Epidemiology of influenza in pregnant women hospitalized with respiratory illness in Moscow, 2012/2013–2015/2016: a hospital-based active surveillance study

This was a prospective, active surveillance hospital-based study conducted during the 2012/2013 to 2015/2016 influenza seasons at CHID#1 (Moscow, Russian Federation). CHID#1 is a unique, specialized department hospital for pregnant women with infectious diseases. The study was performed due to the study site’s participation in the GIHSN international influenza surveillance project and its use of the standardized GIHSN protocols [16].

Patients admitted to the participating hospital were included, after written consent, if they were residents in the predefined hospital’s catchment area, presented with an acute illness possibly related to influenza, were not institutionalized, and were admitted within 7 days of the onset of symptoms. Patients discharged during the previous 30 days were excluded. Swabs were collected within 48 h from patients meeting the inclusion criteria and tested by reverse transcription-polymerase chain reaction (RT-PCR) for influenza. Influenza-positive samples were sub-typed by RT-PCR to identify A(H1N1)pdm09, A(H3N2), B/Yamagata-lineage, and B/Victoria-lineage strains.

The present analysis was limited to women 15–44 years of age admitted with an acute respiratory infection, in line with the age range used by others [3, 17, 18]. All other aspects of patient selection were in accordance with the GIHSN study protocol [16]. RT-PCR was conducted at the World Health Organization National Influenza Centre at the Ivanovsky Institute of Virology (Moscow, Russian Federation) using Amplisens® Ribo-sorb and Ribo-prep (Federal Budget Institute of Science “Central Research Institute for Epidemiology”, Moscow, Russian Federation) or a PREP-NA DNA/RNA extraction kit (DNA-Technology, Moscow, Russian Federation) to extract RNA; a Reverta-L kit (Federal Budget Institute of Science “Central Research Institute for Epidemiology”) for reverse transcription; and kits from Federal Budget Institute of Science “Central Research Institute for Epidemiology” and DNA-Technology to amplify influenza A, A(H1N1), A(H3N2), A(H1N1)pdm09, and B genes.

Socio-demographic and clinical information were collected by face-to-face interviews with patients or attending physicians or by reviewing clinical records. Information collected included socio-demographic characteristics, the major complaint at admission, smoking habits, underlying conditions, vaccination status, pregnancy outcome during the current admission, clinical course, and major diagnosis at discharge. Registered pregnancy outcomes included abortion (terminated at < 20 weeks gestational age), stillbirth (≥ 20 weeks gestational age with no heartbeat or respiratory effort), delivery (birth at any gestational age with heartbeat or respiratory effort), live preterm birth (< 37 weeks of gestation), low birth weight (< 2500 g), perinatal death (i.e. during the mother’s current admission), and caesarean delivery. Main discharge diagnoses were recorded by the physician using ICD-10 codes.

Statistical analyses were restricted to women recruited in periods with continuous influenza circulation, defined as ≥2 admissions positive for influenza in ≥2 consecutive weeks. Calendar time (admission week) was modeled using restricted cubic splines with four knots. The number of knots was set based on the Akaike information criterion [19]. The odds ratio (OR) of admission with influenza was calculated by bivariate logistic regression using the category with the lowest value as the reference. To estimate the significance of differences among groups, chi-squared, likelihood ratio, t, and nonparametric K-sample tests were used. For comparisons of continuous variables among multiple categories, equality of medians and one-way analysis of variance were used with Scheffe correction for multiple comparisons. Likelihood ratio tests were used to check for confounding, interaction, linearity, and clustering. The adjusted odds ratio (aOR) of influenza among pregnant women was estimated by multivariate logistic regression using minimal sufficient adjustment by variables identified as confounders by causal diagrams (e.g. age, underlying conditions, smoking habits, admissions in the previous year, time to swab, season, and epidemiological week at admission). The goodness of fit of the models was assessed using Akaike and Bayesian information criteria. Conditional plots [20] of the average predicted probability of influenza positive admission and of pregnancy outcome during admission were used to assess complex relationships between the pregnant women’s age in years, their infant’s gestational age, underlying conditions, influenza infection, and infection by A subtype or B lineage. The predicted probabilities of either influenza infection or pregnancy outcomes during admission were adjusted by age, smoking habits, chronic underlying conditions, admission in previous 12 months, pregnancy trimester, time to swab, and calendar time (season-week) as restricted cubic splines. All p-values were two-tailed. A p-value < 0.05 was considered to indicate statistical significance. Heterogeneity in the effects of risk factors were quantified using the I² test, with heterogeneity defined as an I² > 50%. All statistical analyses were performed with Stata/SE version 14 (College Station, TX, USA).

