Social media-based civic engagement solutions for dengue prevention in Sri Lanka: results of receptivity assessment

Abstract

This article focuses on a novel social media-based system that addresses dengue prevention through an integration of three components: predictive surveillance, civic engagement and health education. The aim was to conduct a potential receptivity assessment of this system among smartphone users in the city of Colombo, the epicenter of the dengue epidemic in the island country of Sri Lanka. Grounded in Protection Motivation Theory (PMT) and using a convenience sampling approach, the cross-sectional survey assessed perceived severity (PSe), perceived susceptibility (PSu), perceived response efficacy (PRE), perceived self-efficacy (PSE) and intention-to-use (IU) among 513 individuals. The overall receptivity to the system was high with a score of >4.00 on a five-point scale. Participants belonging to younger, better educated and higher income groups reported significantly better perceptions of the efficaciousness of the system, were confident in their ability to use the system, and planned to use it in the future. PMT variables contributed significantly to regression models predicting IU. We concluded that a social media-based system for dengue prevention will be positively received among Colombo residents and a targeted, strategic health communication effort to raise dengue-related threat perceptions will be needed to encourage greater adoption and use of the system.

Introduction

Dengue, the mosquito-borne infectious disease that has troubled Sri Lanka for over 40 years has reached hyperendemic status in the island country. This happened on the back of two major epidemics in 2004 and 2009 [1], and most recently in the period 2012–2014 when nearly 125 000 dengue cases were reported [2]. The epicenter of the dengue epidemic in Sri Lanka lies in the capital city of Colombo that, in 2013, accounted for nearly 57% of dengue cases [3].

Even as its threat continues to grow, prevention and management of dengue in Colombo is faced with a myriad challenges. For instance, public health inspectors (PHIs)—the last mile in the country’s public health delivery system—who are responsible for epidemiological surveillance are overburdened. Reports suggest the coverage of PHIs has dropped in recent in years from 1 per 10 000–1 per 40 000 people [4]. This problem is compounded by manual mechanisms of conducting surveillance (like identifying breeding sites), paper-based reporting of dengue cases to the hospitals, and coordinating with other epidemiological staff to treat breeding sites. Then, mapping of dengue hotspots during a dengue outbreak is done as the outbreak unravels (in a reactive manner), as opposed to using advance predictive models to come up with proactive mapping that can help both authorities and the public undertake preventive actions in advance. Lastly, in a country that is boasting rapidly increasing rates of internet and cellular service penetration [5, 6], community education and outreach about dengue continues to be executed using outdated media channels like pamphlets and brochures. Consequently, the capacity of public health institutions to persuade the public to practice healthy behaviors to protect themselves from dengue is limited, and needs upgrading and strengthening through digital interventions.

In response to these challenges, researchers at the Centre of Social Media Innovations for Communities (COSMIC) have implemented a tablet-based integrated social media system [7] for dengue prevention among the public health workforce in Sri Lanka, and aim to launch a similar version through a mobile application among the general population. In preparation for the launch of this application, a survey was conducted in December 2013 to assess its potential receptivity by the general public.

The paper is organized as follows. We commence with a review of emerging social media innovations for infectious diseases, and highlight gaps in existing solutions. We then present a description of our social media system for health inspectors and explain the relevance of adapting the concept for use by the general public. Then, we enunciate our theoretical framework, describe our methodology and analyses, present findings and discuss implications from theoretical and applied perspectives.

