Delivery Strategies for Postpartum Care

A Systematic Review and Meta-analysis

Abstract

OBJECTIVE: 

To systematically review the effects of postpartum health care–delivery strategies on health care utilization and maternal outcomes.

DATA SOURCES: 

We searched Medline, EMBASE, CENTRAL, CINAHL, and ClinicalTrials.gov for studies in the United States or Canada from inception to November 16, 2022.

METHODS OF STUDY SELECTION: 

We used duplicate screening for studies comparing health care–delivery strategies for routine postpartum care on health care utilization and maternal outcomes. We selected health care utilization, clinical, and harm outcomes prioritized by stakeholder panels.

TABULATION, INTEGRATION, AND RESULTS: 

We found 64 eligible studies (50 randomized controlled trials, 14 nonrandomized comparative studies; N=543,480). For general postpartum care, care location (clinic, at home, by telephone) did not affect depression or anxiety symptoms (low strength of evidence), and care integration (by multiple types of health care professionals) did not affect depression symptoms or substance use (low strength of evidence). Providing contraceptive care earlier (compared with later) was associated with greater implant use at 6 months (summary effect size 1.36, 95% CI 1.13–1.64) (moderate strength of evidence). Location of breastfeeding care did not affect hospitalization, other unplanned care utilization, or mental health symptoms (all low strength of evidence). Peer support was associated with higher rates of any or exclusive breastfeeding at 1 month and any breastfeeding at 3–6 months (summary effect size 1.10–1.22) but not other breastfeeding measures (all moderate strength of evidence). Care by a lactation consultant was associated with higher breastfeeding rates at 6 months (summary effect size 1.43, 95% CI 1.07–1.91) but not exclusive breastfeeding (all moderate strength of evidence). Use and nonuse of information technology for breastfeeding care were associated with comparable rates of breastfeeding (moderate strength of evidence). Testing reminders for screening or preventive care were associated with greater adherence to oral glucose tolerance testing but not random glucose or hemoglobin A_1c testing (moderate strength of evidence).

CONCLUSION: 

Various strategies have been shown to improve some aspects of postpartum care, but future research is needed on the most effective care delivery strategies to improve postpartum health.

SYSTEMATIC REVIEW REGISTRATION: 

PROSPERO, CRD42022309756.

In Brief

Various strategies improve delivery of some aspects of postpartum care, but more research is needed into the most effective strategies to improve postpartum health.

Postpartum health care aims for birthing people to thrive during their transition to parenthood. However, up to half of postpartum individuals in the United States do not receive any routine health care, with significant disparities across groups marginalized by racism, geographic location, immigrant status, and socioeconomic status.^1–5 These barriers are further exacerbated because many new mothers have difficulty traveling to clinics. Remote visits with information technology may help, but such visits are not routine and may affect care quality.

Even for those with health care access, coordination is often suboptimal. Care is provided by multiple professionals (eg, obstetricians, primary care physicians) and peer supporters (eg, doulas), each focusing on specific postpartum care aspects (eg, contraception, chronic health conditions, mental health, breastfeeding), often resulting in care fragmentation.^6,7 Care timing also may not meet current needs. In the United States, among the 65% of pregnancy-related deaths that occurred postpartum in 2017–2019, 12% occurred within 6 days after delivery, 23% from days 7 to 42, and 30% from 43 days to 1 year.⁸ Moreover, many challenges (eg, breastfeeding difficulties) present within 2 weeks postpartum. Although the American College of Obstetricians and Gynecologists recommends an initial interaction with the care professional within 3 weeks postpartum, followed by ongoing care as needed and a comprehensive visit by 12 weeks,⁶ optimal timing for postpartum care is not known.

It is unclear how postpartum outcomes are affected by important care aspects, including where, when, how, and by whom care is provided. We therefore conducted a systematic review for the Agency for Healthcare Research and Quality (AHRQ) and the Patient-Centered Outcomes Research Institute to inform the American College of Obstetricians and Gynecologists’ new guidance on postpartum care within 1 year postpartum.⁹ In this summary article based on the full systematic review report available on the AHRQ website,⁹ we address the following question: What health care–delivery strategies affect health care utilization and improve maternal outcomes within 1 year postpartum? Relevant delivery strategies included where, how, and when health care is provided; who provides health care and support; care coordination; information technology; and interventions targeting health care professionals.

SOURCES

We used AHRQ systematic review methodology and report this systematic review in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-analyses.^10,11 We engaged diverse clinical and methodology experts and patient representatives to finalize the research questions and methods.⁹ We registered the protocol through PROSPERO (CRD42022309756).

We searched for published studies from inception through November 16, 2022, in Medline, EMBASE, CENTRAL, and CINAHL and for unpublished studies in ClinicalTrials.gov using terms for postpartum, delivery strategies, and insurance coverage (for both research questions in the full report, available at https://effectivehealthcare.ahrq.gov/products/postpartum-care-one-year/research⁹) (Appendix 1, available online at https://links.lww.com/AOG/D279). We did not use any language restrictions. We also scanned the reference lists of relevant systematic reviews for additional studies.

