Study Design
This study employed a retrospective analysis of secondary data obtained from the Pakistan Maternal Mortality Survey (PMMS) 2018–19, a component of the Pakistan Demographic and Health Survey (PDHS) program. The PMMS is a specialized survey focusing on maternal mortality and its determinants that is conducted periodically to provide comprehensive insights into maternal health outcomes in Pakistan.
Data Source
The data for this study were extracted from the PDHS 'Verbal Autopsy' dataset (PKVA7ASV). The PDHS is a nationally representative cross-sectional survey conducted by the National Institute of Population Studies (NIPS) in collaboration with ICF International and sponsored by the US Agency for International Development (USAID). The survey encompasses all provinces of Pakistan, including the Merged Areas, Gilgit Baltistan, and Azad Jammu and Kashmir, ensuring broad coverage of the country's demographic and health indicators.
Data collection and sampling
Data collection for the PDHS was conducted using a stratified cluster sampling method. This approach involved dividing Pakistan into strata based on geographical regions and urban‒rural classification. Clusters were then selected from each stratum using probability sampling techniques, ensuring proportional representation of different regions and population groups. The 'Verbal Autopsy ' dataset included detailed information on demographic characteristics, maternal health indicators, contextual and causal factors behind maternal deaths, and other relevant variables for women of reproductive age.
Study population
The study population consisted of maternal deaths recorded in the PMMS 2018–19 dataset. Maternal deaths were defined as deaths occurring during pregnancy, childbirth, or within one year postpartum, regardless of the cause. This includes deaths related to direct obstetric causes such as hemorrhage, sepsis, and eclampsia and indirect causes such as preexisting medical conditions aggravated by pregnancy.
A subanalysis group utilized in this study was pregnancy-related deaths. These were defined as deaths that were directly or indirectly associated with pregnancy and childbirth. These deaths occur during pregnancy, during childbirth, or within one year postpartum as a result of complications related to pregnancy, childbirth, or their management.
Data Preprocessing
Handling of Missing Values
Patients lacking a specified place of treatment were removed from the dataset to ensure clarity and accuracy in subsequent analyses. This filtering process resulted in a final sample of 1,177 patients for analysis. Variables with excessive missing values were identified and filtered out to maintain data quality. Missing values for the remaining variables were addressed using appropriate methods, such as imputation or exclusion, to ensure the integrity of the dataset.
Data Cleaning
The dataset underwent a thorough cleaning process to identify and rectify any inconsistencies or errors. This included standardizing variable formats, correcting data entry errors, and removing duplicate entries.
Data Analysis
Descriptive frequencies and percentages were computed to summarize demographic characteristics and pregnancy-related factors among the mothers who died. This involved tabulating key variables such as age, education level, and place of treatment to provide an overview of the study population. Further statistical analysis included the following steps:
- Stratified analysis: A stratified analysis was performed to explore associations between various factors and death at either the patient’s home or at a health facility.
- Contingency table analysis: Contingency table analysis was conducted to investigate the relationship between reaching a health facility and first-day mortality.
- Three-delay model analysis: A three-delay analysis was undertaken to understand the temporal aspects of maternal care-seeking behavior and provision.
Further details of each analysis are provided in their relevant portions in the results section. All the data analyses were conducted using IBM SPSS version 25.
Ethical considerations
Ethical approval for the PDHS was obtained from the National Bioethics Committee of Pakistan under reference number Ref: No.4-87/NBC-346/19/. Since this study involved secondary analysis of existing data, additional ethical approval was not needed. The data were anonymized and deidentified prior to analysis to ensure confidentiality and privacy.