This study was approved by the Ottawa Health Science Network Research Ethics Board, Protocol # 20120229-01H. Individual participant consent was not sought because the study involved the secondary use of population-wide health care administrative data.
Study population and data sources
Our source population included all children born in Ontario who received newborn screening between April 1, 2006 and March 31, 2010. The false positive cohort included children who received a positive newborn screening result for MCADD but were ultimately determined to be unaffected. In Ontario, screening for MCADD uses an algorithm involving C8 acylcarnitine (octanoylcarnitine) as a primary analyte, with C6 (hexanoylcarnitine), C10 (decanoylcarnitine), C10:1 (decenoylcarnitine), C8/C10 ratio, and C8/C2 ratio as secondary markers [19]. Infants who screen positive (potentially affected) are referred to one of five regional Newborn Screening Treatment Centres, which are based at pediatric tertiary care centres in the province. The Treatment Centre works with the infant’s primary health care provider to contact the parents and arrange for diagnostic evaluation, and is also responsible for on-going follow-up and management for affected children. The diagnostic evaluation typically includes plasma acylcarnitine profiling, urine organic acid analysis, and testing for mutations in ACADM [19]. Neonates are classified by the Treatment Centre and Newborn Screening Ontario medical staff review results. Infants are classified as true positive if they have a disease-associated genotype (e.g., homozygous for the c.985A > G mutation), and/or have persistent abnormal plasma acylcarnitines, and/or hexanoylglycine detected on urine organic acids analysis. Infants with normal metabolic profiles on diagnostic testing are defined as having received false positive screening results.
A primary comparison cohort included all children with negative newborn screening results for all disorders during the same time period. A secondary comparison cohort included 10 controls with negative screening results matched to each child with a false positive result for MCADD, based on sex, calendar year of birth, urban-rural status of the child’s residence, and an area-based indicator of socioeconomic status. The purpose of the matched comparison cohort was to address potential residual confounding by factors that may be associated with access to health services. Individuals were excluded from the study if they were ineligible for health care coverage at the time of birth or deceased within 24 h following birth (a bloodspot sample must be collected at > 24 h of age in Ontario to be considered satisfactory).
Newborn screening diagnostic confirmation data were securely linked to the provincial health care patient registry at the Institute for Clinical Evaluative Sciences, and then to administrative databases encompassing health service visits from April 1, 2006 through March 31, 2012. Physician encounters were identified using the OHIP Claims Database, which captures services provided by Ontario physicians who bill OHIP on a fee-for-service basis; and services provided by most other Ontario physicians who work in capitation payment models [22]. Emergency department (ED) visit data were retrieved from the Canadian Institute for Health Information’s National Ambulatory Care Reporting System, covering nearly all ED visits in Ontario [20]. Inpatient hospitalization data were obtained from the Canadian Institute for Health Information’s Discharge Abstract Database, covering all acute inpatient facilities in the province [21].
Potential confounding variables
Additional variables were ascertained from the hospitalization database at the time of birth (sex, birth weight, gestational age, season of birth), and Census/geographic data linked by the child’s postal code (socioeconomic status, urban-rural status). We grouped children into low (< 2500 g) and normal/high (≥ 2500 g) birth weight categories. We dichotomized gestational age to preterm (< 37 weeks) and term/post-term (≥ 37 weeks). Season of birth was categorized as January–April, May–August, or September–December.
A proxy measure of socioeconomic status was defined as the neighborhood-level income quintile, based on average household income data from the 2006 Canadian Census, linked to a child’s residential postal code at birth [23]. Neighborhoods were Census “dissemination areas”, with populations of approximately 400–700 persons; income quintiles were assigned across dissemination areas within larger regions [24]. We merged the two lowest and three highest quintiles to define lower and higher socioeconomic status. Urban-rural status of the child’s residence at birth was defined using the Rurality Index for Ontario, based on population size and density and on travel time to higher levels of hospital care [25]. We defined a rural community using a score of ≥ 40, the criterion used for rural physician eligibility in Ontario [26].
Utilization outcomes
We included each original health service encounter within the study period (physician visits, ED visits, and hospitalizations). If a child had multiple billed procedures within a single physician visit, these were considered as one visit. However, if a child saw multiple physicians on the same day, these were considered separate encounters. Each ED visit was a separate encounter as was each inpatient hospitalization.
Statistical analysis
Study datasets were linked using unique encoded identifiers and analyzed at the Institute for Clinical Evaluative Sciences; cell sizes < 6 were not reported due to privacy policies. Counts and percentages were calculated and chi-square tests used to examine bivariate associations between sociodemographic characteristics and cohort membership.
The number of physician visits, ED visits, and hospitalizations during the study period were summed for each child. The length of follow-up for each individual was the time elapsed between the date of birth and the earliest of 3 possible end points: the date of OHIP eligibility loss (mainly related to emigration from Ontario), the date of death, or the last date of follow-up for the study. Unadjusted visit rates and incidence rate ratios (IRR) were calculated to compare the false positive and screen negative cohorts.
Using the Vuong test as a criterion [27], we chose negative binomial regression modeling to compute IRRs for health services comparing the false positive with the screen negative cohorts while adjusting for confounders. Influential observations were identified [28, 29] and truncated to the 99th percentile for each service type. Models were stratified by age at the time of visit (< 1 year of age and ≥ 1 year of age). As a sensitivity analysis to address potential residual confounding by premature birth, we re-ran final models restricted to children with term births (≥ 37 weeks’ gestation). Finally, as a post-hoc sensitivity analysis, we re-ran the final model for physician visits excluding the first month of life, when some visits were likely related to resolving a positive screen as a false positive. This 1-month period was based on the clinical experience of metabolic physicians in Ontario. Analyses were performed using SAS® version 9.3 (SAS Institute, North Carolina, USA).
