Health of singleton neonates in Switzerland through time and crises: a cross-sectional study at the population level, 2007-2022

This study uses routinely collected data from the FSO. The data set covers all births that occurred between 1987 and 2022 in Switzerland, at the monthly and municipality level. Due to data-anonymity rules, it was not possible to obtain data at the weekly level. We used data from 2007 because gestational age was only systematically recorded from that year onwards. The data were provided by the FSO in a fully anonymized form upon a contractual agreement. Information regarding language region, urbanity and altitude of maternal municipality, as well as the annual number of permanent Swiss residents were obtained from the FSO websites [52, 53]. According to the Human Research Act, no ethical clearance is required when working with fully anonymized governmental data [54].

Between 2007 and 2022, the database contained 1’517’751 births. Of these, we kept births that took place in Switzerland and for which gestational age was recorded (n=1’365’805). We then excluded births from mothers without a permanent resident status (n=43’401, comprising of short-term residents, asylum-seekers or persons working in Switzerland but living in another country), multiple pregnancies (n=46’666), and missing birth weight (n=263). Cases where either birth weight <500 grams (g) or gestational age <22 weeks were also excluded (n=633), since these are classified as miscarriages and not stillbirths. This implies that cases where birth weight was <500g were kept as long as gestational age was >22 weeks, because these classify as stillbirths. We also excluded missing or unlikely birth length (<20cm or ≥65cm, n=216) and birth weight values (<100g or ≥7 500g, n=11), and those with a large discrepancy between birth weight and gestational age (birth weight >2’000g and gestational age <23 weeks, or birth weight <500g and gestational age >35 weeks, n=21). Births from mothers aged over 50 years old were excluded as well (n=146), because they have much higher risks of adverse pregnancy outcomes [55]. They are also much more likely to give birth by caesarean section [56], and our dataset does not include information about delivery mode. The final dataset for the analysis of stillbirth analysis included 1’274’449 births. From this dataset, 4’863 stillbirths were excluded to explore birth weight and PTB outcomes, resulting in 1’269’586 livebirths. A flowchart of exclusion steps can be seen in Supplementary (Suppl.) Figure S1.

We examined three outcome variables: birth weight (continuous, in grams), PTB rate (<37 completed weeks of gestation) and stillbirth rate (either gestational age ≥22 weeks or birth weight ≥500g). Birth weight is usually measured in the first hour after birth using a calibrated scale. For more than 20 years, ultrasound scan has been routinely performed to determine gestational age in almost all pregnancies in Switzerland before 12 weeks of gestation. Information on birth weight and gestational age is collected by the hospitals and midwives by filling in a form [57] which is then entered into the FSO database.

The following covariates which were expected to be associated with the outcomes were used in the analyses. Maternal age (in years) was included as a continuous individual-level variable. In addition, we included continuous ecological variables based on the municipality in which the mother was living at the time of delivery: altitude (in meters above sea level (MASL) based on mean altitude of all residential buildings of the municipality), and mean Swiss neighbourhood index of socioeconomic position (SSEP 2.0) developed by the Institute for Social and Preventive Medicine at the University of Bern [58]. The SSEP is a high-resolution area-based index that allows studying SEP when individual-level information is missing. In our dataset, SSEP ranges from 23.6 to 86.7 index points, with higher index indicating higher SSEP. Regarding categorical variables, maternal nationality was categorized as Swiss, African, Asian, non-Swiss European, Northern American, Southern and Central American, missing and Oceanian. As the proportion of women from Oceania was very small, it was not considered a separate category and was included in the “missing” category. Civil status was categorized as single vs. married/in a registered partnership, parity as 1, 2, 3 or >3 and neonatal sex as male vs. female. At the ecological level, we included urbanity (urban vs. rural region) and language region (German/Romansh-, French- or Italian-speaking Switzerland) as categorical variables, based on maternal municipality of residence using official FSO categorisations. Language region roughly reflects cultural, behavioural, as well as dietary, smoking or alcohol consumption patterns [50].

