Parents and adolescents preferences for asthma control: a best-worst scaling choice experiment using an orthogonal main effects design

Potential attributes and levels related to asthma control were initially identified through a literature review. Attributes were discussed in two parent focus groups and one adolescent focus group. Parents and teens were not related to each other. Focus group participants presented their views on factors they perceived to be important in asthma control. Focus groups were audiotaped and transcripts were coded for recurrent themes. Factors important for asthma control clustered around symptoms, medication changes, school and work absences, physical activity limitations, emergency services, season and susceptibility to colds. A draft BWS instrument with five asthma control attributes was created for pilot testing [1‐3, 17].

As a full factorial design which would encompass all possible combinations of attribute levels would not be feasible, a fractional design that included a plausible array of choice options was created using Sawtooth™ software (Sequim, WA). A subset of these choice tasks was selected to adhere to BWS design principles to maximize design efficiency. This included displaying i) orthogonality which ensures that differences in the levels of each attribute vared independently over choice sets, and ii) balance, to confirm that all levels appeared with equal frequency in the questionnaire [15]. Using an orthogonal main effects plan design, a questionnaire with 18 choice tasks was created. To avert respondent fatigue, this questionnaire was divided into two blocks or versions, each with nine choice tasks. Each task contained the same five attributes, with levels that varied across tasks. Each task asked respondents to choose the most and least preferred items (Table 1, Additional file 1). The questionnaire took 10 to 20 minutes to complete.

The draft instrument was pilot-tested in 18 English-speaking parents of children with clinically diagnosed asthma and unrelated adolescents with asthma. The pilot test resulted in minor revisions. The final instrument contained 5 attributes with 3 levels each. The attributes and levels were: Night-time symptoms (none; 3 days per week; 5 days per week), Wheezing or tightening of chest (no chest tightening or wheezing; chest tightening or wheezing but it is manageable [does not worsen]; chest tightening or wheezing and is bothersome [may worsen]); Changing medication (no changes needed to medication; more doses or adding on another asthma medication needed; adding oral steroids for 5 days needed), Emergency visits (no Emergency room visits; 4 Emergency room visits per year; 10 Emergency room visits per year) and Participation in physical activities (no physical activity limitations; 2 limitations per month; 10 limitations per month). Numeric values for levels were designated rather than ordinal values (e.g. never, occasionally and often) so that respondents would interpret each level in the same way. The use of extreme values (e.g. Night-time symptoms 5 days per week, Emergency room visits 10 per year) is useful in BWS experiments to enable measurement of negative as well as positive preferences. A reasonable time interval was selected for each attribute (e.g. week, month or year). For example, an interval of one month was selected for activity limitations as children typically participate in organized physical activity on a weekly basis over several months and parents and teens could therefore express their preferences for the indicated levels without difficulty. Similarly, parents have been shown to be reliable reporters of asthma emergency room visits over one year [18] and choosing this interval allowed for the specification of a wide frequency range.

The target populations were parents with a child with clinically diagnosed asthma between 2 and 12 years of age or adolescents with asthma aged 12 to 16 years who had a prescription for an asthma controller in the previous year. Children with significant respiratory, cardiac, neurological and congenital conditions, cancer, musculoskeletal abnormalities and psychological or mental health impairments were excluded. Respondent groups were unrelated and the same family could participate in only one of the two surveys. Focus group and pilot study participants were excluded. Parent and adolescent respondents were recruited from two outpatient asthma clinics, an asthma education clinic and through the Asthma Society of Canada (ASC), a community advocacy organization.

Potential subjects were randomly allocated to one of two BWS questionnaire blocks. Packages containing an information sheet, the BWS instrument, a parent-completed demographics questionnaire, and a parent-completed child health questionnaire were mailed to interested participants. The demographics questionnaire captured child age, child sex, parental education, whether the reporting parent was born in Canada, whether the child had access to a drug plan and household income. These items were selected to permit a characterization of the socioeconomic status of the sample and facilitate comparison with other studies. As colds were identified as an important factor for asthma worsening during the focus groups, questions related to colds were included in the health questionnaire. These questions used a 6-month recall interval shown to be reliable in previous studies [18]. For adolescents, their parents completed the health and demographics questionnaires only. The questionnaires could be completed by mail, phone or online.

