INTRODUCTION

Dengue is a mosquito-borne viral disease that affects around 390 million people (more than 20% of the global population) annually, with reported cases in nearly every continent [1, 2]. Since 2007, annual dengue deaths have increased by 65.5%, from 24,500 to 40,500 deaths in 2017 globally [3]. While the disease mainly affects lower- to middle-income countries, ongoing globalisation and climate change contribute to increasing dengue transmissions to previously dengue-free regions in Asia, Europe and North America [4, 5]. Fortunately, dengue mortality can be significantly reduced with timely and appropriate treatment.

Currently, dengue treatments are supportive, and no antiviral treatment exists [6]. Likewise, no vaccine for dengue has been proven to be efficacious or safe for public use to date [7-9]. Therefore, without curative treatments for dengue, treatment relies on individuals to seek timely and appropriate care when they suspect contracting the disease. While the mortality rate of untreated severe dengue cases is around 20%, supportive care diminishes the risk of mortality to below 1% [6, 10].

Despite good prognosis with supportive care, many patients delay medical care for dengue [11, 12]. Some studies attributed this delay to low health literacy [13, 14]. Hence, efforts to improve health literacy using educational programmes have been undertaken to improve health-seeking behaviour [15-17]. However, the number of dengue mortalities continues to increase, suggesting that health literacy may not be the only factor causing delayed care seeking. This necessitates a more comprehensive understanding of the factors affecting healthcare seeking for dengue. Currently, there is a lack of literature documenting the wide range of factors that affect patients’ healthcare-seeking behaviour for dengue. Thus, this systematic review aimed to identify and synthesise the factors influencing healthcare seeking for dengue.

MATERIALS AND METHODS

This qualitative meta-synthesis was guided by the 2020 Preferred Reporting Items for Systemic review and Meta-Analysis (PRISMA) statement and includes a PRISMA checklist report (Supplemental File A) and flow diagram [18, 19]. The protocol was registered with PROSPERO (CRD42021233852).

Eligibility criteria

This review included qualitative, quantitative and mixed-method original research articles that reported any factors influencing dengue healthcare seeking. All qualitative study designs were included, but only cross-sectional and retrospective cohort studies were included in this review for quantitative studies. Interventional studies such as randomised controlled trials were excluded as they aimed to change health-seeking behaviour rather than reporting the current behaviour. Mixed-methods studies were included if one of the components (qualitative or quantitative) reported relevant data.

All studies included in the review focused on dengue infection. Studies that included dengue and other infectious diseases were included if the variables and outcomes between the diseases were well differentiated. Besides that, the studies had to discuss formal healthcare seeking (provided by healthcare professionals) as an outcome and the associated factors. Studies covering a broader scope, such as dengue prevention and control, were included if factors to healthcare seeking were covered. We excluded studies that reported ‘attitude on dengue fever’ as they did not truly reflect care seeking. The study population could either involve community members or patients of dengue endemic regions. No restriction was applied to the year and language of the publication.

Information sources and search strategy

We used three methods to obtain articles for this review: database searches, field expert consultation and backwards reference searching.

Five databases, namely PubMed, EBSCO's CINAHL Complete, EBSCO's Psychology and Behavioral Sciences Collection, Web of Science and Scopus, were searched. The search strategy was devised using the Population, Intervention, Control and Outcome (PICO) framework [20]. Since this review did not include interventional studies, only the population (dengue) and outcome (health-seeking behaviour) concepts were included in the search strategy. Study design was used as the third concept to focus on interested study designs. The search strategy was reviewed and revised by the research team on several rounds. The final search strategy comprised keywords and ‘Medical Subject Headings’ (MeSH) (or its equivalent in other databases), as well as the title and abstract search tag. The final search strategy can be found in Supplemental File B. All references were extracted from the databases on Dec 13, 2020.

A list of field experts was derived using a literature analysis tool (PubReMiner, University of Amsterdam), fed with the same search strategy used for PubMed. The top 20 field experts (dengue health-seeking behaviour) were contacted via e-mail requesting literature recommendations. All experts were contacted in February 2021.

A backward reference search was carried out by examining the references of included studies from the database. This was done after completing the full-text assessment of the articles. The reference list was extracted using an online metadata extractor (CERMINE, University of Warsaw) [21]. This process was repeated for studies included from reference mining to search for additional relevant studies.

