Abstract
Nonadherence to inhaled medication leads to poor asthma control and increased healthcare utilisation. Many studies exploring adherence determinants have been conducted, but summaries of the evidence are scarce. We performed a systematic review of observational research on determinants of asthma inhaler adherence among adults.
We searched for articles in English reporting quantitative observational studies on inhaler adherence correlates among adults in developed countries, published in EMBASE, Medline, PsychInfo and PsychArticles in 1990–2014. Two coders independently assessed eligibility and extracted data, and assessed study quality. Results were summarised qualitatively into social and economic, and healthcare-, therapy-, condition- and patient-related factors.
The 51 studies included mainly examined patient-related factors and found consistent links between adherence and stronger inhaler-necessity beliefs, and possibly older age. There was limited evidence on the relevance of other determinants, partly due to study heterogeneity regarding the types of determinants examined. Methodological quality varied considerably and studies performed generally poorly on their definitions of variables and measures, risk of bias, sample size and data analysis.
A broader adoption of common methodological standards and health behaviour theories is needed before cumulative science on the determinants of adherence to asthma inhalers among adults can develop further.
Abstract
Major opportunities for strengthening evidence on determinants of nonadherence to asthma inhalers: methods and theory http://ow.ly/DY4vr
Introduction
The introduction of inhaled medication as the primary treatment for asthma has led to substantial improvements in asthma control [1, 2]. However, uncontrolled asthma is still common and represents a considerable burden to patients and society [3, 4]. An important reason for poor asthma control and, consequently, increased healthcare expenditure is suboptimal adherence to the prescribed regimen [5–7]. To date, few adherence interventions evaluated in asthma treatment have been found to be (cost-)effective [8–10]. A systematic review of observational evidence on adherence determinants could help identify the patients most at-risk for nonadherence and the key drivers of nonadherence that can be modified in adherence interventions.
Although several narrative reviews on determinants of adherence to asthma medication have been conducted [11–18], only two systematic reviews on observational research are available. Both examined adherence to inhaled corticosteroids (ICS): one focused on children [19], the other exclusively evaluated the role of illness and treatment perceptions in adults [20]. Neither examined the quality of the methodology of included studies, which is important in interpreting empirical evidence [21–23]. To our knowledge, no comprehensive systematic review of factors related to adherence to inhaled medication in adults with asthma has been published to date.
The objective of this study was to synthesise the current observational evidence on determinants of inhaler adherence in asthmatic adults through a systematic review, including a critical appraisal of the methodological quality of the studies, and develop recommendations for future research in this domain.
Methods
Literature search and study selection
EMBASE, Medline, PsychInfo and PsychArticles were searched for manuscripts published between January 1, 1990 and June 26, 2014 with keywords on asthma, adherence, persistence, compliance, concordance, determinant, cause, influence, barrier and facilitator (Supplementary material 1). Eligibility was determined using the following criteria: peer-reviewed article in English; reporting an empirical quantitative observational study (cross-sectional or longitudinal designs); presenting results on adult (aged >18 years) asthma patients living in developed countries [24]; investigating one or more predictor of adherence to inhaled asthma medication; and describing the adherence measurement procedure. The selection was initially based on the information in the title and abstract; if inconclusive, the entire manuscript was examined. Two reviewers (A.L. Dima and O. Cunillera) examined the search results independently. Disagreements were reconciled by a third reviewer (M. de Bruin) and through consensus.
Data extraction
Two coders (A.L. Dima and O. Cunillera) extracted information on: study characteristics (objectives, methodology, country, language, setting, sample size, age, sex, asthma severity and type of inhaled medication studied); adherence behaviours and determinants (definition, measurement and psychometrics); and statistical data (type of analysis and results reported). The data extraction procedure was piloted on articles not included in the review. Each coder extracted data from 50% of the papers. The accuracy of the recorded information was verified by the other coder, and disagreements were discussed and reconciled.