Eight investigators independently screened each title and abstract using Abstrackr.¹² We rescreened (in duplicate) all accepted abstracts in full text. At both stages, discrepancies were resolved through team discussion or by a third investigator.

STUDY SELECTION

We included studies of postpartum individuals within 1 year after giving birth, which we defined as a live birth, fetal death or stillbirth, or induced abortion that occurred at 20 or more gestational weeks. To maximize relevance to the U.S. decision-making context, we included only studies conducted in the United States or Canada because of geographic proximity and similarity to the United States in terms of health care–delivery strategies. We included studies addressing general postpartum care or specific intervention targets (contraceptive care, breastfeeding care, or screening or preventive care) (see Supplemental Box 1, available online at https://links.lww.com/AOG/D338). For each intervention target, we used the Template for Intervention Description and Replication framework¹³ to categorize delivery strategies into where health care is provided (eg, hospital, clinic), how health care is provided (eg, dedicated postpartum care visit, as part of well-child visit), who provides health care or support (eg, obstetrician–gynecologist, midwife, doula), health care coordination (eg, patient navigators), information technology use (eg, bidirectional telemedicine), and interventions targeting health care professionals (eg, reminders) (see Supplemental Box 1, https://links.lww.com/AOG/D338).

Box 1.

List of Outcomes and Potential Effect Modifiers of Interest

OUTCOMES

• Health care utilization outcomes
• Attendance at postpartum visits*
• Unplanned care utilization (eg, unplanned readmissions, emergency room visits)*
• Adherence to condition-specific screening or testing (eg, blood pressure monitoring, glucose tolerance testing) or treatment*
• Transition to primary care professional for long-term care*
• Clinical outcomes (outcomes include incidence, prevalence or continuation, severity, and resolution as appropriate)
• Maternal mortality*
• Symptoms or diagnosis of mental health conditions (eg, anxiety, depression, substance use)*
• Patient-reported outcomes
• ▪ Quality of life (using validated measures)*
• ▪ Perceived stress*
• ▪ Pain
• ▪ Sleep quality
• ▪ Fatigue
• ▪ Sexual well-being and satisfaction
• ▪ Awareness of risk factors for long-term ill health
• Physical health or medical outcomes
• ▪ Postpartum onset of preeclampsia or hypertension
• ▪ Infections (eg, mastitis, wound infections)
• ▪ Severe maternal morbidity
• Cardiovascular disorders (eg, cardiomyopathy)
• Cerebrovascular disorders (eg, stroke)
• Bleeding
• Venous thromboembolism
• Other
• Interpregnancy interval
• Unplanned pregnancies
• Contraceptive initiation and continuation
• Breastfeeding intention, initiation, duration, and exclusivity
• Reduction in health inequities (eg, by race, ethnicity, geography, disability status)
• Harms
• Health inequities*
• Reported discrimination*
• Overutilization of health care
• Patient burden regarding postpartum care

POTENTIAL EFFECT MODIFIERS

• Patient-level factors
• Age
• Race and ethnicity
• Gender identity
• Sexual identity
• Physical disability status
• Education level
• Socioeconomic status
• Immigration status
• Refugee status
• Barriers to transportation to health care facility
• Paid family leave policies (eg, presence vs absence, different durations of leave)
• Access to internet (for virtual care or telehealth questions)
• Substance use or substance use disorder
• Type of insurance coverage (insured vs uninsured, private vs public [eg, Medicaid], insurance coverage of postpartum care, Medicaid insurance coverage extension or expansion)
• Presence vs absence of disorders of pregnancy (eg, hypertensive, cardiovascular, gestational diabetes mellitus) or peripartum complications that increase risk of postpartum complications
• Preterm vs term delivery
• Live birth vs stillbirth or spontaneous abortion or induced abortion
• Number of infants (eg, singleton vs twins or triplets)
• Presence vs absence of a supportive partner
• Infant health (eg, neonatal intensive care unit admission, congenital anomalies)
• Setting factors
• Country (United States vs Canada)
• Geographic location (urban vs suburban vs rural)
• Different levels of neighborhood vulnerability (eg, social vulnerability index)
• Volume of facility or hospital (high vs low)
• Type of facility or hospital (private vs public, community vs tertiary, academic vs nonacademic)
• Racial and ethnic concordance between provider and patient
• Language concordance between provider and patient

* Outcomes prioritized for assessment of strength of evidence and for making conclusions.

Outcomes included health care utilization, clinical outcomes, and harms (Box 1). We report outcomes prioritized by stakeholder panels (Box 1) and outcomes for which we could conduct meta-analyses. We evaluated whether results vary by patient-level (eg, age) and setting (eg, neighborhood vulnerability) factors. We included randomized controlled trials (RCTs) with 10 or more participants per group and prospective or retrospective nonrandomized comparative studies with 30 or more participants per group and analyses that accounted for confounders.

