Study
Past history of depression or anxiety
Pregnancy, depression or anxiety
General supports
LEs or adversity
Partner support
Neuroticism (personality vulnerability)
IPV in pregnancy
aO’Hara and Swain (1996)
Meta-analysis-
Ns 1000–2000;
variable F/U period
a0.57
a0.75
a−0.63
a0.6
a0.13
a0.4–0.5
aRobertson (2004)
Meta-analysis-
Ns 2000–3000;
variable F/U period
a0.58
a0.75
a > 0.6
‘large ES’
a0.61
a0.39
Milgrom (2008)
Prospective study-
Ns 8000–10,000;
F/U EPDS 6 week p-p
b1.7 (95 % CI 1.39–2.07)
p < .0001
b1.39 (95 % CI 1.12–1.73)
p < .01
Nonsignificant AOR
Daily hassles
b0.79 (95 % CI 0.64–0.97)
p < .05
b0.68 (95 % CI .59–.89)
p < .01
High support vs. moderate
Perfectionism
b1.26 (95 % CI 1.04–1.52)
p < .05
N/A
Howard (2013)
Meta-analysis-
Pooled ORs: Ns 400–500
variable F/U period
N/A
N/A
N/A
N/A
N/A
N/A
c3.1 (95 % CI 2.7–3.6)
Recent reviews support an association between prenatal anxiety, stress and depression and adverse birth outcomes, including preterm and small for gestational age birth (Dunkel Schetter and Tanner 2012; Grote et al. 2010). While many studies do not distinguish between treated and untreated mental illness, evidence based on a recent systematic review suggests that unmedicated prenatal depression increases the risk for preterm birth (in four out of four studies; Davalos et al. 2012) and low birth weight (in four out of four studies; Davalos et al. 2012) compared to non-depressed controls. Comparisons of infant outcomes in medicated versus unmedicated depressed pregnant women are less clear, with studies showing similar increases in risk of preterm and small for gestational age delivery, in both medicated and unmedicated women (Oberlander et al. 2006), and greater risk among unmedicated depressed women (Dayan et al. 2006). Prenatal and postnatal anxiety and depression have also been implicated in suboptimal development across the spectrum of domains of cognitive, psychomotor, socio-emotional development and behaviour among infants (Kingston et al. 2012b) and school-aged children (Kingston and Tough 2014). Long-term effects of prenatal and postnatal depression, stress and anxiety have also been reported. For example, studies have found greater risk of mental disorder among 4–13-year-old children of mothers with anxiety during pregnancy compared to those without (12.3 % vs. 6.8 %; O’Donnell et al. 2014) and epigenetic changes in the DNA of 8-year-old children of women exposed to a natural disaster (and thus severe emotional stress) during pregnancy (Cao-Lei et al. 2014). Studies comparing the outcomes of adolescents of mothers with and without postnatal depression have reported poorer educational performance at age 16 (boys; Murray et al. 2010), depression at age 16 (boys and girls, 41.5 % vs. 12.5 %; Murray et al. 2011) and externalising problems and impaired social competence at age 16 (boys; Korhonen et al. 2012, 2014) among those whose mothers had post-partum depression (PPD).
Although few studies have examined the effect of chronic mental illness on child outcomes (Kingston et al. 2012b), one study reported that infants of mothers who experienced depressive symptoms during pregnancy and post-partum had greater risk of suboptimal cognitive development compared with those whose mothers were depressed during the post-partum period alone (Beckwith et al. 1999). In terms of severity of symptoms, reviews report associations between a range of symptom severity and adverse child outcomes from mild to severe (Glover 2014; Kingston et al. 2012b).
Indeed, recent findings using the data from Australia’s longitudinal pregnancy cohort study, the Maternal Health Study, indicate that children of women with high levels of depressive symptoms (EPDS 10-14) and those with subclinical symptoms (EPDS 6-8) had greater odds of emotional-behavioural difficulties (24 % and 19 % respectively), compared with women with no symptoms (7 %; Giallo et al. 2015).