Quality rating
Two coders (A.L. Dima and G. Hernandez) rated methodological quality based on six criteria adapted from the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) guidelines, which are considered key requirements for observational studies [25, 26]. Scoring was performed on a four-level response format, from no information reported to adequate reporting of appropriately used methodology (Supplementary material 2). The studies were judged on clarity of methods and pertinence in six domains: 1) selection of participants (e.g. sampling strategy, eligibility criteria and methods for assessing eligibility); 2) definition of variables (i.e. outcomes, determinants and confounders); 3) description of data sources and measurement procedures for all variables; 4) addressing potential sources of bias (e.g. medical surveillance, recall, or response bias); 5) sample size justification (e.g. power analysis, multiple comparisons correction); and 6) data analysis (e.g. data preparation, controlling for confounding and data collection, and sensitivity analyses). Disagreements were discussed and reconciled.
Data analysis
The data on study characteristics and adherence measurement were summarised descriptively. The results on the relationships between adherence determinants and behaviours were grouped separately for reliever (e.g. short-acting β2-agonists (SABA)) and controller (e.g. ICS) medication as they relate to different recommendations (daily versus as needed use). Controller adherence was examined separately for the three stages of adherence [27]: 1) starting treatment (initiation); 2) accuracy of medication use (implementation); and 3) continuing treatment (persistence). Determinants were classified using the five dimensions of the World Health Organization (WHO) taxonomy [26, 27]: 1) social and economic factors, 2) healthcare team and system-related factors, 3) condition-related factors, 4) therapy-related factors, and 5) patient-related factors; each with additional sub-dimensions. We summarised results regarding the statistical significance and direction of relationships for all studies. Adjusted results obtained by multivariate analyses were prioritised over unadjusted when available.
Metric properties of the six study quality items were investigated. Reliability was assessed by estimating inter-rate agreement with weighted kappa, considered appropriate for ordinal scores [28], and interpreted based on established thresholds for poor, fair, moderate, good and excellent agreement (0.20, 0.40, 0.60 and 0.80) [29]. A Mokken scaling and correlational analyses were performed on consensus scores to evaluate structural validity and examine the relationships between criteria. Total quality scores were computed adding scores on the criteria with adequate metric properties; studies were classified as higher versus lower quality via median split. Statistical analyses were performed with SPSS version 21 (IBM Corp., Armonk, NY, USA) and the R-project (www.R-project.org) mokken package [30, 31].
Results
Study selection
The database search identified 2878 unique articles (fig. 1). The two coders agreed on the selection of 213 articles as potentially relevant (Cohen's κ=0.60). The third coder reviewed 235 disagreements and selected 86 additional articles. Thus, 299 articles were reviewed to confirm they fulfilled all inclusion criteria. 213 articles were excluded based on title and abstract, and a further 35 articles were excluded after full manuscript examination. Finally, 51 studies were included in the review. The reasons for exclusion are presented in figure 1.
Flow diagram of article selection process.
Study characteristics
Characteristics of studies are showed in table 1. Most studies were conducted in European countries (n=22) or the USA (n=19). Settings of studies were diverse, and included: primary and secondary care; pharmacies; general population; and various prescription and insurance claims databases. 11 studies used existing databases, while 40 studies collected data directly from patients. 32 studies focused solely on adults (aged >18 years), while 19 studies included adults and children. Sample sizes ranged from 34 to 292 738 participants (median (interquartile range) 204 (906)). Most studies included more females than males. Asthma severity was reported in 16 studies and ranged from mild to severe asthma.
Study characteristics of empirical studies on inhaled medication adherence in adults with asthma
20 studies focused on adherence to ICS only, eight assessed adherence to inhaled asthma medication as a generic treatment category and 23 studies focused on various types of medication, including ICS and long-acting β2-agonists (LABA) or SABA, either in monotherapy or in fixed (ICS/LABA) or free (ICS+LABA) combinations. Two studies analysed repeated measures of adherence in longitudinal cohort designs, prospectively [65] or retrospectively [81]. All other studies collected data cross-sectionally, retrospectively or prospectively (n=22, n=16 and n=12 studies, respectively) and analysed relationships between determinants and single adherence measures.