One investigator extracted data and assessed risk of bias for each study into the Systematic Review Data Repository Plus,¹⁴ which were verified by a second investigator. For risk of bias, we used items from the Cochrane Risk of Bias¹⁵ and the Risk of Bias in Nonrandomized Studies of Interventions tools.¹⁶

For dichotomous outcomes, we preferred risk ratios (RRs). For continuous outcomes, we preferred net mean differences for outcomes measured both at baseline and after intervention and mean differences for outcomes measured only after the intervention. When eligible studies reported only other measures of association, we extracted those data. For nonrandomized studies, we extracted only adjusted analyses. We considered categorical outcomes with a 95% CI that extends beyond both 0.50 and 2.0 imprecise. When three or more studies reported results from similar analyses, we conducted pairwise random-effects model meta-analyses using R.

We assessed strength of evidence per the AHRQ Methods Guide, considering risk of bias, consistency, precision, directness, and sparsity.¹⁰ For each prioritized outcome, we assigned a high, moderate, low, or insufficient strength of evidence rating.

RESULTS

Our electronic searches found 25,973 citations (Fig. 1). We screened 589 full-text articles and included 64 studies (80 articles,^17–96) in the current systematic review. Follow-up in the earliest study began in 1989 and in the last study ended in 2021. Although four studies ended follow-up in 2020, only one involved substantial follow-up during the coronavirus disease 2019 (COVID-19) pandemic (2021). The 64 studies included 50 RCTs and 14 nonrandomized comparative studies and comprised 543,480 postpartum individuals (range 34–154,163) (Appendix 2, available online at https://links.lww.com/AOG/D279). Fifty-five studies (86%) were in the United States; nine were in Canada.

Fig. 1.: Literature flow diagram. NRCS, nonrandomized comparative study; RCT, randomized controlled trial.

Participant average ages were similar across studies (range 17–34 years) except for two adolescent studies^57–60,94 (Appendix 3, available online at https://links.lww.com/AOG/D279). Three studies enrolled exclusively Black individuals, and one enrolled exclusively Hispanic individuals. With few exceptions, participants across the other studies were racially diverse. Other than two studies of participants with cesarean deliveries, 57–88% of deliveries were vaginal. Other than 12 studies of participants with term deliveries, 3–27% of deliveries were preterm. Sixteen studies excluded pregnancies resulting in stillbirths, spontaneous or induced abortions, or neonatal deaths. The intervention targets were general postpartum care (18 studies), contraceptive care (11 studies), breastfeeding care (29 studies), and screening or preventive care (six studies) (Table 1, https://links.lww.com/AOG/D339, and Appendix 4, available online at https://links.lww.com/AOG/D279). Table 2, available online at https://links.lww.com/AOG/D340, lists the outcomes reported in all 64 studies.

Among the 50 RCTs, we rated five as low risk of bias, 25 as moderate risk, and 20 as high risk, generally related to lack of blinding and incomplete outcome data (Appendix 5, available online at https://links.lww.com/AOG/D279). Among the 14 nonrandomized comparative studies, we rated nine as moderate risk of bias and five as high risk, related to confounding and lack of blinding (Appendices 6 and 7, available online at https://links.lww.com/AOG/D279).

In the following sections, organized by intervention target and strategies for delivering each intervention target, we summarize results for all prioritized and meta-analyzed outcomes (Appendices 8–29, available online at https://links.lww.com/AOG/D279) and state only conclusions that were feasible on the basis of at least low strength of evidence (Table 3, available online at https://links.lww.com/AOG/D341).

Eighteen studies (12 RCTs, six nonrandomized comparative studies) in 27 articles^{20,23,24,26,32,33,37,40,43,53,54,57–62,68,70,72,73,76,77,84,88–90} addressed general postpartum care. Six studies addressed where care is provided, four addressed how, three addressed when, eight addressed who, and two addressed care coordination.

For where general postpartum care is provided, one RCT (Dodge et al^32,40) and one nonrandomized comparative study (Arias et al²⁰) reported inconsistent results regarding attendance at postpartum visits (Appendix 10, https://links.lww.com/AOG/D279). Dodge et al^32,40 reported that participants randomized to home visits from nurses and those randomized to usual care had comparable attendance. However, Arias et al²⁰ reported that virtual visits during the COVID-19 pandemic were associated with greater visit attendance than before (adjusted odds ratio [aOR] 1.90, 95% CI 1.47–2.46). The RCT (Dodge et al^32,40) reported that participants randomized to home visits from nurses had more emergency room visits than participants randomized to usual care (mean difference 0.21, 95% CI 0.01–0.40); however, the numbers of hospitalizations per participant were comparable (Appendix 8, https://links.lww.com/AOG/D279). The nonrandomized comparative study (Arias et al²⁰) reported that compared with before the COVID-19 pandemic, virtual visits during the pandemic were associated with higher depression screening rates (aOR 4.61, 95% CI 3.38–6.28) but comparable oral glucose tolerance testing (OGTT) rates (Appendix 12, https://links.lww.com/AOG/D279).

