This systematic review and meta-analysis will answer the following questions:

1.   What is the overall perinatal mortality rate and how does this vary in different contexts (geographic location, study setting) in East Africa?

2.   What are the causes of perinatal mortality in East Africa?

3.   What are the risk factors for perinatal mortality in East Africa?

The design and implementation of this systematic review will adhere to the Preferred Reporting Items for Systematic Review and Meta-Analysis Protocols (PRISMA-P) 2015 statement ¹⁷ and reporting of findings will follow the Preferred Reporting Items for Systematic review and Meta-Analyses (PRISMA-2020) updated guideline. ¹⁸ The protocol for this review was registered on PROSPERO ( CRD42021291719).

Studies for this systematic review and meta-analysis will be selected based on the criteria specified below.

We will include both published and unpublished studies, that report perinatal mortality (stillbirth and/or early neonatal death), its causes and/or risk factors of perinatal mortality. No restrictions will be imposed on language of publication, sex, or ethnicity of participants. This study will include studies that have been conducted in East Africa and published between January, 2010 – June, 2022.

We will exclude studies which are reviews, or published outside of the study area and time period (before 2010 and after June, 2022). We will also exclude studies that focus on specific populations (e.g., high-risk mothers). Studies will be excluded if extracting data is not feasible after appropriate attempts to seek the full text and contact the corresponding author where needed. We will exclude studies that are limited in methodology (inappropriate statistical analysis or methods used to control confounders).

Population: All births (both livebirths and stillbirths) with ≥500g/≥22 weeks of gestation and newborn deaths within the first week after birth (0–6 days) ¹⁹ .

Exposure: Determinants or risk factors of perinatal mortality. The determinants or risk factors are characteristics or exposures that increase the likelihood of perinatal mortality. These may be related to distal, underlying, or proximal determinants.

Comparison: The reported reference group for each determinant or risk factor in each study (e.g., perinatal mortality in mothers with antenatal care versus mothers with no antenatal care).

Outcome: Perinatal mortality rate, which is defined as “the total number of deaths of a fetus with birth weight of 500 grams or more or a gestational age of 22 completed weeks of age or more until the 7 ^th day after delivery per 1000 live birth” ²⁰ . The other outcomes for this study are the causes and risk factors of perinatal mortality.

All observational studies (cross-sectional, case-control, prospective cohort and retrospective studies) and community-based trials which reported the magnitude of perinatal mortality and/or its cause or risk factors will be included.

Studies that have been conducted in East Africa, which encompasses the following countries; Burundi, Comoros, Djibouti, Eritrea, Ethiopia, Kenya, Madagascar, Malawi, Mauritius, Mozambique, Rwanda, Seychelles, Somalia, Somaliland, South Sudan, Sudan, Tanzania, Uganda, Zambia, and Zimbabwe according to United Nations ²¹ will be considered. Both community-based and facility-based studies will be used for this study. This study will be conducted from October 2022 to February 2023.

All studies published from January 1, 2010 to June 30, 2022 and all articles reported in any language will be considered for this study. We will consult a specific language expert for translation for studies in a language the authors do not speak.

All studies that fulfill the eligibility criteria will be considered regardless of their publication status (published, and unpublished or grey literature). To access unpublished reports of likely high relevance and quality, up-to-date data will be requested from large surveillance studies, forming an investigator group.

The databases searched to identify published research articles will be Medline, Web of Science, EMBASE, Global Health, SCOPUS, Cochrane Library, CINAHL, HINARI, African Index Medicus, African Journals Online (AJOL), and WHO African Regional Office (AFRO) Library. In addition to this, a manual search will be performed to retrieve unpublished studies and grey literature via Google Scholar, Google and institutional repositories of higher education institutions, which are found in East Africa and outside the region that have joint projects in East Africa. An investigator group from large studies with ongoing surveillance in the region will be requested from Child Health and Mortality Prevention Surveillance (CHAMPS) and Health and Demographic Surveillance sites (HDSS) to facilitate the inclusion of the most up-to-date data. Three CHAMPS networks in the region, namely Harar and Kersa, Ethiopia, Siaya and Kisumu, Kenya, and Manhiça, Mozambique ¹ , and HDSS from Ethiopia (Harar and Kersa, Dabat, Butajira, Arba Minch and Gilgel Gibe HDSS), Kenya (Nairobi, Kilifi, Mbita, Kombewa and Kisumu HDSS), Malawi (Blantyre, Karonga HDSS), Mozambique (Chokwe and Manhica HDSS), Tanzania (Magu, Rufiji, Bagamoyo, Ifakara, Korogwe, Moshi and Pemba HDSS), Uganda (Awach; Gulu, Iganga/Mayuge, Kyamulibwa, Rakai and Toro HDSS), Zambia (Lusaka HDSS) ^{22, 23} will be asked to join the investigator group if they have appropriate data, which they are able to contribute.

