Eligibility and inclusion in the HOPE cohort

Healthcare nurse service is universal in Denmark and free of charge, with several visits during the initial eight months postpartum [21]. Screening for PPD is frequently completed at the visit eight weeks postpartum, most often using the EPDS [21]. Permission from the Danish Patient Safety Authority was obtained to receive these questionnaires for the period January 1, 2015, to December 31, 2021, and from the Danish Data Protection Agency to establish a research project linked with register data hosted by Statistics Denmark. Permission from the Danish Patient Safety Authority was granted for the 88 out of 98 municipalities who use a specific software system, NOVAX, for storage of their healthcare nurse data, on the condition that the municipalities gave consent to provide data to the cohort.

The leading healthcare nurses in the 88 eligible Danish municipalities were contacted by email by the corresponding author, MMZK, followed by one or several phone calls from October 2021 through March 2022, to obtain consent to access their EPDS data. Of the 88 municipalities, 67 (76%) gave consent to provide data for the cohort (Fig. 1). Of the 21 municipalities not participating, 10 declined, 9 did not respond to the request, and 2 did not store EPDS data in NOVAX, after all.

Fig. 1

Map of the participating, non-participating, and non-eligible Danish municipalities. Blue = participating municipalities, red = non-participating municipalities, grey = non-eligible municipalities

A dataset comprising EPDS data from the 67 included municipalities was transferred from NOVAX to a secure server at Aarhus University, representing 198,021 unique mothers or fathers (241,422 questionnaires). All Danish residents are assigned a unique 10-digit Danish Civil Registration number at birth or immigration, and using the Danish Civil Registration System, data were merged with the personal ID numbers on Statistics Denmark [22]. To account for variation in assessment time points and potential delays with electronic registration of the EPDS, we restricted to questionnaires conducted within 12 weeks postpartum. Consequently, we excluded 11,505 mothers or fathers (20,793 questionnaires) due to either an unidentified or invalid ID number or not having a registered birth within 12 weeks of filling in the questionnaire. Further, 43,107 unique persons (46,866 questionnaires) were excluded for the present study, as the respondent was registered as the father of the child. This subset of data will be investigated in detail in a separate project. Of the remaining 143,409 unique mothers (173,763 questionnaires), a further 614 (3,545 questionnaires) were excluded as we restricted to questionnaires being complete, containing exactly 10 questions, and limited to only one unique questionnaire within 12 weeks after childbirth. Consequently, our final sample comprised 142,795 unique mothers of 170,218 unique childbirths, and thus 170,218 unique EPDS questionnaires, all of whom were included in the HOPE cohort (Fig. 2).

Fig. 2

Flowchart of the HOPE cohort

Data in the HOPE cohort

Using the unique ID numbers for each participant in the Danish Civil Registration System, each mother with an EPDS screening included in the HOPE cohort was linked with information from several registers through Statistics Denmark [22]. An overview of these registers and accompanying information is presented in Table 1.

Table 1 Overview of data in the HOPE cohortFollow-up

Using the personal ID number, continuous future updates of the register data are planned, with the potential of adding supplemental and prolonged follow-up information.

Data access

Data are stored on a secure server at Statistics Denmark, and restrictions apply to availability of this dataset. Access can potentially be granted through collaboration with the HOPE research group and will be considered on a case-by-case evaluation by members of a steering group. Future studies conducted on the HOPE cohort must focus on perinatally related topics due to the permission data relies on.

Characteristics of the HOPE cohort

We compared characteristics in the HOPE cohort to the source population. Since inclusion into the HOPE cohort was restricted to women with an EPDS screening conducted within 12 weeks postpartum from January 1, 2015, to December 31, 2021, the cohort members gave birth from October 2014 through December 2021. Accordingly, the source population was defined as all births by women residing in Denmark during the same period.

Variables included in the comparison were obtained, defined, and categorized as follows: Information on PPD diagnosis was retrieved from the National Patient Registry and the National Prescription Register, defined as either an in- or outpatient contact to a specialized psychiatric treatment facility with a depressive episode (International Classification of Diseases, version 10 (ICD-10) codes: F32-33) or a redeemed antidepressant prescription (Anatomical Therapeutic Chemical (ATC): N06A) within 6 months postpartum, to account for delay in PPD treatment, and categorized as yes or no [23, 25]. From the Civil Registration System, we obtained information on maternal age at delivery categorized as < 20 years, 20–24 years, 25–29 years, 30–34 years, 35–39 years, 40–44 years, and ≥ 45 years, calendar year of delivery categorized separately for each year from 2014 to 2021, and country of origin categorised as Denmark or foreign/unknown [22]. Cohabitation status was derived from the Population Statistics Register and coded as cohabitating or not cohabitating. Information on the highest attained maternal education was obtained from Statistics Denmark’s Register of Education and categorized as mandatory, short, medium, and high according to the International Classification of Education [30, 33]. The Medical Birth Register informed on parity categorized as 1, 2, 3, and ≥ 4 children, and singleton births categorized as yes or no [26]. The Obstetric Comorbidity Index is an index for assessing the risk of morbidity or mortality in obstetrics and has been validated in a Danish context [34, 35]. Information needed to calculate the Obstetric Comorbidity Index was obtained from the Medical Birth Register and the National Patient Register [23, 26]. The index was defined according to Bateman and categorized according to previous work as 0, 1, 2, or ≥ 3 [34, 36, 37]. Information on maternal hospital depression diagnosis within one year prior to delivery was obtained from the National Patient Registry and defined as a depressive episode (ICD-code: F32-33) within one year prior to delivery and categorized as yes or no [23]. Information on maternal antidepressant use within one year prior to delivery was obtained from the National Prescription Register and defined as a redeemed antidepressant medication prescription within one year prior to delivery (ATC-code: N06A) and categorized as yes or no [25]. Personal history of psychiatric disorders was obtained from the National Patient Registry and defined as any in- or outpatient visits to psychiatric facilities with psychiatric diagnosis (F00-F99) from 1995 to delivery in the index mother [23]. Information on family history of psychiatric disorders was obtained from the National Patient Registry and defined as any in- or outpatient visits to psychiatric facilities with psychiatric diagnosis (F00-F99) from 1995 to delivery in the index mothers’ parents [23]. Information on gestational diabetes mellitus was obtained from the Medical Birth Register and defined according to ICD-10 (ICD-10 code: O24) and categorized as yes or no [26]. Information on macrosomia was obtained from the Medical Birth Register and defined as a birth weight of > 4.0 kg. and categorized as yes or no [26, 38]. Information on smoking during pregnancy was obtained from the Medical Birth Register and categorized as yes or no [26]. Information on babies born small for gestational age (SGA) was obtained from the Medical Birth Register and defined as a birth weight below the 10th percentile restricted to children with a birth weight > 300 g and < 6400 g and born within gestational week 23–45, and categorized as yes or no [26, 39]. Information on maternal pre-pregnancy body mass index (BMI) was obtained from the Medical Birth Register and calculated as weight in kilograms divided by the square of height in meters and defined according to the World Health Organization’s (WHO) pre-defined BMI groups: <18.5, 18.5–24.9, 25-29.9, ≥ 30 restricted to women with a BMI > 12 and < 50 [26, 40]. Information on acute caesarean section (C-section) was obtained from the National Patient Registry and the Medical Birth Register and defined according to ICD-10 (ICD-10 codes: DO821, DO843, DO829, DO842 and procedure codes: KMCA10A, KMCA10E) and categorized as yes or no [23, 26].

