We conducted a mixed methods systematic review following results-based convergent design [23]. We adhered to Cochrane methodology [24] and used the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines to report our review findings [25]. The review is registered at PROSPERO (CRD42022344224), and the protocol is published [26].

We developed a search strategy involving the review authors and a senior librarian from the University of Edinburgh. We searched MEDLINE, Embase, Scopus, PsycInfo, CINAHL, and Cochrane Library (S1 in Multimedia Appendix 1) from 2001 (when internet accessibility surged with third-generation cellular technologies) [27] to July 2022. We updated the search in September 2023. We placed no language restrictions on the database search and kept the option to translate relevant quantitative studies to English [28]. However, we only considered qualitative and mixed-methods studies written in English due to the potential loss of nuance in translation [29].

After the search, all identified records were saved in EndNote 20 (Clarivate Analytics), with duplicates removed using SRA Deduplicator software [30]. In total, 2 authors (NU and MH) independently screened titles, abstracts, and full-text papers for eligibility against the criteria (S2 in Multimedia Appendix 1), using Covidence [31]. Any discrepancies were settled through team discussion (HP, VH, KM, and JS). The search results are presented in a PRISMA flow diagram.

Our primary (clinical) outcomes from quantitative data were asthma control or quality of life measured with any validated tools and acute attacks (eg, emergency department visits, hospitalizations, or unscheduled care) [32-34]. We were interested in between-group differences at the first follow-up assessment postintervention. Where multiple assessment tools for an outcome were reported, we used the most frequently reported validated measure (eg, asthma control test [ACT] [32], childhood-asthma control test [C-ACT] [35]) in the meta-analysis.

We included studies that explored the views and experiences of patients and professional stakeholders on asynchronous digital health (with or without other modes of communication) for asthma care.

NU and MH independently extracted quantitative data, and NU and DS independently extracted qualitative data from the included studies. HP checked the accuracy of data across text, tables, and meta-analyses.

Textbox 1 summarizes the methods of data analysis used to address our review objectives.

We assessed statistical heterogeneity through visual inspection of forest plots, chi-square, and I² tests for our clinical outcome [38]. We used a random-effects model to account for heterogeneity in the study intervention, population, and settings. The limited number of studies (<10) in the meta-analysis precluded creation of a funnel plot to assess publication bias [39].

Our a priori planned subgroups were age (child or adult), high- or low-income countries, asthma severity, and intensity of intervention, and we planned to include an additional subgroup if that were suggested by the qualitative synthesis. We undertook a sensitivity analysis of our meta-analysis comparing the intervention with usual care, excluding studies at high risk of bias and those that could potentially influence results due to study design or inadequacies in data reporting.

We contacted the author(s) of included studies to collect any incomplete or missing data but did not perform any statistical calculation to impute missing data into the meta-analysis.

We used the revised Cochrane risk-of-bias tool [40] and assessed risk of bias for clinical outcomes within the randomized control trials (RCTs) that were included in the meta-analysis. We assessed the methodological quality of the remaining quantitative studies with diverse study designs using the Downs and Black checklist [41-43]. We used the critical appraisal skills program for qualitative studies [44,45] and mixed methods appraisal tool for mixed methods studies to appraise their methodological quality [46].

Using Cochrane’s GRADEpro GDT software [47], we assessed confidence in evidence for the clinical outcomes following the 5 GRADE (Grading of Recommendations, Assessment, Development, and Evaluations) criteria (risk of bias, consistency of effect, imprecision, indirectness, and publication bias) [24]. We used the Grades of Recommendation, Assessment, Development, and Evaluation-Confidence in the Evidence from Qualitative Reviews (CERQual) approach to assess confidence in synthesized qualitative findings using the interactive summary of qualitative findings tool [48,49].

We identified 11,034 records from 6 databases (Figure 1). After deduplication, 5662 titles and abstracts were screened, leading to the assessment of 85 full-text articles. We included 30 studies (31 unique papers) after combining 2 reports from the same study [50-80]. Out of the 1446 records screened in the search update, 4 full-text articles were assessed, but none met the eligibility criteria for inclusion in the review. The list of excluded articles and reasons is detailed in S4 in Multimedia Appendix 1.

