 52 Family status, an umbrella term which includes marital status, cohabiting with a spouse or partner and 53 parenthood, has consistently been associated with the occurrence and severity of depression (2,14). 54 For example, depression has been associated with lower odds of ever getting married (odds ratios 55 ranging from 0.6 to 0.8) (15-17). A longitudinal study found an 11% reduction in the probability of 56 getting married for children reporting depressive symptoms, compared to their siblings (18). Large 57 meta-analyses of studies in Chinese older adults have shown that the prevalence of depression is 58 substantially higher amongst non-married (including divorced, unmarried and widowed) individuals 59 (19,20). International data follow a similar pattern (21), with married individuals reporting higher 60 levels of well-being across cultures (22). In women, not being married is associated with a higher 61 incidence of postnatal depression (23). Married individuals tend to be healthier, happier and less 62 depressed (24). Based on these findings it has been suggested that marriage is a protective factor 63 which shields individuals from stressful circumstances and provides confidant support (25). Findings 64 from twin studies which attempt to control for genetic predisposition to depression have shown that 65 divorced, widowed and never-married twins had higher rates of depression, indicating a protective 66 effect of marriage (26-28). Similar associations have been found for physical health, with married 67 individuals having a lower risk of cardiovascular disease (29,30). Concerning non-marital 68 cohabitation, data from the Health and Retirement Study suggested that non-married cohabiting 69 individuals had a higher risk of depression compared to married individuals, but lower compared to 70 never-married, divorced or widowed individuals (31). 71 72 Parenthood has been associated with lower rates of depression, with one study showing that 73 depressive symptoms decrease as the number of children increases (32). Another study reported no 74 association between depression and parenthood in general, but positive associations with certain types 75 of parenthood (for example, in single parents and parents of young children) (33). However, these 76 associations might be modified by marital status -with single parents reporting more depressive 77 symptoms -as well as by cultural and ethnic factors (33,34). Single mothers in particular suffer from 78 high rates of depression and it has been suggested that this association is mediated by high levels of 79

Study description 114
The UK Biobank is a prospective study of approximately 500 000 men and women, aged 115 recruited between 2006-2010. Individuals who were living within a 25-mile (~40 km) radius of one of 116 22 assessment centres across England, Scotland and Wales were invited to participate. At the baseline 117 assessment, participants provided informed consent and reported sociodemographic, lifestyle and 118 medical history factors through touch-screen questionnaires and nurse-led interviews. They also 119 provided biological samples (blood, urine and saliva), physical measurements (e.g. height and weight) 120 and consented to their data being linked to their health records. The study has been described in detail 121 in previous publications (41,42).

Family status 131
Two variables were included for family status: cohabitation status and number of children. 132 Participants who answered "Husband, wife or partner" in response to the question "How are the other 133 people who live with you related to you?" were classified as "living with a spouse or partner". The 134 number of biological children was based on "Number of live births" for women and "Number of 135 children fathered" for men. We categorised participants into five groups: "childless", "parent of one", 136 "parent of two", "parent of three" and "parent of four or more". 137 138

Depression 139
Lifetime depression was assessed using the depression module of the CIDI-SF and defined according 140 to DSM-5 criteria for major depressive disorder. Lifetime severe depression was defined as meeting 141 criteria for lifetime depression, endorsing all non-core symptoms on the CIDI-SF and reporting a lot 142 of impairment in normal functioning. Current depression was defined as meeting criteria for lifetime 143 depression and reporting at least five symptoms, including at least one core symptom, on the Patient 144 Health Questionnaire 9 (PHQ-9) more than half the days (several days for recent thoughts of suicide 145 or self-harm) over the past two weeks (45,46). Current severe depression was defined as meeting 146 criteria for current depression and having a sum score of 15 or more on the PHQ-9. See Supplement 2 147 and (44). Covariates 150 Potential confounders of the association between family status and depression were identified from 151 the baseline assessment data (sex, marital separation/divorce in the two years prior to assessment, 152 death of a spouse/partner in the two years prior to assessment, migrant status, highest educational or 153 professional qualification, annual gross household income, employment status, Townsend deprivation 154 index, smoking status, long-standing illness, disability or infirmity, neuroticism, participation in 155 leisure/social activities, loneliness, ever had same-sex intercourse, lifetime number of sexual partners 156 and body mass index) and from the MHQ (age at completing the questionnaire, alcohol use, adverse 157 childhood experiences and traumatic life events) (Supplement 3). We also included a PRS based on 158 summary statistics from the most recent depression GWAS, excluding UK Biobank participants (39) 159 (see Supplement 4). 160 161