We will search the electronic databases above, based on the following concepts: perinatal mortality (stillbirth and/or early neonatal mortality), causes of perinatal mortality, stillbirths or early neonatal mortality, risk factors for perinatal mortality, stillbirths or early neonatal mortality, study design (cross-sectional, case-control, cohort and community-based trial) and location and geographic setting (countries of Eastern Africa), and published covering the time period from January 1, 2010 to June 30, 2022.

The search will be conducted in appropriate search fields of electronic databases, and with sensitive searches that combine text words with indexing terms. Both free-text words (including spelling variants, synonyms, related terms, plurals, acronyms, truncations, wildcards, and proximity operators) and appropriate subject headings will be used. We will use Boolean operators ‘AND’ and ‘OR’ to connect and focus a search by combining subject headings and keywords.

Various combinations of the following key terms will be used to identify papers on the burden of perinatal mortality, its cause and determinants in Eastern Africa: ‘perinatal mortality’, ‘perinatal death(s)’, stillbirth(s), stillborn(s), ‘fetal death(s)’, ‘fetal demise’, ‘fetal mortality’, ‘neonatal death(s)’, ‘infant mortality’, and ‘East Africa’ ( Table 1). We will also identify studies that were cited by others (descendent search strategy).

Articles will be searched using different electronic databases and imported to EndNote software version X20 using each of the databases’ citation manager to facilitate review process and exclusion of duplicated studies.

After importing studies to Endnote X20, duplicates will be excluded. Titles and abstracts of remaining studies will be screened by Y.A.A., then abstracts of selected studies will be exported to Covidence review management software for full-text screening ²⁴ . Full-text articles will be independently screened by two investigators (Y.A.A. and N.A), and when there is uncertainty, a third reviewer (L.A.P or A.C.S) will make a final decision. The total number of studies identified, screened, eligible and included in the study will be described, and the reason for exclusion at each stage of the study selection process will be explained. A single failed eligibility criterion is sufficient for a study to be excluded from a review. Results from comparable groups of studies will be combined into a statistical meta-analysis using STATA-17 software ²⁵ .

Required information for the systematic review will be extracted and summarized using the Joanna Briggs Institute and Cochrane data extraction template. Information on the title, author, publication year, study design, study setting (rural vs. urban), study type (community-based vs. hospital-based), sample size, study participants, study period, sampling methods, and outcome of interest (definition of outcomes) will be extracted. When we extract data regarding perinatal mortality rate, we will note the denominator used by studies (live births or total births) and we will also collect the stillbirth and early neonatal death rates separately. Furthermore, we will extract the time of death for stillbirths; antepartum or intrapartum (fresh or macerated) if reported. In addition to this, we will look at the ascertainment of causes and risk factors in each study and the classification system will be recorded where available. The measures of association (odds ratio or relative risk with their respective confidence intervals) for each risk factor will be extracted and included in meta-analyses where feasible.

Perinatal mortality rate: is the sum of stillbirths and deaths in the first week of life (0–6 days) per 1000 total birth (both live and stillbirths) ¹⁹ .

Stillbirth rate: is fetal death at ≥500g/≥22 weeks gestation or ≥1000g /≥28 weeks gestation by WHO for general statistics and international comparison respectively per 1000 total births ²⁶ .

Early neonatal mortality: deaths among live births during the first week (0–6 days) of life ²⁷ .

Causes of perinatal mortality: are any condition/s with a reasonable mechanism likely to lead to the death of the fetus or early neonate and it is classified as the underlying cause, immediate cause, and main maternal cause) ^{28, 29} .

ICD-PM: is the WHO application of ICD-10 to deaths during the perinatal period, which provides a standardized system for classifying perinatal mortality based on time of death as antepartum (before the onset of labor), intrapartum (during labor but before delivery) or neonatal (the first week after delivery), and it also links the contributing maternal condition, if any, with perinatal death ²⁸ .