The study included 1992 women 15–44 years old admitted for acute respiratory diseases between 2012/2013 and 2015/2016. Of these admissions, 1748 were pregnant (Table 1). Influenza was detected in 47.2% (825/1748) of pregnant admissions and 23.8% (58/244) of non-pregnant admissions (OR for influenza = 2.87 [95% confidence interval (CI), 2.10–3.92]; p <  0.001; data not shown). Proportions of pregnant admissions with influenza were similar during the four influenza seasons included (48.7% in 2012/2013, 44.5% in 2013/2014, 52.4% in 2014/2015, and 44.9% in 2015/2016; p = 0.112 by test for homogeneity of equal odds; data not shown). Proportions of non-pregnant women with influenza were not analyzed further because of insufficient numbers. The main comparative assessment and conclusions were made by comparing hospitalized influenza-positive pregnant women with hospitalized influenza-negative pregnant women.

During the four influenza seasons included in this study, influenza was detected during similar periods, although the season varied from as short as 13 weeks in 2014/2015 to as long as 24 weeks in 2015/2016, and the peak occurred as early as week 3 in 2015/2016 and as late as week 11 in 2013/2014 (Fig. 1). Influenza A(H1N1)pdm09 was the dominant strain in 2012/2013, A(H3N2) in 2013/2014, B/Yamagata lineage and A(H3N2) in 2014/2015, and A(H1N1)pdm09 in 2015/2016 (Table 1 and Fig. 1).

Influenza-positive pregnant admissions were slightly older than influenza-negative pregnant admissions (median = 28.5 vs. 28.0 years; OR = 1.02 [95% CI, 1.00–1.04]; p = 0.021) (Table 2). The risk of a positive influenza result increased with each year by 2% (95% CI, 0–4%). When analyzed by subtype/lineage, the probability of influenza infection increased with each additional year by 3% (95% CI, 1–6%; p = 0.012) for A(H1N1)pdm09 and by 9% (4–13%; p <  0.001) for B/Yamagata lineage but was unaffected by age for A(H3N2) and B/Victoria lineage (Table 4). Median ages were higher for pregnant admissions positive for B/Yamagata-lineage viruses (30 years) than for those positive for A(H1N1)pdm09 (29 years; p = 0.01), A(H3N2) (28 years; p = 0.006), or B/Victoria lineage (28 years; p = 0.006) (Table 3). Minor but significant differences in age were found for pregnant admissions by influenza subtype/lineage (p = 0.028).

Pregnant admissions who were positive for influenza more frequently had underlying conditions than those who were negative for influenza (OR = 1.52 [95% CI, 1.16–1.99], p = 0.003; aOR = 1.56 [95% CI, 1.16–2.08]) (Tables 2 and 4). This was confirmed for influenza A(H3N2) and A(H1N1) but not for the two B lineages (Table 4). However, significant differences between influenza-positive and influenza-negative pregnant admissions were not detected for the individual underlying conditions (Tables 2 and 3).

Gestational age distribution was similar in influenza-positive and influenza-negative pregnant admissions (p = 0.872) (Table 2). Pregnant admissions positive for influenza were homogeneously distributed by trimester (p = 0.37 for homogeneity in the distribution of estimates and p = 0.49 for trend, data not shown). The OR of admission with any influenza did not differ between the first and second trimesters (1.19 [95% CI, 0.92–1.53]; p = 0.16) or between the first and third trimesters (1.12 [95% CI, 0.86–1.46]; p = 0.39). Most of the pregnant admissions infected with influenza A(H1N1)pdm09 or B/Victoria lineage were in the first or second trimester, whereas most of those infected with influenza A(H3N2) or B/Yamagata lineage were in the second or third trimester, resulting in a significant difference in strain distribution by trimester (p = 0.005) (Table 3); however, aORs for each strain did not differ between trimesters (Table 4). Likewise, distributions of gestational age at admission differed significantly by virus subtype/lineage, and median gestational ages were lower for admissions positive for A(H1N1)pdm09 (21 weeks) than for admissions positive for A(H3N2) (25 weeks; p = 0.032 [data not shown]) or B/Yamagata lineage (24 weeks; p = 0.027 [data not shown]) (Table 3). Conditional plots did not reveal interactions between trimester and patient age or presence of underlying conditions for the risk of admission with any influenza or with each subtype/lineage (Additional file 1).