The value of perinatal psychosocial assessment and intervention is underscored by the substantial clinical impact of perinatal mental illness on child outcomes. For instance, in their meta-analysis, Grote et al. (2010) report 39 and 49 % increases in risk of preterm and small for gestational age delivery, respectively, among women with prenatal depression compared to those without (Grote et al. 2010). Based on studies using the Avon Longitudinal Study of Parents and Children (ALSPAC), an attributable risk of 10–15 % has been estimated as the contribution of prenatal anxiety and depression to suboptimal child development (Glover 2014; Talge et al. 2007). Furthermore, the lack of specificity of the type of mental illness and child outcome (Glover 2014; Kingston and Tough 2014; Kingston et al. 2012b; O’Donnell et al. 2014) supports assessment of a broad spectrum of illness.
Perinatal mental health problems are severely underdetected and undertreated (Coates et al. 2004; Spitztler et al. 2000). Without standardised psychosocial assessment, as many as 80 % of cases are not detected (Carroll et al. 2005; Mitchell and Coyne 2009). Although the benefit of psychosocial assessment is recognised by health-care providers (Buist et al. 2006; Chew-Graham et al. 2008; Leiferman et al. 2008; Reid et al. 1998), significant barriers to routine perinatal psychosocial assessment and referral have been cited, including most prominently the lack of time, education, and linkages with mental health resources (Byatt et al. 2012; Kim et al. 2010). At the same time, significant barriers deter women from seeking mental health care (Byatt et al. 2012; Flynn et al. 2010; Reay et al. 2011; Sword et al. 2008) and self-disclosing mental health concerns during the prenatal and postnatal periods (Sword et al. 2008; Woolhouse et al. 2009) including stigma, being reassured by family and friends that their emotions are ‘normal’, preferring to address their concerns on their own without professional assistance, perceiving that the health-care provider is disinterested or lacks time, and lacking the knowledge of what emotions constitute a ‘normal’ versus ‘abnormal’ experience (Kingston et al. 2015b, in press). Further barriers impede women from engaging in follow-up assessment or treatment following a positive screen (Kim et al. 2010). Indeed, recent research suggests that only half of those who screen positive during follow-up with a subsequent mental health assessment (Kim et al. 2010) and 30–85 % (Bales et al. 2015; Bowen et al. 2012; Marcus et al. 2003; Reay et al. 2011; Woolhouse et al. 2009) do not engage in treatment. Highlighting the impact of routine psychosocial assessment on treatment engagement, a recent study also found that women who were not asked about emotional health were far less likely to seek formal mental health care during pregnancy (AOR 0.09, 95 % CI 0.04–0.24) or post-partum (AOR 0.07, 95 % CI 0.02–0.13; Reilly et al. 2014). This study also demonstrated the importance of the referral process, as women who had undergone mental health screening but did not receive a formal referral were less likely to engage in treatment in pregnancy (AOR 0.26, 95 % CI 0.15–0.45) and post-partum periods (AOR 0.14, 95 % CI 0.07–0.27; Reilly et al. 2013b). Additionally, the actual process of screening appears to influence women’s decisions regarding treatment uptake. A recent Canadian study found that 78 % of women reported that the ‘way’ a provider asked about their emotional health would have a major impact on whether they sought treatment or not (Kingston et al. 2015b, in press).
