Background

Multimorbidity is having two or more long-term health conditions [1]. Although well studied in other disease area, there is currently sparse literature for multimorbidity in pregnant women. Pregnant women with multimorbidity are at increased risk of adverse outcomes for mother and child [2, 3]. Although maternal death is rare in the United Kingdom (UK), 90 % of women who died during/within a year after pregnancy had multiple health and social problems [4].

Multimorbidity increases health care burden for patients, for instance, needing to attend multiple health care appointments and being on multiple medications [1]. These challenges increase during pregnancy, with the addition of specialist antenatal clinic appointments and monitoring, and concerns regarding how medications may affect the developing foetus.

Despite this, there is a dearth of basic information on the prevalence and types of pre-existing health conditions affecting pregnant women. Better understanding of the epidemiology of multimorbidity in pregnant women could help policy makers and health care providers plan services to prevent women from developing multimorbidity, for early detection and optimal management of health conditions prior to conception, and tailor maternity services to pregnant women with multimorbidity.

In the UK, most people are registered with a general practitioner (GP), the gatekeepers to primary care and specialist referrals. In secondary care, health care utilization administrative data are recorded for reimbursement. Thus, both provide good data sources for multimorbidity and pregnancy research.

This study aims to describe the epidemiology of pre-existing multimorbidity in pregnant women. It also seeks to understand the utility of routine health care datasets in the study of multimorbidity in pregnant women, by using three datasets from different health care settings and across the four UK nations (England, Northern Ireland, Scotland and Wales).

Methods

Study design and study period

This was a cross sectional analysis of the prevalence of pre-existing multimorbidity prior to the start of pregnancy in the UK across three separate databases. We included index pregnancies where the conception date was between 1/1/2018 and 31/12/2018.

Inclusion and exclusion criteria

Women aged 15–49 years with a conception date in 2018 were eligible. Last menstrual period or gestational day 0 was considered the conception date [5]. When a woman had more than one pregnancy episode in 2018, the first recorded pregnancy in that year was included (not necessarily the first ever pregnancy). Women whose data did not meet standard quality checks were excluded (Additional file 1).

Data sources

This study used three datasets from different health settings, covering all four nations in the UK: Clinical Practice Research Datalink, (CPRD, England, Northern Ireland, Scotland and Wales), Secure Anonymized Information Linkage (SAIL, Wales) and Scottish Morbidity Records (SMR, Scotland).

Primary care

CPRD GOLD contains anonymized, longitudinal medical records for over 19 million patients in the UK (England, Northern Ireland, Scotland and Wales) from over 940 participating general practices; it currently covers 4% of UK GP practices and is widely acknowledged to be representative of the UK population [6]. It includes data on demographics, diagnoses and prescriptions [6]. Linkage to area based deprivation index was available for patients in England. Within CPRD GOLD, the CPRD Pregnancy Register is an algorithm that takes information from maternity, antenatal and delivery health records to detect pregnancy episodes and their outcomes [5].

The SAIL databank is a whole population level database in Wales. It is a repository of anonymized health and socio-economic administrative data and provides linkage at an individual level [7]. It holds data for 4.8 million people and covers 80% of Welsh GP practices [7]. Within SAIL, the National Community Child Health Dataset was used to detect pregnancies and was linked to the Welsh Longitudinal General Practice dataset and the Welsh Demographic Service dataset for diagnoses, prescriptions and demographics data respectively.

Secondary care and community prescriptions

SMR data was available from two Scottish regional health boards: National Health Service (NHS) Tayside and NHS Fife [8]. A dataset was created linking the Scottish Maternity Records (SMR02) to data from Hospital Admissions (SMR01), Mental Health Inpatients (SMR04), Accident and Emergency, and the Demography and Death registry. This covered diagnoses and demographic data for all inpatient stays and day cases for residents in the two regions. The dataset was also linked to the Prescribing Information System for data on all medications dispensed in the community. Pregnancies were detected from maternity records or pregnancy-related hospital admissions.

Definition of multimorbidity

Multimorbidity was defined by the presence of two or more pre-existing long-term physical or mental health conditions prior to the index pregnancy. We defined long-term conditions as conditions that have ongoing significant impact on patients, including conditions that are relapsing and remitting in nature.

One of the wider research aims is to mitigate the effect of multimorbidity on adverse pregnancy outcomes. As pregnancy related conditions (e.g., gestational diabetes and pregnancy induced hypertension) will be subsequently studied as maternal outcomes, they were not included in the definition of pre-existing multimorbidity.

