Abstract
Dear Editor,
We thank Drs. Mehta and Sah for their commentary on our article. 1 Their comments highlight important methodological considerations, most of which we believe were addressed in our manuscript. However, we appreciate the opportunity to provide further clarification.
To reduce the global burden of maternal mortality and morbidity, and particularly that in Sub-Saharan Africa, a four-pronged approach is required. Namely, improvements in health system responsiveness, universal provision of skilled birth attendants, health care worker behavioural change and increased awareness by pregnant and birthing mothers of signs and symptoms associated with obstetric complications are essential. Importantly, the last of these prompts healthcare-seeking by symptomatic women, the topic of our manuscript, and is the first step in the cascade of timely and appropriate care that saves the lives of mothers and babies.
Firstly, our use of self-reported data was completely appropriate as the data related to women with symptoms of obstetric complications and not self-reports of diagnosis of obstetric complications as inferred by Drs. Mehta and Sah. To prevent escalation to potentially life-threatening outcomes, pregnant and post-partum women with any concerning obstetric symptoms need to seek healthcare for early diagnosis and management. 2 Our study included analyses of healthcare seeking behaviour by women with symptoms of both life-threatening and non-life-threatening complications. Unlike the general population, pregnant and birthing women who present with “severe headache” and “lower abdominal pain” require urgent medical evaluation. 3 However, Drs. Mehta and Sah did not appreciate the importance of these symptoms in pregnancy but cited a study showing low sensitivity of acute abdominal pain for life-threatening conditions among the general population. They also cited a study by Alshuaylan et al., 4 which was retracted on 25 January 2024. These are both inappropriate and misleading. Severe headache is a key symptom of preeclampsia/eclampsia, while lower abdominal pain is a common symptom of ectopic pregnancy, placental abruption, and preeclampsia. A Delphi consensus identified severe headache as one of the key symptoms of obstetric complications that can be self-assessed by women. 5 We contend that women are also quite able to self-assess abdominal pain.
Secondly, while our analysis did not independently examine the effects of each category of variables (predisposing, enabling, and needs-based factors), the classifications were based on theoretical guidance from Andersen’s 6 model and literature, rather than derived from interaction effects. We did assess data for multi-collinearity and found none. Additionally, we used multilevel modelling to test antenatal care use as an individual-level variable and/or community wealth status and mass media access as community-level variables. As reported in our paper, these analyses highlighted that the ANC effect on healthcare seeking might not be moderated by community wealth status nor access to mass media. The community-level variable’s impact on women’s behaviour was discussed with random-effect and fixed-effect results.
Thirdly, Dr. Mehta and Dr. Sah were concerned about the exclusion of termination-related complications from the analyses of childbirth and postpartum. We did exclude women with termination of pregnancy from analyses on healthcare seeking behaviour for childbirth and postpartum complications but these were not excluded for healthcare seeking behaviour for overall obstetric complications and pregnancy complications. 1 Strictly speaking, women with pregnancy terminations do not go through childbirth and postpartum and therefore should not be included in analyses of healthcare-seeking behaviour perinatally. Therefore, the generalizability of our findings is not compromised because women with pregnancy termination complications were included in the relevant analyses. Drs. Mehta and Sah also cited a paper about maternal mortality due to unsafe abortion. However, that study is not related to maternal mortality due to unsafe abortion; instead, it addressed late request of induced abortion services among women seeking induced abortion, and showed that 21.5% of women requested late pregnancy termination services, that is after 12 weeks of gestation. 7
The fourth point raised was about the high variance at the community level. The high ICC findings necessitated we employ a multilevel model to improve the accuracy of estimates. As reported, the ICC was 63% in the null model but after including individual and community-level variables, the ICC decreased to 34.7% indicating the significant impact of community-level variables. In our paper, the role of clustering, using random effects results (such as the median odds ratio), and fixed effects estimates of community-level variables was explored. To account for community norms, including variation in the roles of traditional birth attendants, we recommended that qualitative research should explore these in healthcare-seeking behaviour during obstetric complications. 1
Finally, the potential for recall, misclassification and social desirability biases was raised. However, our study focuses on maternal healthcare-seeking behaviour for obstetric complications at the population level. Hence, cross-validation with facility-level data is outside the scope of our paper given we are interested in healthcare-seeking behaviour for – rather than the prevalence of – obstetric complications. Additionally, when symptomatic pregnant and postpartum women do not seek healthcare, facility cross-validation is not possible. The data used in this study were collected using a tool validated in the Ethiopian context for which a sensitivity of 97.4% and specificity of 93.3% for healthcare-seeking during immediate postpartum complications has been reported. 8 While we agree social disability bias and misclassification are a concern in self-reported studies, we have already discussed these in the Strengths and Limitations.
In conclusion, we appreciate your interest and taking the time to engage with our research. While your comments raise important considerations, several were already addressed or stated in the limitations section of our paper.
