Risk factors and recurrence of hyperemesis gravidarum: A population‐based record linkage cohort study

Abstract

Introduction

There are limited contemporary population‐based studies on the risk factors for hyperemesis gravidarum (HG), a severe type of nausea and vomiting in pregnancy. This study aimed to determine the prevalence and trend of HG over time, identify risk factors for any and multiple HG health service visits during pregnancy, and investigate HG recurrence across pregnancies.

Material and Methods

This population‐based record linkage cohort study featured births in New South Wales, Australia from 2010 to 2019. Hospital and emergency data collections were used to identify health service visits for HG using relevant diagnosis codes and were linked to the corresponding pregnancy on the birth data set. Outcomes included any HG and multiple HG visits during pregnancy, and HG recurrence across pregnancies. Annual HG prevalence was calculated, and negative binomial regression was used to examine standardized prevalence trends. Risk factors for any HG and multiple HG visits within a pregnancy were examined using Robust Poisson models with generalized estimating equations and Prentice–Williams–Peterson Gap Time models, respectively. Rates and risk of recurrence were calculated for women with a second and third pregnancy.

Results

Of the 955 107 pregnancies, 21 702 (2.3%) were classified as HG. There was an average annual increase of 6.8% (95% CI 5.3–8.3) in HG prevalence. Younger maternal age, multiple pregnancies, and selected preexisting conditions were associated with an increased risk of HG, with the strongest factor being HG in any previous pregnancy (risk ratio 8.92, 99% CI 8.43–9.44). Hyperemesis gravidarum recurrence at the second (28.9%) and third (54.7%) pregnancies was high.

Conclusions

Hyperemesis gravidarum history is the strongest risk factor for HG, which has implications for counseling and care that women receive around pregnancy.

While there were several maternal and infant characteristics associated with hyperemesis gravidarum (HG), the strongest risk factor was a history of HG in a prior pregnancy, which was also evident in the high recurrence rates observed across subsequent pregnancies.

Abbreviations

ART

assisted reproductive technology

BMI

body mass index

CI

confidence interval

HG

hyperemesis gravidarum

NSW

New South Wales

PAF

population attributable fraction

RR

relative risk

Key message.

This population‐based cohort study provides contemporary and clinically useful information for identifying women at higher risk of hyperemesis gravidarum. Our findings suggest that future models of care should support women with this condition, particularly those with a history, to improve their pregnancy outcomes.

1. INTRODUCTION

Hyperemesis gravidarum (HG) is a debilitating complication of pregnancy characterized by constant and severe nausea and vomiting in pregnancy. Reported prevalence estimates vary from 0.3% to 10.8%. ¹ , ² The definition, and thus prevalence, of HG varies across population‐based studies, with some studies classifying non‐specific nausea and vomiting diagnosis codes as HG. While several risk factors have been identified as being associated with HG, there are limited contemporary population‐based studies in this area. ³ , ⁴ , ⁵ , ⁶ , ⁷ , ⁸ , ⁹ Furthermore, risk factor definitions are inconsistent across the literature and the contribution of these to the overall prevalence of HG has not been quantified. The risk of HG recurrence is high and population‐based estimates vary from 15% to 24%. ¹⁰ , ¹¹ , ¹² Recurrence definitions vary, as some studies calculated recurrence after the woman's first recorded pregnancy, ¹¹ while others calculated recurrence after the woman's first HG pregnancy, regardless of whether it was their first overall pregnancy. ¹²

Our research builds on previous studies by providing contemporary population‐based estimates of prevalence and risk factors. We defined HG by selecting diagnosis codes that minimize misclassification and measured recurrence using multiple methods to ensure replicability. The aims of this study were to determine the prevalence and trend of HG over time, to identify factors associated with HG, and investigate HG recurrence across pregnancies.