Social media innovations for infectious disease prevention

Infectious diseases, like social networks, can spread through space and time at a speed and pattern that can pose serious challenges from an epidemiological surveillance standpoint. As we will see below, a number of social media innovations attempting to address different aspects of infectious disease prevention have emerged between 2008–2013. Google Flu Trends (GFT) was among the first initiatives to use online search data across 40 flu-related terms to generate predictive graphs of where flu was headed. Although the initial paper reporting the results of this experiment was widely hailed [8], the veracity of GFT’s findings and the larger role of big data in predicting disease outbreaks have been most recently called in to question by reports published in Science [9], Nature [10] and the Harvard Business Review [11]. Freifield et al. [12] developed an integrated system that collects online news articles, expert-curated accounts and validated expert alerts to stay abreast of disease outbreaks. More recently, Freifield et al. [13] reported on the development of a system that would allow community-based reporting of disease incidents through mobile phones. The concept of crowdsourcing—where the general public contributes information about disease outbreaks—has simultaneously been utilized in a range of other initiatives such as Frontline SMS and Ushahidi [13, 14]. Smartphone applications (or mobile apps) are now increasingly commonplace in health communication interventions, as they provide the ability to access hard-to-reach populations and disseminate messaging using a variety of interactive, personal modes [15, 16]. In terms of creating awareness during health emergencies and outbreaks, public health agencies use the power of platforms such as Twitter and Facebook, to track disease spread, as well as to gauge popular sentiment [17–19]. As such, most social media innovations in the public health domain to date have been designed to address one aspect of the preventive health canvas, such as epidemiology or health education or disease modeling. This is in spite of the fact that the very nature of social media allows an integration of such functionalities in a single system, and can be made accessible to multiple health system stakeholders (public health agencies, community organizations, individuals) in an integrative format.

Mo-buzz: an integrated social media application for dengue prevention

Considering the gaps in dengue prevention and the limitations of social media innovations that address different aspects of the problem in silos, we designed an integrated social media system called Mo-Buzz that comprises three integrated components for (i) predictive surveillance, (ii) civic engagement and (iii) health communication.

The predictive surveillance component helps to predict dengue outbreaks using an algorithm and computer simulation that is based on historic and current, dengue-related data. The predictions are made available to public health authorities and the general public in the form of hotspot Geographical Information System (GIS)-based maps, on their Android smartphones and/or tablet devices. The purpose is to forewarn both these stakeholders of dengue outbreaks in order to facilitate quicker and more efficient resource planning among public health authorities, and persuade the general public to practice personal protective behaviors that might reduce their risk.

The civic engagement component is built on the concept of crowdsourcing. Traditionally, communication about public health issues has been top-down, from health authorities to the general public. Disease surveillance during a public health outbreak has been the sole preserve of epidemiological departments. Our endeavor was to create a two-way communication platform that could help public health departments obtain real-time intelligence about disease spread and risk factors from the general public in a manner that could facilitate rapid response with minimal delay. Mo-Buzz allows the general public to report dengue symptoms and/or post pictures of potential dengue mosquito breeding sites. Such reports and postings are automatically geo-tagged and sent to the health authorities with a click of a button, thereby stimulating the first stage of response by health authorities.

The health communication component comprises two sub-modules. The static module consists of educational materials about dengue pertaining to dengue transmission, symptoms, prevention and treatment. The dynamic module refers to geographically targeted alerts (based on predictions in the hotspot maps) and messaging that is customized to the kind of report sent by the user. A more detailed elucidation of the rationale and description of the three components of Mo-Buzz is available elsewhere [7].

In December 2013, the research team launched phase 1 of Mo-Buzz for PHIs in Colombo, Sri Lanka. In phase 2, we conducted formative research among the general public in Colombo to assess their potential receptivity to such a system.

Theoretical framework

The potential receptivity to Mo-Buzz among its targeted beneficiaries, the general public, can be examined through a range of classical behavior change theories postulating different pathways to explain and predict health behavioral performance. We review four theoretical models that emerge from the cognitive perspective on health behavior [20] and are most relevant to our intervention. The Health Belief Model (HBM) [21], among the oldest of such theories, suggests that an individuals’ engagement in health promoting behavior is driven by their beliefs about the enormity of the health problem, and their perception of acquiring the illness or disease in accordance with their perceptions regarding the benefits of and barriers to the health behavior (or solution) being recommended. According to the HBM, cues-to-action activate the individuals’ motivation to perform the behavior even as the individuals evaluates their own ability to execute the said behavior (self-efficacy). Past critiques of the HBM have emphasized the lack of definitions of its key constructs and its additive model of effects, where all the constructs together bear upon behavioral performance [22, 23]. The Theory of Planned Behavior (TPB) [24], on the other hand, suggests individuals’ beliefs and attitudes towards the behavior (behavioral beliefs), their perceptions about others’ beliefs that they should perform the behavior (normative beliefs) and their beliefs about their own ability to perform the behavior (control beliefs) influence behavioral intention (the most proximal determinant of behavioral performance). Lastly, the Social Cognitive Theory (SCT) [25] suggest a dynamic triad where behavioral change is suggested to be a function of reciprocal relationships between the environment, personal factors and an evaluation of the attributes of the behavior itself. Despite its comprehensive nature, testing the SCT in its whole commands an inclusion of all its 10 determinants which, in a field study such as ours, was impractical given logistical and time constraints.

