Abstract
Background: Primary care physicians, who manage the care of most children with asthma, often do not optimally assess disease control, prescribe asthma controller medications, or provide family asthma education. We developed a pediatric asthma clinical pathway embedded in an electronic medical record (EMR) for use by primary care practices, and we sought to evaluate its effects on prescription and use of asthma controller medication for children with asthma.
Methods: We conducted a cluster randomized controlled trial, enrolling primary care practices in Alberta that used Wolf or Med Access EMRs, and managed at least 50 children with asthma. The multifaceted intervention included an EMR-based pathway for pediatric asthma, Web-based education modules for physicians, and train-the-trainer sessions for practice staff to provide patient education. The control intervention was standard care. We extracted study data from participating practices’ EMRs, and standard emergency department and hospital administrative data sets. The primary and main secondary outcomes were improvement in the proportion of children prescribed and dispensed controller medications, respectively.
Results: Eleven practices were randomly assigned to each of the intervention and control groups. The intervention did not significantly change the proportion of children prescribed (mean difference 4.3%, 95% confidence interval [CI] −2.0% to 10.5%) or dispensed (mean difference −0.1%, 95% CI −7.1% to 6.9%) controller medications.
Interpretation: Our multifaceted intervention did not improve the proportion of children in primary care who were prescribed or dispensed a controller medication for asthma. These results suggest that such interventions may require active alerts and targeting of walk-in and urgent care clinics to have a meaningful impact on clinical practice.
Trial registration: Clinicaltrials.gov NCT02481037.
As many as 1 in 7 children in Canada has asthma.1–3 These children often experience poor disease control, with emergency department visits, hospital admissions, missed school days, and low quality-of-life ratings.3–11 Evidenced-based guidelines highlight that appropriate prescription of controller medications, guided by standardized assessments, and asthma education to increase child and parental understanding and compliance, can achieve better disease control and reduce emergency department visits and hospital admissions. 12–17 However, evidence suggests that primary care physicians, who manage the care of most children with asthma, often do not optimally assess disease control,16,18,19 prescribe controller medications,12,13 or provide family asthma education.20–22
We developed an evidence-based electronic medical record (EMR) pathway to facilitate optimal primary care asthma management practices, such as standardized assessment of asthma control, provision of asthma education by primary care support staff, and algorithms for guiding use of asthma controller medication.23 We sought to determine whether use of the pathway (called the Primary Care Asthma Pediatric Pathway [PCAPP]) by primary care practices, as compared with standard care, would increase the prescription and use of controller asthma medications for children with asthma.
Methods
We conducted a cluster randomized controlled trial (RCT) in 22 primary care practices over 2 years (preintervention June 1, 2016, to May 30, 2017; postintervention June 1, 2017, to May 30, 2018). The study protocol was previously reported.23 We followed the Consolidated Standards of Reporting Trials Statement, extension to cluster randomized trials.24
Clusters
Because the PCAPP was embedded in the EMRs of primary care practices (cluster), we sought to enrol a diverse range of practices across the province of Alberta. We identified potentially interested practices through the Canadian Primary Care Sentinel Surveillance Network, the Alberta Health Services Respiratory Health Strategic Clinical Network, and the Alberta Medical Association. To be eligible to participate, primary care practices had to use either Wolf or Med Access EMRs (Telus Health) and had to manage the care of a minimum of 50 children with asthma. We established how many children with asthma each practice had in accordance with a method developed by Cave and colleagues,25 with minor modifications to medications and clinical features resulting in a slightly broader case definition. Children had to have been 1 to 17 years of age and, in their EMR, have asthma medications prescribed, or have International Classification of Diseases, Ninth Revision codes for asthma (493.0 to 493.2) in their problem list, visit diagnoses, or billing codes (Appendix 1, available at www.cmaj.ca/lookup/doi/10.1503/cmaj.241694/tab-related-content).
For primary care practices that met our study criteria, the study team (H.S. or A.C.) approached practice leads, toward the end of the preintervention period, to obtain consent, until 22 practices were enrolled. Physicians within each practice were also approached and consented, and are subsequently referred to as “study physicians.” Nonstudy physicians include any other practising physician in Alberta. Because the study intervention was a clinical pathway for use by clinicians within practices, the research ethics boards waived individual patient consent,26 and instead parents whose children received asthma care from participating practices were informed about the study through posters displayed in each practice’s waiting room and offered the opportunity to have their child’s data excluded.