Exclusion criteria 162
Individuals with missing data or who responded "prefer not to answer" or "do not know" were 163 excluded from all analyses. The lifetime number of sexual partners was truncated at the 99th%ile (> 164 50). We also excluded individuals with self-reported psychosis or mania on the MHQ. Of the 165 individuals who did not meet criteria for depression as described above, we excluded any participant 166 who had self-reported any other mental disorder, if they were currently taking antidepressant

Statistical analysis 172
Our main analysis focused on lifetime depression (i.e. having ever met diagnostic criteria for major 173 depressive disorder). We used logistic regression to estimate the association between family status 174 and depression. For each of the two explanatory variables (cohabitation status and number of 175 children) we first fitted crude models without adjustment for potential confounders. In subsequent 176 models we (i) adjusted for age and sex and (ii) adjusted for age, sex, all other non-genetic covariates, 177 the depression PRS, six ancestry-informative principal components as well as batch and assessment 178 centre. We also fitted an additional model that included both cohabitation status and number of 179 children as well as all covariates (termed the "fully adjusted model"). values, multiple testing correction was performed using the Benjamini & Hochberg false discovery 183 rate approach (49). For each explanatory variable, we calculated its statistical significance adjusted for 184 a 5% false discovery rate, by taking into account the variable's p-values from the four main 185 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 7, 2020.

Sensitivity analysis 201
To assess whether our findings were consistent across different depression phenotypes, we substituted 202 current depression for lifetime depression in the fully adjusted model. We also restricted analyses to 203 individuals with severe depression for both lifetime and current depression (Supplement 2). 204 205 Finally, we repeated the main analysis after excluding individuals with postnatal depression which 206 might confound the association between family status and depression. The association between 207 depression and childbirth, which is associated with hormonal changes and circumstances during 208 pregnancy and birth (50), should be considered separately from that of parenthood in general. Since 209 we were interested in the effect of parenthood and not childbirth per se, we excluded women who 210 reported depression that was possibly related to childbirth.

Mendelian randomisation 215
Exploratory mendelian randomisation (MR) analyses were performed to test for any causal genetic 216 associations between lifetime depression and family status. Our primary MR analyses were carried 217 out using Generalised Summary-data-based Mendelian Randomisation (GSMR) (51). The GSMR 218 method tests for putative causal association between a risk factor and an outcome using summary-219 level data from GWAS analyses (see Supplement 8) and requires a reference sample with individual 220 level genotypes for linkage disequilibrium (LD) estimation. In the present study we used the 1000 221 Genomes Project as reference sample, with a clumping r 2 threshold of 0.05 and an FDR threshold (to 222 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 7, 2020. . https://doi.org/10.1101/2020.07.07.20148023 doi: medRxiv preprint shrink chance correlations between SNP instruments to zero) of 0.05 (52). The HEIDI-outlier method 223 threshold for detecting pleiotropy was set to 0.01. This is a robust method for detecting and 224 eliminating genetic instruments that have pleiotropic effects on both the exposure and outcome (51). 225

226
Due to the small number of SNPs significant above the standard p < 5 × 10 −8 threshold, we lowered 227 the threshold in the present study to p < 5 × 10 −7 for lifetime depression and number of children, and p 228 < 5 × 10 −6 for cohabitation status. This resulted in 12, 21 and 18 SNPs that could be used as genetic 229 instruments, respectively. We performed bi-directional MR analyses between number of children and 230 lifetime depression, and between cohabitation status and lifetime depression. 231

