Study design
We conducted a retrospective population-based study, including all suspected cases of CZS registered in RESP and as a comparison group the population of live births without a record of CZS, born in Brazil from January 1, 2015, to December 31, 2018. Both groups were followed up from birth until death, 36 months, or December 31, 2018, whichever came first.
Data sources
Cases were notified to RESP when considered to have microcephaly and/or central nervous system (CNS) alterations confirmed or suspected to be associated with congenital ZIKV infection since 2015. We obtained data on all notifications to RESP and extracted data on final CZS classification (confirmed/probable, inconclusive or unconfirmed), head circumference, imaging findings, and maternal reporting of a rash during pregnancy [4]. Suspected cases are those notified to RESP. They were notified because they meet one or more of the criteria defined for notification at one point during the period [8]:
-
1)
microcephaly was defined as HC of 33 cm or less for term boys and girls, which was reduced to 32 cm on December 12, 2015, and reduced again (March 2016 following the World Health Organization (WHO) recommendation) to 31.9 for term boys and 31.5 for girls or more than two standard deviations below the mean for age and sex (according to INTERGROWTH 21st standards for those born at fewer than 37 gestation weeks, or WHO standards for those born with 37 gestation weeks or more);
-
2)
central nervous system changes suggestive of congenital infection detected from neuroimaging tests (accepted imaging were cranial computed tomography, brain magnetic resonance, or transfontanellar ultrasound); two or more neurological, visual or auditory manifestations;
-
3)
newborns or fetuses from mothers who reported a fever and/or skin rash during pregnancy, likely or confirmed to be due to ZIKV infection.
After notification, all suspected cases were investigated by local epidemiological surveillance teams and classified as confirmed, probable, inconclusive or unconfirmed. Suspected cases were considered [8]:
-
a.
confirmed/probable when they had signs and symptoms consistent with CZS regardless of laboratory confirmation or maternal symptoms.
-
b.
unconfirmed if they had compatible clinical symptoms that, after clinical and epidemiological investigation, were attributed to having another cause; for example, microcephaly related to restricted intrauterine growth or genetic diseases.
-
c.
inconclusive if there was insufficient information for proper classification.
The Live Births Information System (SINASC) registers data from the Declaration of Live Births. It is legally required document completed by the health worker who assists the delivery [11]. It covers close to 100% of all births in Brazil. We extracted information from SINASC [12] on the date and place of birth, gestational age at birth, birth weight, and newborn sex.
From the Mortality Information System (SIM), which records deaths and provides death certificates (a legally required document that must be completed by the physician who attests the death), we obtained information on the date of death. As of 2011, SIM was estimated to cover more than 96% of all deaths in Brazil [12].
Comparison group: We included a population of live births born in the same period as cases without a linked record of CZS to provide a baseline for the mortality rate comparisons.
All the data were provided by the Brazilian Ministry of Health to the Center of Data and Knowledge Integration for Health (CIDACS) in 2020.
Linkage process
Data from the 3 sources, SINASC (births), SIM (deaths), and RESP (suspected cases of CZS) were linked. The name of the mother, maternal date of birth or age (when the date of birth of the mother was missing), and place of residence were used as matching variables. Linkage was done at CIDACS in a strict data protection environment and according to ethical and legal rules using CIDACS-RL software. CIDACS-RL is a novel record-linkage tool that applies a combination of indexing and searching algorithms developed in-house to link large-scale administrative datasets [13, 14]. The linkage is processed two by two (SINASC-RESP and SINASC-SIM), then the information from the three datasets are organised in one single database.
Procedures and definitions
We included all suspected cases of CZS with a complete epidemiological investigation and excluded cases that did not link with a birth registry from the SINASC. Live births who died during the study period were identified by linking SINASC with SIM.
We then classified the suspected cases according to HC and birth weight using the Z scores estimated according to INTERGROWTH 21st standards by sex and gestational age. The newborns were classified as having a normal head circumference (NHC) when between + 2 and -2 standard deviations (SD) from the mean; microcephaly when below two and above or equal three SD of the mean; severe microcephaly when more than 3 SD below the mean; and macrocephaly when more than 2 SD above the mean. According to birth weight, we classified as appropriate for gestational age (AGA) between + 2 and -2 Z SD of the mean; small for gestational age (SGA) more than 2 SD below the mean and large for gestational age (LGA) defined as more than 2 SD above the mean [15]. We also categorised cases according to imaging findings (not abnormal, abnormal compatible with congenital infections, abnormal other) and according to the presence of a maternal rash during pregnancy.
Statistical analyses
Anthropometry at birth, imaging findings, maternal rash, region and year of birth were categorised in the final epidemiological classification using numbers and percentages for categorical data and means and SD for continuous variables. Differences in characteristics were summarised using the chi-squared test and analysis of variance used to compare means considering a 5% significance level (p-value < 0.05).
Latent class analysis was used to cluster unconfirmed cases into classes [16,17,18]. The resultant classes represented probabilistic groups of patients with similar combinations of characteristics, based on anthropometry at birth (HC and birth weight for gestational age and sex), imaging findings, maternal reported rash, region, and year of birth [16]. Characteristics were described according to each cluster using numbers and percentages and chi-squared tests. Kaplan–Meier curves were plotted, and Cox proportional hazards models were fitted to determine the association of clusters of unconfirmed cases, confirmed cases and the comparison group with mortality up to 36 months. In addition, the association of the individual characteristics was used to fit the model. Mortality in each cluster was assessed using the same Cox proportional hazards models. Data analyses were performed in Stata version 15.0.
This study is reported as per the Reporting of studies Conducted using Observational Routinely-collected Data (RECORD) guideline (Supplementary material 1).