Intervention and control
The core component of the multifaceted intervention was a primary care asthma pediatric pathway.27 The pathway was developed over 3 years by the Asthma Working Group of the Alberta Health Services Respiratory Health Strategic Clinical Network, 25 and built into 2 commonly used EMRs.28 The pathway was designed to assess asthma control, to provide therapeutic recommendations, and provide patient resources (written asthma action plans and prescriptions). The pathway algorithms and example screenshots from the EMR are shown in Appendix 1.
The other parts of the multifaceted intervention included a Web-based professional teaching module and train-the-trainer sessions for practice staff to provide patient education (Appendix 1). The Web-based teaching modules were designed to ensure primary care practitioners understood the underlying rationale and evidence behind the development of the pathway).29,30 The Web-based learning modules were designed to be asynchronous, allowing both physicians and allied health team members to complete the modules at their own pace. The modules focused on evidence-based management of pediatric asthma in primary care and incorporating the pathway into their practice. Train-the-trainer sessions were used to educate allied health team members to enable them to provide asthma education to parents and children with asthma.31,32
The control was standard care. Although this was highly variable, none of the participating practices routinely included formal assessments of asthma control; standardized algorithms for diagnosing asthma, managing exacerbations, or follow-up visits; medication dosing guidelines; or in-practice asthma education.
A Theoretical Domains Framework study was conducted to facilitate a theory-based approach to design our implementation strategy.33 The strategies we used to ensure primary care physicians provided optimal evidenced-based care included incorporation of the pathway into the EMR to facilitate its use and the development of Web-based professional education learning modules for physicians and other staff to improve their confidence in managing asthma in children. The Theoretical Domains Framework study also identified ways to optimize the above implementation strategies.33
Randomization and blinding
The consented 22 primary care practices were stratified according to academic status (academic v. nonacademic) and geographic location (urban v. rural) before randomization. We used postal codes to determine rurality, and university affiliation of physicians to determine academic status. After stratification into 4 blocks, the Microsoft Excel RAND function (2016) was used to randomly allocate practices to control or intervention groups within each block. The treatment allocation for each practice was generated and kept by an individual who was not a member of the study team. When the study team was ready to commence implementation at a given practice, they were informed of the practice allocation. Primary care practice physicians and staff were aware of their allocation status.
Retrospective data extraction, data linkage, and statistical analyses were carried out by personnel who were blind to study objectives and practice allocation. Specifically, a trained abstractor was granted access to each practice EMR and extracted the data elements outlined below for children who met our definition of asthma outlined above. These data were securely transferred to the Health Quality Council of Alberta (HQCA), and an HQCA analyst (J.P.) extracted and linked Alberta health administrative data from their data repository to the study EMR practice data. The linked master data set was deidentified and provided to a biostatistician (Q.M.D.) for analysis. Patient identifiers were not available to any other study team members.
Data sources and linkage
As noted above, study data came from participating practices’ EMRs and health administrative data sets. Practice-based EMR data included individual patient-level data, such as patient demographic characteristics, encounter history, medication prescriptions, comorbid diseases, and provision of asthma action plans. All asthma medication prescriptions were extracted and classified as reliever (e.g., salbutamol) or controller (inhaled corticosteroid, montelukast), or combination (e.g., inhaled long-acting β-agonist, inhaled corticosteroid). Health administrative data sets included the National Ambulatory Care Reporting System (emergency department visits), Discharge Abstract Database (hospital admissions), Alberta physician claims (practitioner visits), and Pharmaceutical Information Network (pharmacy dispensing). Pharmacy dispensing data from the Pharmaceutical Information Network provided patient-level dispensed medications from all providers in the province. Dispensed controller medications were categorized as having been prescribed by study physicians or nonstudy physicians.