Our study was thus guided by the Protection Motivation Theory (PMT, [19]) which suggests that individuals’ intention to perform a behavior is intrinsically driven by the need to protect them from the health threat under consideration. The PMT’s fundamental argument supports the facilitation of behavioral change by appealing to an individual’s fear through the magnitude of harm depicted, the probability of the event’s occurrence and efficacy of the protective response [26]. The underlying principle of PMT is situated appropriately within the three components of Mo-Buzz that highlight the existing and oncoming threat of dengue and provide a protective response to deal with this threat using the civic engagement and health education components.

The PMT postulates that the intention to perform protective behavior is largely guided by two main processes: threat appraisal (comprising perceived severity and perceived susceptibility) and coping appraisal (response efficacy and self-efficacy). In the evaluation of fear appeals, threat appraisal refers to an individual’s perceptions about the level of endangerment provided by the threat (or disease) [26]. These perceptions refer to their evaluation about the severity of, and their susceptibility to, the threat. The construct of perceived severity is particularly relevant in the Mo-Buzz context given the enormity of the dengue epidemic in Sri Lanka where nearly 125 000 people were affected from 2012 to 2014 with more than half of the disease burden reported from the capital city of Colombo (where this study was conducted). Similarly, the relevance of perceived susceptibility to dengue among the general public in Colombo emanates from the four DENV serotypes [27] that have been cocirculating in the Sri Lanka for more than 30 years. People’s perceived vulnerability to dengue has also been shaped by the fact that dengue occupies a prominent position in the local social discourse which is also shaped by media coverage of dengue outbreaks, and news about dengue prevention programs initiated by the Colombo Municipal Council.

Coping appraisal refers to an evaluation of the effectiveness of the response called response efficacy and an assessment of one’s own ability to undertake or perform the protective behavior in question or self-efficacy. In the case of Mo-Buzz, an individual’s intention to use Mo-Buzz (behavioral intention) would be shaped by their perception about the effectiveness of this system to protect them from dengue, and their confidence to use this mobile-based interface and understand the information offered by the application. Our study thus focuses on exploring two overarching research questions:

• RQ1: What is the level of potential acceptance of Mo-Buzz among the general population in Colombo, Sri Lanka?

• RQ2: In the context of Mo-Buzz, what are the demographic differences in dengue-related threat and response perceptions among the general population in Colombo?

Methodology

Survey instrument

The 10-min survey questionnaire contained two sections. The first section captured demographic variables including age, gender, marital status, ethnicity, highest educational level, and monthly household income. This was followed by a brief description of the functionalities of Mo-Buzz. In the second section, participants rated their agreement to a series of statements pertaining to PMT constructs. Adapted from Milne et al. [28], each construct was captured through a three-item five-point Likert scale.

Perceived severity was operationalized as perceptions about the intensity of the threat posed by dengue, and captured with statements like: ‘The thought of dengue scares me’. Perceived susceptibility was operationalized as perceptions about the level of dengue risk to oneself, and captured with statements like: ‘I feel I am at high risk of getting dengue now’. Perceived self-efficacy was operationalized as an individual’s confidence in his/her abilities to be able to use the functionalities or features of Mo-Buzz and was captured with statements like: ‘I am confident of being able to use this dengue mobile application.’ Perceived response efficacy was operationalized as perceptions about the potential effectiveness of the Mo-Buzz application in protecting oneself from the risk of dengue, and was captured with statements like: ‘Being alerted about dengue beforehand will help me protect myself from dengue’. Intention-to-use Mo-Buzz was captured with statements like: ‘I intend to protect myself from dengue using the methods highlighted in this mobile application.’

