Papers

Psychosocial and psychological interventions for prevention of postnatal depression: systematic review

BMJ 2005; 331 doi: http://dx.doi.org/10.1136/bmj.331.7507.15 (Published 30 June 2005) Cite this as: BMJ 2005;331:15
  1. Cindy-Lee Dennis (cindylee.dennis{at}utoronto.ca), assistant professor1
  1. 1 University of Toronto, Faculty of Nursing, 50 St George Street, Toronto, ON, Canada M5S 3H4
  1. Correspondence to: C-L Dennis
  • Accepted 1 February 2005

Abstract

Objective To assess the effects of psychosocial and psychological interventions compared with usual antepartum, intrapartum, or postpartum care on the risk of postnatal depression.

Data sources Medline, Embase, CINAHL, Cochrane central register of controlled trials, Cochrane pregnancy and childbirth group trials register, Cochrane depression, anxiety, and neurosis trials register, secondary references and review articles, and experts in the field.

Study selection All published and unpublished randomised controlled trials of preventive psychosocial or psychological interventions in which the primary or secondary aim was a reduction in the risk of postnatal depression. All trials recruited pregnant women or new mothers less than six weeks postpartum. Eligible studies were abstracted, assessed for methodological quality, and pooled with relative risk for categorical data and weighted mean difference for continuous data.

Results Fifteen trials with 7697 women were included. Although there was no overall statistically significant effect on the prevention of postnatal depression in the meta-analysis of all types of interventions (15 trials, n = 7697; relative risk 0.81, 95% confidence interval 0.65 to 1.02), these results suggest a potential reduction in postnatal depression. The only intervention to have a clear preventive effect was intensive postpartum support provided by a health professional (0.68, 0.55 to 0.84). Identifying women “at risk” assisted in the prevention of postnatal depression (0.67, 0.51 to 0.89). Interventions with only a postnatal component were more beneficial (0.76, 0.58 to 0.98) than interventions that incorporated an antenatal component. In addition, individually based interventions were more effective (0.76, 0.59 to 1.00) than group based interventions (1.03, 0.65 to 1.63).

Conclusions Diverse psychosocial or psychological interventions do not significantly reduce the number of women who develop postnatal depression. The most promising intervention is the provision of intensive, professionally based postpartum support.

Introduction

Postnatal depression is a major health issue for many women from diverse cultures.1 2 Although longitudinal and epidemiological studies have yielded varying prevalence rates, a meta-analysis of 59 studies reported a prevalence of 13%,3 with most cases starting in the first three months postpartum.4 This morbidity has well documented health consequences for the mother, child, and family. Women who have postnatal depression are significantly more likely to experience future episodes of depression,5 and infants and children are particularly vulnerable because of impaired maternal-infant interactions and negative perceptions of infant behaviour.

The cause of postnatal depression remains unclear,6 with extensive research suggesting many contributory factors. Epidemiological studies and meta-analyses of predictive studies, however, have consistently identified the importance of psychosocial and psychological risk factors3 6 7—such as life stress,3 79 marital conflict,3 710 maternal self esteem,7 11 and lack of social support.3 7 9 1115 A comprehensive review suggested that in women with postnatal depression, psychosocial and psychological treatment may be suitable.16 As such, it is theoretically possible that these interventions may also prevent postnatal depression, as many of the known risk factors are present during pregnancy and the immediate postpartum period. There have been two critical reviews of preventive trials17 18 and one systematic review that examined diverse interventions to reduce “probable depression” in the postnatal period.19 However, no systematic review has examined the overall preventive effect of psychosocial and psychological interventions or determined which characteristics are most beneficial.

I assessed the effects of such interventions compared with usual antepartum, intrapartum, or postpartum care on the risk of postnatal depression. This systematic review is based on a full review published in the Cochrane Library.20

Methods

Searches

I searched the Cochrane pregnancy and childbirth group trials register. This database contains trials identified from quarterly searches of the Cochrane central register of controlled trials, monthly searches of Medline and hand searches of 30 journals and the proceedings of major conferences. In addition, the Cochrane depression, anxiety, and neurosis trials register, Medline (1966-2004), Embase (1980-2004), and CINAHL (1982-2004) were all independently searched. Secondary references and review articles were scanned and experts in the field were contacted. Trials published in all languages were considered.

Selection

Published and unpublished studies were eligible if they were randomised controlled trials; were methodologically strong based on a validity assessment; evaluated a psychosocial or psychological intervention in which the primary or secondary aim was a reduced risk of postnatal depression; and included pregnant women and new mothers less than six weeks postpartum. I excluded studies if they incorporated a quasi-randomised design; recruited women identified with symptoms of depression, or solely evaluated an educational intervention. For this review, a psychosocial or psychological intervention incorporated various non-pharmaceutical strategies that were delivered antenatally or within the first month postpartum, or both, by a health professional or layperson.