232
To validate the findings from GSMR, a secondary MR analysis was carried out using the Two-Sample 233 MR package in R (53). The same genetic instruments were used for each GWAS as for GSMR 234 analyses. The same bi-directional analyses were performed using the Inverse Variance Weighted 235 (IVW), MR-RAPS and MR-Egger methods (54). These have been described as robust MR methods 236 and are commonly used as sensitivity analyses when performing MR (54,55). 237 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Descriptive statistics 239
Of the 157 389 MHQ respondents, 126 315 were retained after genetic quality control (see 240 Supplement 4). After excluding individuals with missing data on depression, family status or 241 covariates, 52 078 participants were included in our main analysis ( Figure 1). The average time 242 between the baseline assessment and completion of the MHQ was 8.11 years (SD = 0.88). 243 244 Table 1 illustrates the characteristics of the full and analytical sample. The percentage of individuals 245 in the analytical sample who were cohabiting with a spouse or partner was higher than that of the UK 246 general population (93% compared to 69% of those aged 45 and above) (56). There were notable 247 differences in rates of lifetime depression between individuals who reported cohabiting with a spouse 248 or partner and those who did not (27% and 50%, respectively) ( Figure 2). However, lifetime 249 depression rates were fairly consistent across parental categories, ranging from 27% for parents of two is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted July 7, 2020. in the main analysis. However, amongst individuals not living with a spouse or partner, we found that 295 parenthood was consistently associated with higher odds of lifetime depression. For example, parents 296 of three who were not cohabiting had 122% higher odds of lifetime depression compared to childless 297 individuals not living with a spouse or partner. For parents of two children and parent of three 298 children there was also evidence of an interaction in the full analytical sample (both pinteraction < 0.001). 299 However, there was no evidence of an interaction between cohabitation status and having one child or 300 four or more children (pinteraction = 0.07 and 0.68, respectively). 301 302 Table 3 presents the results of the sensitivity analyses. Excluding women who reported postnatal 304 depression (n = 1 935) from the analytical sample resulted in almost identical results for cohabitation 305 status. However, associations between number of children and lifetime depression were attenuated, 306

Sensitivity analyses 303
and there was no longer evidence of an association between having three children and lifetime 307 depression (OR = 1.02, 95% CI 0.94-1.10). Examining lifetime severe depression, current depression 308 and current severe depression yielded similar results for cohabitation status. With respect to number 309 of children, the magnitude of associations and corresponding uncertainty estimates increased, and 310 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted July 7, 2020. . https://doi.org/10.1101/2020.07.07.20148023 doi: medRxiv preprint there were some inconsistencies in findings for current and current severe depression. There was no 311 evidence that having two children was associated with depression across the different phenotypes. 312 313

Mendelian randomisation 314
Our findings show that the genetic instruments for number of children were associated with increased 315 odds of lifetime depression (OR = 1.98, 95% CI 1.23-3.20, p = 0.005). The genetic instruments for 316 cohabitation status were associated with approximately 23% lower odds of lifetime depression, 317 although the effect was not statistically significant (OR = 0.77, 95% CI 0.36-1.65, p = 0.498). We did 318 not find evidence that the genetic instruments for lifetime depression were associated with number of 319 children (OR = 1.004, 95% CI 0.99-1.02, p = 0.581) or cohabitation status (OR = 1.006, 95% CI 0.99-320 1.02, p = 0.428). These findings are presented in Figure 8.  . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

329
In the present study we showed that living with a spouse or partner was associated with substantially 330 lower odds of depression. This association remained after extensive adjustment for potential 331 confounders and was robust to sensitivity analyses. Additional adjustment for number of children had 332 little impact on the association between cohabitation status and depression. 333

334
Most individuals in the present study who experienced marital separation or divorce or the death of a 335 spouse or partner in the two years prior to the baseline assessment were not cohabiting, and this 336 accounted for part of the association between cohabitation status and depression. Additionally, 74% of 337 individuals not living with a spouse or partner were female, who typically report higher rates of 338 depression, and adjusting for sex slightly attenuated the association between cohabitation status and 339 depression. Low annual household income, adverse childhood experiences and ever having had same-340 sex intercourse were more frequent amongst participants not living with a spouse or partner, which is 341 line with previous studies (57,58). When we adjusted for these and other potential confounders, we 342 still found a substantial association between cohabitation status and depression. 343