There were substantial differences between studies in operationalisation and measurement of both adherence determinants and behaviours (Supplementary material 3). Of the 68 adherence behaviour assessments (several studies used multiple measures) (table 2), 31 relied on patient reports, 24 accessed medical records (prescription and refill data), seven employed electronic monitoring, four used canister weighting, one used dose counters and one requested physician reports. 15 of the patient-reported adherence assessments applied validated questionnaires, such as the Medication Adherence Rating Scale [38] and Revised Asthma Adherence Scale [83], while the remainder used self-constructed nonvalidated questionnaires.
Definition and measurement of adherence behaviours in the studies reviewed
As most results focused on implementation of controller medication, we chose to summarise these both graphically and in the text (figs 2 and 3). The results on controller initiation and persistence and on reliever use were limited and, therefore, are only described textually.
Determinants of controller implementation. Number of positive, nonsignificant and negative relationships with adherence indicators for determinants with three or more results identified. HCT: healthcare team; ns: nonsignificant; MDI: metered-dose inhalers; DPI: dry-powder inhalers; HRQoL; health-related quality of life. #: versus other; ¶: versus white.
Determinants of controller implementation (results from higher quality studies). Number of positive, nonsignificant and negative relationships with adherence indicators for determinants with three or more results identified. ns: nonsignificant. #: versus other; ¶: versus white.
Determinants of controller medication adherence
Initiation
Determinants of controller initiation were examined in one study that reported a higher probability of non-initiation for younger patients, females, African–American ethnicity (versus white), and with fewer SABA fills in the preceding year [55]. No associations were found with socioeconomic status, comorbidity, costs of treatment and various healthcare utilisation indicators.
Implementation
We identified 544 results in 47 studies, of which 457 relationships between a determinant and an adherence measure could be assessed in terms of significance and direction of relationship. Figure 2 provides details on the WHO determinant sub-dimensions with at least three results. As different measures of adherence may lead to different associations with determinants, we distinguished between objective measures, medical records and subjective reports with each type of measurement. Results from higher quality studies are presented in figure 3. Determinants with less than three results are only described briefly in the text.
Social and economic factors were investigated in 15 studies. Adherence was related to higher income in three out of eight reported results [34, 40, 53–55, 57–59]; more prescription coverage in one out of four results [34, 40, 45, 59]; lower treatment costs in two out of seven results [47, 54, 55, 61, 77]; and lower perceptions of social norms in one out of three results [68, 72, 77]. Several other variables were identified in fewer than three results and were found to be unrelated to adherence: geographical area [47]; urban location [59]; immigration status [52]; crime rate in area of residence [54]; social modelling [68]; and social support [40, 68]. Minority status was related to adherence in one result [34], and employment status in one out of two results [52, 59].
Eight studies examined healthcare team and system factors, with education provision relating to adherence in three out of four results [32, 45, 67]. Several other variables were examined in fewer than three results: lower adherence was linked to inability to get an appointment when needed in one result [61], to patient–provider communication in one out of two results [34, 40], and to the time interval being registered with the same prescriber in one result [81], while receiving a prescription from a specialist versus a generalist was unrelated to adherence [59].
Therapy-related factors were investigated in 18 studies. Adherence was mostly unrelated to the number of drugs in the treatment regimen (three out of four results; [63, 70, 78]), the number of daily doses (five out of seven results; [39, 47, 64, 67, 78]), and having reliever inhalers prescribed (four out of five results [34, 47, 48, 64]). Using dry-powder inhalers (DPIs) versus metered-dose inhalers (MDIs) was linked to adherence in two out of four results [66, 67]. Some variables examined in a single result were unrelated to adherence: prescribed use of peak flow meter or action plan [45]; treatment duration [67]; using various other drugs [44, 48, 52, 57, 64]; using autohalers versus other MDIs [39]. Other single result variables were related to higher adherence: using diskus DPIs versus diskhaler DPIs [49]; using ultrafine versus large-particle formulation [76]; not using a spacer [52]; and receiving more refills in a prescription [47]. Three studies compared ICS/LABA regimens with different types of alternative regimens and reported better adherence to ICS/LABA compared to ICS and/or LABA and/or SABA [62], and compared with ICS in monotherapy or in combination with LABA or montelukast [46], but no differences in intentional or accidental nonadherence between ICS/LABA and ICS+LABA regimens [64].