While personal and system-related barriers deter women from self-initiating discussions regarding concerns with their mental health, the vast majority responds to provider-initiated care (Sword et al. 2008; Woolhouse et al. 2009), and fewer than 4 % refuses provider-initiated screening (Austin et al. 2010; Chew-Graham et al. 2009; Miller et al. 2009). Overall, studies across Canada, 44 (Kingston et al. 2015b, in press; Reid et al. 1998) Australia (Bales et al. 2015; Bowen et al. 2012; Reay et al. 2011), and the United States (Chew-Graham et al. 2009; Miller et al. 2009) report high levels of acceptability of mental health screening by midwives, nurses, family physicians and obstetricians among pregnant and post-partum women. Importantly, vulnerable women also report high acceptability, including those with high depression scores at the time of screening (Gemmill et al. 2006), women of non-English-speaking background (Matthey et al. 2005), those having a previous diagnosis or treatment history for mental illness (Kingston et al. 2015b, in press) and those experiencing IPV (Matthey et al. 2005). Indeed, a recent Canadian study found that 99 % of pregnant women who have not been screened would be comfortable with provider-initiated screening, and 97 % of those who had been screened reported the same (Kingston et al. 2015b, in press). In this same study, demographics, type of provider and history of diagnosis or treatment for mental illness were unrelated to whether pregnant women found screening acceptable or not. Finally, regarding provider views of feasibility of perinatal mental health assessment, providers in settings that have implemented an infrastructure for routine psychosocial assessment as part of a system of assessment-referral-care have found it to be a feasible, effective approach (Flynn et al. 2010; Mitchell and Coyne 2009; Reay et al. 2011; Sword et al. 2008). Thus, a key message from this evidence is that opportunistic screening that occurs, as a component of routine prenatal and postnatal care, is highly acceptable to women and providers and considered to be a feasible approach to perinatal mental health care.
The Debate Around Depression Screening: Harms, Benefits and Cost-Effectiveness
While the benefits of screening are intuitively appealing, the harms – in particular the cost to women through over-detection and incorrect labelling and the cost to the system in terms of resourcing – are less easy to quantify but just as important to consider. The current evidence base is inconclusive as to whether the benefits of perinatal depression screening outweigh the risks (Myers et al. 2013). Indeed a clear resolution may not be attainable because a perfect evidence base is not feasible in a field where translation of research into practice has to walk a fine line between slavish reliance on randomised controlled trial (RCT) findings and the demands of strict systematic literature reviews – and the exigencies of the real-world clinical setting. While these questions are complex as they involve understanding of both psychometric and health economic parameters and modelling, it is incumbent upon us to clarify the key parameters of the debate as much as possible.
Psychometric Parameters of Screening Tools
In order to discuss the psychometrics of commonly used tools, some key definitions are in order. A screener is considered adequate if both its sensitivity and specificity are at least 80 %. Sensitivity is a measure of ‘true positives’ – i.e. women correctly identified on a depression screener, e.g. the EPDS (Cox et al. 1987) using a set cut-off – most commonly ten or more for any depression (minor or major) and 13 or more for major depression. Specificity is a measure of ‘true negatives’– i.e. women correctly identified on the EPDS as not having depression using the same cut-off (12 or less on the EPDS for major depression). The positive predictive value (PPV) is the number of cases correctly identified as a proportion of all test-positive women (both true and false positive) on the screener and is also a function of the disorder prevalence. For the EPDS, PPV sits between 40 and 60 % (depending on the input parameters used). In essence, only about one in two positive screeners has an actual diagnosis of depression. At best using an EPDS cut-off of 13 or more with sensitivity of 81 % and specificity of 96 % (derived from the largest study of Murray and Carothers 1990) and point prevalence of 6.8 % for major depression at 6 weeks post-partum, the EPDS yields a 62 % PPV (Milgrom et al. 2011). It is notable that low prevalence conditions (e.g. depression) usually have much better specificity and thus negative predictive value; hence, PPV will never be very high. The NICE guidelines (2014) recommend the use of sensitivity and specificity (rather than PPV) as these are not population dependent and hence more generalisable. Conceptually, the challenge lies in understanding that improved sensitivity is at the cost of specificity and vice versa, i.e. these psychometric parameters vary in relation to each other depending on where the screener cut-off is set (see Table 11.2 for examples with different EPDS cut-offs).