344
We found similar associations when we examined lifetime severe depression, current depression and 345 current severe depression, and the results were consistent across age groups, the sexes, levels of 346 neighbourhood deprivation and depression PRS quintiles. These findings suggest that living with a 347 spouse or partner was associated with lower odds of depression, irrespective of the social, economic 348 or genetic characteristics that we adjusted for. Of particular interest is the absence of any interaction 349 with sex. Previous findings from genetic studies suggested a beneficial effect of marriage exclusively 350 for women (26). There have also been claims that marriage is only beneficial for men, and harmful for 351 women's happiness and wellbeing (59). Our findings do not support such views, since the association 352 between cohabitation status and depression was the same in both sexes. 353

354
The direction of causation was explored in the MR analyses. The results suggest that living with a 355 spouse or partner is causally linked with lower odds of lifetime depression. The effect was not 356 statistically significant, however this may be due to reduced power, given the limited number of 357 genetic instruments that were used. Large-scale genome-wide association studies are required in order 358 to identify more SNPs related to family status. The MR analyses did not provide evidence in support 359 of a causal effect of lifetime depression on family status. These results, although not conclusive, give 360 more credence to the hypothesis that partnership is a protective factor against depression (26-28), as 361 opposed to that of depression-prone individuals tending not to form relationships (18). The finding 362 that the association between cohabitation status and depression was consistent across depression PRS 363 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted July 7, 2020. . https://doi.org/10.1101/2020.07.07.20148023 doi: medRxiv preprint quintiles suggests that living with a spouse or partner is associated with reduced odds of depression, 364 regardless of genetic predisposition to depression. 365 366 It has been suggested that married individuals or those in partnership are less likely to be exposed to 367 unpleasant experiences which may induce depression (25). In the present study we found that the 368 association between cohabitation status and depression remained after controlling for traumatic life 369 events and other risk factors, suggesting that this association could not be fully accounted for by 370 married or cohabitating individuals having fewer adverse experiences. An additional explanation, 371 which is more consistent with our results, is that living with a spouse or partner provides a source of 372 intimacy and confidante support which might reduce the risk of depression (25). It is worth noting that 373 we also adjusted for self-reported loneliness. Therefore, we can infer that the association between 374 cohabitation status and depression is not driven only by the absence of loneliness in cohabiting 375 individuals, but that there is something unique about partnership per se, which is associated with 376 reduced rates of depression. This conclusion is also supported by the fact that individuals who did not 377 live with a spouse or partner were not necessarily living alone but might have been cohabiting with 378 other relatives or unrelated individuals. 379

380
We examined the broad category of "living with a partner or spouse" as no information on marital 381 status was available in the UK Biobank. This is both a strength and potential limitation. We provide 382 evidence that living with a spouse or partner per se, and not necessarily any legal or financial 383 conditions associated with marriage, was associated with lower odds of depression. However, other 384 research has suggested that there are differences in the effect of partnership on depression between 385 married and non-married cohabiting couples (31), and we could not explore this in the current study. 386 387 Associations between number of children and depression were less consistent. When we included 388 only number of children in the model, having one child was associated with slightly higher odds of 389 depression, while having two children or having three children was associated with lower odds of 390 depression. However, the associations between having two or three children and depression were not 391 consistent when we accounted for other variables in the adjusted models. It is therefore likely that 392 potential confounding might explain some of the associations observed in the unadjusted model. For 393 example, having a college degree is associated with higher rates of depression (47) and having fewer 394 children (60), therefore educational attainment is an obvious candidate for confounding the 395 association. 396

397
Our findings provide some support for the association between number of children and depression 398 that was recently observed in genetic studies (61). Having four or more children was associated with a 399 27% increase in the odds of reporting lifetime depression in the fully adjusted model, possibly 400 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted July 7, 2020.

