Socio-economic Factors and Postnatal Depression Relationship

(a) Objectives of the project and any related information

The aim of his project is to study the relationship of socio-economic factors with postnatal depression in Spanish mothers. This will be done both at individual and area–based level. The main hypothesis of the research is that unemployed mothers, with low education and low income have higher risk of developing postpartum depression. Besides a geographical comparison among four different areas within the Spanish cities of Barcelona, Bilbao, Madrid and Seville will be done. An area-based deprivation index will be used for testing the complementary second hypothesis of the study which is that the communities more deprived have higher prevalence of postpartum depression than the less deprived ones.

(b) Work which has led up to the project

Postpartum depression is one the most common disorders suffered from mothers within the first 12 months after childbirth. Several studies places its average prevalence around 10-15% (24) and needs to be considered as a public health problem that can affect, besides to the mother and to her environment, to the emotional development and well-being of the children. Postpartum depressions also needs to be differenced from the baby-blues and the puerperal psychosis, a more severe type of depression. The baby blues is mainly caused for the hormonal alterations and, although might have the same impact on the mood as a depression, the symptoms normally disappear within two weeks after giving birth without any treatment. The puerperal psychosis affects on average to a 0.1 – 0.2 % (24) of mothers and hospitalisation is usually required. The postpartum depression and can last several weeks or months and, if not treated, can lead to a chronic recurrent depression.

The most common symptoms of the postpartum depression are sadness, emptiness, exhaustion, low energy, feeling incapable of taking care of the baby, guiltiness. The signs are similar to any other depression disorder, but with a special focus on the life changes and relationship with the new born.

Regarding the causes of the there are many research that have studied the predictors or risk factors for developing a postpartum depression, and based on two existing literature reviews on the topic (22) (24) the main predictors of postpartum depression could be categorised as follows:

• Physical and biological factors: poor physical health, negative body image and bodyweight.
• Psychological factors: antenatal depression, previous psychiatric illness and childcare stress.
• Social factors: low education level, low income, unemployment and social support.

This study will focus on the social factors and within them, the ones related to the socioeconomic status: education level, income and employment. They can lead to unequal rates in postpartum depression that, as socially determined, could be avoidable.

In the past the relationship of socioeconomic status and depression has been underlined in many studies worldwide (10) (18) (22) but in the particular case of Spain no research that take into account these factors and their impact in postnatal depression prevalence have been found.

Spain is one of the European countries that is suffering the most consequences of the global recession that begun in 2007. The economic crisis is having dramatic impact in the labour market, public sector and therefore in population lives. The socioeconomic status is related with higher psychiatric morbidity, but in an economic crisis context, because of the additional uncertainty about the future, the mental health of the population tends to get worse. There are already studies taking place in Spain that are founding increases in mental health problems since 2007, especially in families that are experiencing unemployment (17).

The current unemployment rate in Spain is 23.2% raising until 24.3% in case of women versus 22.2% in men and up to 50.7% in population younger than 25 years old (14). But these rates are not equally geographically distributed. There are Spanish regions that because of their past productive framework are suffering bigger economic struggles.

As said above no studies that relate postnatal depression and socioeconomic factors in Spain are known, that is why this research will test the association between socioeconomic status and postnatal depression at the individual level and then will compare with Spanish areas with unequal deprivation indexes.

On top of this there are studies that encourage to use both the area deprivation index and individual socioeconomic status, as these two measures make independent contributions to the health outcome (28).

Although the results of this study will not be able to be compared with past records on postpartum depression this could be a starting point for further studies of the impact of the crisis on the mothers’ mental health and about its geographical disparities.

(c) Study design and methods to be used in investigating this problem and potential limitations

Design

A longitudinal cohort study will be conducted for this research. Because of the nature of the outcome this is the most appropriate type. The onset of the postpartum depression is within 12 months after birth, and the longer periods of evaluation will predict higher prevalence (24). A single point of collection of data would minimise therefore the results.

