We used data from the 2017–2018 cycle of the Canadian Community Health Survey (CCHS). The CCHS is a cross-sectional survey comprising a representative sample of Canadians 12 years of age and older, excluding those living in institutions, on Reserves, and in certain regions of Québec, as well as those serving in the Armed Forces (< 3% of the population) [23]. The CCHS used a complex multi-stage sampling design, with the sample ≥ 18 years accrued from the Labour Force Survey sampling frame and the 12 to 17-year-old sample accrued from the Canadian Child Tax Benefits file. Questionnaires were administered using computer-assisted interviews and were 35 to 45 min long. We included all 15 to 55-year-old women who gave birth in the previous five years and had complete data on the CCHS questions of interest. Analyses were conducted at the University of Toronto Research Data Centre; research ethics board approval was not required due to the use of deidentified secondary data.

Disability was ascertained using the Washington Group Short Set on Functioning (WG-SS), which uses the WHO’s International Classification of Functioning, Disability, and Health as a framework to collect data on difficulties a person has performing basic universal actions [24]. Questions assess difficulties in six domains—i.e., seeing, hearing, walking or climbing stairs, remembering or concentrating, self-care, and communicating—with response options of “no difficulty”, “some difficulty”, “a lot of difficulty”, and “cannot do at all”. Women were classified as having a disability if they reported “some difficulty”, “a lot of difficulty”, or “cannot do at all” in ≥ 1 domains. Women without a disability were the referent. We further examined disability by severity (i.e., mild or moderate/severe) and the number of domains impacted (i.e., 1 or ≥ 2).

Our main outcomes were: (1) breastfeeding non-initiation, defined by a negative response to the question, “Was [your last child] breastfed or given breastmilk even for a short period of time?” [12]; (2) early cessation of any breastfeeding prior to 6 months in response to the question, “How long did you breastfeed or give breast milk to [your last child]?” [12]; and (3) early cessation of exclusive breastfeeding prior to 6 months in response to questions about the timing of addition of other liquids (i.e., milk, formula, water, juice, tea, or herbal mixture) or solids (i.e., cereals, mashed up or pureed meat, vegetables, or fruits) to the feeds [12]. We also examined cessation of breastfeeding by 12 and 24 months, as secondary, longer-term outcomes.

As covariates, we measured (1) sociodemographic characteristics and (2) health behaviours and comorbidities. Sociodemographic characteristics were maternal age (i.e., 15–24 and 35–55 vs. 25–34 years), marital status (i.e., single, widowed, separated, or divorced, vs. married or common-law), level of education (i.e., high school diploma or less, vs. some post-secondary education or more), annual household income level (i.e., < $40,000, $40,000 to $79,999, vs. ≥ $80,000 CAD), and immigration status (i.e., born outside of vs. in Canada). Health behaviours and comorbidities were smoking status (i.e., daily or occasionally, vs. not at all), body mass index (BMI; i.e., overweight or obese, vs. normal or underweight), and diabetes mellitus and chronic hypertension diagnosed by a health professional (i.e., present, vs. absent).

We described the proportions of women with and without disabilities with each baseline characteristic and compared them using standardized differences [25].

We then used Modified Poisson regression [26] to estimate the prevalence ratios (PR) and corresponding 95% confidence intervals (CI) of breastfeeding non-initiation, early cessation of any breastfeeding by 6 months, and early cessation of exclusive breastfeeding by 6 months, comparing women with and without disabilities. The main multivariable models adjusted for maternal age, marital status, level of education, annual household income level, and immigrant status. Health behaviours and comorbidities that could explain hypothesized disparities between women with and without disabilities (i.e., smoking status, BMI, diabetes, and chronic hypertension) were added to the models in a second step, as these were considered possible pathway variables (Additional file 1).

We conducted several additional analyses. First, we modelled the secondary outcomes of breastfeeding cessation by 12 and 24 months. Second, we examined disability by severity (i.e., moderate/severe and mild disabilities, vs. no disability) and the number of action domains (i.e., ≥ 2 and 1 domains, vs. no disability). Third, we restricted the sample to women who had their baby ≤ 24 months before data collection, to maximize accuracy of breastfeeding recall [27] and reduce the chance that disability onset followed pregnancy. Finally, using descriptive analyses, we compared women with and without disabilities on the reason for non-initiation (“bottle feeding is easier”, “breastfeeding is unappealing”, “medical condition in the mother”, “other”) and early cessation of exclusive breastfeeding (“not enough breast milk”, “baby was ready for solid foods”, “inconvenience or fatigue due to breastfeeding”, “difficulty with breastfeeding”, “medical condition in the mother”, “medical condition in the baby”, “planned to stop at this time”, “child weaned him or herself”, “returned to work or school”, “other”).

All analyses were weighted using a CCHS-assigned weight representing the individual’s contribution to the total population, wherein weights accounted for the CCHS multi-stage sampling design and were adjusted for population projections of age and sex strata within each province as well as survey non-response. The original size of the sample was maintained by dividing individual weights by the average size of the weight in the sample [23].

SAS version 9.4 (SAS Institute Inc., Cary, NC) was used for all analyses.