Inclusion/exclusion criteria

This review focuses on studies examining the private sector’s involvement in delivering quality healthcare to pregnant people, mothers, newborns and children (aged nine years and under) in low-income and middle-income countries. Using the World Bank Atlas method for the current 2020 fiscal year, this review will include countries classified as having low-income economies, lower-middle-income economies and upper-middle-economies.8 The population, interventions, control, outcomes, timeframe, setting (PICOTS) criteria, detailed in table 1, provide an overview of our inclusion criteria. Quantitative, qualitative and mixed-methods studies will be considered, provided that they report on at least one of the following critical outcomes:

• Maternal morbidity or mortality.

• Newborn morbidity or mortality.

• Child morbidity or mortality.

• Components of quality of care, defined by WHO as ‘the extent to which health care services provided to individuals and patient populations improve desired health outcomes. In order to achieve this, health care must be safe, effective, timely, efficient, equitable and people-centred’. (p1046)9

• Experience of care, including respectful care.

• Service utilisation.

Journal articles and grey literature (eg, reports, book chapters), both peer-reviewed and non-peer-reviewed, will be eligible for inclusion. In recognition of the rapid increase in public-private collaborations for health during the late 1990s,10 items must be published from 1 January 1995 to 30 June 2019. Items will be eligible if published in English, French, German or Italian. For inclusion, quality must be a specific focus among the interventions being implemented, such as quality assessment, quality improvement, clinical quality, perceived quality or the delivery of quality care. Interventions addressing universal health coverage will also be considered as focusing on quality, since quality health services are a fundamental feature of universal health coverage.11

As we are focused on service delivery, we are limiting the private sector to providers who deliver direct medical care (eg, private health facilities, private health providers, civil society organisations delivering care, charities delivering care). Thus, private sector entities that do not deliver direct medical care will be excluded. For example, we will exclude private organisations offering social service support to orphaned children, private pharmaceutical providers (including pharmacies) and private health insurance companies.

We will exclude studies reporting on aggregated service delivery data (ie, public sector and private sector outcome data combined), as we are only interested in outcomes resulting from service delivery by the private sector. Studies addressing allied fields like family planning, malaria and HIV/AIDS will be excluded unless they include specific MNCH interventions (eg, postpartum family planning, integration of family planning with maternal health services, abortion services, treatment of malaria in children, prevention of mother-to-child transmission of HIV) as a specific focus or aim. If commentaries, reviews and editorials do not introduce new evidence, then we will exclude them in favour of the original study or studies being discussed. Study protocols will also be excluded.

Search strategy and terms

We will search the following eight electronic databases that were assessed for likely coverage, availability and relevance of literature on the private sector:

• Cumulative Index to Nursing and Allied Health.

• EconLit.

• Excerpta Medica Database.

• International Bibliography of the Social Sciences.

• Popline.

• PubMed.

• ScienceDirect.

• Web of Science.

While we considered the LILACS (Latin American and Caribbean Health Sciences Literature) database, our test search returned no results. A test search in Open Grey returned one extraneous result, so we excluded that database as well. Based on discussions with experts and an initial assessment of the electronic databases, we suspect that a significant body of literature on the private sector is published as grey literature or programmatic reports. Thus, we will make a concerted effort to supplement our electronic database searches with expert-recommended articles and reports that we will obtain by sending a standardised request for literature to a list of experts and working groups identified in collaboration with colleagues at WHO.

Given our interest in locating grey literature, we will also conduct a targeted search of publications on two websites: Health Care Provider Performance Review (HCPPR) and the Maternal healthcare markets Evaluation Team (MET) at the London School of Hygiene & Tropical Medicine. HCPPR is a database of over 700 studies from a systematic review on the effectiveness of strategies to improve healthcare provider performance in low-income and middle-income countries.12 HCPPR’s search strategy included extensive searching of document inventories for unpublished studies. MET conducts multidisciplinary research on the role of public and private health sectors in delivering maternal healthcare. Its website includes both peer-reviewed literature and grey literature. We acknowledge risk of bias in locating studies. Our efforts to locate evidence on the provision of quality healthcare by the private sector may be limited, specifically when such evidence is only disseminated internally where it is beyond public reach.

Our searches will be conducted using combinations of search terms detailed in table 2. These terms aim to capture all studies examining the private sector, quality of care and MNCH. The terms were developed and tested for sensitivity, and they will be adapted to the basic search particulars (eg, wildcards (*), capacity for complex searches) of each electronic database.

