This is a retrospective study focusing on the perinatal outcomes of twin pregnancies in Finland in the years 1987–2014. The data were collected from the national Medical Birth Register and the Care Register (Finnish Institute for Health and Welfare, Finland), separating data on twin pregnancies. The Medical Birth Register contains, from 1987 onwards, data on all births in Finland, including information on deliveries and newborns until the age of seven days. Data on ART use and induction of labour are available since 1990, and data on neonatal diagnoses from 1996 onwards. The Hospital Discharge Register, also called the Care Register for Health Care, includes data on diagnoses, procedures and interventions received by the patients. Information on hospital outpatient care is available since 1998. Linkages between the registers can be performed by using the mother’s and child’s unique personal identity codes. Using International Statistical Classification of Diseases and Related Health Problems (ICD-9 in 1987–1995, ICD-10 since 1996) and Nordic Medico-Statistical Committee (NOMESCO) Classification on Surgical Procedures (NCSP) since 1997, the following data were retrieved and analyzed: the incidence of twin pregnancies (spontaneous and by ART), preterm deliveries (< 37 weeks of gestation), the onset of labour (spontaneous and induced), gestational age, sex, birthweight, Apgar scores at 1 and 5 min, umbilical artery pH, PNM, early neonatal mortality and neonatal morbidity. Intrauterine growth restriction and small-for-date are reported by the same ICD-code and thus, could not be separated in the registers. The mode of delivery is analyzed in further detail in our study on maternal complications in twin pregnancies [12].
Perinatal mortality rate was defined as death in the perinatal period (from 22 weeks of gestation up to 6 days postpartum) per 1000 children born dead or alive, and early neonatal mortality rate as death up to 6 days postpartum per 1000 children born alive. Stillbirths were defined from 22 weeks of gestation onwards. Neonatal morbidity includes diagnoses up to 28 days postpartum.
Data concerning the following perinatal/neonatal diagnoses were collected using corresponding ICD 10-codes: different levels of asphyxia (P20, P21.0, P21.1, P21.9), neonatal respiratory distress (P22), pneumothorax (P27.1), other respiratory distress (P28), retinopathy (H35.1) and sepsis caused by different pathogens (P36.0, P36.1, P36.2, P36.3, P36.8). Cerebral haemorrhages were collected by corresponding codes: (P52.0, P52.1, P52.2, P52.3, P52.4, P52.5, P52.6, P52.8) and data on necrotising colitis (P77), bowel perforation (P78.0), seizures (P90), hypoxis-ischemic encephalopathy grade I-III (P91.00, P91.01, P91.02), cerebral ischemia (P91.08), cystic periventricular and other leukomalasia (P91.1, P91.2), other and unspecified cerebral disturbance (P91.8, P91.9), were also collected. The proportions of low (< 2500 g) and very low (< 1500 g) birthweight and levels of prematurity were gathered from data on birthweight and gestational age at delivery.
In addition, the incidences of the following parameters were analyzed: neonatal antibiotic treatment, phototherapy, intubation, respiratory support, NICU stay, and children at home by the age of seven (7) days. All data were separated for twin A and B when possible and compared to corresponding singleton data from the same period, when applicable.
The data from the Medical Birth Register are available since 1987, except for ART, induction of labour and neonatal interventions since October 1990, and neonatal diagnoses since 1996. Five-minute Apgar scores were not recorded in the registers in 1990–2003, thus lacking during that period. Chorionicity was not recorded in the registers during the study period and thus could not be analyzed. Although the Finnish registers are considered reliable, the quality of reporting may vary causing possible bias. Twin parturients’ characteristics, and pregnancy and delivery complications have been presented in detail elsewhere [12].
Statistical analyses
The data were analyzed using SPSS (IBM SPSS Statistics for Windows, Version 24.0, Armonk, NY, IBM Corporation) and RStudio (RStudio: Integrated Development Environment for R 1.1.447, Inc., Boston, MA 2016). Binomial regression for aggregate data was used to study predictors and dependencies of different variables. To further test the predictive properties of different models created, chi-squared test was used. A p-value <0.05 was considered statistically significant, and a 95% confidence interval was used. One sample t-test was used to compare the means of the variables and two-sample z-test was used to compare sample proportions. To analyze differencies between groups, crosstabs were also executed. Microsoft Excel 2010 and SPSS 24.0 were used to create figures, graphs and trend lines. The results are expressed in percentages, and the means, ranges and standard deviations are reported when appropriate.