Data from parent and adolescent samples were analysed separately. Demographic and health data were analysed with descriptive statistics using SAS version 9.2 (Carey, NC). Conditional logit regression with choice as the dependent variable was used to analyse the BWS data and generate regression coefficients for attribute levels using Latent Gold Choice, version 4.5 (Belmont, MA). A main effects model in which the regression coefficients represent the strength of preference for the specified level for each group was constructed. This analysis allows a comparison of preferences for all levels of all attributes relative to a single reference level. Ten physical activity limitations per month was designated as the reference level because its utility fell between more preferred attribute levels (e.g. no night-time symptoms) and less preferred levels (e.g. ten emergency visits). A standard error, p value and 95 % confidence intervals (CI) were calculated for each coefficient based on all choices within each group, without adjustment for clustering of preferences at the respondent level. The main effects analysis did not assess statistical interactions between attribute levels and sociodemographic variables.

The study was approved by the Research Ethics Boards of The Hospital for Sick Children (Sick Kids), the University of British Columbia-Providence Health Care Research Institute, and the William Osler Health System, Canada (adolescent study only). Completion of the questionnaires signified informed consent. Participants were not asked to consent to making their data public.

Parents in both groups were comparable on most characteristics (Table 2). A higher proportion of parent group respondents received a university education compared to parents of adolescents. There was a trend toward greater respiratory health services use in the parent group. Ninety percent of the children of the parent sample were using a controller with or without other medications compared to 85 % in the adolescent sample. Twenty-six percent of the children of parent respondents used combination long-acting beta agonist plus inhaled corticosteroid medications compared to 59 % of adolescents, however the former group represents younger children for whom combination inhalers may not be indicated. The sample displayed a symptom frequency, rate of exacerbations and use of urgent care indicative of moderate to severe asthma for the children of parent participants as well as the adolescents.

Parents’ reports of physical activity limitations and the effects of colds on their child’s asthma were similar for young children and for adolescents (Table 3). Eighty percent of parents of young children and 76 % of parents of adolescents with asthma reported their children being equally or more physically active than other children of the same age, but just over half reported that their children catch colds, flu or respiratory infections more easily. A majority of both groups reported worsening of asthma due to colds.

The regression analysis of parents’ preferences are presented in Table 4 and plotted in Fig. 1. The results indicate that parents had the strongest positive preference for the absence of night-time symptoms (β = 2.09, p < 0.00001), which was greater than their preference for the absence of wheezing or chest tightening, no changes to medications, no emergency room visits and no physical activity limitations. Parents’ strongest negative preference (least favourable) was for 10 emergency room visits per year (β = −2.15, p < 0.00001) which was much less preferred than frequent night-time symptoms, bothersome wheezing or chest tightening, the need to add oral steroid medications or ten physical activity limitations per month. The spread between the most and least preferred levels for each attribute is an indication of the relative importance of that attribute to respondents, given the level options specified in the BWS questionnaire. The wide magnitude for emergency room visits suggests that parents strongly distinguished between zero and ten emergency room visits per year. Their preferences between the highest and lowest levels of night-time symptoms, wheezing or chest tightening and physical activity limitations were similar. Parents expressed the least difference in their preferences with regard to the need to make changes to their child’s asthma medications. While parents expressed a negative preference for the addition of an oral corticosteroid (β = −0.20), they were only slightly more averse to this than to night-time symptoms three days per week (β = −0.18) or two physical activity limitations per month (β = −0.14) None of these three coefficients was statistically significant.

Adolescents’ preferences for asthma control attributes are presented in Table 4 and Fig. 2. While the strength of their preference for averting night-time symptoms was similar to parents, adolescents demonstrated the strongest positive preference for the absence of physical activity limitations (β = 2.17, p < 0.00001), which was greater than their preference for the absence of night-time symptoms, the absence of wheezing or chest tightening, no changes to medications, and no emergency room visits. Adolescents’ positive preference for averting emergency room visits (β = 1.90, p < 0.00001) was similar to parents (β = 1.82, p < 0.00001). Adolescents’ strongest negative preference was for ten physical activity limitations per month (β = −1.97) closely followed by ten emergency room visits per year (β = −1.83, p < 0.00001) which was less preferred than frequent night-time symptoms, bothersome wheezing or chest tightening or the need to add oral steroid medications. The wide spread between upper and lower levels for physical activity limitations and for emergency room visits suggests that these two attributes were most important to adolescents, given the options available. Their preferences for levels of night-time symptoms and wheezing or chest tightening were similar. Adolescents had a stronger preference for no changes to medications compared to parents but appeared more indifferent than parents with regard to adding more doses or another medication (β = −0.03) and adding an oral corticosteroid (β = 0.02).