Study selection process

Search results were imported into a citation manager (EndNote 20, Clarivate Analytics), and duplicates were removed. A de-duplicated list was uploaded to a citation screening tool (Abstrackr, Brown University), where titles and abstracts were screened for inclusion. Subsequently, full-text screening was performed using EndNote 20. Non-English studies were translated using an online translator tool (Google Translate, Alphabet Inc.) and validated by two native speakers of the language in question. The study selection (both title/abstract and full-text assessment) was performed independently by two researchers (T. C. N and J. Y. T), with discrepancies resolved through discussions mediated by a third researcher (C. H. T). The exact process was repeated for articles obtained through reference mining and experts. No studies were excluded automatically using computational screening tools.

Study risk of bias assessment

The Joanna Briggs Institute's (JBI) Critical Appraisal Tools were used to appraise the quality of included studies. The JBI's Checklist for Qualitative Research was used for qualitative studies, while the Checklist for Analytical Cross-Sectional Studies was used for the quantitative studies [22, 23]. The appraisal of all articles was conducted independently by two researchers (T. C. N and J. Y. T). As the qualitative synthesis did not involve meta-analysis, the appraisal was done only to report the quality of the studies. No article was excluded based on the outcome of the appraisal.

Data extraction, synthesis and analysis

The best fit meta-synthesis approach was used in this review, where extracted data were charted according to a preliminary framework with new themes subsequently formed if any data could not fit the framework [24]. The data extraction forms were created in spreadsheets (Excel 2019, Microsoft) to chart the data on study characteristics and the primary outcomes that influence healthcare seeking in dengue. A preliminary framework was created for the primary outcomes based on McLeroy's social-ecological model, which consisted of the individual, interpersonal, organisational, community and policy domains [25]. Two sets of the form were created for qualitative and quantitative studies. The data extraction forms (Supplemental File C) were pilot tested with three articles and finalised.

All extracted data existed in the form of barriers or facilitators to dengue healthcare seeking. Two researchers (T. C. N and J. Y. T) first independently extracted, verbatim, the barriers and facilitators and charted them in the data extraction form. Any factor extracted from qualitative studies was charted depending on how the sentence was originally phrased in the article. For example: ‘lack of knowledge is a reason why people did not seek healthcare when having dengue’ was charted under the barrier column. For quantitative studies, the factors compared groups (e.g. lower education vs. higher education) and were charted as barriers and facilitators according to their effect on healthcare seeking. The p-values (from multivariate, else univariate analysis) were also extracted and charted in the data extraction form. Any discrepancies in data extraction were resolved through discussions mediated by a third researcher (C. H. T). One study reported the p-value for healthcare seeking in multiple settings (hospital vs. clinic vs. health centre vs. self-treatment vs. folk healer) separately, and we had to combine the raw data of hospital, clinic and health centre and re-run the study's statistical test to represent healthcare seeking in a modern context.

For data synthesis, the verbatim texts were reduced to summarised statements and sorted according to similar themes to form the factors under each domain [26]. Factors from both qualitative and quantitative (significant factors, p < 0.05) studies were combined to form a comprehensive framework on factors influencing healthcare seeking for dengue.

Additionally, the number of studies that reported each factor was counted to present the most reported factors influencing healthcare seeking in dengue patients. Separately, a sub-analysis among quantitative studies was done. For each factor, the number of quantitative studies that reported the factor as significant and the number of quantitative studies that reported the factor as non-significant were counted. Finally, the ratios of studies significant to non-significant for each factor were presented to compare the differences.

RESULTS Study selection

Our search strategy yielded 7436 articles from the five databases, and 3794 articles remained after deduplication (Figure 1). After screening the titles and abstracts, full texts of 184 articles were retrieved and assessed. Sixteen articles were included based on full-text assessments. Reasons for exclusion included no clear link between the factors and healthcare seeking (3607 articles), defunct articles (three articles), non-primary articles (one article), irrelevance to dengue (six articles), interventional studies (three articles) and inexplicit reporting of study topic (158 articles). In addition, there were 16 articles forwarded by experts and 749 articles identified through reference mining, of which 4 additional articles were included, rendering a total of 20 articles included in the meta-analysis.

A PRISMA Flow Diagram reporting a final inclusion of 20 studies in this review

Study characteristics

The publication year of the 20 included studies ranged from 1997 to 2020, and 15 were published between 2011 and 2020 (Table 1). Ten studies used a qualitative design, seven used a quantitative design and three used a mixed-methods design. The studies were frequently conducted in Asian countries (11 of 20 studies), although Colombia contributed the most of all countries (5 of 20 studies). Fifteen studies were conducted within the community, and 10 involved adults above age 18. Most studies (15 of 20) did not report any theoretical frameworks. Among those who did, the health belief model was most common (3 of 20 studies). Specific details of each study can be found in Table 2 (qualitative studies) and Table 3 (quantitative studies).

