We performed a retrospective population-based cohort study using administrative datasets linked using unique encoded identifiers and analyzed at the Institute for Clinical Evaluative Sciences (ICES). We considered singleton term infants born in an Ontario hospital between April 1, 2002 and March 31, 2013, whose mother was an Ontario resident at the time of the birth. Maternal-newborn pairs were identified from the ICES MOMBABY dataset, which links the inpatient records of delivering mothers and their newborns in the Canadian Institute for Health Information (CIHI) Discharge Abstract Database (DAD), housed at ICES. Our study cohort was further limited to infants who had at least 1 postnatal visit with a family physician within 190 days of age in the Electronic Medical Record Administrative data Linked Database (EMRALD™), also housed at ICES. The EMRALD™ dataset was created using data from participating family practice EMRs across Ontario [14]. Specific to our study, EMRALD™ contains data for well-baby visits to family physicians, including the date of each visit, the infant’s corresponding age, anthropometric measures, feeding practices, developmental milestones and physical exams. Multi-fetal pregnancies were excluded as they are more prone to indicate preterm birth, small-for-gestational age, and specialized pediatric postnatal care.
To obtain information about feeding practices, we formulated an algorithm combining free text searches and structured field searches from the Rourke Baby Record, which is described at http://www.rourkebabyrecord.ca/pdf/RBR%202017%20Ontario%20English%20-%20Black%20171004.pdf and http://rourkebabyrecord.ca/default.asp. The Rourke Baby Record is a standardized and commonly used method for family physicians in Canada to record well baby visits in newborn and infant medical records. The content of the search algorithm is listed in the Additional file 1. The electronic search algorithm was used to abstract information about mode of infant feeding from all EMRs with a Rourke Baby Record. To find Rourke Baby Records the search algorithm looked in the free text progress notes made by the family physician for mention of a Rourke form. The name of the form can vary, so as many variations were accounted for, such as “Rourke”, “Well Baby Visit”, “Well Baby Check Up”, “Newborn Visit”, “1 month visit”, etc. If a Rourke Baby Record was found, the algorithm to search for documentation of feeding was then applied. Those EMRs without a Rourke Baby Record were manually abstracted for similar details about infant feeding. We included records up to 750 days of age in order to capture historical information on type and duration of feeding recorded in non-Rourke entries, as well as Rourke Baby Records with exclusive breastfeeding documented beyond 6 months, from which we inferred exclusive breastfeeding at earlier time points.
Three trained abstractors performed the manual EMR abstraction. The initial charts were also reviewed by a content expert (ALP) to correct for any inconsistencies. Both intra and inter-rater reliability were assessed for 5 % of the charts.
Typically, well baby visits to primary care providers occur shortly after birth and at 2 months, 4 months and 6 months, according to the Ontario infant immunization schedule [15]. As such, we chose these time points to assess the rates of exclusive breastfeeding. The denominators for the rates at 2, 4 and 6 months included infants with any visit at ≥ 60 days, ≥ 122 days, and ≥ 182 days of age, respectively. We chose the visit closest to but not preceding the target age for each time point. For children who did not have visits at all three time points, we estimated feeding status based on that documented at future visits. For example, if a child had a visit at 60 days then their 2 month feeding status was determined from that visit. If, however, their next visit was not until 182 days or more, then we determined if they were currently exclusively breastfeeding (from a Rourke record) or were previously exclusively breastfeeding (from a progress note) and we inferred that they were exclusively breastfeeding at 4 and 6 months.
To assess possible misclassification of exclusive breastfeeding as a result of inferring feeding status from future visits, we performed a complete case analysis of infants who had documented feeding at all three time points. Exclusive breastfeeding duration was calculated based on the date of birth and the date of the latest visit with confirmed exclusive breastfeeding.
Newborns to immigrant mothers were determined by linkage to the federal Immigration, Refugees and Citizenship Canada Permanent Resident Database held at ICES, which has records for permanent residents who immigrated to Ontario from 1985 to 2012. Neighborhood income quintile (Q) was determined by residential postal code at the time of birth, derived from Statistics Canada census data. Rurality was determined by the Registered Persons Database (RPDB) and maternal age at birth, parity and birthweight were determined by linkage with the MOMBABY database.
Rate ratios (RR) and 95% confidence intervals (CI) were calculated for each characteristic, comparing each group to that with the lowest rate (the reference group).
Statistical analysis was performed using SAS for UNIX, Version 9.4 (SAS Institute, Cary, NC), and EMR data analysis was performed in SQL Server 2012. The study was approved by the Sunnybrook Health Sciences Centre Research Ethics Board.