Data source
Data used in our study on the occurrence of omphalocele between January 1996 and September 2010 were retrieved from the national birth defects surveillance database maintained by the hospital-based Chinese Birth Defects Monitoring Network (CBDMN). Although the database we used is not freely available, we have obtained the permission from the the Division of Maternal and Child Health Services, National Health and Family Planning Commission of China, to use it. The CBDMN included approximately 460 hospitals during the period 1996–2005 throughout 116 cities or counties in 31 provinces, municipalities, or autonomous regions of China. Approximately 300 new member hospitals in 220 cities and counties were added to the CBDMN in 2006, but we did not use data from these hospitals in this study because of concerns regarding consistency of the data source and the relatively low data quality. Surveillance subjects in the CBDMN consisted of all live births, stillbirths and terminations of pregnancy for fetal anomalies at 28 weeks’ gestation or more. If the gestational age at birth was unknown, infants with birth weight greater than 1000 g were also included in the monitored subjects [
15,
16]. The maximal diagnosis time for birth defects was within 7 days after birth. Cases with omphalocele that were born or induced in member hospitals were required to be registered in the CBDMN.
Ethical approval for our study was provided by the Ethics Committee of West China Second University Hospital, Sichuan University (The Granted number: 2010015).
Data collection
A three-level (county, provincial, and central) surveillance network and clinical expert groups were established to undertake the data collection [
16]. In member hospitals of the CBDMN, each neonate (or terminated fetus) was required to be examined immediately after birth by trained health care professionals, to screen for congenital anomalies. Each case of an abnormality required confirmation by experts in the departments of pediatrics or obstetrics or ultrasound experts at member hospitals. Cases in which abnormalities had been confirmed by a prenatal diagnosis were reconfirmed by experts after birth. When the diagnosis of a case was unclear, the staff (usually the nurse) responsible for birth defect monitoring at the hospital collected more details (e.g., medical records, photos of the case) to be used for rediagnosis by the higher-level expert group. For each birth defect case, the staff was responsible for gathering information (e.g., family socioeconomic and demographic information, clinical features, and obstetric items) through interviews with the mothers or medical record reviews. Additionally, the number of maternal age-specific, residential-specific (urban and rural), and sex-specific births were also collected monthly [
17]. The data were regularly entered into the online reporting system for maternal and child health (MCH) surveillance (
http://zhibao3.mchscn.org) by specialized staff at the county-level MCH hospital.
Statistical analysis
Cases of omphalocele in this study were diagnosed in accordance with the International Classification of Diseases, Tenth Revision (Q79.2). All isolated, multiple cases of omphalocele were included in our analysis. The prevalence proportion was used to describe the occurrence of omphalocele. This value was expressed as the number of omphalocele cases in newborns at 28 weeks’ gestation or more per 10,000 births (including live births and stillbirths). Three associated factors (residential area, geographic region, and maternal age) were included in further analyses. Residential area was categorized as urban (cities, urbanized areas or neighborhood communities) or rural (villages or countryside), according to the last place the mother had resided for at least the previous 12 months. Region refers to the mother's residence location. In China, areas north of the 35th parallel north were classified as the northern region, and areas to the south comprised the southern region. Maternal age was divided into five age groups: <20 yrs, 20–24 yrs, 25–29 yrs, 30–34 yrs, and ≥35 yrs.
Negative binomial cyclical regression models were used to analyze long-term trends and seasonal fluctuations in the occurrence of omphalocele between January 1996 and September 2010. The negative binomial model was selected because omphalocele is a relatively rare event; 72.4% of the region-, residential-, and age-specific number of cases was 0 in a given month. The basic form of the models is expressed as M1:
$$ \ln \left({\mathrm{d}}_{\mathrm{j}}\right)= \ln \left({\mathrm{N}}_{\mathrm{j}}\right)+{\upalpha}_0+{\upalpha}_1T+{\displaystyle \sum_0^k}\psi \cos \left(2k\pi \omega {t}_j-{\uptheta}_k\right)+{\upbeta}_1{\mathrm{X}}_1+{\upbeta}_2{\mathrm{X}}_2+\dots +{\upbeta}_{\mathrm{p}}{\mathrm{X}}_{\mathrm{p}} $$
(M1)
where j is the time period (j = 1,2,…,j), d
j is the number of omphalocele cases in the period j, N
j is the number of births in period j, and α
0 is the logarithm of the baseline hazard function.