2. MATERIAL AND METHODS

2.1. Study population

This population‐based record‐linkage cohort study utilized linked administrative health surveillance data. The study population included all women who gave birth in New South Wales (NSW), Australia's most populous state, from 2010 to 2019 based on the NSW Perinatal Data Collection. The Perinatal Data Collection captures all livebirths, and stillbirths of at least 20 weeks gestation or at least 400 g birthweight occurring in NSW public and private hospitals as well as home births. Linked hospital and emergency data were used to identify health service visits during pregnancy. Hospital admissions were obtained from the NSW Admitted Patient Data Collection, which captures all inpatients in all public and private hospitals in NSW. The Admitted Patient Data Collection records include diagnosis and procedure (a principal code and up to 50 additional codes for each) information relevant to the episode of care. Emergency Department presentations were obtained from the NSW Emergency Department Data Collection, which captures visits to public hospital Emergency Departments in NSW and includes a principal Emergency Department diagnosis code. A number of studies have provided information on the validity of the birth and hospital data sets. ¹³ , ¹⁴ , ¹⁵

Data linkage was carried out by the NSW Centre for Health Record Linkage using probabilistic methods based on personal identifying information. ¹⁶ De‐identified data were provided to the researchers, with each mother and child assigned a unique project identification number to allow for linkage and follow‐up of their birth, hospital, and emergency data across the study period.

Hyperemesis gravidarum was identified using relevant diagnosis codes (ICD‐10‐AM, ICD‐9, or SNOMED; Table S1) at any diagnosis position and classified by at least one hospital admission or emergency presentation for HG at any time during the estimated pregnancy period up to the baby's date of birth. The estimated pregnancy start date was calculated by subtracting the gestational age (and adding 14 days to allow time from conception) from the baby's date of birth. Due to potential misclassification, if a pregnancy included health service visits with a recorded diagnosis for nausea and vomiting (nausea and vomiting in pregnancy; Table S1) and no visit with HG diagnosis, it was excluded from the analyses.

2.2. Outcomes and risk factors

Study outcomes included: any HG in pregnancy, multiple HG health service visits during pregnancy, and HG recurrence across pregnancies. For HG recurrence, the study cohort was restricted to women whose first observed pregnancy was in 2010–2017 to allow adequate follow‐up time. Three methods were used to define HG recurrence. ¹¹ , ¹² The index pregnancy was defined as the first pregnancy in the study period, where HG was identified, and recurrence was calculated as the proportion of women who had HG in any pregnancy following the index pregnancy. Then recurrence in the second and third pregnancies was calculated for women with at least two and three pregnancies, respectively.

Potential risk factors for HG were identified via clinical consultation and review of previous literature. ⁴ , ⁵ Sociodemographic, lifestyle and health conditions, and infant characteristics were ascertained from the birth data including year of birth, maternal age in completed years, and country of birth grouped according to the Standard Australian Classification of Countries. ¹⁷ Maternal residence (Statistical Area 2) was used to derive geographical remoteness category ¹⁸ and socioeconomic status, as measured by the Index of Relative Socioeconomic Disadvantage in quintiles. ¹⁹ Maternal parity, smoking during pregnancy, and body mass index (BMI) were also sourced from the birth data. Body mass index scores were calculated from height and weight recorded in the first trimester, after excluding potential biologically implausible values, ²⁰ and then categorized using World Health Organization cutoffs (underweight: <18.5 kg/m², normal weight: 18.5–24.9 kg/m², overweight: 25.0–29.9 kg/m², obese: ≥30 kg/m²) for mothers aged over 18 years and age‐specific cutoffs ²¹ , ²² for mothers aged less than 18 years. Infant sex and plurality were identified on the birth data and combined to allow for previous interaction found between these. ⁵ Maternal preexisting health conditions, including chronic hypertension, diabetes, asthma, autoimmune disorders, and psychiatric disorders were defined by identifying relevant hospital admissions any time prior to the estimated start date for the current pregnancy and on the birth data where available (applies to prior hypertension and diabetes; Table S1). Assisted reproductive technology (ART) was defined as hospitalizations for any recorded use in the 90 days prior to the current pregnancy.

2.3. Statistical analyses

Prevalence of HG was calculated as a proportion of all pregnancies in the study period, with variation by sociodemographic risk factors assessed. Negative binomial regression was used to examine standardized prevalence trends over time. To analyze factors associated with HG in pregnancy, robust Poisson models were fitted using generalized estimating equations with a log link and exchangeable correlation to account for clustering between pregnancies. Crude relative risk ratios (RR) and 99% confidence intervals (CI) for HG associated with each risk factor were estimated. An alpha level of 0.01 was chosen to account for the multiple statistical tests being undertaken. Given there was no primary exposure of interest and that many of the risk factors were potentially correlated or interrelated with lifestyle and health conditions, no multivariable analysis was conducted to avoid producing any biased effect estimates. ²³ Missing values for risk factors were coded and analyzed as separate category when there were >1% missing on a particular variable; otherwise, they were excluded from the analysis of interest. Sensitivity analyses assessing the risk of HG as defined by hospital visits only were also conducted as a proxy for HG severity, that is, to determine whether risk factor patterns differed for women who required higher levels of care.