Study population and sample

Pregnant women that are 18 years old or older and who are registered in the Spanish maternity services and live in Barcelona, Bilbao, Madrid or Seville will be invited to participate in the study

The exclusion criteria will be individuals with psychiatric illness in the previous year.

The sample size was calculated based on equivalent measures found in existing literature regarding the socioeconomic individual exposures (income, employment status and education) (LITERATURE) and in an area-based deprivation index and their association either with postnatal depression or similar outcomes. The desired power of the sample (90%), the potential non-responders and the loss over the course of the follow up was also considered in the calculations.

The area-based deprivation index that will be used in this study was created in 2001 in Spain (8) in order to identify the socioeconomic conditions of the measured areas. The information needed for feeding the index is available in the National Census Institute (INE) and could be updated with the data of 2014. This index allows to identify the more disadvantaged areas within a city. Although it was associated in its origin with rates of mortality, it was created with the aim of studying wider range of social inequalities in health in Spain.

This area-based deprivation index is created from the following socioeconomic indicators: manual workers, unemployment, temporary workers, total low education, and youth low education.

The geographical units for the composition of the index are the census tracts of the cities of Barcelona, Bilbao, Madrid and Seville. (2.358 in Madrid, 1.491 in Barcelona, 510 in Seville y 288 in Bilbao). The index will be divided in 4 quartiles from the more deprived to less deprived measure. In each city one census tract for each quartile will be selected.

The sample will be selected through multi-stage cluster sampling. The census tract will be the primary sample unit. Then sample of individuals will be selected from a primary care centres where pregnant women living in each one of the tract are registered.

Four primary centres in each city will be selected.

SAMPLE SIZE & THE POPULATION BETTER DEFINED

Data collection

Spain has a universal health system, everyone has the right and free access to it. When a women becomes pregnant it is registered and monitored by her assigned general practitioner, gynaecologist and paediatric medical doctors, during and after her pregnancy, in the primary care centre of her neighbourhood.

Every pregnant women in the centres selected will be invited to participate in the study, with the exclusion criteria of women who had any psychiatric disorder in the previous year.

They will be informed about the study in their first visit to their GP and appointments for filling in the questionnaires during their next visit and during pregnancy will be planned. 3 questionnaires will be used during the 4 interviews scheduled.

During pregnancy: Baseline questionnaire with socio-demographic questions, employment status and type, income, education, marital status, number of children and address of residence.

Three months after delivery: Social support questionnaire and the Edinburgh Postnatal Depression Scale questionnaire

Six months and twelve months after delivery: Edinburgh Postnatal Depression Scale questionnaire

All the questionnaire will be self-reported. The Edinburgh Postnatal Depression Scale is a 10 items questionnaire used to screen postpartum depression. The validated Spanish version will be used (9)

The social support questionnaire is the Spanish abbreviated version (6 items versus 19) of the MOS Social Support Survey (23).

Statistical analysis

The main outcome of the study is postpartum depression defined as a categorical variables derived from the results of the Edinburgh Postnatal Depression Scale. The cut-off point of the validated Spanish version for a positive outcome is 11.

Cases will be considered when women report positive outcome the 3 times of follow-up against women who reported zero, once or twice (non cases).

Main exposures are level of income, education, and employment status (socioeconomic status measures), and area-based deprivation. Other covariates selected for the baseline and social support questionnaires will be included as possible confounders or effect modifiers.

The sample characteristics will be describe through univariate and bivariate statistics. Multivariable logistic regression will be used for testing the association between main exposures and outcomes adjusted for the others covariates.

Initially each main exposure will be modelled individually with the outcome, only age-adjusted. Secondly each exposure it will be adjusted by other covariates, then by covariates and other socioeconomic status exposures and the area-based deprivation. Finally the model will be fully adjusted with all exposures and covariates together.

The statistical software STATA will be used.

Other

Potential limitations

As in all the longitudinal studies there is the risk of loss during the follow–up. This is already considered in the calculation of the sample size.

The self-reported questionnaires could lead to the common limitations of these types of tools: response bias, the restrictive nature of the scale-based questionnaires, understanding, lack of introspective ability etc.