Condition-related factors were investigated in 26 results, with nonsignificant results regarding asthma duration (nine results [34, 35, 38, 41, 45, 52, 61, 67]), pulmonary function (six out of eight results [34, 40, 45, 51, 57, 58]), and presence of current symptoms (19 out of 22 results [34, 35, 41, 43, 45, 48, 52, 57, 58, 61, 62, 64, 70, 79, 82]). Asthma exacerbations showed 13 nonsignificant [34, 40, 48, 55, 57, 67, 73, 81], but also five positive [36, 55, 73, 81] and six negative associations [52, 67, 70] with adherence. Higher health-related quality of life was associated with better adherence in four out of 11 results [45, 57, 62, 64, 67, 70], and higher asthma severity was linked to better adherence in five results [48, 68, 71, 78, 81], compared to one negative [81] and six nonsignificant results [40, 52, 64, 70, 71].
Patient-related factors were investigated in 40 studies. Patient demographics such as age and sex were included in numerous studies. Older age related to better adherence in 16 out of 28 results [32, 34, 35, 38, 40, 41, 45, 47, 52–55, 57, 58, 61, 63, 64, 67, 69, 70, 72, 73, 78, 81, 82]. Sex showed 24 nonsignificant results [34, 38, 40, 45, 48, 52, 54, 55, 57–59, 62–64, 67, 68, 70, 71, 73, 79, 82], with females showing better adherence in three results [41, 47, 53] and males in another three [61, 72, 78]. Being of white ethnicity was linked with better adherence in five out of 10 results [40, 48, 54, 55, 57, 59, 61, 70, 73, 78], while participants with higher education levels were more adherent in four out of 10 results [34, 38, 40, 45, 48, 52, 53, 57–59].
Few studies found significant roles of variables related to patients' general health status. Smoking status was consistently unrelated to adherence [40, 48, 52, 57, 58, 63, 64, 71], as was depression [40, 45, 57, 58]. Higher comorbidity was associated with better adherence in two out of eight results [47, 48, 54, 55, 57, 63], while less healthcare utilisation was linked to better adherence in two out of 11 results [34, 38, 40, 55, 70]. Asthma knowledge was found to be unrelated to adherence [32, 53], while medication knowledge was reported to be related to adherence in only one out of five results [34, 40, 61, 77]. Asthma beliefs (i.e. perceptions of the asthma impact in terms of severity, consequences, timeline, etc.) showed inconsistent relationships with adherence, with eight positive results [35, 36, 41, 53, 72], 10 nonsignificant results [35, 38, 41, 53, 57, 58], and one negative result [38].
The role of treatment beliefs was studied extensively. Stronger beliefs in the necessity of using inhalers were associated with better adherence in 14 out of 16 results [38, 40, 53, 60, 61, 65, 69, 74, 77, 82], beliefs in their effectiveness in four out of seven results [35, 40, 52, 53, 77], and more broadly-framed positive beliefs in inhaler usefulness or benefits in one out of three results [34]. Having fewer concerns about medication was related to better adherence in nine out of 17 results [38, 40, 60, 64, 65, 68, 72, 74], lower perceived side-effects in two out of four results [72, 77], lower beliefs that medication in general is overused in one out of three results [60, 77], and stronger beliefs in inhaler necessity relative to concerns in two out of three results [68, 69, 72]. Readiness to use inhalers showed positive associations to adherence in three results [37, 61], indicators of self-efficacy in four out of nine results [32, 35, 40, 57, 65, 68], and stronger adherence routines in three results [53, 68, 72]. A better ability to perceive changes in asthma symptoms was related to adherence in three of five results [51, 58], while lower confidence in the ability to monitor symptoms was related to adherence in one of three results [41, 53].