Determinants or risk factors: are characteristics associated with, but not obviously causal for, stillbirths or early neonatal deaths, such as advanced maternal age ³⁰ . ( Figure 1)

The perinatal mortality rate will be the primary outcome measure; it is calculated by dividing the number of fetal deaths after 22 weeks of gestation or weighing more than 500g and neonatal deaths in the first week after delivery by the total number of births (stillbirths and live births) that have been included in the study (sample size) ^{5, 19} . The second outcome is the determinants (risk factors) that are associated with perinatal mortality among the study subjects; factors associated with perinatal mortality will be socio-demographic and economic factors, maternal factors, fetal factors, health service-related factors, and environmental factors. The third outcome measure of the study will be the causes of perinatal mortality, which can be classified as the underlying cause (perinatal asphyxia or hypoxia, infection or sepsis, preterm birth complications, congenital anomalies), immediate cause (perinatal asphyxia or hypoxia, infection or sepsis, preterm birth complication, and birth trauma), and main maternal factors (complications of placenta, cord or membranes and maternal medical or surgical conditions, mainly associated with hypertensive disorder of pregnancy) of death ^{1, 31} ( Figure 1). Furthermore, if any other classification systems were utilized in the included studies, we will also attempt to synthesize them.

We will assess the methodological quality of included studies using the Joanna Briggs Institute quality assessment tool for observational and trial studies ³² . The Joanna Briggs Institute critical appraisal tool for cross-sectional, case-control and cohort studies comprises 8, 10 and 11 questions, respectively. The tool supports an assessment of sample representativeness of the target population, participant recruitment, adequacy of the sample size, detailed description of the study subjects and study setting, appropriate method of the statistical analysis, objective criteria in the measurement of the outcome variable and identification of subpopulation, reliability, and identification of confounding variables ³⁴ .

Each item for each study will be judged as Yes (1) and No (0). When the information provided is not adequate to make a judgment for a specific item, we will grade that item with a ‘No’ (0). Each study will be graded depending on the number of items judged ‘Yes’ (1) as low-risk bias (≥ 7), medium-risk bias (5 to 6), or high-risk bias (≤ 4) for cross-sectional studies, low-risk (≥8), medium-risk (5 to 7) and high-risk (< 5) for case-control studies, and low-risk (≥ 9), medium-risk (6 to 8) or high-risk (≤ 5) for cohort studies, and trials will be treated as cohort studies.

The study selection processes will be summarized using a PRISMA flow diagram, and for studies which are excluded the reason will be described and explained ¹⁷ . A narrative synthesis will be used to summarize all studies included in the study and characteristics like study population, cause of perinatal mortality, and identified risk factors will be summarized in a descriptive table.

If appropriate perinatal mortality rates from different studies with a common definition of perinatal mortality will be pooled together to provide a single summary (pooled perinatal mortality rate) estimate using STATA-17 software. Further, we will calculate the pooled risk ratio for the risk factors of perinatal mortality using the random effect model as it assumes that the observed variation of effect size is because of real differences ³⁵ . The syntax “metaforestplot” will be used to generate forest plots with their corresponding weights, as well as the pooled rate across studies and its corresponding 95% Confidence Intervals (Cl).

To examine the magnitude of the variation between studies statistical heterogeneity test will be assessed by Higgins’ I ². The I ² test measures level of statistical heterogeneity between studies; the values of <25 %, 25–50 %, 50–75 % and >75% are to mean very low, low, medium and high heterogeneity respectively ³⁶ . Since heterogeneity is expected in this study because of the differences in perinatal mortality rate across different settings, random effect model will be used. If heterogeneity is significant (I ²>50%), sub-group analysis, meta-regression or meta-analysis will be conducted to investigate sources of heterogeneity and if meta-analysis is not possible, a narrative synthesis will be conducted.

Sub-group analysis will be conducted based on study design, study type (community-based or facility-based), publication status (published or unpublished), study setting (rural vs. urban), geographic stratification, publication year, and study quality score (low or high score).

Sensitivity analysis will be performed to assess the robustness of a pooled estimate. We will use the single study omission analysis to test the robustness of a pooled estimate, and a study will be considered to have no influence on the pooled prevalence if the pooled estimate without it lies within the 95% confidence limits of the overall pooled prevalence ^{37, 38} . Furthermore, although most countries use the definition of perinatal mortality for international comparison, which is fetal deaths after 28 weeks of gestation or weighing more than 1000g till 7 ^th day after birth, in the primary analysis we will try to capture studies that were done using the WHO definition for general statistics (≥500g or ≥22 weeks of gestation). Then, the analysis will be repeated using the perinatal mortality definition for international comparison.

We will inspect funnel plots subjectively and Egger’s test objectively to assess publication bias. subjectively and evidence of publishing bias will be suggested by an asymmetrical funnel plot. Evidence of publishing bias will be suggested by an asymmetrical funnel plot and a p-value < 0.1 ^{39, 40} .