To assess risk factors for multiple visits for HG within one pregnancy, the Prentice‐Williams‐Peterson Gap Time model, stratified by visit number and with a robust sandwich estimator, was used. The outcome was the time since the end of the previous hospital admission or emergency presentation. Crude hazard ratios (HR) and 99% CIs were estimated.

Hyperemesis gravidarum status at each pregnancy, patterns of recurrence, and interpregnancy intervals, defined as the number of completed months between the end of one pregnancy and the estimated start date of the next pregnancy, were determined for each woman. The proportion of women without a second birth and the interpregnancy interval between the first and second birth were compared between women with and without HG. Lasagna plots were produced to visualize patterns of HG status across pregnancies. ²⁴ Risk of HG in the second and third pregnancies was calculated using multivariable robust Poisson models given the pattern of HG in the previous pregnancies and adjusting for risk factors (described above) and interpregnancy interval in the latest pregnancy.

Category‐specific population attributable fractions (PAF) were calculated from the relative risk estimates from the relevant regression models to estimate the proportion of the HG prevalence that can be attributed to a specific risk factor category (eg age < 20 years). ²⁵ Statistical analyses were conducted using SAS Enterprise Guide 7.1 and plots were produced using R version 4.3.1.

3. RESULTS

There were 955 107 pregnancies to 630 609 women between 2010 and 2019 in NSW. Among these, 21 702 were classified as HG, yielding a prevalence of 2.3%. The prevalence increased from 1.7% in 2010 to 2.8% in 2019, with an average annual increase of 6.8% (95% CI 5.3–8.3). Hyperemesis gravidarum prevalence varied for different socioeconomic groups, with higher rates for women aged <20 years (4.8%), those born in African or Middle Eastern countries (3.2%), living in regional or remote areas (2.9%) and in the most disadvantaged groups (2.8% and 3.1% for 1st and 2nd quintile respectively) (Figure S1).

3.1. Risk factors for HG

After excluding nausea and vomiting in pregnancy pregnancies, there were 945 805 pregnancies to 627 040 women. Table 1 shows the characteristics of pregnancies with and without HG, and the risk of HG associated with each characteristic. Of the sociodemographic characteristics, the highest risk of HG was associated with women aged less than 20 years of age and attributed 3.8% of the HG prevalence (RR 2.53, 99% CI 2.33–2.74; PAF 3.8%) with other factors having modest association (Table 1). Pregnancies of a singleton female baby were associated with 1.2‐fold increased risk of HG compared to singleton male pregnancies (RR 1.19, 99% CI 1.15–1.23) and contributed to 8.4% of all HG. Multiple pregnancies were associated with double the risk of HG (eg multiple female: RR 2.04, 99% CI 1.70–2.44) but made very small contribution to overall risk (PAF 0.5%). Of maternal lifestyle and preexisting conditions, most were associated with a moderate increased risk of HG, including smoking (RR 1.24, 99% CI 1.17–1.32; PAF 2.3%), underweight (RR 1.27, 99% CI 1.12–1.42; PAF 1.2%), overweight (RR 1.11, 99% CI 1.04–1.19; PAF 2.5%), and obese (RR 1.23, 99% CI 1.15–1.32; PAF 3.8%) BMI categories, preexisting diabetes (RR 1.32, 99% CI 1.14–1.52; PAF 0.4%), asthma (RR 1.85, 99% CI 1.66–2.06; PAF 1.4%), autoimmune disorders (RR 1.54, 99% CI 1.34–1.76; PAF 0.7%) and psychiatric disorders (RR 2.51, 99% CI 2.32–2.73; PAF 3.2%). ART was associated with reduced risk (RR 0.80, 99% CI 0.69–0.93; PAF −0.1%) and hypertension was not associated with HG.

TABLE 1.

Maternal characteristics associated with hyperemesis gravidarum (HG) for pregnancies resulting in births from 2010 to 2019.