Numerous other patient-level variables were examined in fewer than three analyses, most with nonsignificant results: general health status and body mass index [57]; marital status [48]; number of causal attributions for asthma [38]; extent of attributing asthma to internal causes [41]; general health self-efficacy [65]; self-control [45]; and various personality and medical history characteristics [34, 39, 45, 52, 58, 62, 68, 69, 71, 73, 74]. Several exceptions referred to better adherence in people who consider medication as less harmful (two results [60]), display lower neuroticism, higher agreeableness and conscientiousness (one out of two results [69, 74]), and believe more strongly that their asthma can be controlled [38, 41]. Several single results showed better adherence in people with a family history of asthma [71], asthma onset at younger age [58], lower impulsivity [62] and high literacy [80]. Other single findings suggested that more adherent people attribute their asthma more to external factors [41], believe that God is less in control of their health and attribute more control to physicians [73], perceive themselves less vulnerable to side-effects, report higher intention to use inhalers [72], have better inhaler use skills [79], are more satisfied with the device [70], prefer to use inhalers rather than pills [32], have no preferences regarding daily inhaler dosage [75], believe more strongly in participating actively in care [36], and report no symptom improvement due to herbal drugs [52].
Persistence
Controller-persistence determinants were investigated in three studies, and results are presented below. Patients receiving prescriptions from a specialist, using MDIs, having a lower recommended dose, having once-daily dosing frequency, having used LABAs in the previous year, and having had previous asthma-related hospitalisations were more likely to persist using single ICS treatment during 1 year, while adolescents and patients with more than twice daily dosing frequency were more likely to discontinue [56]. For ICS/LABA therapy, persistence was less likely for adults compared to children, for people with longer therapy duration, higher daily dose, and having used antibiotics in the previous year [56]. Patients using ICS/LABA were more likely to persist with therapy compared to those using ICS+LABA, as were male patients, older patients, those receiving social assistance, those with lower daily dosage, those receiving prescriptions from a specialist, and those using more medications currently and in the previous year [50]. Time to discontinuation of ICS/LABA therapy was longer for male patients, older patients, those paying moderately for treatment, having more refills included in the first prescription, having prescriptions for other conditions, and having had relievers prescribed before the start of the study [47].
Determinants of reliever use
Reliever use recommendations were examined in three studies. Reliever overuse (as indicator of nonadherence to reliever recommendations) was linked to increased symptoms in two out of three results [43, 57], to older age in one out of two results [42, 57], and to lower education, higher self-perceived asthma severity and lower general health status in one result [57]. Other factors were unrelated to overuse (e.g. sex, ethnicity, socioeconomic status, smoking status and various health status indicators).
Study quality
The 51 studies received relatively good quality scores regarding participant selection methods and measurement of variables, with 19 and 14 studies receiving the maximum score, respectively (table 3, Supplementary material 4). Scores were considerably lower on appropriateness of data analysis, measures taken to protect against bias, study size justification and clarity of definitions for the variables included. Common limitations in reporting patient selection were omitting methods of sampling and checking eligibility, and not specifying response rates. The concept definitions often overlapped with the description of measurement methods, or only variable labels were reported. Many studies did not describe measurement methods for all main variables. The majority of studies did not mention any source of bias, and none gave a clear sample size justification or reported optimally on study size decisions. Some studies reported power computations for unspecified analyses, did not correct for multiple comparisons, dichotomised adherence scores without giving a valid rationale, did not control for potential confounders, and offered unclear descriptions of statistical procedures. Inter-rater agreement for the six quality rating criteria (table 3) was poor to moderate, but all discrepancies were resolved through discussion between the two coders. Participant selection methods, measurement of variables, clarity of variable definitions and appropriateness of analyses formed a homogenous scale, with a homogeneity±se of 64±0.07. Performance on the two remaining criteria (addressing bias and justifying sample size) was only weakly related to the quality scores on the other four criteria (item properties not shown for brevity).
