Youth aged 8-17 years were eligible for participation in this 2-month study. The inclusion criteria were as follows: (1) high-risk asthma, defined as patient having emergency department or hospital visit for asthma in the previous 12 months or the child’s primary asthma provider in the outpatient setting having the opinion that the youth is high risk for future asthma visits to the emergency department or hospital owing to concern for noncompliance or poorly controlled asthma; (2) prescribed a controller and rescue medication compatible with Bluetooth inhaler cap; (3) caregiver or youth own and are able to use a smartphone compatible with Bluetooth devices; (4) English speaking; (5) identified primary care provider with at least one visit in the last year; and (6) ≥1 identified caregiver present for enrollment. This study was approved by the Institutional Review Board at the Medical University of South Carolina.

Our analysis included descriptive statistics for demographics, comorbid diagnosis, medications, the usability survey results, and percentage of days with asthma data acquired (EMA data or any asthma data including inhaler use or EMA). We analyzed the percent of days with data acquired before and after the app redesign using t test. We further investigated differences in the EMA reporting frequency by age, gender, and weekend versus weekday reporting with Pearson correlation coefficients or t tests. We used generalized linear mixed models with daily reports of asthma symptoms and emotional state predicting the outcome of daily adherence to controller medication accounting for the random effect of individual participants to explore whether the adherence behavior may be associated with individual EMA responses [17,23]. The models included a time variable for each study day to account for the longitudinal nature of our data with repeated measures of symptoms and EMA responses over time. This analysis enabled more accurate correlation between daily factors as it accounted for clustering on the patient level (eg, individual enrollee lower or higher adherence). Encounters for asthma in the emergency department or hospital during the 2-month study period were explored with participant-level data on reported asthma symptoms to determine whether preceding signals exist for an acute care encounter. Statistical analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC, USA, 2013).

In this study, we enrolled 14 youth aged 8-16 years and their parents (Table 2). Of the 20 patients screened for enrollment, 14 (70%) participated in this study. While 4 patients screened did not meet the inclusion criteria (age <8 years, not on a controller medication, uncertain access to cellular service because of leaving country of residence for the majority of study period, or noncompatible smartphone), 1 participant consented but medication changes immediately after consenting no longer met the inclusion criteria. Of patients who were referred and met the inclusion criteria, only 1 declined participation citing they preferred to avoid research studies. Partial EMA or adherences data were used for 3 patients because of smartphones being reported as lost or damaged.

Among 14 participants, 13 were defined as high-risk asthma because of emergency department or hospital visit within past 12 months, and 1 was included because of the consideration of high risk by the primary provider. Overall, 14 patients reported comorbid diagnoses, including attention deficit hyperactive disorder (3/14), allergic rhinitis (2/14), eczema (3/14), food allergies (1/14), gastroesophageal reflux disease (1/14), obstructive sleep apnea (1/14), and vasovagal syncope (1/14). All patients were receiving both controller and rescue medications. Additional medications at the time of enrollment included antihistamines (5/14), nasal steroids (3/14), leukotriene receptor antagonists (1/14), proton-pump inhibitors or H₂ blockers (2/14), oral corticosteroids (4/14 total with 3/14 having a 3- to 5-day course of oral prednisone and 1/14 oral dexamethasone), stimulant medications (2/14), and Omalizumab (1/14). The overall controller medication adherence was 30%.

Multimedia Appendix 2 presents usability survey responses. All youth and parents agreed or strongly agreed with the statement “I feel comfortable with a doctor or a nurse monitoring my health information using mobile technology.” During the exit interview, 1 parent stated, “I had no challenges, at all” “it was easy, and I’m not technology savvy.” Another parent highlighted the benefits of smartphone use over using computers through her statement of “...a lot of people, with the phone, they carry that everywhere. With your computer, most of the time you’ve gotta wait ‘til you get home and check it’.” Previous studies have also demonstrated African American adolescents, and their parents prefer technology interventions that are not based on a home computer platform [21]. Our iterative design process used enrollee feedback to inform the redesign of our technology-based approach for enrollees 8-14 to maximize the participant usability. Post-hoc analysis to determine the difference in the usability survey response before and after redesign was difficult because of the small cohort and frequent zero response in disagree or neutral categories. We did find the percent of enrollees who responded with agree or strongly to the overall look and feel of the App is visually appealing (question 3) increased nonsignificantly from 60% to 100% (P=.09).