Assessment of validity

The methodological quality of each trial was assessed according to the recommendations of the Cochrane Collaboration and examined the generation of allocation sequence; allocation concealment; blinding of outcome assessors; completeness of follow-up data; and intention to treat analysis. Two reviewers independently assigned a quality rating to each trial; results were compared and differences discussed until agreement was obtained.

Abstraction of data

Two reviewers independently extracted data and included study design; participants (number and characteristics); intervention type, mode, onset, duration, and provider; outcomes measured; and results. Wherever necessary, unpublished or missing data were requested from the trial's corresponding author and data were double entered into Review Manager version 4.2.3 (Cochrane Collaboration software).

Quantitative data synthesis

While the primary meta-analysis was based on the occurrence of postnatal depression (however measured by trialists), several depression rating scales or cut-off points were incorporated. To address the potential measurement differences, I made direct comparisons using a fixed effect model between trials using the same rating scale and cut off. Meta-analyses were performed using relative risks as the measure of effect size for binary outcomes and weighted mean differences for continuous outcome measures, both with 95% confidence intervals.

Heterogeneity was investigated by calculating I2 statistics,21 and if this indicated a high level among the trials included in an analysis (I2 > 50%), a random effects meta-analysis was used. Where I found high levels of heterogeneity I used sensitivity analyses, excluding the trials most susceptible to bias based on the following quality assessment: those with unclear allocation concealment, high levels of losses or exclusions after randomisation, or uncertain on no blinding of outcome assessment. A priori subgroup analyses estimated the effect of intervention type (for example, psychosocial and psychological), intervention mode (for example, individual v group based), intervention onset (for example, antenatal and postnatal v postnatal only), and sample selection criteria (for example, targeting women with specific risk factors v a general population).

Results

The search identified 155 studies, of which 99 were excluded as non-experimental. Of the 56 trials retrieved for a more detailed evaluation, 15 studies were of treatment interventions for postnatal depression. The 41 remaining potentially appropriate trials were examined for inclusion/exclusion criteria and methodological quality. I excluded 26 trials because of a quasi-experimental design (n = 4),2225 poor methodological quality (n = 3),2628 not a psychosocial or psychological intervention (n = 12),2940 the prevention of postnatal depression was not the primary or secondary objective (n = 5),4145 and the intervention was for the treatment of antenatal depression (n = 2)46 47 (table 1).

Table 1.

Characteristics of excluded studies

View this table:

Study characteristics

The 15 trials in the meta-analysis incorporated 7697 women and were published between 1995 and 2003 (table 2). Most trials were conducted in Australia and the United Kingdom; two trials were conducted in the United States48 49 and one in China.50 Seven trials targeted women believed to be at additional risk of postnatal depression,4854 while the eight others enrolled women from the general population.

Table 2.

Characteristics of included studies

View this table:

Types of interventions

The studies were subgrouped into categories to examine specific types of psychosocial interventions—such as antenatal and postnatal classes,52 53 55 professional51 56 and lay57 home visits, continuity of care,58 and early postpartum follow-up (for example, postpartum care initiated earlier than usual practice)59—and psychological interventions, such as debriefing50 54 6062 and interpersonal psychotherapy.48 49 The interventions were provided by various professionals, including physicians,59 nurses,5052 midwives,5356 58 6062 and other healthcare providers.48 52 In one trial, the intervention was provided by lay women recruited from the community.57 In most studies, the control group was reported to have received usual antenatal/postnatal care, which varied both between and within countries.

Definition of postnatal depression

In all trials but one,49 postnatal depression was defined as a score above a specified cut-off point on a self reported measure. Most studies (10 out of 15) used an Edinburgh postnatal depression scale score > 12 to indicate postnatal depression. Two additional trials used the Edinburgh postnatal depression scale but incorporated a different cut off score; one study used a 10/11 cut off52 while another selected a 11/12 cut off.55 Several studies also reported mean scores using this measure.48 51 5557 59 60 The Edinburgh postnatal depression scale does not diagnose postnatal depression (as this can be accomplished only through a psychiatric clinical interview) but rather it is the most commonly used instrument to assess postpartum depressive symptoms.7 Two trials used the self reported hospital anxiety depression scale,50 61 and two studies used a semistructured diagnostic interview (structured clinical interview for DSM-IV).48 49 The timing of the outcome assessment varied considerably between studies, ranging from three61 to 2448 53 55 57 59 60 62 weeks postpartum; one trial also included a 52 week assessment.62

