Depressive symptoms measured using the Edinburgh Postnatal Depression Scale in mothers and partners in the ALSPAC Study: A data note

Depression is a leading cause of disability and is associated with a number of adverse offspring outcomes when it occurs in parents. Depression is present in men and women at different rates, and recent research suggests that symptom profiles between the sexes may differ. Longitudinal data are needed to answer remaining questions about the long-term course, gender differences, antecedents and outcomes of depression. The Avon Longitudinal Study of Parents and Children (ALSPAC) is a large birth cohort study in England which administered one of the most commonly used depression instruments, the Edinburgh Postnatal Depression Scale (EPDS) at 11 timepoints in mothers and at 10 timepoints in their partners. In addition to repeated measurements of the EPDS, ALSPAC has a wealth of participant data on biological, social, demographic, and lifestyle factors. The purpose of this data note is to introduce potential users of the data to the characteristics of the EPDS in ALSPAC, as well as some key considerations when using the data.


Introduction
Depression is one of the common mental disorders and is a leading cause of disability worldwide 1 . Previous research has also shown that individuals suffering from depression are at increased risk of transmitting this and other mental health problems to their offspring [2][3][4] . Understanding how depression unfolds over time as well as its risk factors in general population samples is therefore an urgent public health priority.
Population wide studies consistently show that depression is more common in women than in men 5 . Depression in the perinatal period has received an increasing amount of research attention. Research indicates that depression in both mothers and fathers in the time before and shortly after the birth of a child is common 6 and is linked with a range of negative offspring outcomes in longitudinal studies 3,4,[7][8][9][10] .
The Avon Longitudinal Study of Parents and Children (ALSPAC) is a birth cohort study of over 20,000 women, partners and their children that started in the early 1990s. Data collection is still ongoing and includes a wealth of repeated measurements of personality characteristics, mental health, biological measurements, lifestyle factors and sociodemographic information. One of these repeated measurements is of depressive symptoms via the Edinburgh Postnatal Depression Scale (EPDS) 11 , the most widely used perinatal depression instrument which has been translated into over 60 languages 12 . Because the EPDS was originally developed to measure depression in the perinatal period, somatic depression symptoms (e.g. fatigue or changes in appetite) were omitted. Despite its original intended use, the EPDS is also used to assess depressive symptoms outside of the perinatal period [13][14][15][16] and is administered to fathers as well 4,17 .
Over a dozen published ALSPAC studies have used EPDS data in mothers, partners, or both mothers and partners 8,16,18-28 . One early finding using ALSPAC data is that depressive symptoms were just as common during pregnancy as in the postnatal period 29 . Another ALSPAC study found that adolescents of mothers who had been depressed during and shortly after pregnancy were at significant risk for a diagnosis of depression 10 . More recently it was demonstrated that the prevalence of depression in pregnancy is higher (25%) in the second generation of ALSPAC mothers than in the first (17%) 30 .
There are still many unanswered questions regarding depression in mothers and partners. The aim of this data note is to describe the longitudinal data on maternal and partner depression available from ALSPAC to facilitate future research on depression.

Methods
The ALSPAC sample The Avon Longitudinal Study of parents and Children (ALSPAC) is a longitudinal birth cohort that recruited pregnant women residing in Avon, UK with expected dates of delivery between 1 April 1991 and 31 December 1992. The initial number of pregnancies was 14,541, which resulted in 14,062 live births and 13,988 children alive by the age of 1 year. Further details on the representativeness, cohort profile and recruitment have been published elsewhere 31-33 . The ALSPAC study website contains details of all data available through a fully searchable data dictionary and variable search tool.
Fathers were not initially enrolled in the study directly 34 . Instead, mothers were sent a questionnaire with the option of passing it on to him to complete. Consequently, no information on the number of mothers who invited their partners to participate is available. At least one questionnaire was returned by 75% of the partners of enrolled women.
Starting in 2014, study data were collected and managed using REDCap (Research Electronic Data Capture) version 7.4.9 hosted at the University of Bristol 35,36 . REDCap is a secure, web-based software platform designed to support data capture for research studies.

Ethical approval and consent
Ethical approval for the study was obtained from the ALSPAC Law and Ethics Committee and the local research ethics committees. Full details of the approvals are available from the study website.
Written informed consent for the use of data collected via questionnaires and clinics was obtained from participants following the recommendations of the ALSPAC Ethics and Law Committee at the time.