Enrollees had asthma data collected, either inhaler use or EMA, on 89% of days with a significant increase when comparing enrollees before and after the redesign (86%-92%, P=.01). EMA data were obtained on 20% of study days for all enrollees and significantly increased after the redesign (10%-40%, P=.006). The redesign increased the data acquisition, and this might represent improvements in the ease of use. No significant differences were observed in the frequency of EMA reports by age (P=.48), gender (female 27% and male 13%, P=.28), or weekend versus weekday (14% and 22%, respectively, P=.09).

EMA reports during the study period frequently included coughing (42/110, 38%), wheezing (8/111, 7%), chest tightness (9/109, 8%), stress (12/109, 11%), not relaxed (43/107, 40%), not happy (51/111, 46%), angry (11/111, 10%), and bored (57/109, 52%). In this population of high-risk youth with asthma, symptoms of wheezing or chest tightness were never reported beyond a little with only coughing being infrequently reported as a lot. Multimedia Appendix 1 provides a detailed description of responses. A total of 10 new asthma symptoms episodes were detected in the cohort over 2 months.

Our exploratory generalized linear mixed models indicated that the participants’ controller medication adherence might increase on days when asthma symptoms (9.2%; 95% CI 0-18.5) or boredom were reported compared with days when reports demonstrated no asthma symptoms or boredom (Table 1). We investigated the impact of imputing missing report days as no symptom days and found a similar effect size but with a narrower CI (9.9% adherence; 95% CI 4-15). We explored this further and found the adherence was higher on days with reports of nonsymptoms than on nonreport days (35.1% vs 25.2% adherence, P<.001), suggesting nonreport days were similar to nonsymptom days but might represent days with even lower adherence. As nonreport days had significantly lower adherence, we excluded them from the analysis of contextual factors that impact the adherence to avoid overestimating the significance of symptoms on the adherence.

One patient experienced an emergency department visit while enrolled in the 2-month study compared with 4 acute care visits in the cohort 2 months pre-enrollment (Table 3). The 4 pre-enrollment acute care visits included one intensive care unit admission. In addition, 1 patient with an emergency department visit during the study period reported asthma symptoms in the days preceding the emergency department visit. During the 6 months pre-enrollment, our study cohort had 5 visits to the acute care setting, including the abovementioned intensive care unit visit compared with 3 acute care visits in the 6 months postenrollment.

Smartphone asthma monitoring was feasible and acceptable as reported by children and their families in a population of youth aged 8-16 years with high-risk asthma. The iterative design approach successfully used youth and family engagement to increase data attainment during the study period. Despite being a small-scale pilot trial, the smartphone approach resulted in high-frequency data attainment and began demonstrating statistical associations between reported asthma symptoms and emotional state and participants’ daily adherence patterns. Understanding the reasons for daily adherence behavior among high-risk youth with asthma might lead to better-tailored interventions that maximize the adherence behavior. Nevertheless, larger studies should be performed to determine the generalizability of the findings.

This feasibility trial was designed to test the acceptability and feasibility among patients, including the collection of real-time asthma data. It was not powered to find significant and generalizable differences in the adherence or other clinical outcomes. Additionally, the adherence rate (30%) reported in this trial, which was comparable to other studies, might have been artificially increased because of the Hawthorne Effect. Participants were aware of the medication monitoring and the intervention permitted patients to manually enter medication use when the Bluetooth devices ceased functioning. This approach was adopted as it was preferred by families. One parent noted their child had her inhalers at her grandparents one weekend and was away from the smartphone. Ultimately, the inhaler caps stored this information and the information was captured once in proximity to the smartphone. Thus, larger studies validating youth response to these symptom questions compared with other measures of risk, such as asthma control and emergency department use, are warranted. We cannot confirm the presence or absence of asthma symptoms reported with EMA, but previous studies suggested that this approach is valid in multiple disease states, including asthma [19,20,23]. Furthermore, individual participants may experience symptoms differently because of different body awareness, and this could impact their symptom reporting, particularly their severity of reported symptoms. More so, as a pilot study, we permitted the youth and parents to decide whether the child was to report symptoms alone or in collaboration with the parent. We used interviews to assess how the approach worked for them. Youth would frequently report symptoms alone but had discussions with their parent, which was consistent with the literature suggesting youth value a shared asthma experience with their caregiver and these ages are transitioning to the primary responsibility of asthma care. Unfortunately, we did not have an objective assessment to determine the degree of parental involvement in reporting, but this will be added to a subsequent larger study. Finally, this pilot study was conducted at 2 centers in a high-risk population; therefore, larger studies are needed to determine whether the results are generalizable to other populations.