Study quality: frequencies and inter-rate agreement for quality criteria#
Discussion
This systematic review aimed to qualify and synthesise the observational evidence on determinants of inhaled medication adherence in adults with asthma. In the 51 studies included, patient-related factors associated with controller implementation were the most frequently studied, and healthcare team and system factors the least. The more robust evidence linked stronger treatment necessity beliefs to better implementation. The few studies assessing controller initiation and persistence mainly suggest a possible influence of therapy-related factors and patient demographics. Studies on reliever use were scarce, with reliever overuse related to several patient-related factors. This limited evidence offers only provisional guidance for developing inhaler adherence interventions. Furthermore, the findings regarding each adherence determinant and behaviour should be interpreted with caution and within each study context due to the heterogeneity among studies. Our review reveals important knowledge gaps that need to be addressed in the future, and also highlights crucial methodological limitations that can inform researchers regarding concrete steps to take for accumulating sound evidence in future studies.
Regarding the results on determinants of controller use implementation, the substantial focus on patient-related determinants was noted in previous reviews in asthma [19, 20] and in other chronic conditions [85–87], and reflects an interest in both identifying at-risk groups and understanding patient perspectives as proximal determinants of patient behaviours. Demographic and clinical characteristics and patients' knowledge of asthma and of medication were generally unrelated to controller use, except a possible higher risk of nonadherence in younger adults. Treatment necessity beliefs were consistently related to better controller implementation but moderate evidence exists on the role of other positive treatment beliefs and concerns. These results confirm a previous review on treatment beliefs [20] and support the relevance of addressing patients' views regarding their condition and treatment in adherence interventions.
Determinant categories not related to patients were studied substantially less and should be prioritised in future research. Condition- and therapy-related factors seemed unrelated to controller implementation behaviours or showed inconsistent results. Among these factors, several medical outcomes, such as asthma exacerbations, severity or symptoms, showed contradictory results, suggesting that their relationships with adherence might vary depending on other parameters, which would need careful examination. Despite the relevance of social and economic factors identified in previous reviews [85–87], only financial information was examined more extensively but showed inconsistent results. Limited data were available on the influence of the social environment in adults with asthma, despite the key role of social factors identified in children's asthma management [19] and in adherence to other long-term treatments for chronic conditions in general [85, 88]. Healthcare team and system factors were rarely studied, although the improvement of health services for chronic conditions is currently a priority [89] and adherence-enhancing interventions usually include changes in the structure of healthcare delivery [10]. This highlights the need for further research on the structure and content of adherence support in routine clinical care, which can have a major impact on patient behaviours and treatment success rates [90, 91]. Future studies could also benefit from adopting broader theoretical approaches that also explore factors beyond the individual patient level, such as the Precede-Proceed framework, which would facilitate behaviour change intervention design [92].
The barriers to evidence consolidation identified during the present review raise an important question: what methodological standards would future studies apply to obtain quality evidence on determinants of inhaler adherence? Table 4 summarises nine main barriers and several recommendations for improvement, formulated considering the existing methodological advice for observational research [26] and adherence research [93] in order to invite further dialogue on this topic. The first barrier identified was the substantial study heterogeneity, not only in sample characteristics but also in variable selection, definition, measurement, study design and statistical analyses. Secondly, the studies lacked a unifying theoretical approach which led to differences in variable selection and, thus, to many determinants being examined only in single studies, often without a theoretical justification. Finally, the results gave limited insight regarding causal influences, as only two studies involved repeated measures of adherence [65, 81] and only 17 studies measured determinants before adherence. Moreover, many studies showed limitations in the six quality criteria assessed, although several studies performed well (Supplementary material 4). To address these barriers, we endorse the practical recommendations provided in STROBE [26] and provide brief advice based on STROBE and our experience in this review. Theoretical frameworks and taxonomies of adherence behaviours and determinants are available [27, 94, 95] and should be used more extensively. Conducting research on common theoretical and measurement foundations would allow the field to progress from identifying bivariate or multivariate associations in heterogeneous prediction models towards testing more homogeneous and comprehensive causal models.