Sampling strategy

This study was conducted in collaboration with the University of Colombo School of Computing (UCSC), the Colombo Municipal Council (CMC), and Mobitel. The target population for our study was potential users of mobile applications, in other words individuals with access to smartphones and internet connectivity. We adopted a convenience sampling approach by utilizing the auspices of our collaborators to approach those sites where we were most likely to find substantial numbers of smartphone users. Our study sites included the official premises of Sri Lanka’s national television channel, the official headquarters of a national bank and a national telecom provider, the telecom provider’s outlets in two malls in the city, and one higher education institution.

Data collection

Each participant was requested to read and sign on an informed consent form that prefaced the survey questionnaire. This form described the purpose of the study, the duration of the survey, and their right to choose the questions in the survey that they would like to answer. The paper-based survey was administered in English but language transition (through a translator) was provided to any participant needing assistance. Each participant received an incentive in the form of a Mobitel voucher worth 200 Sri Lankan Rupees. In the absence of a centralized ethics review committee for behavioral research in Sri Lanka, we sought the official permission from each of the institutions or organizations where data were collected. The informed consent process conformed to international standards where participants were explained their rights and had the freedom to opt out of the survey at any point.

Data analysis

Data were analyzed on statistical software SPSS v.20 in three main stages. First, simple frequency counts with percentages, and means analyses were used to generate the respondent profile and summary scores of PMT constructs respectively. Second, demographic analyses of PMT constructs were conducted using Analysis of Variance (ANOVA) with Tukey’s HSD post-hoc tests (significance at P < 0.05). Finally, multivariate analyses comprised of hierarchical linear regression modeling where we evaluated the value of PMT constructs in explaining intention-to-use Mo-Buzz. With behavioral intention as our dependent variable, Model 1 comprised demographic factors alone, and Model 2 comprised both demographic variables and the four PMT constructs (perceived severity, perceived susceptibility, perceived self-efficacy, and perceived response efficacy).

Results

A total of 513 responses were collected during a 2-week period in December 2013. The demographic profile of our respondent pool is presented in Table I.

Table I.

Demographic profile of survey respondents in Colombo, Sri Lanka (N = 513)

Categories	Frequency (n)	Percentage (%)
Gender
Male	289	56.34
Female	223	43.47
Age
18–30	128	24.95
31–40	220	42.88
41 and above	160	31.19
Ethnicity
Sinhalese	363	70.76
Muslim	62	12.09
Sri Lanka Tamil	49	9.55
Indian Tamil	30	5.85
Marital status
Single	148	28.85
Married	278	54.19
Separated/divorced/widowed	86	16.77
Highest Educational Level
Secondary and below	84	16.37
Certificate or diploma	239	46.59
University and above	181	35.28
Monthly Household Income
Rs 25 000 and below	107	20.86
Rs 25 001–50 000	144	28.07
Rs 50 001–75 000	130	25.34
Rs 75 001 and above	124	24.17

Respondent profile

Our study sample comprised nearly 55% male and 45% female participants. Roughly consistent with the generic Sri Lankan population, the ethnic distribution of our sample comprised an overwhelming majority (∼71%) of Sinhalese with the Tamils and Muslims accounting for the rest. Age-wise distribution showed that that the 31–40-year age bracket found greatest representation (∼ 43%), followed by participants from the >41 and 18–30 age brackets.

A majority of our participants were married (∼54%), with single (∼29%) and separated/divorced/widowed (∼17%) individuals accounting for the rest. Nearly 35% of our participants had been educated at the university-level or above, while nearly 47% had received a certificate or diploma. In terms of monthly household income, our participant profile was nearly equally distributed with participants from all four income groups accounting for 20–30% of the total sample.