An exhaustive list of long-term health conditions was first identified from existing literature [4, 9, 10], in particular based on the work commissioned by Health Data Research UK on multimorbidity conceptualization [10] and health conditions that were leading indirect cause of death in the UK maternal mortality report (MBRRACE) [4]. This list and phenome definitions were refined and harmonized through workshops with our research advisory group, consisting of patient and public representatives, clinicians from general practice, obstetrics, maternal medicine, psychiatry, public health, and data scientists. Seventy nine health conditions were selected on the following basis: (i) prevalence; (ii) potential to impact on pregnancy outcomes; (iii) considered important by women; and (iv) recorded in the study datasets.

Diagnoses of these 79 long-term health conditions were determined from Read Codes version 2 (primary care datasets) and the International Classification of Disease 10th version (secondary care datasets) [11]. The validity of diagnostic coding has previously been shown to be good in primary care records and generally health conditions under payment for performance schemes, such as Quality Outcomes Framework, are well coded [12]. Code lists and phenome definitions used are available in Additional files 2 and 3.

Data analysis

The primary analysis was the prevalence of pre-existing multimorbidity in pregnant women. The denominator was the total number of index pregnancies identified in 2018, regardless of the pregnancy outcome. Additional analysis was performed for multimorbidity with at least one mental health conditions and active multimorbidity. Active multimorbidity limits common transient/episodic conditions (e.g., mental health, dermatological and atopic conditions and headaches) to those that were active in the 12 months preceding index pregnancy (Additional file 3).

Multivariable logistic regression was performed to examine the association of multimorbidity with maternal age (five-yearly categories), ethnicity, deprivation quintiles (patient level Index of Multiple Deprivation [IMD] for all three datasets), latest maternal pre-pregnancy body mass index (BMI) categories, latest pre-pregnancy smoking status, and gravidity (total number of pregnancies up to and including index pregnancy). Obesity was considered a covariate (BMI categories) instead of a health condition. For CPRD, practice level IMD was available for all four nations, but patient level IMD was only available for England, therefore, the regression analysis was limited to England. We then described the prevalence of individual health conditions, and the prevalence of mutually exclusive multimorbidity combinations.

Missing data were assigned to separate categories and included in the analyses. Sensitivity analysis was performed for CPRD (England), where missing ethnicity was imputed with data from linked hospital administrative data, and missing patient level IMD was imputed with practice level IMD.

Study results were presented for each dataset separately. Data were not combined as there was a possibility of patient overlap between CPRD (Wales, Scotland) with both SAIL (Wales) and SMR (Scotland). Deduplication was not possible as the datasets are anonymized, and only aggregated data were exported within the permission of the data access approval.

Post hoc analysis

As our study found no association of recorded multimorbidity with social deprivation, we conducted a post hoc analysis in the CPRD cohort, with the list of conditions used to define multimorbidity in a seminal paper that found this association [13]. We also examined the association of selected health conditions with deprivation and ethnicity. Guided by our patient representatives, we analysed the prevalence of multimorbidity for selected health conditions to illustrate the burden of multimorbidity. The selected health conditions were: i) the top ten most common individual health conditions in this study, and ii) leading causes of maternal deaths [4].

Analysis was performed using STATA 16 and R. The study is reported in accordance with the RECORD guideline (Additional file 4).

Results

Study population

Overall, there were 37,641 (CPRD), 27,782 (SAIL) and 6099 (SMR) pregnant women aged 15–49 years included in the analysis in 2018. Additional Fig. 1 presents the flow chart for the study population selection. The characteristics of the study cohort is presented in Table 1 and Additional Table 1. Most of the study participants were 20–34 years old, of White ethnicity, of normal weight or were overweight pre-pregnancy and were non-smoker pre-pregnancy. Linkage to area-based data for patient level IMD was available for 75% of the CPRD study cohort in England. There were more affluent women in the primary care dataset but vice versa for SMR.

Table 1 Baseline characteristics of pregnant women in CPRD (UK), SAIL (Wales) and SMR (Scotland) in 2018
Full size table

Multimorbidity

The prevalence of pre-existing multimorbidity in pregnant women was 44.2% (95% confidence intervals [CI] 43.7 to 44.7%), 46.2% (45.6 to 46.8%) in CPRD and SAIL respectively (primary care dataset) but was halved in SMR’s secondary care and community prescription dataset, 19.8% (18.8 to 20.8%).