Outcomes
We measured outcomes retrospectively for all children who met the definition of asthma and received care during the pre- and postintervention periods at one of the participating practices. The primary outcome for the study was the proportion of children with asthma who were prescribed a controller medication. Secondary outcomes included the proportions of children, among all those managed with asthma, who were dispensed controller medications as a measure of primary adherence, who had 1 or more emergency department visits for asthma, and who had 1 or more hospital admissions for asthma.
Sample size calculation
Using data from Alberta primary care practices that used Med Access as their EMR, we established that the median number of children with asthma per practice was 70, the proportion currently treated with controllers was 30%, and the estimated intra-cluster correlation coefficient (ICC) for controller treatment was 0.025. To estimate the minimal clinically important difference (MCID) for this trial, we also conducted a survey of physicians, which determined that a 15% absolute change in controller prescription was clinically important. Using the approach of Hayes and Bennett,34 an α value of 0.05, and the above data, we determined that 11 clusters per arm would yield at least 90% power.
Statistical analyses
We conducted all statistical analyses in SAS Enterprise Guide, version 8.3 (SAS Institute). We used descriptive statistics to demonstrate demographic and clinical characteristics of children with asthma in the intervention and control arms pre- and postintervention. We used a difference-in-differences approach to compare estimated within-group effects (i.e., absolute percent difference) pre- and postintervention. Although the analysis plan presented in our protocol specified a posttest-only design, we thought it was important to compare changes in practice, and we therefore opted for a difference-in-differences approach. Almost all children were observed twice (1511/1588), highlighting the need to measure changes in treatment within these children. We analyzed outcomes using generalized estimating equation (SAS proc genmod) models with a logistic link for binary outcomes. These models used an exchangeable correlation structure for subjects nested within clusters (primary care practices) to account for the clustered nature of the data and used the effects of time (pre v. post), treatment (intervention), and the treatment by time interaction. We applied an SAS macro that used marginal methods (%MARGINS) to obtain the differences in differences, expressed as absolute percent differences, that correspond to the effects from our logistic generalized estimating equation model. The main results presented are unadjusted. We conducted a sensitivity analysis in which all models were adjusted with preintervention age, sex, academic, and rural stratum. We also conducted a subanalysis to explore the proportion of dispensed controllers prescribed from study physicians versus nonstudy physicians. To assist researchers conducting research in this area, we provide the ICC, which summarizes the similarities among observations within clusters, for our data. We calculated the ICC using the entire 2-year pre–post study period, from a generalized mixed effects model with a binary distribution and logit link.
Ethics approval
The University of Alberta Health Research Ethics Board — Health Panel (Pro000056585) and the University of Calgary Conjoint Health Research Ethics Board (REB15–2221) approved this study.
Results
Eleven family practices, comprising 44 physicians (59.5% female with a median of 18 [interquartile range 11 to 26] years in practice) were randomly assigned to the intervention group, and 11 practices, comprising 32 physicians (60.6% female with a median of 14.5 [interquartile range 8 to 27] years in practice) to the control group (Figure 1). All practices operated within a fee-for-service model. In 7 clusters, there was 1 physician per cluster, but the other clusters had more than 1 physician.
Flow diagram of clusters and patients pre- and postintervention. See Related Content tab for accessible version.
Two practices and 4 physicians withdrew, and no parents withdrew their child’s data. In the intervention group, 706 children contributed data preintervention and 737 children contributed data postintervention, and in the control group, 805 children contributed data preintervention and 851 children contributed data postintervention (Figure 1). A total of 1511 children contributed data to both the pre- and postintervention periods (with 31 and 46 new children during the postintervention period in the intervention and control groups, respectively); thus, all children evaluated preintervention were also evaluated postintervention, with the preintervention period accounting for 95% of all children evaluated (1511/1588).
Children in the control group were more likely to be younger than 7 years and from academic urban or nonacademic rural practices (Table 1). In the intervention group, 42 children received care from a pediatrician, respirologist, or allergist during the preintervention period, and this number was 39 in the control group. During the postintervention period, these numbers were 14 in the intervention group and 13 in the control group.