Of the 30 included studies, 20 were quantitative [50,51,53-57,60,61,64,65,67-69,72-74,76,78-80], 6 were qualitative [52,58,66,70,71,77], and 4 were mixed-methods studies [59,62,63,75]. These studies, involving individuals with asthma, their caregivers, and health care professionals, were conducted in 9 countries, predominantly from the United States (n=14) [50,52,53,55,56,58-62,66,69,70,79] and the Netherlands (n=9) [65,71-78]. The quantitative studies encompassed various designs, including RCTs (n=14) [51,53,55,57,60,61, 65,67,68,72,74,78-80]. Of the RCTs, a cluster trial was reported in 2 papers [64,65], detailing the main trial outcomes in 1 paper [65], and the process evaluation in the other [64]. The summary of included studies is presented in S5 and S6 in Multimedia Appendix 1. The pool of studies varies for the different objectives, so further characteristics are described by objective in Table 1.

In total, 21 studies documented the use of digital functionalities, with all of them incorporating either online chat (n=12) [50,53,54,57,59,60,65,72-74,76,79] and email (n=12) [51,55,56,59-61,68,69,74,76,78,80], and SMS (n=4) [67,69,74,76] for asynchronous communication between patients or their caregivers and health care professionals. The response time ranged from within 24 hours [51,54] to 48 hours [61,72] and 72 hours [78]. No studies explicitly addressed the role of non-digital support in using digital functionalities. However, in an RCT where both groups used a web portal, similar usage was observed, despite the intervention group receiving additional support through home visits by community health workers [53]. Figure 3 [50,51,53-57,59-61,65,67-69,72-74,76,78-80] illustrates the digital health functionalities used in the included studies, with additional details available in S10 in Multimedia Appendix 1.

The thematic synthesis yielded four themes: (1) acceptability of routine asthma reviews, (2) advantages and disadvantages of asynchronous digital health, (3) implementation barriers and facilitators, and (4) preferred digital functionalities.

Implications from qualitative synthesis (preferred digital functionalities and implementation facilitators) were juxtaposed with trial findings for clinical outcomes and integrated to see if those implications had a role in improving clinical outcomes (Table 2). Out of the 6 RCTs [57,60,61,65,72,74] that showed positive improvement in asthma control, all had monitoring logs [57,60,61,65,72,74], 5 had ease of asking quick questions [57,60,65,72,74], 5 provided tailored asthma information [60,61,65,72,74], 3 had medication reminders [57,61,65], 3 had digital action plans [60,61,72], 1 was linked with the EHR, and 4 had an organized workflow with a specific person designated to respond to patients’ queries [61,65,72,74]. One study in which asthma control deteriorated had monitoring logs only [78]. Three RCTs that reported improvement in quality of life, all had ease of asking quick questions, monitoring logs, tailored asthma information, and organized workflow [51,65,74]. One RCT that reported no change in quality of life also had these features [55]. One study, where quality of life deteriorated, had monitoring logs only [78]. Of 2 of the 5 RCTs [55,57,60,68,72] that reported reduced risk of emergency department visits, 1 had medication reminders [57], and the other was linked with the EHR [60]. Of the 3 RCTs [55,68,72] that reported increased risk of emergency department visits, none had medication reminders nor were linked with EHR. However, 2 out of these 3 RCTs had ease of asking quick questions, monitoring logs, tailored asthma information, and organized workflow [55,72]. There was no specific pattern to differentiate the trials that reported increased risk of hospitalization [55] compared with those that reported reduced risk of hospitalization [60,67,72]. Overall, the ease of asking quick questions, medication reminders, tailored asthma information, and organized workflow emerged as important factors that might positively affect the intervention outcomes.

The Downs and Black scores ranged from 6 to 24 for the other quantitative studies that were not included in the meta-analysis. Methodological quality was good in 3 studies (range: 20-25) [53,79,80], fair (range: 15-19) in 2 studies [69,76] and poor (≤14) in the other 4 studies [50,54,56,73]. Among qualitative studies, 1 had high quality [77], while 5 raised some concerns [52,58,66,70,71]. Of the 4 mixed-method studies, 1 demonstrated high quality [63], and 3 raised some concerns [59,62,75]. Methodological quality assessments for qualitative and mixed-methods studies are detailed in S11 and S12 in Multimedia Appendix 1.