Month and year of birth were used as a continuous numerical variable from January 2007 to December 2022, adding up to a total number of 192 months. We examined four crises, namely heatwaves (2015 and 2018), the Great Recession (2008/2009), the H1N1 influenza pandemic (2009/2010) and the COVID-19 pandemic (from 2020). Heatwaves: the two most significant heatwaves during the studied period occurred in July 2015 and August 2018 [59]. As our temporality is only at the monthly level, we used monthly information and do not consider heatwaves’ duration nor temperatures reached. We assume that the effect of both heatwaves on neonatal health is comparable and therefore we combine the two heatwaves under the variable “heatwave exposure”. This dummy variable takes the value 1 if the mother was pregnant during a heatwave (in any month of pregnancy), and 0 otherwise. For the other three crises, which lasted longer than one month, we created relative continuous exposure variables. Great Recession: the economic crisis lasted six months (2008-10 to 2009-03) [60]. The number of pregnancy months overlapping with the Great Recession was summed up for each birth, then divided by gestational age at birth and normalized between 0 and 1. Flu pandemic: the crisis lasted 4 months (from 2009-10 until 2010-01 [61]). The flu exposure variable was constructed in the same way as for the Great Recession. The Great Recession and flu pandemic variables therefore quantify the duration of crisis exposure, relative to the total pregnancy duration. We chose exposure relative to pregnancy duration rather than absolute exposure (total number of months of pregnancy that overlapped with a crisis): infants with the shortest gestational duration were also exposed to the crisis for a smaller time period. If we do not adjust for pregnancy duration we risk seeing that the longer crisis exposure is, the less risk there is to have adverse pregnancy outcomes (stillbirth, lower birth weight, PTB). COVID-19 pandemic: the pandemic hit Switzerland in early 2020. As a proxy for COVID-19 exposure, we used the number of COVID-19 hospitalisations that occurred in Switzerland. Data on weekly hospitalisations from 2020-02 to 2022-12 were retrieved from the FSO [62] and aggregated at the monthly and national level. The number of hospitalisations that occurred in the country during each month of a woman’s pregnancy was summed, then divided by gestational duration and normalized between 0 and 1. The COVID-19 variable therefore quantifies the intensity of exposure per pregnancy month. The distribution of the crisis variables is displayed in Fig. 1.

We ran additional analyses to investigate the effects of exposure to a crisis occurring in the most vulnerable timings, i.e. first trimester (1^st – 3^rd month) or last trimester of pregnancy [66‐68]. The last trimester generally corresponds to months 7-9. For the few pregnancies that lasted only two trimesters (gestational age <26 weeks, approx. 0.2% of cases), the last trimester then corresponds to pregnancy months 4 to 6. For the Great Recession and the flu, the number of months during the first trimester (0, 1, 2 or 3) which overlapped with either crisis was summed up and normalized between 0 and 1. For COVID-19, the number of hospitalisations which occurred during the first trimester was summed up, then normalized between 0 and 1. The same process was followed for last-trimester exposure. We built two models for each outcome variable: one for first-trimester exposure, and one for last-trimester exposure. Each model was adjusted for the same covariates than the main models. For an example, see Suppl. material (Equation E2).

All models are summarized in Table S14.

R version 4.3.2 was used [69]. GAMs were built with mgcv package [70].

Between 2007 and 2022, 1’517’571 births were recorded in Switzerland. The maternal and neonatal characteristics of this population can be seen in Suppl. Table S1. After exclusion criteria (see Figure S1) we ended up with 1’274’449 singleton births in our analyses modelling stillbirth cases (whole time-period characteristics displayed in Table 1, while yearly characteristics are displayed in Tables S2-3 and Figure S2). The distribution of parity remained stable during the time period (Figure S2 A). We noticed that the share of women giving birth to singletons in German-speaking Switzerland (the country’s biggest area) decreased from 76.5% in 2007 to 71.0% in 2022, while the one from French-speaking Switzerland increased by 6.6 percentage points over the same period (Figure S2C). Most births were given by Swiss mothers, but their share declined from 65.1% to 60.7% between 2007 and 2022. Singleton neonates kept a rather constant mean birth weight, oscillating around 3 325g, with a mean gestational age of 39.3-39.4 weeks (Table S2, Figure S3D-E). Stillbirth rate slightly changed over the years without any discernible pattern (Figure S2G). The mean rate of PTB declined over the years, from 5.9% to 5.1% in 2022 (Figure S2H). LBW rate was at its lowest during the last four years under study (Figure S2I). To study birth weight and PTB outcome, we excluded stillbirth cases (n=4’863), resulting in the 1’269’586 livebirths (Table S3). Yearly birth rate among permanent residents was on an increasing trend between 2007 and 2016, from 9.86 to 10.50/1’000 inhabitants, but then declined until 2020. It increased back in 2021 to 10.28/1’000  inhabitants but reached its lowest rate of the whole time period in 2022 (Table S4, Figure S4).