Barriers and recommendations for a solid evidence base on asthma inhaler adherence determinants
Beyond the practical recommendations for future inhaler adherence studies, our review also highlighted the need to develop consensus on several methodological aspects. The fact that few studies reported on variable definitions, sources of bias and study size suggests that many researchers might not be aware of their importance for observational studies. The latter two aspects were unrelated to the overall study quality, suggesting that even in higher quality studies, bias and sample size are not systematically considered. More discussion is needed among methodologists and researchers to establish their relevance and specify concrete steps to implement them. These results add to previously expressed concerns regarding the lack of validated tools to evaluate quality in observational studies [23], and highlight a general need for further detailing and clarifying methodological guidelines in this area. Our experience with coding quality exposed the difficulties of assessing these broad criteria given the diversity of designs and brief descriptions permitted by space constraints. We would, therefore, encourage adherence-specific methodological guidelines that can be reported in a standard format as supplementary material in published studies.
Our review has several limitations. First, interpreting the summary based on both adjusted and unadjusted results requires caution, as multivariate analyses control for different sets of confounders, while bivariate analyses ignore any additional influences and may reflect biased relationships. We chose to prioritise adjusted over unadjusted data to avoid this, but we acknowledge that the findings may be biased and we recommend the use of theory-based models to provide more valid and replicable results. Secondly, inter-rate reliability for quality scores was low, which may reflect suboptimal study reporting, difficulty of applying the criteria based on the given definitions, or insufficient training of coders. Although the coders were able to reach consensus, these difficulties illustrate the need for more concrete definitions applicable across studies by coders with diverse research backgrounds. Thirdly, we focused our review on developed nations, as the contribution of determinant dimensions on adherence may be different in developing nations, particularly regarding access to care [86], but only 19 studies were excluded based on this criterion. Finally, meta-analyses were not possible due to the substantial heterogeneity; therefore, we opted for a qualitative summary and for identifying methodological improvements that would make future studies more amenable to meta-analytic approaches.
Our findings suggest that adults with asthma implement controller use recommendations better if they believe more strongly in the necessity of using inhalers, and possibly if they hold other positive beliefs and less concerns about using inhalers. Younger adult patients may be more at risk of nonadherence. Other patient-, condition- and therapy-related factors are either mostly unrelated to adherence or partly studied, and little is known about the role of social, economic and healthcare factors. Initiation and discontinuation of controller use and reliever use behaviours were scarcely explored. Moreover, the methodological limitations identified diminish the strength of current evidence. Our key recommendations for further research are to improve methodology and use established theoretical frameworks, which should enable the development of a cumulative evidence base of causes of nonadherence to asthma inhalers among adults.
Acknowledgements
We would like to thank Eric van Ganse (Claude Bernard University Lyon 1, Lyon, France) and Marcel Bouvy (Utrecht University, Utrecht, the Netherlands) for valuable discussions regarding the systematic review process, and Dan Dediu (Max Plank Institute, Nijmegen, the Netherlands) for support with conducting the review and summarising results visually.
Footnotes
This article has been corrected according to the author correction published in the June 2016 issue of the European Respiratory Journal.
This article has supplementary material available from erj.ersjournals.com
Support statement: The research leading to these results has received funding from the European Community 7th Framework Programme (FP7/2007-2013; grant agreement no. 282593). Funding information for this article has been deposited with FundRef.
Conflict of interest: Disclosures can be found alongside the online version of this article at erj.ersjournals.com
- Received May 2, 2014.
- Accepted October 31, 2014.
- Copyright ©ERS 2015
References