Potential receptivity to Mo-buzz

We assessed the potential receptivity to Mo-Buzz (Table II) across three main aspects: the public’s perception about the utility of the system in protecting them from dengue (response efficacy), the confidence in their own ability to use the technology (self-efficacy), and their intention to use the system upon its launch (intention to use). On a five-point scale, we found that respondents’ perception about Mo-Buzz’s efficacy was positive, and the confidence in their ability to use Mo-Buzz was high. Similarly respondent’s intention to use Mo-Buzz after launch was high.

Table II.

Overall potential receptivity to Mo-Buzz using simple means analysis (5 = highest)

Constructs	M	SD
Perceived self-efficacy	4.03	0.53
Perceived response efficacy	4.07	0.43
Intention to use Mo-Buzz	4.06	0.45

N = 513.

Demographic analyses

Bivariate analyses of PMT constructs by demographic factors focused on age, education, income and ethnicity. Our analyses by age (Table III) revealed that participants in the 18–30 group reported significantly higher perceived severity than the other two groups. Similarly, they reported the highest levels of perceived self-efficacy, response efficacy and behavioral intention. Although the oldest age group (≥41 years) reported high levels perceived severity of dengue and perceived response efficacy, their perceived self-efficacy of being able to use the Mo-Buzz application was low and significantly more so than the 18–30 age group.

Table III.

Age-based analyses of PMT variables

Construct	18–30 (a)	31–40 (b)	41 and above (c)	F
	M (SD)	M (SD)	M (SD)
Perceived severity	4.24	4.09	4.02	6.18**
	(0.50)b,c	(0.54)a	(0.50)a
Perceived susceptibility	3.79	3.51	3.63	3.21*
	(0.75)	(.69)c	(0.72)b
Perceived self-efficacy	4.16	4.00	3.99	5.01**
	(0.50)b, c	(0.55)a	(0.48)a
Perceived response efficacy	4.16	4.02	4.07	4.32*
	(0.43)b	(0.45)a	(0.39)
Intention to use Mo-Buzz	4.17	4.02	4.05	4.78**
	(0.44)b	(0.47)a	(0.39)

N = 508. *P < 0.05 **P < 0 .01 ***P < 0.001.

a, b, c, d, e denote P < .05 using Tukey’s HSD post hoc tests.

In terms of education (Table IV), participants who obtained education until the secondary school or below reported significantly lower perceived severity of dengue than the certificate or diploma holders or those who received graduation until university or beyond. Perceived susceptibility to dengue was highest among those who received the most education. Participants from the lowest education bracket also reported the lowest levels of perceived self-efficacy, perceived response efficacy and intention to use Mo-Buzz.

Table IV.

Analyses of PMT variables by education level

Construct	Secondary and below (a)	Certificate and diploma (b)	University and above (c)	F
	M (SD)	M (SD)	M (SD)
Perceived severity	3.93 (0.54)b,c	4.14 (0.51)a	4.16 (0.50)a	8.33***
Perceived susceptibility	3.63 (0.70)	3.75 (0.74) c	3.81 (0.68)b	4.31*
Perceived self-efficacy	3.85 (0.53)c	4.10 (0.49)	4.04 (0.54)a	4.26*
Perceived response efficacy	3.92 (0.39)c	4.10 (0.43)	4.09 (0.45)a	5.92**
Intention to use Mo-Buzz	3.85 (0.46)c	4.12 (0.41)c	4.08 (0.47)a,b	8.49***

N = 504. *p < .05 **p < .01 p < .001.

a, b, c, d, e denote P < 0 .05 using Tukey’s HSD post-hoc tests.

Our analysis by income (Table V) revealed that perceived susceptibility to dengue was significantly higher in the lowest income group than the other three groups, with respondents in the highest household income bracket reporting lowest perceived susceptibility. Perceived self-efficacy, response efficacy and intention to use were uniformly high across all age groups. Intention-to-use Mo-Buzz was the highest among the lowest income group and was significantly higher than the next incremental income group.

Table V.