Table 11.2
Sensitivity and specificity table: EPDS example at different cut-offs
EPDS score | Sensitivity (true positives) | Specificity (true negatives) | Consequences |
|---|---|---|---|
≤16 | .31 Miss many women Many more false negatives | .99 Few false positives | Miss more women Fewer ‘cases’ = less cost PPV not available |
≤13 | .81 | .99 | aPPV = 62 % |
≤10 | Sensitivity = .82 Include more women Fewer false negatives | Specificity = .86 Exclude less women = more false positives | Miss fewer women More ‘cases’ = greater cost PPV not available |
One alternative to the EPDS considered in the NICE guidelines (2014) has been the two questions by Whooley et al. (1997) – During the ‘past month’, have you ‘often’ been bothered by (1) feeling down, depressed or hopeless and (2) little interest or pleasure in doing things? Women are screened positive if they endorse either Q 1 or 2 and screened negative if they endorse neither Q 1 nor 2. While it has 100 % sensitivity, specificity is only 68 %, and there is thus a need to reduce false positives (Mann et al. 2012).
One of the key factors underpinning the debate around screening includes a consideration of screener accuracy. In light of the above discussion, it is clear that improving cost-effectiveness (i.e. reducing false positives and the associated added cost to the service) while improving clinical effectiveness, i.e. minimising false negatives (cost to women of not being detected), is a balancing act impossible to achieve with the use of one screening tool alone. A number of approaches to optimise screening have thus been advocated with some being currently examined (Kingston et al. 2014a). One approach suggested by NICE (2014) is the sequential contingent administration of two depression screeners in a two-step procedure where the two-item Whooley is given to all women and followed by the targeted administration of the ten-item EPDS (i.e. in Whooley-positive women only) with the aim of obtaining optimum sensitivity (100 % on the Whooley), much improved specificity (on the contingent EPDS administration) and potentially reduced cost (NICE 2014). Alternative approaches to reducing the burden of screening include doing a targeted EPDS, i.e. only in women deemed at significant psychosocial risk (at the cost of reduced sensitivity; Canadian Task Force on Preventive Health Care et al. 2013), and repeating the EPDS ~2 weeks later in screen-positive women only (Matthey and Ross-Hamid 2012) though issues of feasibility will arise, given the lack of continuity of care in the primary sector. Finally, another option is that of using a higher than usual EPDS cut-off to define screen-positive women (e.g. score of 16 or more) as per Paulden et al. (2009), leading to excellent specificity (99 %) and reduced system costs but much reduced sensitivity (31 %), i.e. significant under-detection. This approach would not be acceptable in better-resourced settings (NICE 2014); however, it may be acceptable in less economically advantaged settings.
While there is no doubt that postnatal depression will incur significant costs both direct (health care) and indirect or societal (workforce productivity, offspring outcomes), the evidence for clinical effectiveness of depression screening programmes let alone their cost-effectiveness is a much more complex matter to disentangle.
Costs Associated with Untreated Perinatal Depression
Two recent detailed economic reports commissioned by consumer organisations outline the high cost of untreated mental health morbidity in the perinatal period. The Post and Antenatal Depression Association (PANDA 2012) report examined the direct cost to the Australian economy in terms of direct costs for treating perinatal depression and indirect costs of productivity loss (where impact on both mothers and fathers was considered). In total, perinatal depression costs almost $½ billion in 2012 (health-care costs but mostly productivity losses), assuming up to one third of young families are affected. This study did not examine costs associated with suboptimal child outcomes. The London School of Economics report (Bauer et al. 2015) examined direct costs of treating maternal perinatal depression and anxiety and postnatal psychosis and the indirect costs in terms of children’s outcomes; it did not examine loss of productivity. They found a cost of eight billion pounds to the UK economy in 2013 (mostly attached to poorer child outcomes).