TABLE 1. Characteristics of the included studies Characteristics No. of studies (n = 20) Year published 1991–2000 1 2001–2010 4 2011–2020 15 Study design and approaches Qualitative (n = 10) *some studies used multiple approaches Direct social observation 3 Focus group discussions 6 In-depth interviews 10 Quantitative (n = 7) Cross-sectional survey 5 Retrospective study 2 Mixed methods (n = 3) Cross-sectional survey + open-ended questions 1 In-depth interviews + cross-sectional survey 2 Ethnicity Asian 11 Latin American 9 Study country Brazil 1 Cambodia 1 Colombia 5 India 1 Indonesia 2 Laos 1 Malaysia 2 Myanmar 1 Peru 1 Singapore 1 Thailand 2 United States 1 Venezuela 1 Study setting Community 15 Hospital 5 Study population Healthy adults 7 Patients 6 Caretakers 1 Mixed of the above 5 Unreported 1 Age of study population Adults (>18 Years Old) 10 Both 5 Children (<18 Years Old) 1 Unreported 4 Theoretical framework used Health belief model 3 Patient care process mapping approach 1 ‘Path to Survival’ Model (Modelo ‘Camino A La Supervivencia’) 1 Unreported 15 TABLE 2. Study characteristics of articles utilising qualitative methodologies Papers (Source) Disease focus Country (Study area) Setting Ethnicity Participant background Age Qualitative sample size Data collection method Theoretical model Data analysis Wong 2013 (Database) Dengue fever Malaysia (Klang Valley Urban Conglomeration) Community Asian (Malays, Chinese, Indian) Homemakers, Students, Unemployed, Employed 21–70 84 Focus group discussions Health belief model Direct content analysis, grounded theory, interpretive descriptive method Khun 2007 (Database) Dengue fever Cambodia (Kampong Cham Province) Community Asian (Khmer) Women caregivers, village health volunteers Unreported Total: 81 Focus Group: 28 In-Depth Interview: 38 Key Informant: 15 In-depth interviews, focus group discussion, and direct social observation Unreported Thematic analysis, descriptive analysis Pérez-Guerra 2009 (Database) Dengue fever United States (Carolina, Guaynabo and San Juan Municipalities) Community Unspecified Unreported ≥18 59 Focus group discussions Unreported Grounded theory Suarez 2009 (Database) Dengue fever Colombia (Giardot and Melgar Municipalities) Community Unspecified Health promoters, stakeholders, community members Unreported Total: 78In-Depth Interview: 72Focus Group Discussion: 6 In-Depth interviews, focus group discussion and direct social observation Unreported Grounded theory Pinto 2015 (Database) Dengue fever Colombia (Bucaramanga, Cali, Cúcuta, Girardot and Tame Cities) Community Unspecified Adults, caretakers, healthcare personnel Unreported 37 In-depth interviews ‘Path to Survival’ Model (Modelo ‘Camino A La Supervivencia’) Grounded theory Tan 2017 (Database) Dengue fever Singapore (Nationwide) Hospital Asian (Chinese, Malay, Indian, Bangladeshi) Patient, healthcare workers Any