Two RCTs (Dodge et al,^32,40 McCarter et al⁶⁸) reported that visits at home (or by telephone) or at a clinic were associated with comparable depression and anxiety symptoms (Appendix 14, https://links.lww.com/AOG/D279). Dodge et al^32,40 reported that participants with home visits from nurses and those with usual care had comparable rates of possible depression (EPDS [Edinburgh Postpartum Depression Scale] score higher than 10) or anxiety (Generalized Anxiety Disorder-7 score higher than 5) diagnosis (β coefficient −7.70, 95% CI −16.7 to 1.33). Similarly, McCarter et al⁶⁸ reported that, compared with usual care, telephone-based nursing was associated with comparable numbers of participants with scores on the EPDS of 10–12 (RR 0.62, 95% CI 0.24–1.58) and higher than 12 (RR 2.42, 95% CI 0.78–7.45). Because of a high risk of bias, there is low strength of evidence that whether general postpartum care is provided at home (or by telephone) or at the clinic does not affect depression or anxiety symptoms.

For how general postpartum care is provided, one RCT (Polk et al⁷⁷) reported that combined or separate postpartum and well-child visits were associated with comparable 1-month postpartum visit attendance (Appendix 10, https://links.lww.com/AOG/D279). Two RCTs (Laliberté et al,⁶² Hans et al^53,54) reported inconsistent and imprecise results regarding unplanned care utilization (Appendix 11, https://links.lww.com/AOG/D279). Laliberté et al⁶² reported that participants who attended a multidisciplinary clinic within 48 hours postpartum (with additional visits as indicated) and participants with standard postpartum care experienced comparable emergency room visits and hospitalizations rates, although the estimate for readmissions was imprecise. Hans et al^53,54 reported an imprecise estimate for hospitalizations by 3 weeks comparing those who did and those who did not receive case management by community case managers or social services professionals (RR 1.34, 95% CI 0.31–5.89). One RCT (Polk et al⁷⁷) reported that combined or separate postpartum and well-child visits were associated with comparable primary care visit attendance by 1 year (Appendix 10, https://links.lww.com/AOG/D279).

Three RCTs (Koniak-Griffin et al,^57–60 Hans et al,⁵⁴ Laliberté et al⁶²) reported clinical outcomes comparing integrated care (ie, care provided by multiple types of professionals) and nonintegrated care (Appendices 9 and 14, https://links.lww.com/AOG/D279). All three RCTs reported that individuals cared for in integrated and nonintegrated care had comparable depression symptoms. Koniak-Griffin et al^57–60 also reported that participants in the Early Intervention Program (17 home visits and four prenatal classes by public health nurses) and participants with Traditional Public Health Nursing Care had comparable alcohol, cannabis, or tobacco use in the past month and comparable perceived stress (using the Perceived Stress Scale) at 1 year (Appendix 9, https://links.lww.com/AOG/D279). Because of a high risk of bias, there is low strength of evidence that integration of care (eg, combined compared with separate postpartum and well-child visits, multidisciplinary postpartum clinic compared with standard care) does not affect depression symptoms or substance use up to 1 year postpartum.

For when general postpartum care is provided, two RCTs (Bernard et al,^23,24 Pluym et al^37,76) and one nonrandomized comparative study (Chen et al³³) reported visit attendance (Appendix 10, https://links.lww.com/AOG/D279). Bernard et al^23,24 and Pluym et al^37,76 reported comparable 6-week attendance among participants randomized to two visits (2 or 3 weeks and 6 weeks) or one visit (6 weeks). However, Chen et al³³ reported that participants scheduled for a 6-week visit were less likely than participants scheduled for a 2- or 3-week visit to attend a visit by 3 months (aOR 0.42, 95% CI 0.24–0.74). One RCT (Pluym et al^37,76) reported comparable emergency room visit rates among participants randomized to one or two visits (Appendix 11, https://links.lww.com/AOG/D279).

For who provides general postpartum care, one RCT (Dodge et al^32,40) and two nonrandomized comparative studies (Buckley,²⁶ Pan et al⁷³) reported data (Appendix 10, https://links.lww.com/AOG/D279). Dodge et al^32,40 and Buckley²⁶ evaluated care by nurses or nurse practitioners and reported inconsistent results regarding visit attendance. Dodge et al^32,40 reported that participants who did and those who did not receive care from nurses had comparable visit rates. However, Buckley²⁶ reported that compared with no contact with nurse practitioners, visits and phone calls with them were associated with greater attendance (P<.02, adjusted estimate not reported). Pan et al⁷³ reported that community health worker home visits with referrals to social workers (the Baby Love Program) were associated with greater attendance at 2-month visits (aOR 1.46, 95% CI 0.93–2.31).