EPDS administration in ALSPAC
The EPDS was administered to ALSPAC mothers and their partners at 11 and 10 timepoints, respectively. Table 1, Table 2 present a complete list of these data collection timepoints which correspond to the focal child's average age within ALSPAC, as well as the ALSPAC data file and variable names. Nine of the timepoints are the same in both mothers and partners with respect to the focal child's age: 18 weeks' gestation, 8 weeks' post-partum, 8 months, 1 year 9 months, 2 years 9 months, 5 years 1 month, 6 years 1 month, 8 years 1 month, and 11 years 2 months.
Description of the population Due to attrition and other sources of missing data, sample size of mothers with EPDS data at each timepoint varies, from 12,151 at 18 weeks' gestation to 4,107 at 18 years post-partum. In partners (all of whom were male), sample sizes vary from 9,846 at 18 weeks' gestation to 1,951 at 21 years post-partum

Amendments from Version 1
We have added several citations of prior work as suggested by the reviewers and directed readers to ALSPAC's catalogue of publications. We clarified what we present in Figure 4 and Figure � �. We have also fixed some typographical errors.
Any further responses from the reviewers can be found at the end of the article ( Table 3, Table 4). A total of 14,169 mothers have valid EPDS data for at least one timepoint, with 2,854 of these having EPDS data for all 11 timepoints, and 1,476 mothers have data for zero EPDS assessments. In partners, 14,915 have at least one EPDS assessment, with 4,067 having all 10 assessments and 1,476 having zero. As can be seen in Table 5, Table 6 6, missingness on EPDS data appears to be related to several demographic characteristics.

The Edinburgh Postnatal Depression Scale (EPDS)
The 10-item Edinburgh Postnatal Depression Scale (EPDS) 11 assesses depressive symptoms during the prior week. Respondents rate the frequency of each symptom on a Likert-type scale with four response options, the exact wording of which varies depending on the item. See Table 7 for a complete item list with corresponding response options. Responses from all ten questions are coded 0-3 and three items (1, 2, and 4) are reverse coded. All ten items are then summed to give a score ranging between 0 and 30, where higher scores indicate more depressive symptoms. The EPDS does not assess or provide information on the duration or intensity of depressive symptoms 12 .
A score of 13 or higher on the EPDS is often used to indicate probable depression 11,37 . Others have used cut-off points of 12 38,39 and 10 17 . For simplicity, we present descriptive statistics using 10 and 13 as cut-offs. Syntax for creating the scores is available upon request.

Characteristics of the EPDS in ALSPAC
The mean number of EPDS depressive symptoms and the prevalence of probable depression using both the 13 and 10 or greater cut-offs for mothers and partners are presented in Table 3, Table 4. In mothers, the greatest prevalence of depression occurs at 18 years post-partum, while the lowest is at 8 months' post-partum (Table 3). In partners, the greatest proportion with probable depression occurs at 21 years post-partum, while the lowest proportion is at 8 months' post-partum ( Table 4).
Correlations of depressive symptoms between timepoints within mothers and partners are moderate (0.30 to 0.49) to large (< 0.50) ( Table 8, Table 9), providing initial evidence for predictive validity. Correlations between mother and partner depressive symptoms at the same timepoints are small to moderate (0.228-0.294) (Table 10). Data users modelling depressive symptoms over time should use statistical methods to account for these correlations. Internal reliability of the EPDS at each timepoint for both mothers and partners in ALSPAC is good ( Table 3, Table 4).
As can be seen in Table 4 and Table 5, mothers' depressive symptoms and rates of probable depression are consistently higher than their partners'. This is congruent with other research on gender differences in depression 40 . Figure 1 provides a visual representation of the differences in mothers and partners in the average number of depressive symptoms over time. The mean number of depressive symptoms are also presented in histograms at each timepoint for mothers ( Figure 2) and partners (Figure 3).
To help further illustrate these gender differences, Figure 4, Figure 5 present the distribution of the means (on a scale of 0-3) for each specific EPDS item in mothers and partners at all timepoints. In both mothers and partners, a common depressive symptom appears to be unnecessary self-blame. In mothers, being sad/miserable and anxious/worried are common symptoms, while in fathers, anxiety/worry and feeling overwhelmed ("things getting too much") are common.
Some researchers have investigated whether the EPDS total score can be disaggregated into meaningful subscales 41 . Using the ALSPAC data, Coates and colleagues 41 identified a three-factor solution to fit maternal EPDS data at 18 and 32 weeks' gestation and 8 weeks and 8 months post-partum. The three factors were anhedonia (EPDS items 1 and 2), anxiety (EPDS items 3-6), and depression (EPDS items 7-10), similar to what others have found 42,43 .
We examined the fit of these three subscales using EPDS data in ALSPAC (Figure 6) for mothers and partners at 18 weeks' gestation and 11 years 2 months post-partum. Model fit statistics indicated generally good fit at both timepoints in both mothers and partners (Table 11). Correlations between the three subscales within and between parents at 18 weeks' gestation are presented in Table 12. Associations were strongest within rather than between parents.

Related constructs in ALSPAC
Researchers using the EPDS in ALSPAC may want to take advantage of the wealth of information collected on related constructs. In addition to EPDS data in partners at 10 timepoints and in offspring, measures of anxiety, personality, nutrition, IQ, and biological characteristics such as genetic markers are available in ALSPAC. Users of ALSPAC data may consult the searchable database for additional constructs which may be of interest.