Using the GRADE approach, the overall certainty of evidence for the clinical outcomes was judged as very low for asthma control and low for quality of life, emergency department visits, and hospitalization. Downgrading of the certainty was mostly due to the risk of bias and imprecision (S13 in Multimedia Appendix 1). The GRADE-CERQual assessment of confidence revealed that summary review findings of the qualitative studies ranged from low to high quality. Perspectives of routine reviews, advantages and disadvantages of using asynchronous digital health, implementation facilitators, and preferred digital functionalities had moderate confidence. Limitations of using asynchronous digital health revealed low confidence, and implementation barriers revealed high confidence (S14 in Multimedia Appendix 1).

Our systematic review identified 30 studies involving people with asthma (and their caregivers) and health care professionals from 9 countries. Overall, we identified a statistically significant improvement in asthma control and a reduced risk of hospitalization when comparing interventions using asynchronous digital health with usual care. However, no statistically significant differences were observed in quality of life and emergency department visits. Certainty of evidence was very low for asthma control and low for quality of life, emergency department visits, and hospitalizations. Patients liked the convenience of asynchronous digital health, while health care professionals had some reservations. Effective implementation requires an organizational approach. Integrative synthesis highlighted the importance of functionalities such as ease of asking quick questions, monitoring logs, and medication reminders.

A strength of this review is its inclusive design, encompassing various study types and integrating quantitative and qualitative findings. We conducted a comprehensive database search, aided by a senior librarian, and the meta-analysis was reviewed by a senior statistician from the University of Edinburgh. Heterogeneity was a concern for some clinical outcomes in the meta-analysis, but our sensitivity analysis increased the reliability of the pooled estimates. We maintained an openness to non-English papers in quantitative studies and included 1 in Russian [54]. We did not deviate from the methodology outlined in the published protocol [26]. While our GRADE assessment reflected low to very low confidence in clinical outcomes, this was influenced by self-reported outcome measures and diverse tools for asthma control and quality of life. This underscores the need for standardized outcomes in trials [81]. In total, 2 independent reviewers screened titles, abstracts, and full texts, resolving discrepancies through discussion with other authors as needed. While data extraction and analysis were duplicated (NU and MH/DS), the precision of the data was verified by another author (HP). We were aware of the potential impact of reflexivity while analyzing and interpreting the qualitative data. Nonetheless, the involvement of a multidisciplinary author group proficient in qualitative evidence synthesis helped ensure a balanced interpretation of the data.

This systematic review gives confidence that asynchronous digital health can be an effective adjunct to other modes of consultation for asthma care. This could be especially valuable for patients and health care professionals with existing familiarity with digital technologies. Some health care systems have introduced guidelines, such as requiring at least 1 in-person visit before an e-visit, to establish patient-clinician relationships [94]. Successful integration of asynchronous digital health into routine practice must be designed to benefit patients, health care professions and organizations. Efficient task allocation and organization of care are crucial implementation facilitators. User-friendly digital features are essential, including tailored functionalities for 2-way asynchronous communication within set time frames, enabling questionnaire completion, and uploading images and videos (eg, inhaler technique) as necessary. Thus, patients, health care professionals and organizations can harness the benefit of asynchronous digital health for asthma care.

Further research is necessary to understand the organizational context and track the process of arranging, deploying and sustaining asynchronous consultations for asthma care. Moreover, there is a need for research to establish the best practices for implementing asynchronous digital health, engaging patient and public involvement groups, preventing misuse, and assessing the suitability of individuals with asthma or their caregivers and health care professionals across various health care settings. Identifying a low-risk group of individuals who could be managed appropriately through asynchronous digital consultations is another critical aspect requiring investigation.

Our review concludes with low confidence that asynchronous consultation supported by digital functionalities is effective and convenient and can be considered as an option for non-emergency asthma care, showing good acceptability among individuals with asthma and their caregivers. Thus, asynchronous digital consultations present an opportunity for those whose lifestyles or geographical locations impede synchronous consultations but require organization strategies to manage workload.