Two RCTs (Dodge et al,^32,40 Hans et al^53,54) reported on unplanned health care utilization (Appendices 8 and 11, https://links.lww.com/AOG/D279). Dodge et al^32,40 reported that participants with home visits from nurses had more emergency room visits by 1 year than participants with usual care (mean difference 0.21, 95% CI 0.01–0.40). However, the numbers of hospitalizations per participant by 1 year were comparable. Hans et al^53,54 reported that participants with doula home visits and participants with case management by community case managers or social service professionals had comparable hospitalizations by 3 weeks, but the estimate was imprecise.

Four RCTs reported clinical outcomes for comparisons between various care or support providers (Appendices 9 and 14, https://links.lww.com/AOG/D279). Hans et al^53,54 reported that participants with doula home visits or participants with case management by community case managers or social service professionals had comparable significant depression symptom (Center for Epidemiologic Studies Depression Scale score 15 or higher) rates at 3 weeks and 3 months.^53,54 Tandon et al⁸⁹ reported comparable 6-month depression symptom scores using the Quick Inventory of Depression Symptoms comparing participants with usual home visits (visitor type unspecified) and participants with home visits by mental health professionals or by community health workers. Major depression episode rates by 6 months were also comparable, but the effect sizes were imprecise. Edwards and Sims-Jones⁴³ reported that visits and telephone calls from public health nurses, telephone calls from health department clerks, and postpartum education packages were all associated with comparable depression diagnosis rates by 3 months.⁴³ Dodge et al^32,40 reported that participants with home visits from nurses and participants with usual care had comparable possible depression or anxiety rates by 6 months.

For general postpartum care coordination: one nonrandomized comparative study (Tsai et al⁹⁰) reported that providing patients information on their first postpartum appointment while still in the hospital was associated with a greater likelihood of attending visit 1 by 3 weeks (P=.014) and visit 2 by 7 weeks (P=.025) (Appendix 10, https://links.lww.com/AOG/D279). Adjusted estimates were not reported.

Eleven RCTs in 16 articles^{21,29–31,35,36,38,49,51,55,63,65,71,86,95,97} addressed contraceptive care. One RCT addressed how care is provided, nine addressed when, and one addressed who.

For how contraceptive care is provided, one RCT that addressed this (Haider et al⁵¹) did not report any prioritized or meta-analyzed outcomes.

For when contraceptive care is provided, one RCT (Chen²⁹) reported that participants who received depot medroxyprogesterone acetate before discharge and those who received it at 4–6 weeks had comparable EPDS depression symptom scores at 2 months (Appendix 9, https://links.lww.com/AOG/D279). Nine RCTs reported contraceptive use at various time points (3, 6, and 12 months) (Appendix 13, https://links.lww.com/AOG/D279). Meta-analyses suggested that compared with later contraception (typically at 6–8 weeks), earlier contraception (eg, soon after delivery, before hospital discharge, at 2–3 weeks) was associated with greater implant use at 6 months (summary RR 1.36, 95% CI 1.13–1.64; Fig. 2), although combining the two studies largely recapitulated the Morse⁷² findings. Intrauterine device use was comparable at 3 months (Appendix 22, https://links.lww.com/AOG/D279) and 6 months (Fig. 2). Because of the moderate risk of bias, there is moderate strength of evidence that earlier care and later care are associated with comparable intrauterine device use at 3 and 6 months but greater implant use at 6 months.

Fig. 2.: Earlier vs later contraceptive placement: continued contraceptive use at 6 months. RoB, risk of bias; RR, relative risk; IUD, intrauterine device; I ², measure of statistical heterogeneity (percentage of total variability that is attributable to between-study variability).

For who provides contraceptive care, one RCT that addressed this (Simmons et al⁸⁶) did not report any prioritized or meta-analyzed outcomes.

Twenty-nine studies (25 RCTs, four nonrandomized comparative studies) in 34 articles^{17–19,22,25,27,28,39,42,44–48,50,52,56,64,66,67,74,75,78–83,87,91–94,96} addressed breastfeeding care. Eight studies addressed where care is provided, two addressed how, 19 addressed who, seven addressed information technology use, and two addressed interventions targeting care professionals.

For where breastfeeding care is provided, one RCT (Gagnon et al⁴⁷) reported that a single home breastfeeding support visit by a community nurse or a single hospital visit with a nurse 3–4 days after discharge was associated with comparable attendance at the 2-week visit (Appendix 10, https://links.lww.com/AOG/D279). Four RCTs reported unplanned care utilization (Appendix 11). Whether the initial visit was at home or at a clinic was associated with comparable hospitalizations (summary RR 1.38, 95% CI 0.90–2.13) (Appendix 21, https://links.lww.com/AOG/D279), urgent care visits (two RCTs), outpatient visits (one RCT), and emergency room visits (one RCT).