Screening process

Searches and application of the inclusion/exclusion criteria will be conducted according to the PRISMA flow approach. Citation abstracts for all items uncovered by our searches will be exported into EndNote for screening. After first removing duplicates, remaining items will be screened for inclusion on the basis of title and abstract (TIAB). If inclusion/exclusion cannot be determined based on the TIAB, then the item will be pushed forward for full-text screening. SRL will conduct all TIAB and full-text screening, seeking guidance from BM on items considered borderline or problematic.

Data extraction

The data extraction form (online supplementary annex A) will be piloted with five randomly selected studies. Any necessary changes will be discussed among the authors prior to full data extraction. Based on the guidance in our codebook (online supplementary annex B), SRL will extract data for all studies into the data extraction template in Excel. Data will be extracted on the following categories:

• Background information (eg, author, date, setting, study objective).

• Intervention background information (eg, implementing agency, geographic level, study population).

• Intervention details (eg, intervention recipients, nature of intervention, dimensions of quality care).

• Critical outcomes (both quantitative and qualitative):

  • Maternal morbidity or mortality.

  • Newborn morbidity or mortality.

  • Child morbidity or mortality.

  • Quality of care.

  • Experience of care, including respectful care.

  • Service utilisation and efficiency.

• Secondary outcome (both quantitative and qualitative):

  • Infant and/or child growth.

• Evaluation/study details (eg, study type, data type, intervention claims, strategy effectiveness, cost data).

• Study quality (qualitative and quantitative).

Study quality and risks of bias will be assessed in duplicate using quality assessment tools for quantitative and qualitative studies. For quantitative studies, we will use the Effective Public Health Practice Project’s (EPHPP) quality assessment tool that rates public health studies on selection bias, study design, confounders, blinding, data collection methods, withdrawals and dropouts, intervention integrity and analysis.13 The accompanying reviewer’s dictionary provides details on how to evaluate each section (ie, strong, moderate or weak) with the process resulting in ‘excellent’ agreement between reviewers on the study’s final grade.13 14 For qualitative studies, we will use Miltenburg et al’s15 quality assessment tool based on criteria developed by Walsh and Downe.16 This tool rates qualitative studies’ scope and purpose, design, sampling strategy, analysis, interpretation, reflexivity, ethical dimensions and relevance and transferability.16 Like the EPHPP tool, Miltenburg et al’s tool involves rating each section as strong, moderate or weak and then using these ratings to determine the paper’s overall quality rating (ie, strong, moderate or weak).

Data synthesis

Once the data have been extracted, we will analyse the quantitative and qualitative findings separately. Depending on the quantitative studies and data available in the final inventory, a meta-analysis may be appropriate to examine the outcomes of private sector interventions on the provision of quality MNCH care. Data extracted on intervention details (see online supplementary annexes A and B, variables ‘Nature of Intervention$1’ through ‘Nature of Intervention$6’) will be reviewed for their suitability in meta-analyses. We will examine the effect of each type of private sector intervention affecting the delivery of care (eg, introduction or change to on-site support for quality improvement, introduction or change to data systems) on our six critical outcomes. When appropriate, we will conduct meta-analyses and generate forest plots in RevMan V.5.3.5 in non-Cochrane mode.17 For interventions in which the study sample size is large and interstudy heterogeneity is minimal, we will use fixed model effect; otherwise, we will use random effect size.18 The meta-analysis will be limited to randomised control trials and observational studies rated by the EPHPP tool as strong or moderate quality, and meta-biases of included studies will be assessed using ROBIS, a tool for assessing risk of bias in systematic reviews.19 If high heterogeneity exists between study designs, interventions, populations and definitions, then we will instead present the quantitative findings using a narrative synthesis to report the data thematically with tables of descriptive quantitative statistics and study outcomes.

For the qualitative data, we will enter the verbatim findings into NVivo for a three-step thematic synthesis.20 Each line of text will initially be coded based on its content and meaning. All codes will be included in a codebook with new codes added when appropriate. The codes will then be related together to develop categories and linkages between the data. After reviewing the data for analytical themes, we will make inferences about the provision of MHCH care by the private sector. We suspect that this type of synthesis will be particularly well suited to data on patients’ experiences of care by the private sector and existing mechanisms for engaging the private sector in delivering quality MNCH services.

Patient and public involvement

The design of this systematic review protocol did not involve patients, though we hope the final inventory of studies included in this review will report on patients’ experiences of care. The need for this systematic review was initiated by discussions with public sector health officials in Network countries.