Screening for depression in women during pregnancy or the first year postpartum and in the general adult population: a protocol for two systematic reviews to update a guideline of the Canadian Task Force on Preventive Health Care

Prevalence

Depression is the most common cause of disability worldwide, with over 300 million people now living with depression, an increase of more than 18% between 2005 and 2015 [6]. Estimates of prevalence for depression vary by characteristics such as age and sex. For example, women are more likely to suffer from major depressive disorders than men [7, 8]. Many studies report depression rates based on results from self-reported screening questionnaires, rather than validated diagnostic interviews, but this is known to exaggerate rates substantially and to blur distinctions between low- and high-prevalence groups [9]. The 2012 Canadian Community Health Survey-Mental Health used the diagnostic interview technique among 25,113 individuals and reported annual prevalence for major depressive disorder (MDD) of 3.9% (95% CI 3.5–4.2%) and lifetime prevalence of 9.9% (95% CI 9.3–10.5%) [10]. It also reported an annual and lifetime prevalence of MDE among Canadians at 4.7% and 11.3%, respectively [10]. Another 2012 Canadian national health survey reported the highest rate of a MDE was among 15–24-year-olds, with 7% having had depression in the past year, compared to 5% in people aged 25–64 years, and 2% in those 65 years and older [11].

Risk factors

There are several risk factors that have been associated with depression in adults. Socio-demographic risk factors include sex, age, marital status, low socioeconomic status, and low education level [8, 12–14]. In Canada, the largest difference between sexes is in the 15–24 age range, with the difference diminishing and nearly disappearing at more advanced ages [15]. Additionally, married and never-married individuals experience less depression than those who are separated, divorced, and widowed [8]. Other factors such as trauma early in life (e.g. neglect or sexual abuse), chronic disease (e.g. cancer, cardiovascular disease), previous history of depression, and a family history of depression have also been linked to depression [8, 16, 17].

Consequence of depression

Depression affects a person’s physical health and well-being and impacts psychosocial functioning (e.g. personal relationships, employment). A review by Evans et al. [18] conclude that there may be a bidirectional link between depression and disease, as depression might be an etiologic factor for new disease (e.g. stroke) and also might affect the course of existing chronic diseases such as diabetes mellitus. Depression can affect work performance through absenteeism and presenteeism (decreased work productivity while at work), which is a large cost to employers in terms of productivity [2]. In addition, many depressed individuals are unable to enter the workforce. On a population level, it also has a large societal impact through increased health service utilization, decreased work productivity, increased burden on family members, and increased resource costs related to disability [19]. In the 2003 Canadian Community Health Survey, the total economic burden of mental illness (including health service utilization, long- and short-term work loss, and health-related quality of life) was said to be $51 billion dollars [19]. More recently, direct healthcare costs associated with MDD were determined using a population-based cohort study in Ontario, Canada. The age- and sex-adjusted annual per-capita cost among those with MDD was higher than the comparison group (those without MDD or psychological distress) [$3914 (95%CI $2943–4888) vs $3206 (95%CI $2820–3591)], and the population-wide excess cost for those with MDD was $256 million (prices converted to CDN $ from reported USD) [20].

Although effective interventions to reduce the effects of depression exist, individuals need to be identified to benefit from these interventions. The Mental Health Commission of Canada reports that almost half of those who feel they have suffered from depression or anxiety have not seen a doctor about this problem [21]. In addition, among those who have been diagnosed accurately, many do not receive minimally adequate treatment [21, 22].

Current recommendations

In 2013, the Canadian Task Force on Preventive Health Care (CTFPHC) recommended to not routinely screen for depression (this was based on very-low-quality evidence; see Additional file 2). There is disagreement in recommendations between Canada, the USA, and the UK. Neither the CTFPHC nor the United Kingdom National Screening Committee (UKNSC) recommended screening, whereas the US Preventive Services Task Force (USPFTF) recommended screening based on prioritization of linked evidence of effective follow-up and treatment of screen-identified individuals. Additional file 2 provides additional on how the USPSTF recommendation differs from Canada and the UK, followed by some speculation as to why [23].

Prevalence

Although estimates of the prevalence of major depression should be based on validated diagnostic interviews, many studies report depression rates based on results from self-reported symptom questionnaires and other non-valid methods [9]. An Agency for Healthcare Research and Quality SR reported that the period prevalence of major depression during pregnancy (conception to birth) was 12.7% (95%CI 7.1–20.4%) [24]. However, this is based on one primary study. The period prevalence from birth to 3 months postpartum was 7.1% (4.1–11.7%) [24]. A recent US study in which women were interviewed, and diagnosis made using the DSM-IV criteria, found the 12-month period prevalence of MDD to be 8.4% among women who were currently pregnant or had been pregnant in the past 12 months, 9.3% among postpartum women, and 8.1% among non-pregnant women [25]. It should be noted that the prevalence for postpartum women could include time in which they were pregnant, as it covers the previous 12 months.