Four RCTs reported mental health outcomes (Appendices 9 and 14, https://links.lww.com/AOG/D279). Participants receiving the initial breastfeeding support visit at home or the pediatric clinic experienced comparable depression symptoms (three RCTs) and anxiety symptoms (two RCTs). Because of the high risk of bias, there is low strength of evidence that whether breastfeeding care is provided at home or at a pediatric clinic does not affect hospitalization, other unplanned care utilization, depression symptoms, or anxiety symptoms.

For how breastfeeding care is provided, the two nonrandomized comparative studies that addressed this (Rozga et al⁸³ and Witt et al⁹⁶) did not report any prioritized or meta-analyzed outcomes.

For who provides breastfeeding care, one nonrandomized comparative study (Falconi et al⁴⁶) reported that participants who did and those who did not receive doula support had comparable emergency room visit or hospitalization rates by 1 and 2 months (Appendix 11, https://links.lww.com/AOG/D279). One nonrandomized comparative study (Falconi et al⁴⁶) reported that participants who did and those who did not receive doula support had comparable severe maternal morbidity or mortality by 2 months (Appendix 15, https://links.lww.com/AOG/D279) and anxiety or depression by 2 months (Appendix 14, https://links.lww.com/AOG/D279).

Eighteen RCTs reported various breastfeeding outcomes, comparing care provided by different groups (Appendices 16–20, https://links.lww.com/AOG/D279).

• Peer support (10 RCTs). Meta-analyses suggested that peer support was associated with higher any breastfeeding rates at 1 month (summary effect size 1.13, 95% CI 1.03–1.24; Fig. 3) and 3–6 months (summary RR 1.22, 95% CI 1.06–1.41; Fig. 4) and higher exclusive breastfeeding rates at 1 month (summary RR 1.10, 95% CI 1.02–1.19; Fig. 5) but not 3 months (Fig. 5). Meta-analyses also suggested that peer support was not associated with nonexclusive breastfeeding rates at 1 month (Appendix 23, https://links.lww.com/AOG/D279) or 3 months (Appendix 24, https://links.lww.com/AOG/D279).
• Doulas (one RCT). One RCT (Edwards et al^44,52) reported that comparable proportions of participants with and those without home visits by doulas breastfed for less than 1.5 months, 1.5–4 months, and more than 4 months (Appendix 19, https://links.lww.com/AOG/D279). However, participants with doula visits were less likely to not initiate breastfeeding by 4 months (RR 0.74, 95% CI 0.54–1.02) (Appendix 20, https://links.lww.com/AOG/D279).
• Midwives (one RCT). One RCT (Porteous et al⁷⁸) reported that compared with participants without midwife visits, participants with midwife visits were more likely to breastfeed at 1 month (RR 1.47, 95% CI 1.12–1.92; Appendix 16, https://links.lww.com/AOG/D279) and to exclusively breastfeed at 1 month (RR 2.35, 95% CI 1.36–4.06; Appendix 17, https://links.lww.com/AOG/D279). However, comparable proportions of participants with and without midwife visits breastfed nonexclusively at 1 month (Appendix 18, https://links.lww.com/AOG/D279).
• Lactation consultants (seven studies). Meta-analyses suggested that lactation consultant care was associated with comparable any breastfeeding rates at 1 month (four RCTs; Appendix 25, https://links.lww.com/AOG/D279) and 3 months (Appendix 26, https://links.lww.com/AOG/D279) but higher rates at 6 months (summary effect size 1.43, 95% CI 1.07–1.91; Fig. 6). Meta-analyses also suggested that lactation consultant care was associated with comparable exclusive breastfeeding rates at 1 or 3 months (Appendix 27, https://links.lww.com/AOG/D279).

Fig. 3.: Peer support vs no peer support: any breastfeeding at 1 month. RoB, risk of bias; ES, effect size; RR, relative risk; I ², measure of statistical heterogeneity (percentage of total variability that is attributable to between-study variability); aOR, adjusted odds ratio.

Fig. 4.: Peer support vs no peer support: any breastfeeding at 3–6 months. RoB, risk of bias; RR, relative risk; I ², measure of statistical heterogeneity (percentage of total variability that is attributable to between-study variability).

Fig. 5.: Peer support vs no peer support: exclusive breastfeeding at 1 and 3 months. RoB, risk of bias; ES, effect size; I ², measure of statistical heterogeneity (percentage of total variability that is attributable to between-study variability).

Fig. 6.: Lactation consultant vs no lactation consultant: any breastfeeding at 6 months. RoB, risk of bias; ES, effect size; RR, relative risk; I ², measure of statistical heterogeneity (percentage of total variability that is attributable to between-study variability); aOR, adjusted odds ratio.

Because of the moderate risk of bias, there is moderate strength of evidence for these conclusions regarding who provides breastfeeding care:

• Compared with no peer support, peer support is associated with higher rates of any breastfeeding at 1 and 3–6 months and of exclusive breastfeeding at 1 month but with comparable rates of exclusive breastfeeding at 3 months and nonexclusive breastfeeding at 1 and 3 months.
• Compared with no lactation consultants, care by a lactation consultant is associated with higher rates of any breastfeeding at 6 months but not at 1 or 3 months. Lactation consultant care is probably associated with comparable rates of exclusive breastfeeding at 1 or 3 months.