The social support questionnaire is a reduced version because this study wants to focus in the socioeconomic risk factors of postpartum depression. It was included because social support is considered also an important predictor of postpartum depression. More extensive version could be included in future studies.

Also, further analysis that include structured interviews to measure the outcome could be performed. However the positive results of this questionnaires for finding significant associations it is validated by multiple previous studies (CITATION).

Study organisation

The principal applicant is the main coordinator of the study, has extensive experience in social epidemiology and is specialised in socioeconomic determinants on health. It is also a lecturer in statistic in for medical science and will be responsible of the data analysis.

The co-applicant is a UCL member of the social epidemiology department and a visiting lecturer of the Basque Public University (UPV) in Spain. It will be responsible of the coordination and communication with the Spanish team.

The local co-applicant was a member of the research group who developed the area-based deprivation index used on this study and a professor on social epidemiology in the UPV. It will coordinate the Spanish team who will conduct the field work.

The research assistants will conduct the field work and the logistics and communications with the primary care centres.

One research assistant will be recruited in each city.

(d) Timetable using Gantt chart or similar diagram

(e) Ethical issues

All participants will be informed and will need to sign a written consent prior to any analysis of the data.

All the data will be anonymous and treated confidentially following the current Spanish and UK laws of Protection of Data.

Ethical approval will be submitted to the UCL and the UPV.

I am still a bit confused with sample calculation:

For example in the paper below, that is measuring social support and PPD as a binary outcome.Which effect should I focused in? If I calculate the sample size from it, would I alsoneed to use in my study the same questionnaire they are using in this paper and same follow up time?

http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4390689/

ANNE – here is the useful info that I found in the paper (copied and pasted direct from paper)

• ​Incidence of depression was calculated in women who were not depressed at baseline based on proportions of new cases at follow-up in that sample.
• 55 of the 386 without antenatal depression had depression at follow-up, indicating case incidence of 13.9%.
• table 1 – unadjusted association between education/income and incidence of depression (ie new events). So suggests OR=0.49 (low versus high ie high –> low OR=2) and even steeper for income
• But these are unadjusted so after adjusted the measure of effect would probably be attenuated ie smaller.
• If there is no better data, then you could use this, for examplelow educ vs high educsampsi0.174 0.093, p(0.9) but see if you would have power to look at low –> middle education etc.
• For income as above using numbers from table 1

And for the sample calculation of area based deprivation and PPD, I could use papers of association between income inequality and PPD? OR What about this one?

http://www.ncbi.nlm.nih.gov/pubmed/24392759

The undjusted results are:

low-SES community 26.2% (104/397) had depression, compared with 14.8% (24/162) high-SES community

If I do calculation in STATA

sampsi 0.26 0.14, p(0.8)

My sample size would be for each group

N1= 190

N2= 190

But when I use my are based deprivation index I might use different percentiles to categorise lower and higher deprived areas (four at least) What would be the sample size in this case for each percentile?

ANNE – if you use quartiles for deprivation, then you would need to consider not just low –> high, but as for education low –> middle, then middle –> high, and high –> higher.

If incidence is 26% in highest deprivation and14% in lowest, then if you think the association is linear, then you can estimate inicidence in intermediate groups e.g. 26, 22, 18, 14%.

So you need to choose the sample size for these

e.g. sampsi 0.26 0.22, p(0.9) etc.

After all these sample size calculations, choose the largest. Then in your proposal just report that you based sample size on the sample size per group needed to find the smallest difference between SE groups. remember if your sample size calculation says 3200 per group, and you have 4 groups, then your sample size with be 3200×4. You will also need to include extra in the sample because there will be non-responders eg 40%. Also maybe 20% loss over the course of your follow up.

For example, if number per group is 3200, and 4 groups possible, and 60% response and 20% loss during follow-up,then you will need (3200 x 4) / (0.6 x 0.2).

You also asked if you need to use the same measures as the paper uses if you use if for sample size calculation. As long as you state thatyour measures are comparableit is okay.

Q10 REFERENCES