Asthma symptoms are known to vary over time because of numerous factors, including environmental exposures, medication adherence, physical exertion, and the natural progression of the disease. Mulvaney et al demonstrated that EMAs focused on asthma are associated with the established measures of asthma control (eg, Asthma Control Test) but also noted the difficulty in obtaining patient self-reported medication use [20]. Our study adds to this research by using newer technology for passive collection of medication use data, Bluetooth inhaler devices, and demonstrating an association of contextual factors, such as asthma symptoms and emotional state by EMA, with the adherence device data. The calculated adherence over days, weeks, months, and years is a series of discrete events occurring in patients’ daily life that can be affected by mood and motivation. Research in other chronic diseases suggests that day-to-day factors such as control belief, mood, and social support affect daily adherence, often mediated by daily motivation [17]. Our pilot study furthers this investigation into a population of high-risk youth with asthma and suggests daily asthma symptoms and mood may impact their adherence; this suggests asthma symptoms, if recognized, may be a motivational tool for adherence in this population. Our finding associating boredom with higher adherence deserves further exploration. Boredom is typically considered a negative emotion but may also represent a time when youth are better able to adhere to medications as they are less preoccupied and distracted by other daily activities. Further studies may consider assessments of these and other momentary factors in the home environment among a larger cohort.

Ultimately, the recurrence of new asthma symptom episodes, which we observed 10 times, may be an antecedent to an emergency department or hospital visit. Although this study was not powered to determine the association of recurrent asthma symptoms and emergency department visits, it is noteworthy that our 1 patient with an emergency department visit reported recurrent asthma symptoms the day before their presentation to the acute care setting.

The recruitment process was designed to enroll high-risk patients with asthma to develop a technological tool specific to the unique needs of this population. Our participants were predominantly minority with comorbid diagnosis, insured by Medicaid, and experienced numerous acute care visits before enrollment. During KII conducted postintervention, all participants enrolled in this study indicated they had regular access to smartphones, although children who were at younger ages frequently used their parents’ phones. The near-universal adoption of smartphone technology has the potential to improve access to care for high-risk children with asthma who can be difficult to reach and maintain in primary care practices.

Our technology design process included an iterative approach based on the frequent engagement pretrial and during the clinical trial’s final visit. The redesign elements (eg, 4-digit automated log-in) determined by engagement were not resource-intensive but did result in a marked improvement in data accrual, supporting the importance of patient engagement when designing an mHealth platform [29-31]. The frequent rate of data accrual combined with survey and interview data consistently conveyed ease in the use of this approach. The concept of monitoring symptoms and medication use in real time in a patient’s home environment could result in more timely care delivery based on current symptoms with fewer barriers to care. Ultimately, approaches to integrate technology to improve care processes for chronic disease, once monitoring is successful, are limitless. Huckvale et al noted common weaknesses in many asthma apps that should be considered in the design of future research; these include lack of consistency with best practice guidelines, ability to protect privacy, accounting for local variation in care, and demonstration of proven clinical efficacy [32,33].

Patients and families reported that using smartphones for real-time asthma monitoring was acceptable. The youth’s daily medication adherence demonstrated a tendency to increase during days with reports of asthma symptoms or boredom. Future research should investigate other contextual factors that may impact changes in daily adherence behavior and approaches to effectively use monitoring with provider-informed interventions to improve adherence behavior and clinical outcomes.