Characteristics a	HG (N = 21 702)	No HG (N = 924 103)	RR (99% CI)	PAF (%) b
	n (%)	n (%)
Sociodemographic factors
Year of birth
2010–2014	9172 (42.3)	466 482 (50.5)	1.00 (reference)
2015–2019	12 530 (57.7)	457 621 (49.5)	1.35 (1.30–1.40)	14.8
Age (years)
< 20	1197 (5.5)	22 952 (2.5)	2.53 (2.33–2.74)	3.8
20–24	4506 (20.8)	108 110 (11.7)	2.03 (1.93–2.14)	11.0
25–29	6638 (30.6)	246 801 (26.7)	1.36 (1.30–1.42)	8.8
30–34	6132 (28.3)	321 218 (34.8)	1.00 (reference)
35–39	2659 (12.3)	182 014 (19.7)	0.78 (0.74–0.83)	−4.4
40 +	565 (2.6)	42 863 (4.6)	0.71 (0.63–0.79)	−1.4
Remoteness category
Major cities	15 631 (72.0)	717 539 (77.6)	1.00 (reference)
Regional/remote	5700 (26.3)	186 858 (20.2)	1.39 (1.33–1.45)	7.4
Missing	371 (1.7)	19 706 (2.1)	0.88 (0.77–1.01)	−0.2
Socioeconomic disadvantage
1—Most disadvantaged	6178 (28.5)	203 525 (22.0)	1.28 (1.21–1.35)	5.8
2	4170 (19.2)	146 176 (15.8)	1.20 (1.13–1.27)	3.1
3	4160 (19.2)	177 565 (19.2)	1.00 (reference)
4	3553 (16.4)	167 897 (18.2)	0.92 (0.86–0.97)	−1.6
5—Most advantaged	3265 (15.0)	209 019 (22.6)	0.68 (0.64–0.72)	−7.7
Missing	376 (1.7)	19 921 (2.2)	0.83 (0.72–0.95)	−0.4
Country of birth
Australia/New Zealand	15 640 (72.3)	607 173 (66.1)	1.00 (reference)
Asia/Pacific	3256 (15.0)	192 877 (21.0)	0.66 (0.63–0.70)	−7.6
Europe/Russia	670 (3.1)	48 695 (5.3)	0.54 (0.48–0.60)	−2.5
Africa/Middle East	1767 (8.2)	52 903 (5.8)	1.28 (1.19–1.37)	1.6
Americas	311 (1.4)	17 609 (1.9)	0.68 (0.58–0.79)	−0.6
Pregnancy factors
Baby sex
Singleton male	9958 (45.9)	469 997 (50.9)	1.00 (reference)
Singleton female	11 154 (51.4)	440 773 (47.7)	1.19 (1.15–1.23)	8.4
Multiple male	187 (0.9)	4431 (0.5)	1.96 (1.63–2.35)	0.5
Multiple male/female	196 (0.9)	4092 (0.4)	2.23 (1.87–2.66)	0.6
Multiple female	196 (0.9)	4492 (0.5)	2.04 (1.70–2.44)	0.5
Lifestyle and health conditions in pregnancy
Smoking during pregnancy
Smoked	2515 (11.7)	87 310 (9.5)	1.24 (1.17–1.32)	2.3
Did not smoke	19 050 (88.3)	831 865 (90.5)	1.00 (reference)
Body mass index categories c
Underweight (<18.5 kg/m2)	526 (5.2)	15 981 (4.4)	1.27 (1.12–1.42)	1.2
Normal weight (18.5–24.9 kg/m2)	4899 (48.3)	190 176 (52.0)	1.00 (reference)
Overweight (25.0–29.9 kg/m2)	2419 (23.9)	84 129 (23.0)	1.11 (1.04–1.19)	2.5
Obese (≥30 kg/m2)	2003 (19.8)	63 132 (17.3)	1.23 (1.15–1.32)	3.8
Missing/invalid	292 (2.9)	12 549 (3.4)	0.92 (0.79–1.07)	−0.3
Conditions prior to pregnancy
Hypertension d	993 (4.6)	42 209 (4.6)	1.00 (0.92–1.09)	0.0
Preexisting diabetes d	337 (1.6)	10 715 (1.2)	1.32 (1.14–1.52)	0.4
Asthma d	678 (3.1)	15 425 (1.7)	1.85 (1.66–2.06)	1.4
Autoimmune conditions d	422 (1.9)	11 081 (1.2)	1.54 (1.34–1.76)	0.7
Psychiatric disorders d	1194 (5.5)	19 777 (2.1)	2.51 (2.32–2.73)	3.2
ART use e	271 (1.2)	14 811 (1.6)	0.80 (0.69–0.93)	−0.3
Pregnancy history
No previous pregnancies	9626 (44.4)	391 196 (42.3)	1.00 (reference)
Previous pregnancy—no history of HG	9066 (41.8)	523 232 (56.6)	0.72 (0.69–0.75)	−18.9
Previous pregnancy—with history of HG	3004 (13.8)	9365 (1.0)	8.92 (8.43–9.44)	9.4

Abbreviations: ART, assisted reproductive technology; CI: confidence interval; HG, hyperemesis gravidarum; PAF, population attributable fraction; RR, relative risk.