Risk factors

There are many risk factors for depression during pregnancy, including younger age, a history of depression, exposure to domestic violence, increased life stressors, a lack of social support, unintended pregnancy, lower income, lower education, smoking, single status, and poor relationship quality [26, 27]. Prior depression is the greatest risk factor for postpartum depression. Nevertheless, for women who experience postpartum depression, it is a first episode among 40% [28]. Other postpartum risk factors include untreated depression or anxiety during pregnancy, experiencing a stressful life event during pregnancy, having a traumatic birth experience, preterm birth or infant admission to neonatal intensive care, low levels of social or partner support, experiencing domestic violence, low socioeconomic status, obstetric complications, low birth weight, and breastfeeding problems [27].

Consequence of pregnancy and/or postpartum depression

While the prevalence of depression in women during pregnancy and the first year postpartum may be similar to that for other women [24], depression has specific negative short- and long-term effects on maternal health, child health and development, and on the overall health of families [29]. Depression during pregnancy is associated with unhealthy behaviours including poor self-care, poor nutrition, increased use of tobacco and alcohol, lower prenatal care seeking, and poorer maternal-fetal bonding [30, 31]. Postpartum depression may lead to difficulties with infant care, a decrease in breastfeeding initiation, and poor-quality mother-child interactions including mutual touching, smiling, and vocalizations, and compromised mother-child bonding [30, 32, 33]. Negative outcomes for infants in mothers with prenatal and postpartum depression may also include preterm delivery, lower birth weight, cognitive, emotional, social, neural functioning or developmental delay [34–37].

Almost half of Canadians with depression have not seen a primary care provider about their depression [38]; for depression in pregnancy and postpartum, the number may be even higher [35]. Screening for depression, if effective, would allow for treatment among women who would not otherwise be identified and possibly lessen the negative impacts to the mother, fetus/infant, and family. Several treatment options exist, including psychosocial strategies (e.g. peer support, non-directive counselling and self-care such as exercise), psychological therapies, and antidepressant medications [36, 37]. The last poses the additional challenge of considering the safety of exposure to psychotropic medications to the baby in utero and through breast milk [39].

Current practice and recommendations

Across Canada, there is a lack of consensus on how and when prenatal and postpartum depression screening should occur with different provinces and territories having different approaches. Additional file 2 provides examples on how the provinces of Ontario, British Columbia, Alberta, Nova Scotia, and the territory of Nunavut screen women during pregnancy and postpartum. There is discordance in recommendations between Canada, the USA, and the UK. Neither the CTFPHC or UKNSC recommended screening in contrast to the USPFTF recommendation for screening on results that combined screening with treatment. Additional file 2 provides additional details on why the USPSTF recommendation may differ from Canada and the UK [23].

General adult population

Pregnant and postpartum population

General adult review

The general adult review will update the SR used by the CTFPHC for their previous guideline recommendation [41, 42]. A comprehensive search strategy was developed using the previous SR search strategy as guidance. The search will start from the last search date of the previous review (May 2012). Strategies will utilize a combination of controlled vocabulary (e.g. “Depressive Disorder”, “Mass Screening”, “Adult”) and keywords (e.g. “depression”, “screening”, “adults”) (see Additional file 4 for the OVID multifile search).

Pregnant and postpartum review

There will be no date restriction in the search. Strategies use a combination of controlled vocabulary (e.g. “Depressive Disorder”, “Mass Screening”, “Pregnancy Complications”) and keywords (e.g. “depression”, “screening”, “pregnancy”) (see Additional file 5 for the OVID multifile search).

For both SRs, we will search the grey literature for unpublished documents (e.g. reports, theses, governmental publications) following the Canadian Agency for Drugs and Technologies in Health (CADTH) Grey Matters checklist. The CADTH checklist includes national and international health technology assessment agencies, clinical practice guideline organizations, clinical trials registries, Canadian health prevalence and incidence databases, statistics, search engines, and additional databases. The clinical trial registries listed within the checklist include ClinicalTrials.gov, WHO International Clinical Trials Registry Platform, ISRCTN Registry, CenterWatch, and Clinical Trials Registry India. We will supplement the CADTH checklist by searching the websites of the following organizations: the College of Family Physicians, the American College of Physicians, the American Academy of Family Physicians, the Canadian Nurses Association, the American Nurses Association, the Canadian Psychiatric Association, the Centre for Addiction and Mental Health, the Anxiety and Depression Association of America, and the American Psychological Association. Additionally, for the pregnant and postpartum review, we will search the following websites: the Society of Obstetricians and Gynaecologists of Canada, the American College of Obstetricians and Gynecologists, the Royal College of Obstetricians and Gynaecologists, and the Royal Australian and New Zealand College of Obstetricians and Gynaecologists, and the Canadian Association of Midwives.