Characteristics of patients in the intervention and control practices, pre- and postintervention
In the intervention group, asthma controller medications were prescribed to 202 (28.6%) children preintervention and 212 (28.8%) postintervention, and in the control group, they were prescribed to 212 (26.3%) children preintervention and 189 (22.2%) postintervention. There was no significant difference between study arms in the change in proportions of children prescribed controller medications from pre- to postintervention (mean difference 4.3%, 95% confidence interval [CI] −2.0% to 10.5%) (Table 2). Hence, the lower and upper CI ranged between a potentially small decrease and a modest increase, which was smaller than the MCID of 15% from the physician survey. A subgroup analysis that included only those children who were prescribed reliever medications by a practice physician showed similar results (mean difference −5.6%, 95% CI −18.0% to 6.9%). Additionally, the calculated ICC was 0.04, which was higher than the value we used to calculate the sample size. A higher ICC means outcomes within clusters are more similar than anticipated, reducing the effective sample size and power to reject the null hypothesis.
Prescribed and dispensed asthma controller medications in the intervention and control practices, pre- and postintervention
Similarly, no significant difference existed between study arms in the change in proportions of children who were dispensed controller medications by a pharmacist from pre- to postintervention (mean difference −0.1%, 95% CI −7.1% to 6.9%) (Table 2). As nonstudy physicians prescribed approximately 50% of controller medications, we conducted a post-hoc analysis of the proportion of dispensed controller medications that were prescribed by study physicians only. Among children who were dispensed medications, the proportion prescribed by a study physician increased from pre- to postintervention in the intervention group (41.5% v. 51.6%), but not in the control group (46.5% v. 45.6%) (mean difference 10.9%, 95% CI 0.3% to 21.4%) (Table 2).
We found no significant difference between study arms in the change in proportions of children that had 1 or more emergency department visits from pre- to postintervention (mean difference 1.7%, 95% CI −1.9% to 5.2%). Owing to small numbers, this analysis was not conducted for hospital admissions (Table 3).
Emergency department visits and hospital admissions among children from the intervention and control practices, pre- and postintervention
Sensitivity analyses involving multivariable models with sex, age, and the practice stratum confirmed there was no difference in the proportion of children who were prescribed or dispensed a controller medication, or who visited the emergency department from pre- to postintervention. In particular, the effect of the intervention on prescribing controller medications did not differ significantly for children younger than 6 years compared with children aged 6 years and older (marginal mean difference 2.8%, 95% CI −1.5% to 7.0%) (Appendix 1, Table 1). Multivariable analysis for the number of hospital admissions was not conducted owing to small numbers.
Interpretation
We evaluated the impact of a multifaceted intervention, including an EMR-based pathway for pediatric asthma, Web-based education modules, and train-the-trainer sessions for patient education, on the change in proportions of children who were both prescribed and dispensed a controller medication pre- and postintervention. We found that our intervention, compared with standard care, did not improve the proportion of children who were prescribed or dispensed a controller medication for asthma. We also did not observe any change in the proportion of emergency department visits pre- and postintervention. A strength of this study includes the novel data linkages that were established between practices’ EMR data sets and provincial health system records to allow for the examination of important patient and system outcomes.
Our findings contrast with those of several studies suggesting that asthma decision-support systems can successfully improve the number of prescriptions for controller medications,35,36 asthma control,36 and asthma action plan delivery,35,36 as well as those of a systematic review of clinical decision-support systems, which showed that most studies achieved small-to-moderate improvements in care processes. 28 Two of the studies referenced above35,37 show potential factors that may have resulted in such differences. Both of these studies included the use of active alerts (e.g., “The patient’s asthma appears to be OUT OF CONTROL”)32 or reminders to providers to enhance the uptake of decision-support tools. Additionally, Tamblyn and colleagues37 used “smart-analytics,” whereby real-time point-of-care clinical data were used to monitor disease status. It is possible the absence of active alerts and reminders in our intervention may explain, at least in part, our negative results. In fact, in our qualitative evaluation,38 the absence of active alerts and reminders was reported as a barrier to pathway uptake.