Methodological quality

Randomisation was often with consecutively numbered, sealed, opaque envelopes.50 53 54 57 58 61 62 Four trials used various forms of computer based randomisation.51 52 55 56 Two trials incorporated a central computerised randomisation service accessed by telephone,59 60 and one trial used a block randomisation procedure using a random numbers table48; one trial was cluster randomised.56 Allocation concealment was unclear in one trial.49 All but two trials48 49 completed a power analysis, and data were analysed with an intention to treat approach. Outcome data were collected by assessors blinded to group allocation48 51 52 or mailed questionnaires; one study did not identify the outcome assessor.49 Five trials had a follow-up attrition rate > 20% at final study assessment.50 55 56 57 59 Follow-up in all these trials included mailed questionnaires. Based on susceptibility to bias (for example, unclear allocation concealment, high levels of exclusions or attrition after randomisation, or unblinded outcome assessment), I excluded six trials as appropriate during the sensitivity analysis for outcomes with high levels of heterogeneity (I2 > 50%).49 50 5557 59

Quantitative data synthesis

Postnatal depression at last assessment

Variously defined—My main outcome measure was postnatal depression at the last study assessment. Although there was no statistically significant beneficial effect on the prevention of postnatal depression in the meta-analysis of all types of interventions (15 trials, n = 7697; relative risk 0.81, 95% confidence interval 0.65 to 1.02) (fig 1), these results suggest a potential 19% reduction in postnatal depression. There was significant heterogeneity among these trials (I2 = 68.8%). The removal of trials at risk of bias, however, resulted in no substantial change to the conclusion. I found a similar non-significant effect when I calculated a weighted mean difference (WMD) among the trials that provided a mean score on the Edinburgh postnatal depression scale (seven trials, n = 3300; WMD −0.06, −0.37 to 0.26) (fig 2).

Figure1

Postnatal depression at final assessment (variously defined) among studies evaluating interventions versus normal care in the prevention of postnatal depression

Figure2

Mean depression scores at final assessment among studies evaluating interventions versus normal care in the prevention of postnatal depression

Edinburgh postnatal depression score >12—To examine potential measurement differences, I used a random effects model to directly compare trials that used the Edinburgh postnatal depression scale with the recommended 12/13 cut-off score63 and found no preventive effect (10 trials, n = 6126; 0.91, 0.73 to 1.15).

Postnatal depression at 8, 16, and 24 weeks

Variously defined—To assess the short and longer term effects of the preventive interventions, I categorised assessments of postnatal depression at 0-8 weeks postpartum (short term effect); 9-16 weeks (intermediate effect); and 17-24 weeks (longer term effect). Results showed a short term reduction in the development of postnatal depression (eight trials, n = 4091; 0.65, 0.43 to 1.00). The effect weakened at the intermediate period (eight trials, n = 3326; 0.80, 0.56 to 1.12) and disappeared after 16 weeks (seven trials, n = 4314; 1.02, 0.87 to 1.19).

Edinburgh postnatal depression scale score >12—When I included only those trials that used an Edinburgh postnatal depression scale score > 12 as the outcome measure, there were no statistically significant short term (six trials, n = 3452; 0.90, 0.65 to 1.25), intermediate (five trials, n = 2369; 0.72, 0.49 to 1.06), or longer term (six trials, n = 3598; 1.00, 0.84 to 1.19) effects.

Subgroup analyses

Type of intervention—I found no preventive effect with antenatal and postnatal classes (two trials, n = 311; 1.02, 0.61 to 1.72), lay home visits (one trial, n = 481; 0.89, 0.62 to 1.27), and early post-partum follow-up (one trial, n = 475; 0.91, 0.56 to 1.48). I did, however, find a positive trend related to continuity of care (one trial, n = 935; 1.34, 0.97 to 1.85) and a clear beneficial effect with home visits provided by a health professional (two trials, n = 1663; 0.68, 0.55 to 0.84). Among psychological interventions, there was no preventive effect in relation to interpersonal psychotherapy (two trials, n = 72; 0.31, 0.04 to 2.52) but a positive trend in relation to debriefing in hospital (five trials, n = 3051; 0.57, 0.31 to 1.04).

Mode of intervention—Analysis of 11 trials evaluating individually based interventions showed a benefit in preventing postnatal depression at the last study assessment (n = 6642; 0.76, 0.59 to 1.00). When I excluded trials susceptible to bias, the direction of the effect remained the same but the confidence interval widened (seven trials, n = 3667; 0.68, 0.43 to 1.09). Of the four trials that evaluated group based interventions, there was no apparent reduction in depressive symptoms at last study assessment (n = 1055; 1.03, 0.65 to 1.63).

Onset of intervention—Studies in which the intervention began antenatally and continued postnatally failed to reduce the likelihood of women developing postnatal depression (four trials, n = 1283; 1.21, 0.93 to 1.59). However, there was a preventive effect in those trials evaluating a postnatal only intervention (10 trials, n = 6379; 0.76, 0.58 to 0.98).