Evidence for Clinical Effectiveness of Depression Screening (Perinatal) RCTs (See Table 11.3)
Table 11.3
Summary of RCTs examining the clinical effectiveness of screening programmes for PND
Study (recruit time) | Ns (% screen +) | ‘Screen +ve’ definition | Intervention details | 10 outcomes | Results at 6–12 months |
|---|---|---|---|---|---|
Morrell (2009) UK Cluster RCT (6 weeks p-p) | 2277 vs. 1172 IV 404 (17.7 %) CAU 191 (16.3 %) | EPDS ≥12 | Nurse home visits CBT Additional EPDS at 8 week | EPDS ≥ 12 at 6 months p-p | 34 % (IV) vs. 46 % (CAU) are ‘depressed’ (EPDS) at 6 months and 12 months AOR 0.6 (95 % CI .38–.96; p = .028) |
Yawn (2012) US Cluster RCT (5–12 weeks p-p) | 1163 vs. 913 IV 399 (35 %) CAU 255 (35 %) | ≥10 EPDS or PHQ-9 ≥10 (‘depression’) | GP training Nurse home visit Algorithm for managing screen + ve; case manage and medication (if functional impairment on PHQ-9) | PHQ-9 score ↓ by 5 or more points 6 and 12 months p-p | 45 % (IV) vs. 35 % (CAU) are ‘not depressed’ (PHQ-9) at 12 months AOR 1.74 (95 % CIs 1.0–2.86; p < .01) |
Leung (2011) Hong Kong RCT (2 months p-p) | 231/group IV 73 (32 %) CAU 14 (6 %) : *as per next column | ≥10 EPDS, or Q10 +ve or *depressed on ‘clinical impression’ | Nurse counselling or psychiatry referral For both arms-only difference was IV group given EPDS | EPDS <10 at 6 months p-p | ES 0.34 (95 % CI .15–.52; p < .001) |
The small RCT evidence base, although encouraging, is at best of low to moderate strength in favour of the clinical effectiveness of depression screening programmes. Thombs et al. (2014), in a systematic review examining the impact of using a depression screener in addition to care as usual (CAU), found that only one of the three relevant RCTs (Leung et al. 2011) met all review criteria as well as controlling for the care components common to the two groups. This study examined the value of administering the EPDS to CAU compared to the CAU condition alone (which consisted of informal enquiry about current depression by the postnatal nurse, followed by nurse counselling or referral to mental health, as needed). The addition of that simple EPDS at 2 months post-partum was associated with a significant reduction in EPDS scores at 6 months post-partum (effect size 0.34, Leung et al. 2011). Thombs et al. (2014) however discounted this study as being a likely false positive finding – based on the study being biased (halo effect and selective reporting of favourable results) and underpowered. Like the UK National Screening Committee (Hill 2010), the review authors emphasised the potential harms of screening and advised against the benefit of depression screening postnatally while recommending that clinicians ‘be aware’ of possible depression in their clients. It is notable that the two cluster RCTs (excluded by Thombs et al. 2014), examining depression screening as part of an ‘enhanced’ care programme, found improved clinical outcomes compared to CAU (see Table 11.3), using better more robust methodology. Yawn et al. (2012) in a US study found an adjusted odds ratio of 1.74 (95 % CI 1–2.86; p < .01) in favour of improved depression scores at 12 months in the intervention group. Morrell et al. (2009) in a UK study reported very similar findings. The Yawn et al. (2012) and Morrell et al. (2009) studies echo the positive effect sizes of depression screening and collaborative (enhanced) care RCTs undertaken in the general population (Gilbody et al. 2006). Chaudron and Wisner (2014) in their comment on the Thombs et al. (2014) review, caution against hasty discounting of depression screening on the basis of a narrow interpretation of the findings and lack of carefully balancing the risks and benefits of implementing screening in a ‘real-life’ clinical setting.
