Effectiveness of mobile health interventions to improve nasal corticosteroid adherence in allergic rhinitis: A systematic review

Abstract Background Mobile health interventions (MHI) offer the potential to help improve nasal corticosteroid (NCS) adherence in allergic rhinitis (AR). The aim of this systematic review was to summarise the current evidence on the effectiveness of MHI for improving NCS adherence in AR. Methods We systematically searched MEDLINE, Embase and the Cochrane Central register of Controlled Trials (CENTRAL) for randomised controlled trials filtered for publication dates between 2010 and 2021. We evaluated the effects of MHI aiming to improve NCS adherence on self‐management outcomes in AR and comorbid conditions. Two reviewers independently screened potential studies, extracted study characteristics and outcomes from eligible papers and assessed risk of bias using the Cochrane Risk of Bias tool 2.0. High heterogeneity precluded meta‐analysis. Data were descriptively and narratively synthesised. Results Our searches identified 776 individual studies of which 4 met the inclusion criteria. These studies were heterogeneous with respect to participant, intervention and outcome characteristics. We considered all outcome‐specific overall risk of bias assessments to be of high risk of bias except for two studies examining NCS adherence which received ‘some concern’ grades. The three studies which reported on NCS adherence found that MHI were associated with improvement in NCS adherence. Significant MHI‐associated improvement in symptoms or disease‐specific quality of life was found in one study each, whilst no study reported significant differences in nasal patency. Conclusions Whilst MHI showed potential to improve NCS adherence, their effect on clinical outcomes varied. Furthermore, robust studies with longer intervention durations are needed to adequately assess effects of MHI and their individual features on NCS adherence and clinical outcomes.


| BACKGROUND
Allergic rhinitis (AR) is one of the most common diseases globally, estimated to affect over 400 million people; it typically persists throughout life. 1,2 Because of nasal symptoms (nasal itching, sneezing, rhinorrhoea and nasal congestion), often associated ocular symptoms (itching, tearing and redness of the eye; allergic rhinoconjunctivitis [ARC]) and other related symptoms (itching of the palate, postnasal drip and cough), AR significantly impairs sleep quality and cognitive function, increases discomfort, irritability and fatigue and ultimately reduces disease-specific quality of life (QoL). 3 In addition, AR is strongly associated with comorbidities such as asthma 4,5 and chronic rhinosinusitis (CRS). 6 As a result, AR causes substantial direct and indirect costs associated with medical expenses and reduction in work and school performance, respectively. 3 Nasal corticosteroids (NCS) are widely recognised as the most effective medication class for controlling AR symptoms and mitigating their deleterious effects on disease-specific QoL. [7][8][9] NCS are the mainstay of AR, ARC and CRS treatment.
NCS usually need to be taken throughout the entire period of allergen exposure to optimally reduce nasal inflammation and AR symptoms 8,10 ; however, NCS adherence remains poor and inconsistent for many. 11 A myriad of underlying factors, including variables related to disease, patient, treatment, physician-patient relationship and healthcare system contribute to non-adherence. 12,13 However, forgetfulness remains one of the principal barriers, 11,14 suggesting that both intentional and unintentional non-adherence coexist, in turn necessitating diverse and multifaceted strategies and interventions to effectively improve NCS adherence, 11 as with other long-term conditions. 15 Rapid advances in mobile technologies have ushered mobile health (mHealth) to the fore as a potential tool to improve NCS adherence through the use of a multitude of features that principally promote healthcare professional-to-patient and patient-to-patient communication, patient empowerment, monitoring and education. 16 Whilst mHealth represents an intriguing prospect for improving NCS adherence, little clinical research currently exists on its efficacy and benefits. 17 Moreover, to our knowledge no systematic review has embarked on collating and evaluating current clinical research data.

| OBJECTIVES
To examine whether mHealth interventions (MHI) for improving NCS adherence in AR and comorbid conditions (ARC and CRS) were effective in improving NCS adherence and clinical health outcomes (symptoms and disease-specific quality of life) compared to usual care not including MHI.

| Protocol and registration
The systematic review is registered with, and the corresponding protocol is available from, the PROSPERO database with registration number: CRD42020198879.

| Eligible studies
Only randomised controlled trials (RCTs) were eligible for inclusion in the systematic review, including cluster RCTs, wait-list controlled RCTs and cross-over RCTs. Quasi-experimental trials were excluded.