Grey literature searching will be confined to what can be accomplished within 1 week of searching by one person, for pragmatism, and will be restricted to English and French language documents.

Meta-analysis

We will determine whether clinical and methodological heterogeneity exists among studies, prior to performing a meta-analysis. If it is determined to be appropriate, based on clinical similarity between studies and that the body of evidence is not at high risk of bias, data will be meta-analysed using random effects models for effect measures such as risk ratios and risk differences. If it is determined that meta-analysis is not appropriate, the range of effects will be presented. For time-to-event data, the hazard ratio will be pooled using the generic inverse variance method.

Unit of analysis errors can occur in studies that employ a cluster design (e.g. a clinical practice) and yet are analysed at the individual level (e.g. patients), potentially leading to artificially precise results and contributing more weight than would be appropriate in a meta-analysis [60]. If empirically derived intra-cluster correlation coefficients are available, we will adjust the analysis to address these errors [61]. For multiple events that may occur in one person (e.g. suicide attempt), we will assume each event represents a unique individual, unless data are presented as events per individual. If we were to encounter a study where there is reason for concern that many events are recorded in a small percentage of research participants, the impact of this study could be evaluated in a sensitivity analysis.

Sparse binary data and studies with zero events

When studies report rare events, a synthesis will be done narratively. For those outcomes (e.g. suicide completion) where at least one intervention group contains zero events, only the risk difference (RD) will be used. For calculating the RD, we will use the median baseline risk for the control group in the included studies, although we may additionally perform sensitivity analyses using differing baseline risks if thought suitable.

Statistical heterogeneity

The Cochran’s Q and I² statistic will be used to assess the statistical heterogeneity of effect estimates among included studies. For the interpretation of I², a rough guide of low (0–25%), moderate (25–50%), substantial (50–75%), and considerable (≥ 75%) will be used [62, 63]. Should considerable statistical heterogeneity exist, we will present all studies in a forest plot, but will not provide the pooled estimate. When the body of evidence is statistically heterogeneous, we will conduct subgroup, sensitivity analysis, and/or meta-regression analyses, where the optimal approach for each variable will be determined once we see how data are reported in studies. We will follow previously published guidance for meta-regression [64]. Meta-regression will be based on random effects models to allow for residual unexplained heterogeneity. In consideration of the low power of the Q statistic, when the number of studies is small and the possibility of detecting unimportant heterogeneity when the number of studies is large, a p value < 0.10 will characterize statistical significance [64]. When the sizes of the included studies are moderate or large, there should be at least 10 studies for a continuous study-level variable. For a categorical subgroup variable, each subgroup should have a minimum of four studies. These numbers serve as the lower bounds for considering meta-regression [64]. When included studies are mostly small in size, univariate meta-regression will be used when an insufficient number of studies are available to conduct multivariable analyses. We will not pool outcomes if there is an I² of > 75%. We will use the p value from the chi-square test as support to interpret the strength of evidence for heterogeneity.

Subgroup analysis

Sensitivity analyses

Sensitivity analyses may be undertaken to restrict analyses to those studies assessed as being of low ROB, based on the overall judgement, and may also be performed to address any decisions made regarding handling of data or to explore statistical heterogeneity. A sensitivity analysis may also be performed on the timing of publication, based on cut-offs as determined by literature and any other design-specific issues we may come across.

Small study effects

If there is a minimum of 10 studies in any meta-analysis, we will assess for small study effects using a combination of graphical aids (e.g. funnel plot) and/or statistical tests (e.g. Egger regression test, Hedges-Olkin) [63]. Funnel plot asymmetry can be used to identify potential bias, as well as signal exaggeration of treatment effects in small studies [65].

Software

The Cochrane Review Manager software version 5.3 will be used to calculate effect estimates and conduct meta-analyses [66]. For all analyses not possible in RevMan v5.3, we will use Comprehensive Meta-Analysis v3.

Grading the quality of evidence and interpretation

We will assess the quality of evidence for individual comparisons and outcomes using the GRADE approach. GRADE tables will be prepared for each of the critical and important outcomes using the GRADE framework to assess each domain (i.e., risk of bias, imprecision, inconsistency, indirectness, and publication bias) [44, 45]. This will be performed by one reviewer. Verification will be completed by a second reviewer. Disagreements will be resolved by consensus or third-party adjudication.