T (year) is the long-term trend of omphalocele occurrence. Seasonal fluctuation is expressed as
\( {\displaystyle {\sum}_0^{\mathrm{k}}}\uppsi \cos \left(2{\mathrm{k}\uppi \upomega \mathrm{t}}_{\mathrm{j}}-{\uptheta}_{\mathrm{k}}\right) \);
ψ is the amplitude of periodic fluctuation;
k is the order of seasonal fluctuation and
k. usually set as 0, 1, and 2 [
18]. If
k is zero, this implies no seasonal fluctuation in omphalocele. Therefore,
k is the number of peaks in occurrence of omphalocele in 1 year, and θ
k
is the position of the peaks.
ω is the length of cycle. In our analysis, we set 12 months in 1 year as equal to a cycle, so ω = 1/12.
t
j
is the seasonal variable: month. Χ
p (p = 1,2,…) represents the risk factors. In our analysis, geographic region, residential area, and maternal age were added to the models as risk factors. Thus, the ratio of omphalocele birth prevalence between the southern and northern regions, adjusted for residential area and maternal age, can be calculated by e
β.
To facilitate parameter estimation of the model, we transformed
ψ cos(2
πωt
j
− θ) into a linear form [
19]. We set
ψ cos(θ) = γ
1,
ψ sin(θ) = γ
2; therefore,
ψ cos(2
πωt
j
− θ) = γ
1 cos(2
πωt
j
) + γ
2 sin(2
πωt
j
). The parameters
ψ and θ can be estimated by formula F1:
$$ \left\{\begin{array}{l}\psi =\sqrt{\upgamma_1^2+{\upgamma}_1^2}\\ {}\theta =\left\{\begin{array}{l}{\mathrm{tg}}^{\hbox{-} 1}\;\left({\upgamma}_2/{\upgamma}_1\right)\kern4em {\upgamma}_1>0,\kern1em {\upgamma}_2>0\\ {}\uppi + {\mathrm{tg}}^{\hbox{-} 1}\;\left({\upgamma}_2/{\upgamma}_1\right)\kern3em {\upgamma}_1<0\\ {}2\uppi +{\mathrm{tg}}^{\hbox{-} 1}\left({\upgamma}_2/{\upgamma}_1\right)\kern2.5em {\upgamma}_1>0,\kern1em {\upgamma}_2<0\end{array}\right.\end{array}\right. $$
(F1)
We set cos(2
kπωt
j
)=c
k
, sin(2
kπωt
j
)=s
k
. Therefore, model M1 is equal to model M2, which was used for the final analysis in our study.
$$ \ln \left({\mathrm{d}}_{\mathrm{j}}\right)= \ln \left({\mathrm{N}}_{\mathrm{j}}\right)+{\upalpha}_0+{\upalpha}_1T+{\displaystyle \sum_0^k}\left({\upgamma}_{\mathrm{k}1}{c}_k+{\upgamma}_{\mathrm{k}2}{\mathrm{s}}_{\mathrm{k}}\right)+{\upbeta}_1{\mathrm{X}}_1+{\upbeta}_2{\mathrm{X}}_2+\dots +{\upbeta}_{\mathrm{p}}{\mathrm{X}}_{\mathrm{p}} $$
(M2)
We used three models (
k set to 0, 1, and 2, respectively) to estimate the long-term trends and seasonal fluctuations of omphalocele nationwide, in the northern region, southern region, urban areas, and rural areas. The likelihood ratio test statistic G
2 was used to explore the significant seasonal fluctuations.
All statistical analyses in this study were performed using SAS 9.3 software (SAS Institute Inc., Cary, NC, USA). The statistical significance level for α was set at 0.05.