^a Missing categories are not included in the analyses for characteristics where the percent missing was <1%. Not all frequencies will sum to the total.

^b PAF is calculated for each specific category and is expressed as a percentage. ²⁵

^c Applies to 2016+ only. Age‐specific categories are used for mothers aged less than 18 years. ²¹ , ²²

^d Refers to conditions that required hospital admission any time prior to the start of the current pregnancy or (in the case of chronic hypertension and preexisting diabetes) were indicated on the birth data for the current pregnancy.

^e As measured by hospital admission for ART purposes in the 90 days prior to start of the current pregnancy.

History of HG in any previous pregnancy was associated with an almost nine‐times increased risk of HG compared to having no prior pregnancies (RR 8.92, 99% CI 8.43–9.44; PAF 9.4%), with 9.4% of the overall HG prevalence attributable to history of HG. Sensitivity analysis restricting to hospital admissions only revealed similar results, although history of HG hospital admission in a previous pregnancy was associated with over 15‐times increased risk of a HG hospital admission in the current pregnancy (RR 15.35, 99% CI 14.15–16.64; PAF 9.6%) (Table S2).

3.2. Multiple HG visits during pregnancy

Of the 21 702 pregnancies impacted by HG, 6443 (29.7%), involved multiple hospital or emergency department visits for HG (Table S3). After accounting for the time to each revisit, pregnancy history was the strongest risk factor (HR 1.40, 99% CI 1.34–1.48) (Table S3).

3.3. HG recurrence across pregnancies

There were 325 356 women whose first confirmed pregnancy resulted in a birth from 2010 to 2017. Among women with HG in their first pregnancy, 42.5% did not have a second pregnancy, compared to 39.8% of women without HG in the first pregnancy (Figure 1A,B; χ²(1) 22.76, p < 0.0001). For women with HG in their first pregnancy, the median interpregnancy interval was 580 days (IQR 355–910), compared to 583 days (IQR 385–884) for the women not impacted by HG in their first pregnancy.

FIGURE 1.

Lasagna plots of hyperemesis gravidarum status by pregnancy number, women with first known pregnancy resulting in birth from 2010 to 2017. (A) All women. (B) Women with at least one hyperemesis gravidarum pregnancy. In these plots, color is used to represent the outcome measure (HG status in a particular pregnancy), with each row representing a woman's measurements (HG, no HG, or no pregnancy) over the column variable, which is pregnancy number. Interpretation note: Where pregnancy number 2 is filled in with white space, this represents the women that do not have a second pregnancy on the data.

Among the 4931 women with an index HG pregnancy, 1490 (30.2%) had HG in at least one of their following pregnancies (Figure S2). There were 192 628 women with at least two pregnancies across the study period (Figure 2A). Of the 4197 women who had HG in their first pregnancy, 1215 (28.9%) had HG in their second. In comparison, only 1.8% of women without HG in their first pregnancy, had HG in their second. After adjusting for characteristics of the second pregnancy, women with HG in their first pregnancy had almost 14‐times the risk of HG recurrence in their second compared to women with no history of HG (aRR 13.75, 99% CI 12.64–14.96; PAF 24.7%). There were 40 618 women with at least three pregnancies (Figure 2B). Of the 245 women who had HG in their first and second pregnancy, 54.7% had HG again in their third. Compared to having no history of HG, this group had almost 30 times the adjusted risk of having HG in their third pregnancy (aRR 29.59, 99% CI 24.35–35.95; PAF 12.7%). For women who had HG in only one of their first two pregnancies, the risk of having HG in their third pregnancy was similarly high regardless of how recent the latest HG was.

FIGURE 2.

Risk of hyperemesis gravidarum by recurrence pattern, women with first known pregnancy resulting in birth from 2010 to 2017. (A) Second pregnancy. (B) Third pregnancy.