Higher exposure to the COVID-19 pandemic during the last trimester was associated with an increase in birth weight (16.8 g, [95%CI 10.7 to 23.0], Table 5). However, heatwave exposure during the last trimester was associated with a lower birth weight (-10.8 g, [95%CI -15.8 to -5.9], Table 5): this is inconsistent with the main birth weight model (Table 2). Still, Cohen’s d effect sizes are rounded to 0, showing that although some associations are significant, their effect sizes are very small. Using LBW as an outcome, we find that only last-trimester exposure to COVID-19 was significantly associated with a slightly lower risk of LBW (Table S10), thus matching with the higher mean birth weight reported in Table 5. The aforementioned finding of a lower PTB risk when the mother was exposed to a heatwave during pregnancy (Table 3) may be mediated by first-trimester exposure (OR 0.95 [95%CI 0.91 to 1.00]), while last-trimester exposure to a heatwave did not appear to affect the risk of PTB (Table 5). On the contrary, intensified exposure to the COVID-19 pandemic during the first three months of pregnancy was associated with a slightly higher PTB risk (OR 1.06 [95%CI 1.00 to 1.12]). Last-trimester exposure to the COVID-19 pandemic was associated with a higher stillbirth risk, with an OR of 1.24 [95%CI 1.02 to 1.50] and a Cohen’s d of 0.12, denoting a small effect.

Results of the models using the 2009 flu pandemic instead of the Great Recession can be seen in Tables S5-8 and Figures S8-10 and indicate that being exposed to the flu slightly reduced the risk of PTB in general (OR 0.89 [95%CI 0.83 to 0.95], Table S6), and in the first as well as last trimesters (Table S8).

Using only first parities in the models (Tables S11-13, Figures S13-14), the results were consistant with the main models including all parities.

The World Health Organization (WHO) reported that global PTB rates were stable from 2010 to 2020, with values around 10% even among HICs [16, 25]. However, PTB rates were decreasing between 2006 and 2014 in the US but also in Norway, as we report here for Switzerland [24]. While changes in the way gestational age is reported can contribute to PTB rate variation [23, 71, 72], obstetric interventions can also influence it. In Switzerland, caesarean-section rate remained consistently higher than 32% between 2007 and 2022 [73], well above the 10% rate recommended by the WHO [74]. However, the rate varied too little to fully account for the declining PTB trend. In 2017, assisted reproductive technology (ART) changed with the introduction of a single embryo transfer [75]. This reduced the occurrence of multiple births, thus decreasing the number of PTB associated with multiple pregnancies [75]. Although we focus on singletons, these changes might have affected their outcomes as well, increasing the number of singleton births born through ART.

Our result on stillbirth rate mirrors data already-published from the FSO: only minor variations of unadjusted rates are reported between 2007 and 2022 [49]. While some reports point towards decreasing trends between 2000 and 2019 in Western Europe and other HICs [27, 28], this decline has recently slowed down [29]; stillbirth rate was even stable in Germany between 2009 and 2012 [76]. Stillbirth rate might be stagnating in Switzerland because of increasing maternal age [29] or prevalence of maternal comorbidities (such as obesity and diabetes), impeding stillbirth rate of declining further.

Birth weight evolution varied a lot across countries around the turn of the 21^st century. Some HICs experienced declining birth weight [17‐20], but it increased in England [21], while Norway and Sweden exhibited an unexplained increase and then decrease of birth weight [19]. We see that birth weight was mostly constant in Switzerland during the studied period.

Crises such as economic depressions and pandemics have a direct burden on the population’s health in terms of mortality and morbidity, but also create social and economic disruptions and traumatic experiences. The highest-risk crises are without doubt armed conflicts and natural disasters, increasing stress among pregnant individuals through life-threatening events and having an uncertain future. These types of crises have also been associated with higher rates of LBW [30]. This paper focuses on population-level crises, and especially on their potential indirect impact on pregnancies, through stress exposure. In our univariable models, we do not see shifts in birth weight, nor in the rates of PTB or stillbirth during any of the crises investigated. When we consider crisis-exposure in more details using multivariable models, exposure to the Great Recession or to COVID-19 were not associated with variations in PTB nor with stillbirth probabilities. However, when we separate the exposure per pregnancy trimester, COVID-19 slightly increased PTB risk for first trimester exposure, and stillbirth risk for last-trimester exposure.