Over 70 % of pregnant women with multimorbidity had mental health condition/s: 31.3% (30.8 to 31.8%), 33.7% (33.1 to 34.2%) and 14.6% (16.0 to 17.9%) of pregnant women had multimorbidity with at least one mental health conditions in CPRD, SAIL and SMR respectively. The prevalence of active multimorbidity was half that of the primary analysis in primary care datasets, 24.2% (23.8 to 24.6%) and 23.5% (23.0 to 24.0%) for CPRD and SAIL respectively, but remained similar for SMR, 17.0% (16.0 to 17.9%). The percentage of pregnant women by the total morbidity count is available in Additional Table 2.

Characteristics associated with multimorbidity

The prevalence of pre-existing multimorbidity by the characteristics of pregnant women is presented in Additional Table 3. In the CPRD England study cohort (n = 13,075), when all characteristics were adjusted for, increasing maternal age and gravidity remained significantly associated with multimorbidity (maternal age 45–49 years, adjusted odds ratio [aOR] 1.8, 95% CI 1.0 to 3.2; gravidity ≥5, 1.7, 1.5 to 1.9); pregnant women with BMI 25 to 29.9 (aOR 1.2, 95% CI 1.1 to 1.3), BMI 30+ (1.6, 1.4 to 1.8), were smokers (1.6, 1.5 to 1.8) or ex-smokers (1.4, 1.3 to 1.6) had higher odds of multimorbidity. However, higher odds of multimorbidity were not observed in pregnant women of ethnic minority groups or from more deprived socioeconomic groups (Table 2). Findings were similar in the sensitivity analysis of CPRD (England) using imputed data for missing ethnicity and IMD (Additional Table 4). In SAIL, the effect sizes of characteristics were generally similar to that in CPRD (England).

Table 2 Logistic regression for pre-existing multimorbidity by women’s characteristics
Full size table

In SMR, after adjusting for all characteristics, higher odds of multimorbidity were observed only in those age 20–24 and 25–29 years, had gravidity of 3+, BMI 30+, were smokers and ex-smokers and those from more deprived socioeconomic groups. The odds of multimorbidity were not higher in ethnic minority groups (Table 2).

Post hoc analysis

Post hoc analysis was performed to explore whether the lack of association of multimorbidity with deprivation in our primary care datasets was, in part, due to the health conditions we used to define multimorbidity. Logistic regression was repeated in CPRD (England) with the list of 31 health conditions used to define multimorbidity in Barnet et al’s seminal paper [13], the adjusted variables were added in a step-wise manner. After adjusting for maternal age, ethnicity and gravidity, increasing levels of deprivation were associated with higher odds of multimorbidity (most deprived quintile aOR 1.30, 95% CI 1.08 to 1.57). This association was attenuated and was no longer significant when raised BMI and smoking status were added to the model (aOR 1.05, 0.87 to 1.27, Additional Table 5, Fig. 1).

Fig. 1
figure 1

Forest plot of odds ratio for multimorbidity and social deprivation. Legend: Multimorbidity was defined using the 31 health conditions in Barnet et al’s study, logistic regression was used to analyse the study cohort in CPRD England (n = 13,075). The reference group was index of multiple deprivation (IMD) quintile 1 (least deprived)

Full size image

To test this hypothesis further, we repeated the logistic regression in the CPRD (England) cohort by removing eight health conditions that were associated with being in less deprived socioeconomic groups. When adjusted for maternal age, ethnicity and gravidity, multimorbidity (defined by 71 health conditions) was associated with deprivation (most deprived quintile aOR 1.26, 1.10 to 1.44). This association was attenuated and was no longer significant when raised BMI and smoking status were added (aOR 1.08, 0.94 to 1.24, Additional Table 6).

Individual health conditions

Table 3 presents the top 20 most prevalent health conditions in our study cohort. The top four most common health conditions across all three datasets were depression, anxiety (both known as common mental health disorders), allergic rhinoconjunctivitis and asthma, with the prevalence of common mental health disorders being consistently around 20%. The full list of prevalence for each health condition is presented in Additional Table 7.

Table 3 Top 20 most prevalent individual health conditions in pregnant women
Full size table

Combinations of multimorbidity

Table 4 presents the top ten most common combinations of multimorbidity, the most prevalent combinations being depression and anxiety in primary care datasets (2.2 and 2.7% of pregnant women in CPRD and SAIL respectively) and common mental health disorders and asthma for SMR (3.2%). The presented prevalence is for mutually exclusive multimorbidity combinations, and therefore prevalence for depression and anxiety will not include women with depression, anxiety and other health condition/s. When only considering physical conditions, the most common combination was asthma and allergic rhinoconjunctivitis (1.7, 2.1 and 2.2% in CPRD, SAIL and SMR respectively).