For information technology use for breastfeeding care, one RCT (Bender et al^22,42) reported that text message–based breastfeeding support and usual care were associated with comparable attendance at visits (Appendix 10, https://links.lww.com/AOG/D279). Similar results were obtained for Black and non-Black participants when analyzed separately.

One RCT (Ahmed et al¹⁸) reported that receiving interactive web-based monitoring for breastfeeding care was associated with comparable EPDS depression symptom scores at 1 and 3 months (Appendix 9, https://links.lww.com/AOG/D279). Six RCTs reported data on breastfeeding. Meta-analyses suggested that information technology use was associated with comparable any breastfeeding rates at 3 months (summary RR 1.00, 95% CI 0.92–1.09; Appendix 28, https://links.lww.com/AOG/D279) and 6 months (summary RR 1.01, 95% CI 0.89–1.14; Appendix 28, https://links.lww.com/AOG/D279) and comparable exclusive breastfeeding rates at 3 months (summary RR 1.28, 95% CI 0.81–2.03; Appendix 29, https://links.lww.com/AOG/D279). Because of the moderate risk of bias, there is moderate strength of evidence that information technology use and nonuse for breastfeeding are associated with comparable rates of any breastfeeding at 3 and 6 months and of exclusive breastfeeding at 3 months.

For interventions targeting breastfeeding care providers, two RCTs that addressed this (Bonuck et al 2014a and 2014b²⁵) did not report any prioritized or meta-analyzed outcomes.

Six studies (four RCTs, two nonrandomized comparative studies) in six articles^{34,41,56,69,85,89} addressed screening or preventive care. Two studies addressed who provides care, three addressed care coordination, one addressed information technology use, and two addressed interventions targeting care professionals.

For who provides screening or preventive care, two RCTs (Kerver et al,⁵⁶ Tandon et al⁸⁹) reported data. Kerver et al⁵⁶ reported that no maternal deaths occurred in either the peer-counseling or no-peer-counseling groups (Appendix 15, https://links.lww.com/AOG/D279). Tandon et al⁸⁹ reported comparable 6-month depression symptom scores (using the Quick Inventory of Depression Symptoms) between participants with usual home visits (visitor type unspecified) and those with home visits by mental health professionals or by community health workers (Appendix 9, https://links.lww.com/AOG/D279). Major depression episode rates by 6 months were also comparable, but the effect sizes were imprecise.

For screening or preventive care coordination, one RCT (Clark et al³⁴) and two nonrandomized comparative studies (Shea et al,⁸⁵ Mendez-Figueroa et al⁶⁹) reported adherence to OGTT (Appendix 12, https://links.lww.com/AOG/D279). Clark et al³⁴ reported that compared with no reminders, providing patient mail reminders was associated with greater adherence by 1 year (odds ratio [OR] 8.7, 95% CI 2.9–25.6), as was providing clinician electronic medical record (EMR) and patient mail reminders (OR 5.2, 95% CI 1.4–19.6). Shea et al⁸⁵ reported that compared with no mail or telephone reminders, providing mail or telephone reminders was associated with greater adherence by 6 months (P=.01; adjusted estimate not reported). Similarly, Mendez-Figueroa et al⁶⁹ reported that telephone reminders were associated with greater adherence by 1.5 months (P<.001; adjusted estimate not reported).

Two studies (Clark et al,³⁴ Shea et al⁸⁵) also reported adherence to other glucose-related testing. For random glucose testing and for hemoglobin A_1c testing, clinician EMR reminders, the clinician EMR and patient mail reminder combination (Clark et al³⁴), and mail or telephone reminders (Shea et al⁸⁵) were not associated with greater adherence.

For other testing, the two studies were inconsistent. For fasting glucose testing, Clark et al³⁴ reported that compared with no reminders, providing patient mail reminders was associated with greater adherence (OR 4.6, 95% CI 1.4–20.0), as was providing clinician EMR and patient mail reminders (OR 5.3, 95% CI 1.9–11.5). However, Shea et al⁸⁵ reported that mail or telephone reminders were not associated with greater adherence. Similarly, for any glucose testing, Clark et al³⁴ reported that compared with no reminders, providing patient mail reminders was associated with greater adherence (OR 5.4, 95% CI 2.1–13.5), as was providing clinician EMR and patient mail reminders (OR 5.5, 95% CI 1.4–21.3). Shea et al⁸⁵ reported that mail or telephone reminders were not associated with greater adherence. Because of the moderate risk of bias, there is moderate strength of evidence that testing reminders are associated with greater adherence to OGTT up to 1 year postpartum but not random glucose or hemoglobin A_1c testing.