Evaluating potential selection bias

A potential limitation for almost all observational cohorts is selection bias. The extensive Danish registries provide an opportunity to examine potential selection bias in the HOPE cohort, as information on the source population is available. We examined if inclusion into the HOPE cohort influenced five well-established perinatal exposure-outcome associations when compared to the source population:

A)

Personal history of psychiatric disorders and risk of PPD identified through diagnosis [4, 41, 42],

B)

Family history of psychiatric disorders and risk of PPD identified through diagnosis [43],

C)

Gestational diabetes mellitus and risk of macrosomia [44],

D)

Smoking during pregnancy and risk of SGA [45], and.

E)

Maternal pre-pregnancy BMI and risk of acute C-section [46].

F)

Variables for the analyses were obtained, defined, and categorized according to variables used for characteristics. Only exception was age since it was used as a continuous variable in the evaluation of selection bias. See Fig. 3 for details on each of the five exposure-outcome associations.

Fig. 3

Details on definitions of exposures and outcomes in analysis A-E

We employed three separate analysis populations for the five analyses: one population for analysis A (analysis population A), one population for analysis B (analysis population B), and a third population for analyses C, D, and E (analysis population CDE).

Analysis population A: Childbirths in the HOPE cohort commenced from October 2014 onwards. Inclusion ended 6 months prior to the end of follow-up to allow for 6 months follow-up of the outcome (PPD). Hence, inclusion ended on June 30, 2021, and follow-up concluded on December 31, 2021. The source population was accordingly defined as all births in Denmark from October 2014 to July 2021. To prevent misclassification of the outcome (PPD), we applied a washout period, excluding all index mothers with a depressive episode (ICD-10: F32-33 or ATC-code: N06A) one year prior to delivery.

Analysis population B: Same population as in analysis population A, but with an extra exclusion of index mothers without information on the index mothers’ mother or father, since family history of psychiatric disorders was of primary interest.

Analysis population CDE: Inclusion was from October 2014 to December 2021 for both the cohort members and the source population. No washout period was applied.

Statistical analysisCharacteristics of the HOPE cohort

We calculated frequencies of selected characteristics in the HOPE cohort and the source population. Prevalence ratios (PR) were then computed by dividing the frequency in the HOPE cohort by the corresponding frequency in the source population, accompanied by 95% confidence intervals (CI). PRs were used to evaluate direction and magnitude of potential selection bias: underrepresentation (PR < 1) or overrepresentation (PR > 1) of specific characteristics among cohort members compared to the source population. Estimation of 95% CIs for PRs was done by calculating standard errors, since the two populations were not independent [47].

Evaluation of selection bias

To estimate each of the five exposure-outcome associations, we utilized logistic regressions applying Generalized Estimating Equations (GEE) methods to account for instances where women were included more than once in the HOPE cohort (e.g., giving birth to more than one child in the period) [48]. Analysis population CDE was further restricted to singletons, as outcomes in analyses C and D focused on birth weight (macrosomia and SGA, respectively). Complete case analysis was conducted within each analysis population. Results were presented as crude and adjusted odds ratios (ORs) with 95% CIs in both the HOPE cohort and the source population. Adjusted analyses included maternal age, parity, and education. Maternal age was included in the model as a continuous variable modelled by a restricted cubic spline, as age was not linear on the log-odds scale. Relative ORs (ROR) were estimated as the ratio between the adjusted OR in the HOPE cohort and the source population. RORs were likewise estimated to evaluate potential selection bias: underestimation (ROR < 1) or overestimation (ROR > 1) of the specific association among cohort members. The 95% CIs for ORs and RORs were calculated following the same procedure as for the calculation of CIs for the PRs [47].

Sensitivity analyses

Two sensitivity analyses were conducted: one restricting analysis population AB to singletons to account for differences between singletons and twins/triplets, and another presenting characteristic of cohort members and the source population restricted to Danish-born women to account for missing information in the Danish registers among non-Danish born women.