4. DISCUSSION

This study provides contemporary findings on HG prevalence, risk factors, and recurrence with key messages summarized in Figure 3. The prevalence of HG associated with women presenting to hospital or the emergency department was 2.3% and increased over time to 2.8% in 2019. Furthermore, almost one in three women with HG had multiple hospital admissions or emergency department presentations during their affected pregnancy. History of HG in previous pregnancy increased a women's risk of HG in (current) pregnancy up to nine‐fold and was attributable to 9.4% of the overall HG prevalence. Other sociodemographic risk factors, preexisting maternal conditions, and infant female sex were also risk factors for HG.

FIGURE 3.

Summary of main findings.

Following first HG, one in three (30.2%) women had HG in any subsequent pregnancy with recurrence increasing with each successive pregnancy. For those with HG in their first pregnancy, 28.9% had HG in their second, and for those with HG in their first and second pregnancies, 55% had HG in their third. Of note, there was a small difference in the percentage of women not having a second pregnancy based on whether they had HG (42.5%) or not (39.8%). However, whether or not a woman had HG in their first pregnancy did not seem to impact the length of time until her second pregnancy. Further, these findings suggest that while there may be some women who avoid having a second pregnancy after HG, those who do have a second pregnancy do not wait any longer to do so compared to those without HG. The strongest driver of HG, both within pregnancy and across successive pregnancies, was having a history of HG in a previous pregnancy, which has implications for the care and counseling that women receive around pregnancy.

The main strength of the study is our large population‐based cohort and ability to follow up women's pregnancies and health service visits over time. Additionally, our outcome definition only included diagnosis codes that were specific to HG, to minimize misclassification. However, a limitation is that we were unable to capture primary or outpatient care for HG in our outcome definition. As such, it is unclear whether our findings adequately represent women who are afflicted with HG vs access to health services. Women who receive adequate treatment in the community may not require secondary care if their symptoms are managed sufficiently, which is important considering that younger and more socially disadvantaged women were at higher risk of HG according to our study. There is some evidence that secondary care‐only data may underestimate HG prevalence, with a study from England reporting HG diagnosed and managed in primary care alone was 2.5% of pregnancies. ²⁶

As is common with population‐level perinatal studies, birth data only captures live births or stillbirths of at least 20 weeks gestation or 400 g birthweight. As such, women with hospital admissions or emergency department presentations for HG that could not be linked to the birth data (~10% of all HG visits identified) were not included in our analyses. Potential reasons these records could not be linked are that these pregnancies may have resulted in a miscarriage or termination less than 20 weeks gestation, or these women gave birth out‐of‐state. As a result, our recurrence rates and estimates of future pregnancies are potentially underestimated. Future work will be undertaken to examine the outcomes, where available, for these cases. Information on treatment was not available at the time of our study, so we could not determine the role that treatments or interventions such as antiemetic medications and intravenous fluids played on the need for hospital admissions or ED presentations for HG.

The estimated prevalence of HG in our population was 2.3%, which was within the range commonly cited in the literature. ¹ Other known population‐based cohort studies from the last decade have reported slightly lower prevalence estimates (1.5% based on hospitalization data and 1.3% based on hospitalization and outpatient data), ³ , ⁵ which may be due to differences in HG definitions utilized and healthcare systems. The reported prevalence estimates are similar to those we observed when classifying HG based on hospital admissions only. Potential reasons why the prevalence of HG increased over the study period could be increased awareness of the condition, improvement in coding practices, or limited access to primary or outpatient care, leading to an increase in the number of women requiring and being diagnosed via hospital or ED services.