The impact of an economic crisis on neonatal health depends on the initial family financial situation. For instance, an individual-level study found an important association between inadequate employment and birth weight decrease [37]. In Switzerland, the unemployment rate almost doubled between the beginning of the Great Recession and the end of 2009, from 2.4% to 4.4% [77]. If the crisis only impacted already socioeconomically disadvantaged families, we might not be able to see an important effect overall. We control for SSEP, an area-based information, but lack individual information on employment status. The Great Recession was associated with more important effects in countries already socioeconomically disadvantaged. In Portugal, a higher LBW prevalence was noted [35], while in Greece, crude stillbirth, infant and child mortality rates gradually increased during the Great Recession [36]. Given that Switzerland ranks among the highest income countries, it was likely less affected by the economic crisis.

Multiple studies reported decreased PTB rates during the COVID-19 pandemic [78], including a meta-analysis on 52 million births from 26 countries [79]. However, some suggest that this may be due to reporting biases, because the association was no longer significant for adjusted rates [78]. In Switzerland, PTB rate slightly declined in 2020 [80], but the year 2019 was already displaying lower rates, indicating that the reduced PTB odds in 2020 is unlikely to be solely attributable to the pandemic. In addition, the previously-mentioned meta-analysis did not identify any change in PTB rate during the COVID-19 lockdown in Switzerland [79]. Our results are thus consistent with the literature and, with the use of individual-level information, add further evidence suggesting that the COVID-19 pandemic only had, if any, a small effect on PTB rates.

Separating the exposure by pregnancy trimesters, we find a higher risk of stillbirth for exposure to COVID-19 during the last trimester, with an OR of 1.24 [95%CI 1.02 to 1.50]. The associated Cohen’s d is 0.12, which is higher than any d’s of the other crises among the trimesters-effect models, but still denotes a very small effect size. A large meta-analysis mentions no change in stillbirth rate during the pandemic (OR 1.08, [95%CI 0.94 to 1.23]), n= 21 studies) [78]. Few studies separated analyses by pregnancy trimester. In the Indian state of Bihar, stillbirth rate increased in a dose-relationship manner with the number of last trimester months that occurred during the pandemic peak [81]. Whether our finding of a slight increase in stillbirth rate results from maternal severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection or increased stress exposure is uncertain. Although the literature on SARS-CoV-2 vertical transmission is conflicting, maternal infection has been reported to cause placental inflammation. A multi-national study linked 68 stillbirths with maternal SARS-CoV-2 infection, and identified placental abnormalities that could have caused placental insufficiency and foetal death [82]. Maternal stress might be another explanation: shortage of obstetric staff and reduced prenatal care, including mothers postponing visits due to fear of getting infected, might have played a role in stillbirth rate, as Khalil et al. suggest [83].

When women were exposed to COVID-19 during pregnancy, birth weight was higher by 11.7g [95%CI 5.5 to 17.9]). Birth weight was higher for last trimester (16.8g [95%CI 10.7 to 23.0]) but not first trimester exposure (-6.3g [95%CI -12.2 to -0.4]). Two meta-analyses reported small birth weight increases during the pandemic (17g [95%CI 7 to 28] [78] and 15g [95%CI 10 to 20] [47]). In Denmark, a similar increase during the first lockdown was noted among singleton term births [84]. The sex- and gestational age- adjusted increase was of 17g [95%CI 3 to 31]. Increasing birth weight can indeed be due to longer pregnancy duration, but we did not adjust the birth weight-outcome model for gestational age due to potential collider bias [63‐65]. To assess whether our result of a higher birth weight could be linked to longer gestational duration, we display yearly rates of term births for each gestational week: babies did not have longer gestation during the COVID-19 years (Table S15). This supports the idea that higher birth weight among mothers exposed to COVID-19 is not due to longer gestation. Our results align closely with existing literature, showing a similar birth weight increase. The implications of such a small increase are uncertain and do not always imply changes in LBW odds [78]. However, we here found that COVID-19 exposure during the last trimester was also associated with a slight reduction of LBW risk (OR 0.94 [95%CI 0.88-1.00].