Considerations for the data
Like all longitudinal studies, missing data due to attrition must be taken into consideration when using ALSPAC data. However, due to the rich nature of information related to missingness in ALSPAC, users will be able to deal with missing data in a competent manner. Table 5, Table 6 list several of the potential characteristics which users can use to account for missing EPDS data. Users should take steps to handle missing data appropriately.

Figure 3. Histograms for the number of Edinburgh Postnatal Depression Scale (EPDS) symptoms in partners at each of the ten occasions in ALSPAC.
Strengths and limitations of the data A major strength of the ALSPAC data is the number of participants and that it is a general population sample. Another strength is the number of time points at which the entire EPDS was administered, thus providing identical assessments over time, and that it was administered not only to mothers but also to their partners. ALSPAC is therefore a rich resource for further exploration of questions related to gender differences in the presentation of depression. One possibility for future research with the ALSPAC data is to investigate whether there are gender differences in the depression symptom profiles of men and women 49 . Some research indicates that when men are classified as being depressed by including symptoms such as risk-taking behaviour, anger attacks, and substance abuse, they are just as likely to meet criteria for depression as women 50 .
A third advantage is that the offspring of the original cohort are now starting to have their own children, and these mothers, as well as their partners, are also being administered the EPDS. Two generations of EPDS data in mothers and partners are therefore available, enabling researchers to answer important questions about intergenerational depression. The data can also be linked to the wide range of other information already collected on mothers, partners and their children.
One limitation of the data is that the EPDS was not designed to measure information on the duration or intensity of depressive symptoms 12 . A further limitation is the lack of consistency in the timing of measurement in mother and partner depressive symptoms in late childhood and adolescence. Both mothers and partners were administered the EPDS at child age 11 years 2 months, but the EPDS was not subsequently given to mothers until child age 18 and to partners at child age 21. The timing of these assessments will therefore somewhat limit questions which can be asked regarding depression in parents of adolescents. Another limitation of the ALSPAC data concerns racial and ethnic diversity. There are insufficient numbers of non-white participants to enable sub-group analyses.

Sarah K. G. Jensen
Boston College School of Social Work, Chestnut Hill, MA, USA This is a well-written and thorough data note that will be helpful for a researcher who wishes to make use of the rich data on maternal and paternal depression within ALSPAC.
Here are a few suggestions: Table 1: Consider adding the scores to the response categories.

○
Description of the population section: I would move tables 6 and 7 to the start of the section. These tables are very relevant to someone who is determining whether or not this dataset can be used to explore their hypothesis. More information about partners may be helpful. E.g., were all partners male? ○   Tables 2 and 3 are very detailed and only relevant to someone who has access to the data set. These do not have to be included here, but may instead be provided as an appendix. It is fine to mention that factor analysis has been conducted to explore subscales/ dimensions within the EPDS. ALSPAC given the many waves of assessment provides an excellent opportunity to explore whether these data patterns consist over time, which could be informative but possibly too much to include here.

○
When mentioning the CFA, it may also be relevant to mention Dr. Ted Barker's work characterizing longitudinal latent classes (LLC) of maternal depression from pregnancy to 33 months (Barker, 2013 1 ) in ALSPAC. He identified three classes of mothers, namely low risk, medium risk, possible clinical depression (10%).
○ It may not be feasible, but it would be helpful to list all previous publications that have used EPDS in ALSPAC. ○ set. These do not have to be included here but may instead be provided as an appendix. We agree with this point, but unfortunately appendices are not possible with Wellcome data notes. ○ Figures 2 and 3 are very detailed and maybe not necessary. These response patterns are to be expected and can be described briefly in the text.
We have decided to keep these, as some readers may like to see visual representations of symptom patterns.
○ I find the unit in Figures 4 and 5 somewhat confusing. How about the number of mothers/ partners who reported each symptom?
We agree that this needed clarifying. Rather than dichotomising answers, we present the mean on a 0-3 scale for each item and have clarified this in the text.

○
It is fine to mention that factor analysis has been conducted to explore subscales/ dimensions within the EPDS. ALSPAC given the many waves of assessment provides an excellent opportunity to explore whether these data patterns consist over time, which could be informative but possibly too much to include here. When mentioning the CFA, it may also be relevant to mention Dr. Ted Barker's work characterizing longitudinal latent classes (LLC) of maternal depression from pregnancy to 33 months (Barker, 20131) in ALSPAC. He identified three classes of mothers, namely low risk, medium risk, possible clinical depression (10%). We appreciate this suggestion and citation but feel it would be beyond the scope of this data note to include this. ○ It may not be feasible, but it would be helpful to list all previous publications that have used EPDS in ALSPAC. This is an excellent idea. We have cited more previous publications and directed readers to the ALSPAC publications database.