Associations between neighbourhood deprivation, ethnicity and maternal health outcomes in England: a nationwide cohort study using routinely collected healthcare data

Background In the United Kingdom, pregnant women who live in the most deprived areas have two times the risk of dying than those who live in the least deprived areas. There are even greater disparities between women from different ethnic groups. The aim of this study was to investigate the role of area-based deprivation and ethnicity in the increased risk of severe maternal morbidity (SMM), in primiparous women in England. Methods A retrospective nationwide population study was conducted using English National Hospital Episode Statistics Admitted Patient Care database. All primiparous women were included if they gave birth in an National Healthcare Service (NHS) hospital in England between 1 January 2016 and 31 December 2021. Logistic regression was used to examine the relative odds of SMM by Index of Multiple Deprivation and ethnicity, adjusting for age and health behaviours, medical and psychological factors. Results The study population comprised 1 178 756 primiparous women. Neighbourhood deprivation increased the risk of SMM at the time of childbirth. In the fully adjusted model, there was a linear trend (p=0.001) between deprivation quintile and the odds of SMM. Being from a minoritised ethnic group also independently increased the risk of SMM, with black or black British African women having the highest risk, adjusted OR 1.84 (95% CI 1.70 to 2.00) compared with white women. There was no interaction between deprivation and ethnicity (p=0.49). Conclusion This study has highlighted that neighbourhood deprivation and ethnicity are important, independently associated risk factors for SMM.

10 codes: obesity or overweight (yes/no), history of pre-existing medical condition (yes/no), history of pre-existing mental health problems (yes/no), substance misuse and smoking (yes/no).Women were assumed not to have the above confounding factors if they were not coded in any of the hospital records from the preceding length of time listed in Table S2 to the start of the index pregnancy.The list of relevant pre-existing medical conditions and mental health problems and the codes for all confounding variables is provided in Table S2.

Outcome
The outcome was defined as the English Maternal Morbidity Outcome Indicator (EMMOI) 6 .This is a composite outcome that includes 17 diagnoses and 9 procedures, adapted from the Australian Maternal Morbidity Outcome Indicator 7 in 2016, which can be used as a single measure of severe morbidity during pregnancy or childbirth using data from HES APC.The list of the relevant diagnoses/ procedures and their codes are included in Table S1.SMM was coded as a dichotomous variable, with a woman being given the score of 1 if she had one of more of these codes during her birth episode or 0 if she did not.

Interactions
Based on existing literature, 8 maternal age at childbirth was deemed to be potential effect modifier a priori, ethnicity was examined as an effect modifier in the relationship between IMD and severe maternal morbidity, and IMD was examined as an effect modifier in the relationship between ethnicity and severe maternal morbidity.
BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s)

Strengths and limitations of HES APC
The key strength of this data source is its universal coverage of all NHS hospital births in England, allowing for a nationwide population study.This reduces selection bias as information on 97% of all births in England are reported in HES APC.Furthermore, the HES APC is administrative data collected for NHS payment purposes.It is therefore easily obtained without need for resource-intensive bespoke collection.It contains both time stamped clinical information as well as demographic variables such as the IMD and ethnicity.
However, as the data is not collected prospectively for the purpose of this research question it falls short on key variables such useful information on individual socio-economic factors.
Furthermore, the clinical coders rely on the quality and detail of discharge summaries which are created by busy clinical staff.Thus, there is a risk of conditions not being coded for, creating false negatives and misclassification if these data are used for research purposes.
Notably in this study, the prevalence of overweight and obese women, and women with preexisting physical or mental health conditions are low compared to other national sources 9 .
Furthermore, every birth record contains optional additional maternity data, not mandated for collection.This leads to large variations in data quality and completeness between hospitals. 10rity is been shown to be unreliably coded in the maternity data with only 59% or trusts having an expected distribution of parity.However, this study used the 'look back' method with eighteen years of data to reclassify multiparous women with a previous delivery code who may have been incorrectly classified as primiparous, and to determine parity in those women for whom parity had not been recorded.
BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s)  BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s)

Figure S1 .
Figure S1.Directed Acyclic Graph (DAGs).*The DAGs only included the variables which were used in the multivariable analysis and were available in the HES APC.

Figure S2 :
Figure S2: Identification of the study population of included primiparous women

Table S7 . Sensitivity analysis after excluding women with parity coded as missing in the 'maternity tail' - 510,439 women (43.30%) and after multiple imputation if data missing on ethnicity (FMI 11.75%) Multiple Imputation - fully adjusted model (Model 3) Excluding women with parity coded as missing in the maternity tail -fully adjusted model (Model 3)
BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s) *See list of conditions included list in TableS2

Table S8 : E-values
BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s)