Strengths and limitations 414
A major strength of the present study was the large sample size (> 50 000 participants), which 415 allowed for high precision in the estimation of associations and for modelling interactions. Moreover, 416 the wealth of information available through the UK Biobank allowed us to adjust for many of the 417 factors that have previously been linked to family status and depression. Of particular importance, in 418 our view, is that we also included a polygenic risk score to adjust for genetic susceptibility to 419 depression. This study illustrates how social science can incorporate genetic data, not only in the form 420 of twin studies, which necessarily have limited sample size, but by using molecular data available for 421 tens of thousands of individuals. 422 423 Several potential limitations need to be considered in evaluating the present study. The UK Biobank, 424 and the MHQ sample in particular, have been shown to not be representative of the UK general 425 population (62). Participants were more likely to have better health and higher socioeconomic status 426 than the average UK citizen (63). This fact is reflected in our sample, where participants 427 disproportionally belonged to higher income, higher education and less deprived groups. In response 428 to these concerns, UK Biobank has released a statement clarifying that, although its data cannot be 429 used to provide representative disease prevalence rates, associations between exposures and outcomes 430 are nonetheless widely generalisable, although there might be some differences in the magnitude of 431 associations(63). Furthermore, the UK Biobank includes middle aged and older adults for whom 432 perceptions of family life may be different from those of younger people. Raising young children 433 might also be different from being a parent of independent adults, in terms of stress and 434 responsibilities. Regarding number of children, we only included biological children and did not take 435 adoptees into account. As such the present findings might not generalise to adoptive parents. The 436 depression PRS that we used currently explains only a small percentage of the variance in depression, 437 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted July 7, 2020. . https://doi.org/10.1101/2020.07.07.20148023 doi: medRxiv preprint which might explain the absence of any interaction with family status in our study. Finally, our 438 findings might not generalise to non-European ancestry groups. 439 440 Several limitations pertaining to the MR analyses should be noted. First, we lowered the p-value 441 thresholds for all GWAS in order to obtain enough SNP instruments to perform MR. This may 442 introduce false positives and type 1 error. However, we also performed GSMR using only those 443 genetic instruments for number of children that were significant above the standard p < 5 × 10 −8 444 threshold (6 SNPs in total) and obtained similar results. Nonetheless, we could not do the same for the 445 cohabitation status and lifetime depression GWAS and thus there is still the potential for a type 1 446 error. As such, these results should be treated as exploratory rather than conclusive. Second, all three 447 GWAS in this study were carried out on complex traits and although the lifetime depression GWAS 448 phenotype was from clinical cohorts, concerns remain about the validity of associations derived from 449 such analyses. However, this issue is difficult to overcome in genetics when studying complex traits. Future studies should aim to examine our findings using official data on marital status and 462 cohabitation status, and to extend the scope of our research to include both biological and adopted 463 children. Moreover, further research is needed in younger cohorts. It would be particularly useful to 464 adopt a longitudinal design, in order to infer whether depressive symptoms predate and select people 465 into or out of partnership and parenthood, or whether differences in odds of depression become 466 apparent after relationship formation. 467 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted July 7, 2020. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted July 7, 2020. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted July 7, 2020. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted July 7, 2020. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted July 7, 2020. . . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted July 7, 2020. . https://doi.org/10.1101/2020.07.07.20148023 doi: medRxiv preprint  . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted July 7, 2020. . https://doi.org/10.1101/2020.07.07.20148023 doi: medRxiv preprint . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted July 7, 2020. . https://doi.org/10.1101/2020.07.07.20148023 doi: medRxiv preprint . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted July 7, 2020. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted July 7, 2020. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted July 7, 2020. . https://doi.org/10.1101/2020.07.07.20148023 doi: medRxiv preprint . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 7, 2020. . https://doi.org/10.1101/2020.07.07.20148023 doi: medRxiv preprint . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 7, 2020. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 7, 2020. 26.4 (4.1) Missing* 250 (0%) Note: MHQ = mental health questionnaire; SD = standard deviation; IQR = interquartile range; GCSEs = general certificate of secondary education; CSE = certificate of secondary education; NVQ = national vocational qualification; HND = higher national diploma; HNC = higher national certificate. Frequencies in the third column refer to the sample of 126 315 individuals who had completed the MHQ and who were retained after genetic quality control. Percentages are rounded to the nearest whole number and might not add up to 100%. * also includes participants who responded "do not know" or "prefer not to answer"; † unless indicated otherwise; ‡ also includes 'other professional qualifications'.
. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 7, 2020. . https://doi.org/10.1101/2020.07.07.20148023 doi: medRxiv preprint