Table 4 Top ten mutually exclusive combinations of multimorbidity in pregnant women
Full size table

Prevalence of multimorbidity in pregnant women with selected health conditions

These examples have been provided to illustrate the burden of using the CPRD (UK) pregnancy cohort in 2018. The featured health conditions were the leading causes of maternal deaths in the MBRRACE-UK report [4].

Cardiovascular disease (ischemic heart disease, stroke/transient ischemic attack, heart failure, atrial fibrillation, congenital heart disease, valvular heart disease, cardiomyopathy, hypertension) affected 2.0% (745/37641) of pregnant women, of whom 80.1% (597/745) had multimorbidity. Less than 1 % (246/37641) of pregnant women had a history of venous thromboembolism, among whom 85.8% (211/246) had multimorbidity. Epilepsy affected 1.4% (543/37641) pregnant women, among whom 80.7% (438/543) had multimorbidity.

Prevalence of selected health conditions by social deprivation and ethnicity

Table 5 presents examples to illustrate the difference in the prevalence of individual health conditions by patient level social deprivation and ethnicity using CPRD (England). Mental health conditions, asthma and epilepsy increased with deprivation. In contrast, some of the common health conditions were more common in the affluent groups, including anxiety, migraine, irritable bowel syndrome, and polycystic ovarian syndrome. For ethnicity, mental health conditions, asthma, migraine, irritable bowel syndrome and psoriasis were more prevalent in White ethnic group; whilst allergic rhinoconjunctivitis and polycystic ovarian syndrome were more prevalent in ethnic minority groups.

Table 5 Prevalence of selected health conditions in pregnant women by social deprivation and ethnicity
Full size table

Discussion

Main findings

This study used contemporaneous, routinely collected datasets to study the epidemiology of multimorbidity (defined as having two or more long-term physical or mental health conditions) in pregnant women in the UK. Two in five pregnant women had pre-existing multimorbidity. One in five pregnant women had multimorbidity that were active in the year before pregnancy. Seven in ten pregnant women with multimorbidity had a history of mental health condition/s. In women with conditions that are known to be leading causes of maternal death [4], four in five had pre-existing multimorbidity. Pregnant women with multimorbidity were more likely to be older, multigravid, smoked or have raised BMI preconception.

Strengths and limitations

This study utilized electronic health records which provided a rich source of data and is generalizable across different settings. It avoided misclassification bias associated with self-reported surveys. However, as with all research that use routine health records, it is subjected to residual confounding and can be limited by the quality and consistency of data entry by clinicians and administrators [14]. We have attempted to improve the accuracy of health conditions ascertainment through the design of our phenome definitions (e.g., using additional age limit and phenomes by prescriptions).

The definition of multimorbidity used in this study was based on simple counting of conditions, without weightings. There is currently no single multimorbidity index that can measure multimorbidity in all settings definitively [15]. The only currently available validated comorbidity index in maternal health research was developed using secondary care data and only included 20 conditions [16], in comparison to the 79 conditions prioritized by our multidisciplinary group and patient representatives. Obesity was analysed as a covariate (BMI categories) in this study; the prevalence of multimorbidity would be higher if obesity was considered a long-term health condition.

Utility of the different datasets

Compared with CPRD and SAIL primary care datasets, the prevalence of non-life-threatening health conditions such as allergic rhinoconjunctivitis, migraine, irritable bowel syndrome and, ultimately, multimorbidity was lower in the Scottish secondary care with linked community prescription dataset.

This is likely to reflect that health conditions seen in primary care encompass the whole severity spectrum. Some common conditions, such as anxiety or depression, may only present to primary care, some of which are managed conservatively without prescribed medications. In contrast, the Scottish secondary care and community prescription database would only capture the severe spectrum of a condition that requires hospital attendance or regular prescriptions and may under-estimate the prevalence of multimorbidity. This confirms that primary care health records may be a more comprehensive data source to study pre-existing multimorbidity in pregnant women, and to identify the target at risk population for preconception intervention.

Similar findings were observed in both CPRD, and SAIL add to the validity of these findings. Whilst CPRD offered the benefit of representing data from all four UK nations, SAIL offered a more complete coverage at a population level in Wales with good follow up throughout an individual’s lifetime even when they change GP practices.