Our negative findings may also be explained by some additional factors. First, the uptake of the EMR-based pathway varied across intervention sites, with clinicians citing challenges with implementation, including a small pediatric asthma population in their practices, limited integration into practice flow, and, as a consequence, difficulty in remembering to use the pathway because its use was required so infrequently.38 Second, just over 50% of children were prescribed their controller medication from nonstudy physicians, likely indicating they accessed walk-in clinics or urgent care services. The high proportion of families seeking care outside of their primary care practice suggests that interventions targeting these providers alone may be limited in their ability to increase controller prescriptions. Walk-in clinics and urgent care services may represent important locations for interventions to be implemented. In contrast to our overall finding of no effect from our intervention, the 10% increase (albeit less than the MCID of 15%) in dispensed controller medication prescribed by a study primary care physician from the preintervention to the postintervention period in the intervention arm, with no effect in the control arm, is notable. It is possible that the increase in dispensed controller medications prescribed by study physicians within the intervention practices is a result of family asthma education causing more families to fill their prescriptions.
Limitations
Abstracting data from clinics’ EMRs meant we relied on clinic documentation practices, which may vary significantly between practices and physicians. For example, the limitations of the data available through the EMRs did not allow us to reliably classify individual children into asthma subtypes, and thus we could not make graded judgments about whether a given child needed to be treated with a controller medication.
Most practices had more than 1 physician per practice. Our analysis was unable to account for the nesting of multiple physicians within clusters.
Our criteria for a minimum of 50 children with asthma may have included a recruitment bias toward physicians with a greater interest in managing childhood asthma, and thus our findings may not be representative of all practices.
Conclusion
Although our intervention did not yield a positive change in the proportion of children who were prescribed or dispensed a controller medication, a comparison with other published literature shows several potential modifications that could result in improved outcomes. These include the use of active alerts and reminders, and, given the interconnectedness of the province’s health care system, targeting walk-in and urgent care services in addition to primary care practices.
Acknowledgement
The other study authors were saddened by the passing of Dr. Andrew Cave, just months before publication of this article in CMAJ, and want to highlight the critical role he played in bringing this project to fruition.
Footnotes
Competing interests: None declared.
This article has been peer reviewed.
Contributors: All authors made substantive contributions to this work. Melissa Potestio drafted the initial work and critically revised the manuscript for intellectual content; she also made substantial contributions to the acquisition and interpretation of data. Heather Sharpe made substantial contributions to the acquisition and interpretation of the data; she also critically revised the manuscript for intellectual content. David Johnson made substantial contributions to the conception, design, acquisition, analysis, and interpretation of data, and drafted revisions of the work revising it critically for intellectual content. Jeremy Grimshaw made substantial contributions to the conception, design, analysis, and interpretation of data, and revised it critically for intellectual content. Peter Faris oversaw the analysis and played a critical role in the interpretation of the data, and revised the work critically for intellectual content. Qiuli Meggie Duan conducted the analysis and made contributions to its interpretation, and revised the work critically for intellectual content. Jody Pow made substantial contributions to the acquisition and interpretation of the data, and revised the work critically for intellectual content. Andrew Cave made substantial contributions to the conception and design of the study, and acquisition, analysis, and interpretation of data, and revised the work critically for intellectual content. All authors (with the exception of Andrew Cave, who died during preparation of this manuscript for publication) approved the final version to be published and agreed to be accountable for all aspects of the work.
Funding: This study was funded in full by the Partnership for Research and Innovation in the Health System (PRIHS) Program, Alberta Innovates (Andrew Cave was nominated principal investigator). The funder had no role in the design or conduct of the study, or writing of the manuscript. Jeremy Grimshaw held a Tier 1 Canada Research Chair in Health Knowledge Transfer and Uptake.
Data sharing: The full study data set is available directly from the corresponding author Melissa Potestio (mlpotest{at}ucalgary.ca).
- Accepted August 19, 2025.
This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY-NC-ND 4.0) licence, which permits use, distribution and reproduction in any medium, provided that the original publication is properly cited, the use is noncommercial (i.e., research or educational use), and no modifications or adaptations are made. See: https://creativecommons.org/licenses/by-nc-nd/4.0/
References
In this issue
Article tools
Jump to section
Related Articles
Cited By...
- No citing articles found.
More in this TOC Section
Similar Articles
Collections
Latest Jobs
-
Kingston, Ontario
-
Oshawa, Ontario
-
Halifax, Nova Scotia
-
Sarnia, Ontario
-
Vancouver Island, British Columbia
-
Richmond Hill, Ontario
