A cross-sectional survey in the form of an online and paper-based questionnaire was carried out between February and August 2017. Our target sample was based on existing data that suggested up to 20% regional variation in the rates of admission to perinatal services (14). As this was an upper limit, we adopted a more conservative criterion of 15% difference between rural (lower) and urban (higher) groups. Our sample size calculation indicated that 348 participants would be needed to detect such difference in the odds of being at risk for mental illness, assuming a 5% alpha level, a 90% power, and a rural-urban ratio of one to one.
This study was given ethical approval by the Newcastle & North Tyneside 1 Research Ethics Committee (212364) and by the governance framework of the children and young people’s service in the county council area where the study took place.
Participants were eligible if they were a) women aged 16 years or above, b) pregnant (at any stage of gestation) or who had had a baby in the last 12 months prior to enrolment into the study, c) able to provide informed consent in English. Due to limited resources, it was not possible to translate materials into other languages. This work was undertaken as part of a wider project aimed at exploring the needs and experiences of women in rural communities in the area, relating to their mental health during and after pregnancy, which involved both quantitative and qualitative research.
Some participants were informed about the study by their health visitor antenatally (28-34 weeks’ gestation) or postnatally (6 weeks post-delivery). In the UK, health visitors are qualified nurses or midwives working with antenatal and postnatal families offering support and advice, both through home visits and at medical centres(15). Health visitors provided women with the study participant information sheet. Those who choose to take part and gave informed consent completed a questionnaire either during or after their health visitor’s home visit, either online or on paper. Completed paper questionnaires were returned to the research team in a sealed enveloped through the health visitor or by post. Other women were recruited at the 20-week antenatal scan appointment, or at postnatal appointments, at one of three participating maternity hospitals in the region, or through social media (e.g. Facebook), local parenting groups, community midwives, or word of mouth. Participants recruited through these secondary methods received a flyer with information about the study and on how to access the online questionnaire. Online questionnaires were completed by participants after providing informed consent. All women had the option to contact the research team for any questions.
The questionnaire (Additional file 1) was piloted for clarity in a sample of 10 women. Questions collected information on the exposure variable (rural vs urban areas), outcome variables (depression and anxiety) and covariates (sociodemographic information and social support). All variables were based on self-report, except rurality and socioeconomic deprivation.
The exposure variable was the rural or urban status of the area in which participants lived. Participants were requested to provide their postcode, which was then used to determine their rural-urban status according to the following official 10-category classification (16): 1) (most urban) urban major conurbation; 2) urban minor conurbation; 3) urban city and town; 4) urban city and town in a sparse setting; 5) rural town and fringe; 6) rural town and fringe in a sparse setting; 7) rural village; 8) rural village in a sparse setting; 9) rural hamlets and isolated dwellings; 10) (most rural) rural hamlets and isolated dwellings in a sparse setting. In this classification, a broader binary distinction between urban and rural areas is possible: an urban area is any between category 1 and category 4; a rural area is any between category 5 and category 10, which refer to settlements below 10,000 people or which are open countryside. The two-level broad classification was used for the main analyses, as any analysis based on a larger number of rural-urban categories would have resulted in small numbers in each category considering our sample size, which would have reduced the power of the analyses. An exploratory analysis was conducted that compared women in the two most rural categories to the two most urban categories to get the clearest picture of the difference between rural and urban status.
Markers of mental health
We used the Edinburgh Postnatal Depression Scale (EPDS) and the Whooley questions to measure risk of depression, and the 2-item General Anxiety Disorder Scale (GAD-2) to measure risk of anxiety. As part of routine care in the UK, these measures are used by healthcare professionals for screening for mental illnesses in perinatal women (17,18).