| Population
All population groups who were prescribed NCS treatment either as monotherapy or in combination with other treatments for both seasonal and perennial AR with/without ocular symptoms (ARC) or CRS were included. Studies that additionally targeted parents or carers of participants (e.g., children) who contributed to NCS treatment adherence were also included. Individuals exclusively prescribed other treatments excluding NCS (e.g., antihistamines or immunotherapy) were excluded. Interventions which exclusively targeted healthcare professionals were excluded.

| Intervention
Studies were included if they delivered interventions with a primary or secondary aim of improving adherence to NCS through the use of MHI. The World Health Organization's (WHO) definition of mHealth was used for this systematic review, namely a 'medical and public health practice supported by mobile devices, such as mobile phones, patient monitoring devices, personal digital assistants (PDAs) and other wireless devices'. 18 Therefore, studies that implemented MHI using mHealth devices, such as mobile phones, smartphones, smartwatches, tablets, PDAs and electronic monitoring devices as an integral part of the intervention were included. Peripheral devices, for example, sensors and sensory wearables and web-based programmes were included as long as they were accompanied by one or more of the above-mentioned primary devices. Studies using primary devices that were not handheld or mobile, for example, landline telephones or stationary computers, were not included.
The MHI could be used alone or be part of a broader multifaceted intervention which could be with or without healthcare professionalto-patient contact (i.e. face-to-face or virtual consultations).
Lastly, studies that exclusively used phone calls or teleconsultations as an alternative to face-to-face consultations were excluded from this systematic review.

| Comparators
We only included studies with a control group of participants who were not provided or did not have access to an MHI for improving NCS adherence. Control groups either received usual care or the same intervention devoid of the mHealth component. Usual care pertained to standard care per guidelines or standard care in the given setting at the time in which the study was conducted. Multiarm intervention studies, such as varying types of MHI, were included as long as one comparator group matched the criteria above.

| Outcomes
The reporting of one or more of the following primary outcomes constituted an inclusion criterion. All relevant study outcomes were extracted upon inclusion.

| Report eligibility criteria
No restrictions were applied to geographical location or type of setting. Studies written in languages other than English were eligible if they could be translated using Google Translate to a standard where study characteristics were clearly discernible.
Only study reports available in full-text versions were included.
Attempts were made to contact study authors to obtain full-text articles when unavailable. All supplementary reports or conference abstracts were excluded. Lastly, due to the fast-paced nature of mHealth research, only studies from 2010 to present were eligible for inclusion.

| Search strategy
The search strategy was formed using the 'pearl-growing' method in

| Study selection
Two review authors (MB and HT) were blinded to each other's verdicts and independently conducted the two-stage screening of titles and abstracts of study reports extracted from the search results, using the developed screening form (Appendix 2 in Supporting Information S1). Initially, studies that clearly did not meet the inclusion criteria based on their titles were excluded whilst abstracts were scanned against the inclusion criteria during the abstract screening phase. All reports that met the inclusion criteria were coded as 'Yes' and otherwise 'No' in Covidence, 19  regarding eligibility or to address uncertainties regarding incomplete or ambiguous methods that required further clarification.
Disagreements were resolved either through discussion or by a third review author (JS) whilst reasons for exclusion were documented. Review authors were not blinded to either study authors, journal titles or institutions. Cohen's kappa for inter-rater reliability for the title/abstract and full-text screening were calculated.

| Data collection process
A data extraction form (Appendix 3 in Supporting Information S1) was developed and inserted into Covidence. The template for intervention description and replication was used to model the data extraction form. 20 Two review authors (MB and HT) independently extracted the data from each included study.
Both review authors (MB and HT) participated in calibration exercises prior to data extraction. The few disagreements that occurred were resolved through discussion and no arbitrator was needed.
The data extraction form was piloted on one of the included studies and modifications were made where appropriate. Extracted data were divided into the following six distinct domains: general study information, methodology, participant details, intervention details, comparator details and study outcomes.

| Data items
The following data items were extracted from the included studies: 1. General study information: author(s), institution(s), sponsorship source(s), conflicts of interest, country and setting.