Indirect effects of the pandemic may be negative, i.e. through stress exposure or dietary changes, associated with a more sedentary life during lockdowns. Home office was frequent during the pandemic, and might have induced changes in nutrition and a lack of exercise, which could explain the slight increase in birth weight. Furthermore, pregnant women’s mental health was reported to have worsened during the pandemic [45, 85], including in Switzerland [46]. On the contrary, pandemic exposure could also have had positive effects: Switzerland had a light lockdown and was less economically affected compared to surrounding countries: there might even have been an improved work-life balance and hygiene, and a limited exposure to pollution. Still, some women of our dataset were directly affected through SARS-CoV-2 infection, with well-documented increased risks of maternal mortality, PTB and LBW [44, 86, 87]. Thus, the effects of the pandemic on maternal and neonatal health may be opposite [84]: if there were both negative and positive effects of the mitigation measures put into place, as well as negative effects of maternal infection, they might have compensated each other. This could explain the overall absence of association between the COVID-19 pandemic and the neonatal health outcomes we investigated, except for a small stillbirth risk increase. COVID-19 vaccination was extended to all pregnant women in September 2021 in Switzerland [88] and might have prevented severe complications and influenced PTB and stillbirth rates. Unfortunately, there exist no vaccination coverage data among pregnant women. Those who gave birth in 2022 might have been vaccinated during pregnancy, which limits our interpretation. We also observe a birth rate decline in the beginning of 2022. This downturn in births was reported in multiple countries, starting in January 2022 [89]. The effect of crises on fertility rates requires further study and was not the objective of the current study.

Exposure to heatwaves during pregnancy was associated with higher birth weight and lower probabilities of PTB and stillbirth. However, exposure during the first trimester was not associated with birth weight, while last-trimester exposure was linked to lower birth weight. The discrepancy between the main model and the trimester-effect models might be explained by the structure of our data: using monthly-level data may not allow investigating heatwave exposure in detail, since a heatwave usually lasts only a few days. Moreover, some pregnancies which started or ended during a heatwave month might have actually not overlapped with the heatwave event. These pregnancies would have thus been miscategorized as “exposed to a heatwave”; this should however concern very few pregnancies. Most studies describe positive associations between heat exposure and PTB [31], but a meta-analysis concluded that the evidence of reduced birth weight is limited [34]. Few studies have focused on stillbirth, but their majority report an increased risk [31, 32]. In low- and middle-income countries, already experiencing the highest PTB and stillbirth rates, higher temperatures were associated with increased risks of both outcomes [33]. As the frequency and severity of heatwaves is likely to increase in the near future, it is crucial to continue to assess their effect on foetal health. The structure of our data does not allow us to conclude on the effect of heatwave exposure on neonatal health.

The effects of the covariates on which we adjusted the analyses are consistent with the literature, with higher risks of PTB and stillbirth when maternal age increases [13, 90], and primiparous mothers having lighter babies [91]. Birth weight was reduced in French- and Italian-speaking regions of Switzerland, and also at elevated altitudes, as shown previously [50]. Babies born from non-Swiss mothers had a higher risk of stillbirth: Turkish mothers living in Switzerland were previously reported to have a 30% higher stillbirth risk [92]. We also show that mothers from Asian and Southern/Central American regions have higher risks of PTB, while those from Europe and Northern America have a lower risk: maternal ethnicity is known to affect neonatal health, probably more through an interplay of environmental and sociodemographic parameters than through genetic determinants [93‐95]. However, we use nationality as a crude proxy for ethnicity, which is limited. Interestingly, we highlight an effect of civil status on neonatal health, increasing stillbirth risk by 38%. Similarly, neonatal and infant mortality were recently reported to be much higher among unmarried mothers in Switzerland [65]. Important covariates that we could not adjust for are maternal comorbidities and smoking, known to be associated with lower birth weight [96] and stillbirth risk [90]. Our data does not include information on maternal overweight (body mass index (BMI) ≥25) and obesity (BMI ≥30) which are major risk factors for stillbirth in HICs [90].

Our study is thus limited by the structure of our dataset and available covariates. Regarding COVID-19, we cannot identify mothers who were infected by the virus: it is thus not possible to disentangle the direct effect of maternal infection from the indirect effect of stress and mitigation measures. We used hospitalisation cases as a proxy of COVID-19 exposure, regardless of waves and circulating viral strains. Regarding spatial differences, it was mostly during the first COVID-19 wave that the Cantons were differently affected [97]. However, few people were actually infected during this wave [98], and the subsequent waves had similar magnitudes geographically [97]. Furthermore, at the population level, we expect the waves to have the same effect on pregnant mothers in terms of stress. The choice of outcomes could also be discussed: the PTB variable does not differentiate between spontaneous and iatrogenic PTB, and these have different determinants [15].

The robustness of our results relies on the very large sample size, covering nation-wide singleton births at the individual level. Our birth weight- , LBW- and PTB-stratified analyses display consistent results. We can thus be confident that birth weight was stable through time in almost every subgroup assessed, and that PTB declined during the same period. Our sensitivity analyses focusing on first parities, and our analyses with the flu pandemic exposure instead of the Great Recession, enhance the reliability of our findings.