Rates of influenza vaccination and antiviral use before admission were ≤ 1% and did not differ between influenza-positive and -negative pregnant admissions (Table 2). Smoking habits and socioeconomic class also did not differ between pregnant admissions positive or negative for influenza.

The probability of laboratory-confirmed influenza decreased with the time between symptom onset and swabbing (Table 2). The probability was lowest in pregnant admissions with samples taken 5–7 days after the onset of symptoms (OR = 0.36 [95% CI, 0.24–0.53]; p = 0.001). The adjusted probability of a positive result decreased 13% (95% CI, 7–18%) for each day between onset of symptoms and swabbing (aOR = 0.87 [95% CI, 0.82–0.93]). Time to swabbing differed significantly between subtypes/lineages, although the median was 2 days in all cases (Table 3).

Fever (p <  0.001), cough (p <  0.001), and myalgia (p <  0.001) were reported as presenting complaints more often in influenza-positive than influenza-negative pregnant admissions (Table 5). Overall, fever, reported by 97.1%, was the most common presenting complaint in pregnant admissions positive for influenza. The aOR for influenza-positive vs. influenza-negative pregnant admissions was 6.34 (95% CI, 4.01–10.03) for fever and 2.76 (95% CI, 2.13–3.43) for cough (Table 6). Cough was a common presenting complaint in pregnant admissions infected with B/Victoria lineage (86.8%) and A(H1N1)pdm09 (82.0%) but less common for those infected with A(H3N2) (69.2%) or B/Yamagata lineage (72.6%) (p <  0.001) (Table 7). Proportions of pregnant admissions reporting all other symptoms (headache, malaise, myalgia, sore throat, and dyspnea) also differed significantly by subtype/lineage. For example, dyspnea was a major presenting complaint in 18.0% of admissions with A(H1N1)pdm09 but in less than 5% of admissions with A(H3N2) or B/Yamagata-lineage and in 6.6of admissions with B/Victoria-lineage.

Hospital admission occurred sooner after the onset of symptoms in influenza-positive than influenza-negative pregnant admissions (p <  0.001) (Table 5). This was particularly the case for pregnant admissions positive for influenza A(H1N1)pdm09 and influenza A(H3N2), where more than half went to the hospital on the first day (Table 7). In contrast, more than half of pregnant admissions positive for influenza B went to the hospital by the second day. As a result, the time to symptom onset differed significantly by strain (p <  0.001).

The overall length of hospital stay was not significantly longer for pregnant admissions positive for influenza than those negative for influenza (Table 5); however, pregnant women hospitalized for > 4 days had a higher probability of laboratory-confirmed influenza than those hospitalized for ≤4 days (OR = 1.62 [95%CI, 1.21–2.16]; p = 0.001; data not shown). The median hospital stay also differed by subtype/lineage (Table 7) and was longer for pregnant women infected with B/Yamagata lineage (6.4 days) than for those infected with influenza B/Victoria lineage (4.9 days; p = 0.025 [data not shown]) or A(H1N1)pdm09 (5.4 days; p = 0.074 [data not shown]) but did not differ for those infected with A(H3N2) (5.9 days; p > 0.05 [data not shown]).

At discharge, the most common diagnosis was other respiratory infections (n = 923; 52.8%) followed by influenza (n = 725; 41.5%) (Table 5). Pneumonia was the main discharge diagnosis in 22 cases (0.1%) and did not differ between influenza-positive and influenza-negative pregnant women.

Only one influenza-positive pregnant admission and only two influenza-negative pregnant admissions were hospitalized in an intensive care unit (Table 5). No deaths were reported.

Pregnancy outcomes recorded during the hospital stay included 177 live births, 53 abortions (pregnancy stopped before 20 weeks), and 23 stillbirths (Table 5). No perinatal deaths were reported (data not shown).

Abortion was more frequent in influenza-negative than influenza-positive pregnant admissions (4.2% vs. 1.7%; p = 0.003), although the frequency did not significantly differ by subtype/lineage (Table 7). Frequencies of other pregnancy outcomes did not differ between strains, but predicted probabilities were higher for stillbirth in women infected with B/Yamagata, for cesarean delivery in women infected with A(H3N2), and for preterm delivery and low birth weight in women infected with B/Victoria (Table 8 and Fig. 2).

Age had a noticeable impact on pregnancy outcomes in influenza-positive admissions. The probabilities of abortion and stillbirth increased with the mother’s age, whereas the probability of live birth decreased with the mother’s age (Fig. 2). Probabilities of preterm delivery, caesarean delivery, and low birth weight were highest in the youngest mothers, although a second smaller increase in probability occurred in women 35–40 years of age.