For information technology use for screening or preventive care, one nonrandomized comparative study (Shea et al⁸⁵) reported that the group who were provided mail or telephone reminders, the group who were provided mail reminders only, and the group who were provided no reminders had comparable adherence to OGTT, random or fasting glucose testing, or hemoglobin A_1c testing at 6 months (no statistically significant P values; no adjusted estimates reported) (Appendix 12, https://links.lww.com/AOG/D279).

In terms of interventions targeting screening or preventive care professionals, one RCT (Clark et al³⁴) and one nonrandomized comparative study (Domingo et al⁴¹) reported adherence to OGTT (Appendix 12, https://links.lww.com/AOG/D279). Clark et al³⁴ reported that compared with no reminders, clinician EMR reminders were associated with greater adherence (OR 8.4, 95% CI 2.4–28.5), and so were combined clinician EMR and patient mail reminders (OR 5.2, 95% CI 1.4–19.6). However, Domingo et al⁴¹ reported that compared with no reminders, physician EMR reminders were associated with comparable adherence.

Clark et al³⁴ also reported adherence to other testing. Compared with no reminders, clinician EMR reminders were associated with greater adherence to fasting glucose testing (OR 4.2, 95% CI 1.4–12.3), and so were clinician EMR and patient mail reminders (OR 5.3, 95% CI 1.9–11.5). Similarly, compared with no reminders, clinician EMR reminders were associated with greater adherence to any glucose testing (OR 4.2, 95% CI 1.4–12.5), and so were clinician EMR and patient mail reminders (OR 5.5, 95% CI 1.4–21.3). However, for random glucose testing and for hemoglobin A_1c testing, neither clinician EMR reminders nor the clinician EMR and patient mail reminder combination was associated with greater adherence.

DISCUSSION

The evidence identified in this systematic review comprised 64 United States– and Canada-based studies that enabled us to make some low or moderate strength of evidence conclusions. For general postpartum care, neither care location nor care integration affected outcomes. For contraceptive care, earlier (compared with later) care was associated with greater implant use at 6 months. For breastfeeding care, although neither care location nor information technology use influenced outcomes, peer support and lactation consultant care were each associated with higher breastfeeding rates. Reminders improved adherence to OGTT but not random glucose or hemoglobin A_1c testing.

Regarding where care is provided, given that location of general postpartum care and breastfeeding care may not affect outcomes, those deciding on locations for these intervention targets should consider maternal preference and care access. Notably, only 1 of the 64 studies, which evaluated virtual care, involved follow-up substantively into the COVID-19 pandemic. Other delivery strategies with comparable outcomes were general postpartum care integration (compared with nonintegration) and information technology use (compared with nonuse) for breastfeeding care. It is possible that the reason some studies found no difference in outcomes may be that the comparator groups were still receiving more attention and reminders about care than the broader U.S. postpartum population.

For some strategies, we found differences in benefits such as for earlier (compared with later) contraceptive care (for implants), for breastfeeding care by peers and by lactation consultants, and for screening coordination through OGTT reminders. These findings suggest that additional care outreach may influence outcomes more than details on the care modality or timing do.

Only 28% of the studies targeted general postpartum care; future research should address general postpartum care. More research is also needed on interventions targeting health care professionals, for which no conclusions were feasible in this systematic review.

Most studies included in this systematic review enrolled predominantly healthy individuals. Researchers should also design studies that include individuals at high risk of complications. How various delivery strategies affect visit attendance and health outcomes in these population subgroups may vary. When these various subpopulations are enrolled in a larger study, subgroup-specific data should be adequately analyzed and reported.

A strength of the identified evidence is that there generally was racial and age diversity of study participants across the studies included in this systematic review. Given the increasing diversity of the U.S. postpartum population, the conclusions from this systematic review apply generally to postpartum individuals in the United States. We followed rigorous standards during each step of this systematic review. We used a rigorous framework (Template for Intervention Description and Replication)¹³ to conceptualize specific delivery strategies for postpartum intervention targets.

A few limitations are noteworthy. Although we found studies in each broad category of delivery strategies, we could not make any high–strength of evidence conclusions. To make meaningful comparisons, we had to parse the evidence separately for general postpartum care and for specific intervention targets (ie, contraceptive care, breastfeeding care, screening and preventive care), and for certain intervention target and delivery strategy combinations, we had to parse the evidence further. Table 1 (https://links.lww.com/AOG/D339) illustrates how the evidence “thinned out.” For example, we analyzed breastfeeding care studies of doulas separately from breastfeeding care studies of lactation consultants. Another limitation is that many prioritized outcomes (including harms) were inadequately reported to merit conclusions.

Despite the broad scope of the systematic review, we could make few conclusions, all at low or moderate strength of evidence. Future research should compare various care delivery strategies, particularly related to interventions targeting health care professionals. Patient-reported outcomes (eg, quality of life, perception of person-centered care) and maternal morbidity and mortality should also be reported. Researchers should report population subgroup-specific data so that decision makers can understand how postpartum care delivery strategies may affect different subpopulations. Such evidence could inform postpartum care strategies to close the wide and important gaps in outcomes by race in the United States.