Postpartum depression

Note

The modeling strategy for this cause may be updated upon release of the GBD 2023 methods appendix, but this is an adequate placeholder strategy until this is determined

Additionally, there is evidence that the incidence of PPD may depend on birth outcome, mode of delivery (vaginal vs. cesarean), and/or preterm status [Silverman-et-al-2017], which is currently not considered in this cause model.

Abbreviations

Abbreviation

Definition

Note

PPD

Postpartum depression

MDD

Major depressive disorder

Disease Overview

The postpartum period is a time of intense transition and can cause new mothers and birthing parents to be vulnerable to psychiatric disorders. Depression episodes can be twice as likely during the postpartum period relative to other times of life and postpartum depression (PPD) can adversely affect the wellbeing of mothers and birthing parents, infants, and other family members.

PPD refers to non-psychotic depressive episodes in the postpartum period that persist for more than two weeks and shares the same diagnostic criteria as major depressive disorder in the DSM-5 [Shorey-et-al-2018]. Specifically, major depressive disorders are marked by depressed mood or loss of interest/pleasure that represent a change from the person’s baseline and impaired functioning observed across social, occupational, and educational domains. Additional symptoms may include excessive sleeping or insomnia; change in eating, appetite, or weight; agitated or slow motor activity; fatigue; feeling worthless or inappropriately guilty; trouble concentrating; and repeated thoughts about death. [GBD-2019-Capstone-Appendix-PPD]

GBD 2023 Modeling Strategy

Todo

Update information to GBD 2023 as necessary when methods appendix becomes available

Postpartum depression is not specifically estimated by the GBD. Rather, we will inform the prevalence of postpartum depression from a recent review and meta-analysis performed by [Shorey-et-al-2018-mncnh]. This review focused on healthy mothers with no prior history of postpartum depression, but found similar results to previous evaluations of postpartum depression prevalence among the broader population. [Shorey-et-al-2018-mncnh] reported that the incidence of postpartum depression was equal to 12% (95% CI 0.04–0.20).

While the GBD does not model PPD, it does model major depressive disorder. The GBD estimated that the average duration of major depressive disorder was 0.65 (95% UI: 0.59, 0.70) of one year (as informed through analysis of longitudinal studies on remission from major depressive disorder and the assumption that 40 years is the maximum duration of the condition), which we will use to inform the duration of postpartum depression in our simulation [GBD-2019-Capstone-Appendix-PPD-mncnh] (page 1013).

Additionally, the GBD models severity distributions and associated disability weights for major depressive disorder, summarized in the table below [GBD-2019-Capstone-Appendix-PPD-mncnh] (page 1013). Notably, the GBD defines major depressive disorder (MDD) as an episodic mood disorder involving the experience of one or more major depressive episode(s). The percent of case severity was determined from the US National Epidemiological Survey on Alchol and Related Conditions (conducted in two waves from 2001-2002 and 2004-2005) and the Australian National Survey of Mental Health and Wellbeing of Adults. Notably, burden due to major depressive disorder in GBD begins in the 1 to 4 year age group.

We will conservatively assume that postpartum depression is not a cause/contributor of maternal mortality due to suicide or other causes in our simulation. While the GBD estimates deaths due to self-harm, these estimates are not specific to the postpartum population and therefore may not be generalizable.

Vivarium Modeling Strategy

Scope

We will not model PPD as a dynamic transition model, but rather a probabilistic condition that begins at the time of birth and persists for some specified duration. The probability of experiencing PPD will be informed by a ratio per birth from the literature. PPD will be modeled as a YLD-only cause, although this is a limitation of our model as PPD may be associated with risk of suicide or infanticide in extreme cases.

Restrictions

The following table describes any restrictions in the effects of this cause (such as being only fatal or only nonfatal), as well as restrictions on the ages and sexes to which the cause applies.

Cause Restrictions

Restriction Type

Value

Notes

Male only

False

Female only

True

YLL only

False

YLD only

True

YLL age group start

10 to 14 (ID=7)

YLL age group end

50 to 54 (ID=15)

YLD age group start

10 to 14 (ID=7)

YLD age group end

50 to 54 (ID=15)

Assumptions and Limitations

  • We are limited in that the rate of PPD in our model is informed from a systematic literature review and meta-analysis that is not location-, year-, or age-specific and the analysis of PPD incidence primarily conducted in high-resource settings.

  • We are limited in that we do not consider mortality associated with PPD in our model.

  • We are limited in that we assume all PPD cases persist for the same average duration of a single MDD episode. This is limited in the sense that duration may vary by MDD severity (for example, a longer duration for more severe cases), which could cause our estimation of YLDs to be biased. Additionally, we assume that duration of PPD is equal to the duration of all MDD episodes.

  • We assume that the GBD MDD severity distribution, which is based on analysis of high-resource settings, generalizes to the severity of PPD in our simulation population of interest.

  • We assume that the onset of PPD occurs immediately following birth. However, the onset of PPD may peak around two or three months postpartum [Shorey-et-al-2018-mncnh].

  • We assume that the incidence of PPD does not vary by pregnancy outcome (incidence is constant across abortion/miscarriage/ectopic pregnancy, stillbirth, and live births)

Cause Model Decision Graph

Although we’re not modeling PPD dynamically as a finite state machine, we can draw an analogous directed graph that can be interpreted as a (collapsed) decision tree rather than a state transition diagram. The main difference is that the values on the transition arrows represent decision probabilities rather than rates per unit time.

Todo

Put an explanation like the following (but with more precision) on some central page (rather than on each individual model page):

To convert the graph to a decision tree, recursively split nodes with more than one incoming arrow until all nodes except the root have one incoming edge. Each time a node is split, all its outgoing edges are replicated, which may lead to additional downstream splits. Equivalently, the tree structure can be implicitly recovered by remembering the path taken to get to each node.