Total 41

Patient: 21

Healthcare Workers: 20

Focus group discussion and in-depth interviews Patient care process mapping approach Thematic analysis Octaviana 2019 (Database) Dengue fever Indonesia (Banyumas Regency) Community Unspecified Homemakers, traders, civil servants, farmers 20–35 9 In-depth interviews Health belief model Content analysis Krisnian 2017 (Expert Recommendation) Dengue fever Indonesia (Bandung City) Hospital Unspecified Patients 17–46 9 In-depth interviews Unreported Thematic analysis Frank 2017 (Database) Dengue Fever Peru (Iquitos City) Hospital Unspecified Caregivers of children under 5 21–56 18 In-depth interviews Unreported Inductive thematic approach Roberto Suarez 2005 (Reference Mining) Dengue fever Colombia (Villavicencio City) Community Unspecified Community members, stakeholders in healthcare Unreported 30 In-depth interviews, focus groups discussion and direct observation Unreported Unreported Elsinga 2015 (Database) Dengue fever Venezuela (Maracay City) Community Unspecified Adults including guardians of children 18–87 105 Cross-sectional survey, open-ended answers Health belief model Not reported Nair 2018 (Database) Dengue Fever, Chikungunya, Malaria, Leptospirosis India (Trivandrum City) Community Unspecified Individuals with disease history Any 30 In-depth interviews, cross-sectional survey Unreported Thematic content analysis Liu 2020 (Database) Dengue fever Myanmar (Monglo Township) Community Asian (Shan) Household heads, village leaders, community health worker and representatives 18–54 18 In-depth interview, cross-sectional survey Unreported Thematic analysis TABLE 3. Study characteristics of articles utilising quantitative methodologies Papers (Source) Disease focus Country (Study area) Setting Ethnicity Participant background Age Quantitative sample size Data collection method Theoretical model Data analysis Okanurak 1997 (Database) Dengue Fever Thailand (Bangkok city and Suphan Buri Province) Hospital Unspecified Patients, caretakers of patients 0.5–14 (three older than 14) 184 Cross-sectional survey Unreported Unreported Vicente 2013 (Database) Dengue fever Brazil (Victoria municipality) Community Unspecified Adults, children Any 5,895 Cross-sectional Survey Unreported Mann–Whitney non-parametric test, chi-square test, Kruskal–Wallis non-parametric test, multiple-comparison tests Wongchidwan 2018 (Database) Dengue Fever Thailand (Bangkok city) Hospital Unspecified Patients ≥15 210 Retrospective study Unreported T-test, Chi-square test Nur Fatini 2017 (Expert Recommendation) Dengue fever Malaysia (Kubang Kerian locality) Community Unspecified Household heads 21–78 218 Questionnaire survey Unreported Linear regression analyses, multivariate analysis mode Casas 2017 (Reference Mining) Dengue fever Colombia (Cali city) Community Unspecified Patients Any 8668 Retrospective study Unreported Unreported Desjardins 2020 (Database) Dengue fever, Chikungunya, Zika Colombia (Cali city) Community Mixed (Mestizo, Caucasian, Afro-Colombian) Student, homemaker, retirees, employed, unemployed ≥18 327 Cross-sectional Survey Unreported General linear model, variation inflation test, Chi-square test Mayxay 2013 (Reference Mining) Dengue fever Laos (Pak-Ngum district) Community Asian (Loom) Adults ≥18 231 Questionnaire survey Unreported Chi-square test, Fisher's exact Test Elsinga 2015 (Database) Dengue fever Venezuela (Maracay City) Community Unspecified Adults including guardians of children 18–87 105 Cross-sectional survey, open-ended answers Health belief model Chi-square test/Fisher's exact, Student's t-test, Wilcoxon signed-rank test, McNemar's test, likelihood ratio test Nair 2018 (Database) Dengue Fever, Chikungunya, Malaria, Leptospirosis India (Trivandrum City) Community Unspecified Individuals with disease history Any 430 In-depth interviews, cross-sectional survey Unreported Pearson's Chi-square test, Fisher's exact test, Binary logistic regression Liu 2020 (Database) Dengue fever Myanmar (Monglo Township) Community Asian (Shan) Household heads, village leaders, community health workers and representatives 18–54 241 In-depth interviews, cross-sectional survey Unreported Chi-square test, multivariate logistic analysis Risk of bias in studies

Overall, the quality of the qualitative studies was acceptable for checklist items other than the reporting of reflexivity of researchers’ roles (2 of 13 studies) (Table 4). The quality of the quantitative studies was also generally acceptable other than for ‘strategies in dealing with confounding factors’ (4 of 10 studies). Details on each study's appraisal can be found in Supplemental File D.

TABLE 4. An appraisal profile of the included studies. Appraisal checklist used Yes (%) No (%) Unclear (%) JBI appraisal checklist (Qualitative Study) (n = 13) 1. Is there congruity between the stated philosophical perspective and the research methodology? 13 (100) 0 (0) 0 (0) 2. Is there congruity between the research methodology and the research question or objectives? 13 (100) 0 (0) 0 (0) 3. Is there congruity between the research methodology and the methods used to collect data? 9 (69.23) 1 (7.69) 3 (23.08) 4. Is there congruity between the research methodology and the representation and analysis of data? 10 (76.92) 1 (7.69) 2 (15.38) 5. Is there congruity between the research methodology and the interpretation of results? 9 (69.23) 1 (7.69) 3 (23.08) 6. Is there a statement locating the researcher culturally or theoretically? 2 (15.38) 11 (84.62) 0 (0)