Effect of sample selected—Trials that selected participants considered to be “at risk” had more success in preventing postnatal depression (seven trials, n = 1162; 0.67, 0.51 to 0.89) than those that enrolled women from the general population (eight trials, n = 6535; 0.87, 0.66 to 1.16).

Discussion

This systematic review of the prevention of postnatal depression shows that there is no clear evidence to recommend the implementation of antenatal and postnatal classes, early postpartum follow-up, continuity of care models, psychological debriefing in hospital, and interpersonal psychotherapy. There is emerging evidence, however, to support the importance of additional professional support provided postnatally. Although one well designed trial suggested that intensive home visits by nurses with at risk mothers was protective during the first six weeks postpartum,51 the benefit was not maintained to 16 weeks. It is noteworthy that the 16 week assessment coincided with a decrease from weekly to monthly visits. Results from a cluster randomised controlled trial showed that flexible, individualised postpartum care by midwives that incorporated assessment tools also had a preventive effect.56 Due to the cluster randomisation process, however, this trial may have been overweighted in the meta-analyses. The effectiveness of postpartum support provided by laypeople remains uncertain.57

The trials can be further classified into different categories depending on the target population: universal interventions are offered to all women, selective interventions are offered to women at increased risk of developing postnatal depression, and indicated interventions are offered to women who have been identified as depressed or probably depressed.64 Though I did not include any trial that evaluated an indicated intervention, the results of a subgroup analysis to examine the effects of universal and selective interventions showed that identifying mothers with risk factors assisted in the prevention of postnatal depression. Currently there is no consistency in the identification of such women, and a review of 16 antenatal screening tools suggests that there is no measure with acceptable predictive validity to accurately identify women who will later develop postnatal depression.65 This may partially explain why interventions with only a postnatal component seem to be more beneficial than interventions that also incorporate an antenatal component. Although I did not examine interaction effects between the five a priori subgroups, the estimates were independent and the sample sizes were large.66

The included trials were of good methodological quality. The reporting of the trials, however, was often not comprehensive, lacking details on the training and qualifications of the intervention providers and descriptions of adherence to the intervention protocol. There was also a failure to present details of the informational element of the interventions and on the background features of the care received by the control groups. While intention to treat analyses were performed, in trials with group sessions compliance was poor.52 53 55

What is already known on this topic

Postnatal depression is a major health issue affecting about 13% of all new mothers

Epidemiological studies and meta-analyses of predictive studies have consistently found that several psychosocial and psychological variables contribute to increased risk

No systematic review has examined the preventive effect of psychosocial and psychological strategies or specific intervention characteristics

What this study adds

This study did not find sufficient evidence that diverse psychosocial or psychological interventions reduce the number of women who develop postnatal depression

Interventions that target at risk women, are individually based, or initiated postnatally are more likely to be beneficial

Interpretation of results

The diversity of preventive interventions and the widely differing study end points should urge some caution in the interpretation of the pooled data. To partially address this issue, I included short, intermediate, and longer term effects where appropriate. Despite this caution and the subgrouping of end points, this review consistently showed that women who received a preventive intervention were statistically overall just as likely to experience postnatal depression as those who received standard care. It is unknown to what extent some of the heterogeneity or non-significant results are related to the measure used to assess postnatal depression. A similar non-significant effect, however, was found among those trials that incorporated the Edinburgh postnatal depression scale.

The long term consequences of postnatal depression suggest preventive approaches are warranted. Translating research on risk factors into predictive screening protocols65 and preventive interventions, however, has met with limited success, as complex interactions of biopsychosocial risk factors with individual variations need to be considered. The results from this review provide physicians and other health professionals with recommendations for future preventive trials. To further investigate postnatal depression as a public health concern, the inclusion of ethnically and socioeconomically diverse women in these research efforts is critical. In addition, all future trials should include an economic analysis of the relative costs and benefits.

Acknowledgments

I thank L Gorman, R Hagan, and C MacArthur for responding to queries.

Footnotes

  • Contributors D Creedy assisted with assessment of trial quality and data entry, J Kavanagh assisted with the search strategy, and E Hodnett provided editorial suggestions.

  • Funding None.

  • Competing interests None declared.

  • Ethical approval Not required.

References

  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
  26. 26.
  27. 27.
  28. 28.
  29. 29.
  30. 30.
  31. 31.
  32. 32.
  33. 33.
  34. 34.
  35. 35.
  36. 36.
  37. 37.
  38. 38.
  39. 39.
  40. 40.
  41. 41.
  42. 42.
  43. 43.
  44. 44.
  45. 45.
  46. 46.
  47. 47.
  48. 48.
  49. 49.
  50. 50.
  51. 51.
  52. 52.
  53. 53.
  54. 54.
  55. 55.
  56. 56.
  57. 57.
  58. 58.
  59. 59.
  60. 60.
  61. 61.
  62. 62.
  63. 63.
  64. 64.
  65. 65.
  66. 66.
View Abstract