Jira ticket: https://jira.ihme.washington.edu/browse/SSCI-2006

Note

Only simulants who survive birth (do not die of a maternal disorder) are eligible to experience an incident case of postpartum depression

digraph ppd_decisions { rankdir = LR; start [label="start"] end [label="end"] alive [label="parent did not die of PPD"] dead [label="parent died of PPD"] start -> alive [label = "1 - ir"] start -> PPD [label = "ir"] PPD -> alive [label = "1 - cfr"] PPD -> dead [label = "cfr"] alive -> end [label = "1"] dead -> end [label = "1"] }

State Definitions

State

Definition

start

Parent simulant has already been through and survived the pregnancy and intrapartum components

PPD

Parent simulant has postpartum depression

parent alive

Parent simulant did not die of postpartum depression

parent dead

Parent simulant died of postpartum depression

end
Transition Probability Definitions

Symbol

Name

Definition

ir

incidence risk

The probability that a pregnant simulant gets postpartum depression

cfr

case fatality rate

The probability that a simulant with PPD dies due to PPD

Data Tables

Data values and sources

Variable

Definition

Value or source

Note

ir

postpartum depression incidence risk per birth

0.12 (95% CI 0.04, 0.20), truncated normal distribution (truncate at 0 and 1). Apply uncertainty as parameter uncertainty, not individual-level heterogeneity

[Shorey-et-al-2018-mncnh]

cfr

case fatality rate of postpartum depression

0

assumption

case duration

amount of time “infected” with PPD, in years

0.65 years (95% UI: 0.59, 0.70) truncated normal distribution truncated at 0 only. Apply uncertainty as parameter uncertainty, not individual-level heterogeneity

GBD 2019, 2021, and 2023 methods appendices for major depressive disorder [GBD-2019-Capstone-Appendix-PPD-mncnh] [GBD-2021-Capstone-Appendix-PPD-mncnh] [GBD-2023-Capstone-Appendix-PPD-mncnh]
Major depressive disorder sequelae

Severity

Percent of cases

Mean disability weight (95% UI)

Asymptomatic

14

0

Mild

59

0.145 (0.099, 0.209)

Moderate

17

0.396 (0.267, 0.531)

Severe

10

0.658 (0.477, 0.807)

The above values all come from the GBD methods appendix for major depressive disorder, and have not changed between [GBD-2019-Capstone-Appendix-PPD-mncnh], [GBD-2021-Capstone-Appendix-PPD-mncnh] and [GBD-2023-Capstone-Appendix-PPD-mncnh]. However, we did change the prevalence of asymptomatic from 13% to 14% to make the numbers add to 100 (presumably they only didn’t due to rounding error) and we have not used uncertainty for the severity split.

Note

Model uncertainty about the mean estimate of disability weight as parameter uncertainty following a truncated normal distribution with bounds of 0 and 1.

Calculating Burden

Years of life lost

There will be no YLLs due to PPD as we assume it is a nonfatal only cause.

Years lived with disability

For simulants who are determined to have an incident case of PPD, use the following instructions to calculate YLDs due to PPD:

  1. Assign a case severity according to the “percent of cases” columns in the “Major depressive disorder sequelae” table

  2. Multiply the case duration (found in the “Data values and sources” table) by the disability weight corresponding to the case severity assigned to that simulant to calculate YLDs due to PPD for that simulant.

Validation Criteria

  • Check population-level incidence rate is as expected

  • Check that all pregnancy outcomes experience PPD at the same incidence rate

  • Check that YLDs are as expected

References

[GBD-2019-Capstone-Appendix-PPD-mncnh] (1,2,3,4)

Appendix to: GBD 2019 Diseases and Injuries Collaborators. Global burden of 369 diseases and injuries in 204 countries and territories, 1990–2019: a systematic analysis for the Global Burden of Disease Study 2019. The Lancet. 17 Oct 2020;396:1204-1222

[GBD-2021-Capstone-Appendix-PPD-mncnh] (1,2)

Appendix to: Global incidence, prevalence, years lived with disability (YLDs), disability-adjusted life-years (DALYs), and healthy life expectancy (HALE) for 371 diseases and injuries in 204 countries and territories and 811 subnational locations, 1990–2021: a systematic analysis for the Global Burden of Disease Study 2021, The Lancet, Volume 403, Issue 10440, 2024, Pages 2133-2161, ISSN 0140-6736, https://doi.org/10.1016/S0140-6736(24)00757-8.

[GBD-2023-Capstone-Appendix-PPD-mncnh] (1,2)

Appendix to: Burden of 375 diseases and injuries, risk-attributable burden of 88 risk factors, and healthy life expectancy in 204 countries and territories, including 660 subnational locations, 1990–2023: a systematic analysis for the Global Burden of Disease Study 2023. The Lancet, Volume 406, Issue 10513, 1873 - 1922

[Shorey-et-al-2018-mncnh] (1,2,3,4)

Shorey S, Chee CYI, Ng ED, Chan YH, Tam WWS, Chong YS. Prevalence and incidence of postpartum depression among healthy mothers: A systematic review and meta-analysis. J Psychiatr Res. 2018 Sep;104:235-248. doi: 10.1016/j.jpsychires.2018.08.001. Epub 2018 Aug 3. PMID: 30114665.

[Silverman-et-al-2017]

Silverman ME, Reichenberg A, Savitz DA, Cnattingius S, Lichtenstein P, Hultman CM, Larsson H, Sandin S. The risk factors for postpartum depression: A population-based study. Depress Anxiety. 2017 Feb;34(2):178-187. doi: 10.1002/da.22597. Epub 2017 Jan 18. PMID: 28098957; PMCID: PMC5462547.