We found a range of maternal and infant risk factors to be associated with HG. Previous population‐based studies also identified younger maternal age, ³ , ⁴ , ⁶ socioeconomic disadvantage, ⁷ maternal country of birth in African regions, ⁶ multiple pregnancy, ³ , ⁴ , ⁵ , ⁶ , ⁸ , ⁹ being pregnant with a female fetus, ³ , ⁴ , ⁵ , ⁶ , ⁸ , ⁹ BMI, ³ , ⁹ preexisting diabetes, ⁴ , ⁵ , ⁷ asthma, ⁴ autoimmune conditions, ⁴ , ⁵ psychiatric disorders ⁴ , ⁷ and previous HG pregnancy ⁵ as risk factors for HG. While a prior study found that ART was a risk factor for any HG, ³ we found that ART was associated with a decreased risk of HG. This may be influenced by our exposure definition, as we can only capture ART procedures that occur in hospital. There is limited research into the relationship between ART and HG, and as such this would be a worthy topic for further exploration. Smoking during pregnancy has previously been found to be protective for or have no association with HG, ³ , ⁴ , ⁹ , ²⁷ which contrasts with our finding. While biological mechanisms have been hypothesized, there is no established pathway between smoking and HG. ²⁸ Further, the maternity population, smoking rates and access and use of healthcare services may influence results. For previous studies, the protective effect may have been driven by women who smoke being less likely to present to the hospital with HG. However, there is some evidence that the opposite effect may occur in Australia and other countries, whereby women who smoke are more likely to have antenatal admissions. ²⁹ , ³⁰ As such, replicating this analysis using other Australian‐based population health data would add useful information to the existing body of research. Previous studies found HG inpatient readmission rates of 17% ²⁷ and 28%, ⁵ the latter being comparable to our rate of 30%, although we included emergency department presentations in our definition of health service use.

HG in a previous pregnancy was the strongest risk factor for HG in the current pregnancy, which is also evident when examining recurrence across pregnancies. Recent studies using similar definitions of recurrence found slightly lower, yet generally comparable, recurrence estimates to our study. For example, recurrence after the HG index pregnancy has been reported at 22% ³ and 24%, ¹² compared to our 30%. Fassett et al. ¹¹ found recurrence estimates of 24% in the second pregnancy and 48% in the third pregnancy, compared to 29% and 55%, respectively in our study.

These findings provide clinically useful information for identifying women at higher risk of HG, both in their first and additional pregnancies. The establishment of the international consensus definition for HG should ensure more standardized screening and diagnosis of HG, as well as inform patient care. ³¹ The determination of specific risk factors for HG and ability to follow‐up women with a history of HG will aid in early detection of the condition and identify women who would benefit from early intervention alongside the new models of care. The provision of recurrence estimates for different scenarios facilitates informed reproductive decision‐making. This may include better education of both women and primary and secondary care providers, development of improved models of care for women at high risk of HG, including holistic and multidisciplinary care. ³²

Future studies should examine the impact of the international consensus definition on HG prevalence. More research is required to investigate how the different types of treatment impact women's experience with HG, including their hospital admissions or emergency department presentations.

5. CONCLUSION

HG has a significant impact on women and their families, often causing physical, emotional, and financial distress, in addition to the burden placed on the health system. Our study demonstrated that while there were several maternal and infant characteristics associated with HG, the strongest risk factor was a history of HG in a prior pregnancy, which was also evident in the high recurrence rates observed across subsequent pregnancies. Providing estimates of the HG burden ensures population‐based evidence is included in the planning, development, and delivery of HG services that serve to minimize its impact. Given the increasing recognition and prevalence of HG over the last decade and the history of HG being the strongest risk factor for HG, future models of care should identify and target women with a previous history of HG to ensure optimal maternal and infant health outcomes.

AUTHOR CONTRIBUTIONS

Diana M Bond, Antonia W Shand, Natasha Nassar conceived the study, and with Sarah Pont designed the statistical analysis plan. Sarah Pont conducted the statistical analysis and drafted the manuscript. Antonia W Shand and Iqra Khan provided clinical expertise. All authors critically reviewed, contributed to, and approved the final version of the manuscript.

FUNDING INFORMATION

This study was funded by the NSW Health Hyperemesis Gravidarum Research Grants Program and NN was supported by the Financial Markets Foundation for Children and Australian National Health and Medical Research Council Investigator Grant (APP1197940) and HZ a UNSW Scientia Fellowship.

CONFLICT OF INTEREST STATEMENT

The authors have no conflicts to declare.

ETHICS STATEMENT

Ethics approval for this study was attained from the NSW Population and Health Services Research Ethics Committee on September 18, 2019 (2019/ETH11532). The committee provided a waiver of consent for participants.

Supporting information

Figure S1.

Figure S2.

Table S1.

Table S2.

Table S3.

Pont S, Bond DM, Shand AW, Khan I, Zoega H, Nassar N. Risk factors and recurrence of hyperemesis gravidarum: A population‐based record linkage cohort study. Acta Obstet Gynecol Scand. 2024;103:2392‐2400. doi: 10.1111/aogs.14966