Our study highlighted a shortfall in the recording of ethnicity and preconception body mass index, and to a lesser extent, smoking status preconception for pregnant women in routine health records. Patient level data for social deprivation was limited by the availability of data linkage in CPRD. Although sensitivity analysis in the CPRD (England) dataset with imputed ethnicity and IMD showed similar findings with the primary analysis, the interpretation of the association analysis should be taken with caution. In SAIL, pregnancy episodes were detected from the National Community Child Health database (NCCHD), and this does not include pregnancies that resulted in early pregnancy loss; hence the gravidity data generated from this database is likely to be an under-estimation. Historical data from the SMR datasets were available from 2005 to 2019. This meant that if a pregnant woman had a history of a health condition prior to this time period, it may not be captured. This limitation is more likely to affect older women in the SMR pregnancy cohort and may partially account for the lack of association of maternal age with multimorbidity. Further limitations of each dataset are outlined in Additional file 5.

Results in the context of what is known

High prevalence of multimorbidity in pregnant women

The current evidence for the prevalence of multimorbidity in pregnant women is scarce and findings vary widely. This ranged between < 10 to 35% [2, 3, 17, 18]. The high prevalence of multimorbidity in pregnant women in this study is concerning as it is associated with adverse outcomes for mother and child [2,3,4]. In the latest MBRRACE UK maternal mortality enquiry report, 90% of maternal deaths up to one year post pregnancy occurred in women with multiple health and social problems [4]. MBRRACE has called for national guidance for the management of pregnant women with multiple morbidities and social factors [19]. Recently, the Ockenden report (a high profile UK independent inquiry of maternity services at a local hospital) highlighted the need for involvement of maternal medicine specialist and maternal mental health services for managing women with complex pregnancies [20].

Clinical implications

Our study provided a current snapshot of how multimorbidity is distributed in the UK in terms of socio-demographics, the health conditions that constitute multimorbidity and the common combinations of health conditions in pregnant women. Mental health conditions were particularly prevalent and contributed to 70% of multimorbidity in pregnant women. Psychiatric causes were amongst the leading cause of maternal death in the UK [4]. Our findings further support the need for integration of mental health services with maternity services and equitable access to perinatal mental health services in the UK [21].

Social deprivation, ethnicity and multimorbidity in pregnant women

Post hoc analysis found that some health conditions were more prevalent in affluent pregnant women (e.g., anxiety, irritable bowel syndrome), potentially masking the association of multimorbidity with social deprivation. When these conditions were removed, multimorbidity was associated with social deprivation but this effect was not observed when BMI and smoking status were also adjusted for. This suggests that smoking and obesity may mediate the relationship between social deprivation and multimorbidity in pregnant women.

Many of the topmost common health conditions were more prevalent in White pregnant women, particularly mental health conditions. This may have contributed to the lack of association of ethnic minority groups with multimorbidity. Previous literature reported that people of ethnic minority are less likely to access/receive mental health support/treatment [22, 23]. In addition, stigma associated with mental health conditions was reported to be higher in ethnic minorities [24].

Both observations mean that there could be health care access issues for some of the common health conditions, especially mental health conditions, for people from socially deprived and ethnic minority groups prior to pregnancy. In addition, it strengthens the importance of addressing smoking and obesity preconception especially in pregnant women with multimorbidity from socially deprived groups. Smoking and obesity are two well-known modifiable risk factors for adverse pregnancy outcomes [25, 26], the added impact of multimorbidity is likely to compound this. Further research is required to quantify this, but interventions addressing smoking and obesity may help reduce adverse outcomes in pregnant women with multimorbidity.

Research implications

Despite the high prevalence of multimorbidity in pregnant women, and the associated adverse outcomes, there is currently a paucity of evidence in this field. The MuM-PreDiCT consortium is a multidisciplinary collaboration across all four nations in the UK, including women with lived experience of multimorbidity and pregnancy. Our next step is to quantify the impact of multimorbidity on pregnancy, maternal and offspring outcomes. This will provide crucial information for women with multimorbidity who are planning a pregnancy and results from the outcome studies may require us to reconsider how we categorize high-risk pregnancy. The ultimate aim is to produce high quality evidence that would guide clinical practice to prevent pregnancy complications and to optimize long-term maternal and offspring health for pregnant women with multimorbidity.

Conclusion

A significant proportion of women enter pregnancy with pre-existing multimorbidity, especially with mental health condition/s. Amongst pregnant women with health conditions known to be leading causes of maternal death, prevalence of multimorbidity was high. Pregnant women with multimorbidity were more likely to smoke and have a raised BMI and support maybe required to address this. There may be health care access inequalities for some health conditions, especially mental health conditions in pregnant women from deprived or ethnic minority groups.