This mixed-methods study followed a sequential explanatory design with data integration in the discussion. Mothers aged 18 + years with an infant aged < 19 months old with a diagnosis of AD and who were comfortable reading and speaking English were recruited from social media as well as allergy and dermatology clinics in Winnipeg, Manitoba, Canada between October 2021 and May 2022. Upon providing informed consent, mothers completed a series of self-report questionnaires to collect quantitative data. Those specific to the present study are described below. A subset of mothers, who agreed to be contacted for further research related to the overarching AD project, also completed semi-structured interviews following the quantitative questionnaires. Qualitative interview questions were developed a priori by content experts and refined subsequent to quantitative data analysis and as the series of interviews progressed.

Mothers were asked to complete several questionnaires that asked about their household’s socio-demographics, medical history, and their infant’s AD severity and sleep quality.

The primary outcomes were sleep quality and quantity. Data on sleep was collected from different questionnaires: an ad-hoc questionnaire focused on the infant’s AD history and management, the Perinatal Anxiety Screening Scale (PASS) and the Patient-Oriented Scoring of Atopic Dermatitis (PO-SCORAD) tool. In completing the AD questionnaire, participants were asked several questions pertaining to sleep quality: “Does your baby’s atopic dermatitis keep you awake at night?” “Does your baby’s AD keep him/her awake at night?”, “Does your baby receive any medicine to help him/her sleep?”. Participants had the option to answer these questions “yes” or “no”. If a participant indicated that their infant’s AD did keep them up at night, they were also asked the cause (i.e., itching, need to treat, stress, worry, or other). Questions from the PASS used in this study included “How often, over the past month, have you experienced being upset about repeated memories, dreams, or nightmares?” which assessed maternal sleep quality and “How often, over the past month, have you experienced difficulty sleeping even when you have the chance to sleep?” which assessed maternal sleep quantity. Possible answers were; not at all; sometimes; often; or almost always. These possible answers were recoded into no (“not at all”) or yes (“sometimes; often or almost always”). In addition to items described above, infant sleep quality was also evaluated using the PO-SCORAD tool, where infant sleeplessness was assessed on a scale from 0 to 10, where 0 was no sleeplessness and 10 was the worst imaginable sleeplessness.

Mothers were asked to report socio-demographic data for themselves and their infant. Maternal characteristics included: mothers’ age, seasonality, ethnicity, education, and annual household income.

Mother-reported infant characteristics included age (in months), sex (male or female), and allergic comorbidities (asthma, rhinitis, and/or food allergy; not mutually exclusive). AD treatment was reported based on yes or no questions regarding: if moisturizers were used on the infant’s skin, if medications had ever been used on the infant’s skin (e.g., hydrocortisone, betamethasone, etc.), and if other medications had been used for the infant’s AD, such as antibiotics.

AD severity was defined using the PO-SCORAD tool which assesses different areas of the skin to determine the extent of AD on the whole body. Due to the COVID-19 restrictions, the assessment had to be done by the infants’ parents. In completing the measure, participants rated the intensity of their infant’s AD symptoms (i.e., redness, swelling, oozing/crusting, scratch marks, skin thickening, and dryness) on a 4-point scale (i.e., none = 0; mild = 1; moderate = 2; severe = 3) [8]. The subjective symptoms, itch, and sleeplessness, were evaluated on a scale from 0 to 10, where 0 was no itch/sleeplessness and 10 was the worst imaginable itch/sleeplessness [8]. The minimum and maximum PO-SCORAD scores are 0 and 103. A total PO-SCORAD score of < 25 was considered as mild AD, 25–50 moderate and a PO-SCORAD > 50 was considered severe AD [9]. Due to a low number of moderate and severe cases each, these two groups were combined in this study.

Descriptive statistics were used to describe the demographic characteristics of the studied sample and to describe mothers’ sleep quality and quantity and infant sleep. For all demographic variables, missing data was excluded. Information from PO-SCORAD on infant sleeplessness was used to compare cases of mild AD to cases with moderate/severe AD. Binary logistic regression was used to determine the association between AD severity and maternal sleep quality as well as quantity in one crude and one adjusted model.

Potential confounders were identified through a directed acyclic graph (DAG). The DAG resulted in the minimal sufficient adjustment sets of: AD treatment, infant age, infant allergic comorbidity, maternal ethnicity, maternal education, and seasonality. To not over-adjust for socioeconomic status (SES) [10], only maternal education was included in the regression analysis (Additional file 1). Owing to missing observations for seasonality, infant allergic comorbidities, and AD treatment, the model was not estimable, and we had to limit the regression model to the following potential confounders in the adjusted model: maternal ethnicity, maternal education, and infant age (Additional file 1).

Odds ratios (OR) with 95% confidence intervals (CI), were estimated. A p-value, < 0.05, was determined a priori as statistically significant. Data analysis was performed using Stata Version 17.0 (College Station, Texas).

Subsequent to quantitative data collection, a self-selected subset of mothers with infants with AD completed qualitative semi-structured interviews. All mothers who were asked to participate in the qualitative arm of the study had agreed to be contacted for future research related to the overarching AD project. A total of 20 mothers were contacted, 13 provided consent, and 10 completed an interview. All interviews were conducted over the phone by trained research assistants who followed a semi-structured guide (Table 1). While this guide provided the foundational questions for the interview, the interviewers were afforded some flexibility to modify the questions to the individual’s circumstances and in relation to the overall quantitative findings. Each qualitative interview was audio-recorded and transcribed verbatim. Transcripts were then analyzed by two research assistants, who worked independently, but concurrently, using thematic analysis (ZH, MB). Both analysts were supervised by two authors (MG, JP) who possess extensive knowledge of qualitative methods [11].

Thematic analysis is a multi-stage process used to identify and examine patterns within qualitative data [12]. Analysts began by familiarizing themselves with the transcripts [11, 12]. Then transcripts were re-read to better understand participants’ meaning [11, 12]. During this process, the analysts applied meaningful descriptions to relevant sections of text, known as codes [11, 12]. These codes were then systematically applied across all transcripts, and finally codes were organized into overarching themes to describe patterns found throughout the data (Fig. 1) [11, 12]. For example, in the current study, the codes sleep loss; sleep quality; tired; nocturnal scratching; stress induced sleep loss; and sleep loss impacts on mood, were categorized into the overarching theme: atopic dermatitis severity negatively affects sleep quality and quantity for both mothers and infants. Constructs are considered saturated when no new constructs or explanations are identified in subsequent interviews, and a consensus has been reached between analysts on all overarching themes [11, 12].

Thematic analysis process

This study was approved by the University of Manitoba Health Research Ethics Board HS24294 (H2020:426).