The Edinburgh Postnatal Depression Scale (EPDS) has been used extensively both in antenatal and postnatal care (19), and has shown to be a valid and reliable tool across a large number of populations and contexts (20). The scale comprises 10 items related to maternal feelings during the past seven days, assessing depressed mood, guilt, anxiety, and suicidal ideation (e.g. “I have felt sad or miserable”). Items are given a score of 0 (lowest frequency or intensity) to 3 (highest frequency or intensity) and then summed to provide a total score, ranging from 0 to 30. A higher score reflects a higher level of depression. The recommended cut-off points for ‘probable major depression’ is 15 (for antenatal women) and 13 (for postnatal women); for ‘at least probable minor depression’, the threshold is 13 (for antenatal women) and 10 (for postnatal women)(19). The latter criterion was adopted in the present study as we did not diagnose or refer women to clinical services, but rather assess how many women were at risk of (minor) depression. This also allowed us to maximise the number of participants in the analysis so that clearer conclusions could be drawn.
The Whooley questions consist of two ‘yes’ or ‘no’ questions about depressed mood (e.g. “During the past month, have you often been bothered by feeling down, depressed or hopeless?”)(21). Answering yes to one or both questions indicates a positive screen. Common in routine clinical work, the Whooley questions have high sensitivity in detecting depression and should be used as a pre-diagnostic tool, in combination with other screening tools (22).
The 2-item General Anxiety Disorder Scale (GAD-2) is used to assess the frequency of symptoms of anxiety (e.g. “Over the last 2 weeks, how often have you been bothered by feeling nervous, anxious, or on edge?”)(23). Respondents indicate how frequent the symptoms are on a Likert-scale from 0 (Not at all) to 3 (Nearly every day). This short scale is well-established in primary care and has shown high specificity to detect a range of anxiety disorders. Scores can vary from 0 to 6; a score of 3 or more indicates a positive screen (23).
We report these three measures separately, as well as combined into a binary measure, i.e., testing positive on one or more of the three measures. As none of the three tools used was intended to be diagnostic, we refer to risk of depression or anxiety in this study, rather than to rates of depression or anxiety.
Data on two covariate variables - socioeconomic status and social support – were collected as these could potentially explain the association between rurality and outcomes, according to previous research(3,9). Socioeconomic status was based on the index of multiple deprivation (IMD) decile, a common indicator of affluence in the UK, which was determined by searching participants’ postcodes on an online tool(24).
Perceived social support was measured by the Multidimensional Scale of Perceived Social Support (MSPSS)(25) and considered as a potential mediating factor on the causal pathway between rurality and perinatal anxiety and depression. This scale consists of 12 statements about the support received from family (4 items), friends (4 items) and a significant other (4 items), e.g. ‘My family really tries to help me’. Participants rated their level of agreement with each statement on a seven-item Likert scale. Item scores were summed to provide total scores, both overall (range 12-84) and for each of the three subscales (range 4-28). Higher scores indicated perception of greater social support. The factorial validity and internal reliability of the MSPSS have been demonstrated in a number of studies, with alpha scores from 0.87 to 0.93 (25–27), including in pregnant women (alpha 0.92)(28).
Data on other variables was also collected, but these were only controlled for in the analysis if they differed between rural and urban groups. Those variables were age, perinatal stage (antenatal or postnatal), ethnicity, education, employment and relationship status. The potential role of these factors in maternal mental health outcomes has been reported in previous studies(29,30). Questions were adapted from previous Census surveys(31). Additional questions asked participants about healthcare and community servicesincluding mental health support.
Questionnaire data were entered directly by participants using an online survey platform(32); data from returned paper-based questionnaires were entered on to the platform by the research team. All analysis was completed in Stata 14(33). For overall scores consisting of sums, participants with one or more items missing were excluded as required by the instrument guidelines. Descriptive statistics and difference tests were performed to characterise the sample. When data were not normally distributed, the median and interquartile range (IQR) were reported.
There was a statistically significant difference in terms of perinatal stage between rural and urban groups; this was the only variable that differed significantly between the two groups. Consequently, perinatal stage was adjusted for in the main analysis (i.e., in addition to socioeconomic status and social support, which were purposively selected).
Multiple logistic regression models were developed in three steps to assess: a) the unadjusted association between rurality and depression and/or anxiety, b) this association adjusting for socioeconomic status and perinatal stage and c) the effect of social support on this adjusted association. The three-step process was used for each of the three outcomes independently. For each model, we reported the odds ratio for depression and/or anxiety associated with living in a rural area, together with the 95% confidence interval and significance level.