Evidence for the cost-effectiveness of routine depression screening followed by appropriate care is so far very limited and based on modelled data only. Paulden et al. (2009), comparing informal enquiry about depression by the health visitor nurse (CAU) to the use of an EPDS screen (using an EPDS cut-off of ≥16) and assuming that a woman scoring ≥16 would have an average of 15 clinical consults (compared to none if CAU), found the ‘intervention’ was not value for money. That is, the incremental cost-effectiveness ratio (ICER) of £41,000 was much greater than the UK willingness to pay threshold in terms of cases of depression averted (£20k–30k/QALY). Two key reasons for lack of cost-effectiveness in this study were the assumption that false positives (on EPDS) would still get referred for full treatment and that such an extensive range of treatment would be accessed in the primary care setting. Campbell et al. (2008), in a NZ modelling study, evaluated the cost-effectiveness of ‘formal case identification’ for postnatal depression using the PHQ-3 (in addition to the routine use of the EPDS in the CAU group). In contrast to Paulden et al. (2009), they found that this formal case identification approach was highly cost-effective for postnatal depression in NZ with an ICER of NZ$3461/QALY well below their willingness to pay threshold (close to the Australian threshold of $50,000/QALY) or the UK 20,000 lb/QALY. The limitations of this study were that costs were mostly/totally based on antidepressant medication and that GPs would always correctly diagnose women (without the cost of false positives).
Contradictory findings in not dissimilar studies highlight the fact that cost-effectiveness studies are predicated on the key assumptions underpinning their modelling. Ideally, we would have real-world cost-effectiveness studies done in perinatal populations that identify both direct (health related) and indirect (societal, i.e. offspring and workforce) costs. In terms of translation of these studies to the clinical setting, extant evidence has been variably interpreted. The SLR-based Australian CPGL clearly recommends all women to have depression screening (using the EPDS and in association with enquiry about mental health history), followed by formal MH assessment as indicated.
In contrast, on the basis of consensus expert opinion, NICE (2014) has recommend ‘case identification’ by means of routine ‘enquiry’ in relation to the woman’s mental health history, past and current and family history. The use of screening tools is optional, i.e. the clinician may ‘consider’ the use of a depression and anxiety screener (for identification of a current episode). Though still somewhat ambiguous, there has been a shift in the revised 2014 guidelines which now suggest universal use of the brief two questions of Whooley (depression), and GAD-2 (anxiety) screeners could be followed by (targeted) the use of a second longer screening tool (e.g. the EPDS and GAD-7) but only if either the Whooley or GAD-2 is positive. The wording around such contingent use of two screeners has been kept deliberately ambiguous in the NICE guideline, given the lack of adequate evidence base for such an approach.
The influential US AHRQ systematic review (Myers et al. 2013) – while not overtly recommending against universal depression screening – implicitly does so by concluding that ‘while currently available screening instruments are reasonably sensitive and specific in detecting postpartum depression, there is insufficient evidence to draw any conclusions about the net balance of benefits and harms of screening for postpartum depression, or about whether specific tools or strategies would result in a more favourable balance (Myers et al. 2013, p 65.)’.
In summary, while universal (or routine) depression screening has many compelling benefits, it remains contentious because of the substantial inherent limitations and risks of depression screeners, economic burden of comprehensive, integrated screening programmes and lack of clear evidence of cost benefit, i.e. the additional costs to the system sitting within an acceptable willingness to pay band. Furthermore, the ambiguities contained in the recent systematic review of the literature (Thombs et al. 2014), the NICE guideline (2014) and AHRQ report (Myers et al. 2013) leave the primary care clinician with a major quandary: they are supposed to enquire about ‘signs’ of depression and make assessment of the ‘risk factors’ without having the tools to assist them in this task.
Psychosocial Assessment: The Road Less Travelled
It is critical that clinicians not separate mental health disorder from the psychosocial context in which it arises, and yet this has been the modus operandi when it comes to mental health screening in the primary care setting. This dichotomy is especially problematic in the perinatal context where psychosocial context and function take on greater prominence as women transition to the parenting role, which relies heavily on psychosocial context.
Related posts:
Maternal Perinatal Psychopathology: Overview
Psychotropic Drugs and the Perinatal Period
Establishing Parent–Infant Interactions
Addressing Familial Violence and Child Abuse
Ambulatory Care: Home-Based Perinatal Interventions
Ambulatories Cares: Parent–Infant Psychotherapy in Perinatal Mental Health
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