2.
Methods: study design, date of study, methods of randomisation, length of follow-up, total study duration, length of 'run-in' period, study centre details, recruitment setting(s) and recruitment methods.

| Risk of bias individual studies
A risk of bias (RoB) assessment for the primary outcomes (NCS adherence, symptoms and disease-specific QoL) in each study was carried out independently by MB and HT using the Cochrane Risk of Bias tool 2.0 22 and Microsoft Excel version 16.37 (Microsoft Corporation). We investigated the effect of assignment to intervention ('intention to treat'). Prior to the assessment, efforts were made to contact study authors to acquire study protocols and trial registry records that were not available to the review authors. The assessment of RoB was conducted using the following domains (as outlined in 1. Bias arising in the randomisation process.

Bias due to deviations from intended interventions.
3. Bias due to missing outcome data. 4. Bias in measurement of the outcome. 5. Bias in selection of the reported outcome.

Overall bias.
For each domain, a series of 'signalling questions' pertaining to the assessment of RoB was answered with either 'yes', 'probably yes', 'probably no', 'no' and 'no information'. An algorithm mapped the recorded answers and proposed a RoB judgement of either 'low risk of bias', 'some concerns' or 'high risk of bias' for each domain. These were overridden by the review authors when deemed appropriate.
Comments and direct quotations from study reports were attached to support answers given to each signalling question. Likewise, justification was provided whenever RoB judgements from the algorithm were overridden. Lastly, the domain-level judgements provided the basis for an overall RoB judgement for each specific outcome being assessed for each study. Review authors were not blinded to study details.
Disagreements were firstly resolved through discussion and secondly via a third review author (AS) for arbitration.

| Data synthesis
Study outcome data were not pooled in statistical meta-analyses due to the clinical heterogeneity of the study characteristics. Instead, the findings were analysed via a narrative synthesis, including tables and figures to aid in data presentation where appropriate.

| Study selection
The search yielded a total of 985 records as shown in Figure 1. A total of 776 records remained after excluding 209 duplicates.
Subsequently, 764 records which clearly did not adhere to the inclusion criteria were removed during the two-stage screening process, thereby leaving 12 publications eligible for full-text review.
Two full-text 24,25 records could not be procured during the process, despite efforts to contact their respective authors, as these were not available through our institutional holdings. However, upon further examination, it was discovered that both were conference abstracts.
Upon completion of the full-text review, four studies [26][27][28][29] were included in the systematic review. The bibliographies were consulted for each of the four included publications, however, no further relevant citations were identified.
The inter-rater agreement during the title/abstract and full-text phases produced a Cohen's kappa of 0.355 (small agreement) and 1.0 (perfect agreement), respectively. The disagreement during the title/ abstract mainly stemmed from differing interpretations of the interventions and population groups. However, these were resolved during the subsequent full-text screening phase.

| Methods
The characteristics of the included studies are summarised in Furthermore, two of the study reports 27,28 mentioned offering training/walk-throughs in using the mHealth platforms prior to study commencement.
Lastly, no peripheral devices (e.g., sensory wearables) were used in any of the included studies and all interventions utilised the participants' own phones throughout the study duration.

| Comparisons
In all but one study, the comparators were patients without access to the mHealth platform used in the interventions. 26,28,29 In the remaining study, 27

| NCS adherence
Three of the four studies reported on NCS adherence 26,28,29 as shown in Table 3. Of these, one study used a participant-reported assessment, using number of days being non-adherent to NCS, 29 whilst the two other studies utilised objective dose-count assessments; one based on the amount of spray puffs remaining at follow-up, 26 the other on canister weight at follow-up. 28 All three trials found strong evidence to suggest that NCS adherence improved among the participants in the intervention groups compared to those of the control groups. More specifically, Feng et al. 26  Wang et al. 29 , using self-reported days being non-adherent to NCS, used odds ratio to measure the difference between the intervention and control groups. Implementing an adherence cut-off at 95%, the intervention group had almost fourfold higher odds of being NCS adherent compared to the control group.

| Symptoms
Two of the four studies reported on symptoms, using the five-item Allergic Rhinitis Control Test (ARCT) questionnaire 28

and Visual
Analogue Scales (VAS) 29 to assess ARC disease control and AR symptoms, respectively.

and the
SinoNasal Outcome Test-20 (SNOT-20). 26 Feng et al. 26 and Pizzulli et al. 28 did not find any statistical evidence to indicate any difference in disease-specific QoL scores between intervention and control groups, using the SNOT-20 (F-test = 0.043, p = 0.988) and AdolRQLQ (output not reported) assessments.
Lastly, Cingi et al. 27 found strong evidence for a positive difference in RQLQ scores between the MHI group and control group using the Mann-Whitney U test (p < 0.001).

| Nasal patency
Two studies reported on nasal patency, measuring nasal airway resistance (NAR) by a rhinomanometer 28,29 as shown in Table 4.   Neither study reported significant differences in NAR between study groups based on Chi-square test, however, specific statistical outputs were not outlined in the study reports.

| Adverse events
No adverse events or harms were reported.

| Risk of bias within studies by primary outcomes
The

| NCS adherence
Among the three trials reporting on NCS adherence, two trials 26,28 were assessed to have an overall RoB of 'some concerns', whilst one 29 received a 'high risk of bias' grade. These were mainly due to exclusion of participants being linked to non-compliance with treatment, 26,29 underreporting of reasons for participant exclusion, 28 inappropriate outcome measurements, 29 lack of analyses adjusting for missing data 28 and inavailability of study protocols. 26

| Strengths and limitations
Our systematic review had a number of strengths and limitations.
The strengths included the use of an expansive search strategy which yielded 776 unique records. Two reviewers independently screened, extracted data and assessed the quality of studies, with the latter being carried out with a novel version of a validated and compendious RoB tool.
However, this study also had some limitations. In general, there was a lack of power due to small participant numbers and short study durations which might especially have had an influence on the varied results for disease-specific QoL, symptoms and nasal patency as longer time periods may be required to see the full effects of sustained uptake of NCS. 30 The clinical importance of the findings was also difficult to ascertain as no information on clinically relevant differences or effect sizes were provided.

| Outcome level limitations
Moreover, due to the considerable heterogeneity across study characteristics and reported outcome measures, it was not possible to perform meta-analyses.
The small number of included studies meant that sensitivity analyses were not feasible, limiting the evidence of the effects of individual mHealth features, mHealth activity and acceptability on adherence and clinical outcomes.
In a constantly evolving mHealth field, no study was conducted beyond 2016, therefore the effects of more contemporary MHI are also less clear.

| Review level limitations
Only three bibliography databases were used for this review, potentially leaving relevant studies undetected. Also, not using country/region-specific databases might have introduced language bias. However, MEDLINE, Embase and CENTRAL are considered the most integral databases to search for reports of trials by the Cochrane review group. 23 Despite contacting the authors, we were unable to retrieve two full-text articles 24,25 during the full-text screening phase, which could be a limitation. However, these were conference abstracts and are not likely to be available in full-text versions.

| Interpretation of results
To our knowledge, no systematic review on MHI effectiveness has been conducted in AR, ARC or CRS. However, our findings are in line with a systematic review by Miller et al. 31 who reported MHI (mobile apps and SMS) were efficacious in improving inhaled corticosteroid (ICS) adherence in asthma compared to usual care, whilst mixed results were found for disease-specific QoL and asthma symptoms. Although the meta-analysis did provide a positive cumulative standardised mean difference for diseasespecific QoL, one of three studies reported no improvement.
Likewise, a systematic review 32 examining the effectiveness of reminder systems (web-based apps and SMS) for ICS adherence in asthma, found similar improvements in ICS adherence whilst no differences were reported for disease-specific QoL and asthma symptoms compared to usual care. As with the current review, both reviews reported low numbers of included studies and short study durations as significant limitations.
As both unintentional and intentional non-adherence exist in both AR and asthma, a multitude of strategies may be needed to effectively improve adherence. 11 In particular, education is seen as a key tool in addressing intentional non-adherence in AR and asthma, 11,33 especially if underpinned by behavioural change models and supported by technology. 34 While educational components were present in all but one of the included interventions, however, the use of behavioural change models were not reported in any of the included studies.
Overall, incorporating a diverse range of interventional components, as included in the current review, including education, self-monitoring, reinforcement, professional support and reminders could improve NCS adherence in both AR and comorbid conditions.

| Implications and recommendations
Our findings indicate that MHI have the potential to improve

| CONCLUSIONS
The current review highlights both the potential effectiveness of MHI for improving NCS adherence in AR, ARC and CRS and a range of methodological issues within the current evidence base and thus a need for future research to fill important evidence gaps. Due to the relative infancy of the field and current research dearth, more robust studies are needed to properly evaluate the long-term efficacy of MHI and their sub-components on NCS adherence and clinical outcomes in AR, ARC and CRS. It will also be important to understand if MHI for these conditions also affect outcomes of comorbid asthma.