Analyses of PMT variables by income

Construct	Rs 25 000 and below (a)	Rs 25 001–50 000 (b)	Rs 51 001–75 000 (c)	Rs 75 001 and above (d)	F
	M (SD)	M (SD)	M (SD)	M (SD)
Perceived severity	4.16	4.08	4.12	4.10	.45
	(0.42)	(0.56)	(0.55)	(0.54)
Perceived susceptibility	3.98	3.69	3.72	3.65	5.12**
	(0.65)b,c,d	(0.67)a	(0.77)a	(0.70)a
Perceived self-efficacy	4.04	4.01	4.10	4.00	.90
	(0.43)	(0.58)	(0.49)	(0.55)
Perceived response efficacy	4.14	4.00	4.07	4.09	2.25
	(0.38)	(0.46)	(0.41)	(0.44)
Intention to use Mo-Buzz	4.18	3.99	4.08	4.05	3.69*
	(0.47) b	(0.46)a	(0.36)	(0.47)

N = 505. *P < 0 .05 **P < 0.01 ***P < 0 .001.

a, b, c, d, e denote P < 0 .05 using Tukey’s HSD post hoc tests.

Ethnicity-based analysis (Table VI) showed that while Sinhalese respondents reported strongest perceptions of dengue severity, Sri Lankan Tamils felt most susceptible to dengue. Perceived self-efficacy and perceived response efficacy were statistically higher among the Sinhalese as opposed to other ethnic groups. Details of the comparisons can be found in Table VI.

Table VI.

Analyses of PMT variables by ethnicity

Construct	Sinhalese (a)	Muslim (b)	Sri Lanka Tamil (c)	Indian Tamil (d)	F
	M (SD)	M (SD)	M (SD)	M (SD)
Perceived severity	4.17	4.02	4.10	3.60	12.20***
	(0.52)d	(0.49)d	(0.48)d	(0.39)a,b,c
Perceived susceptibility	3.71	3.80	3.99	3.73	2.42
	(0.76)	(0.35)	(0.43)	(0.35)
Perceived self-efficacy	4.07	3.95	4.05	3.68	5.93**
	(0.54)d	(0.49)	(0.44)d	(0.31)a,c
Perceived response efficacy	4.10	4.02	4.07	3.79	5.16**
	(0.43)d	(0.44)	(0.46)d	(0.31)a,c
Intention to use Mo-Buzz	4.08	4.08	4.09	3.68	7.88***
	(0.45)d	(0.44)d	(0.45)d	(0.37)a,b,c

N = 509. *P < 0 .05 P < 0.01 ***P < 0.001.

a, b, c, d, e denote P < 0 .05 using Tukey’s HSD post hoc tests.

Multivariate analyses

We conducted linear regression modeling to evaluate the explanatory power of PMT variables in relation to participants’ intention to use Mo-Buzz (Table VII). Our analyses showed that demographic variables alone explain 5% (R² = 0.05) of the variance in intention-to-use with education and monthly household income proving to be significant predictors.

Table VII.

Linear regression on intention to use Mo-Buzz

Model	B	t	Sig	R2	Adj.R2
1
(Constant)	3.88	32.17	0.00***	0.06	0.05
Gender	−0.04	−1.06	0.29
Age	−0.04	−1.45	0.15
Educational level	0.07	5.06	0.00***
Monthly household income	−0.04	−2.14	0.03*
2
(Constant)	1.13	4.94	0.00***	0.35	0.33
Gender	−0.03	−0.81	0.42
Age	0.01	0.23	0.82
Highest educational level	0.03	2.80	0.01**
Monthly household income	−0.03	−1.91	0.06+
Perceived severity	0.07	2.11	0.04*
Perceived susceptibility	0.04	1.75	0.08+
Perceived self-efficacy	0.24	6.80	0.00***
Perceived response efficacy	0.34	8.21	0.00***

N = 497. ⁺P < 0 .10 *P < 0 .05 **P < 0 .01 ***P < 0 .00.1.

Controlling for demographic factors, PMT variables strengthened the explanatory power of the model by 29% (R² = 0.33). In this model, perceived severity, perceived self-efficacy and perceived response efficacy proved to be significant predictors of intention to use.

Discussion

This study tested the potential receptivity of a novel social media-based dengue prevention system among the general public in Colombo and attempted to generate evidence that informs future social media interventions for infectious diseases prevention in the developing world. We identified specific attitudinal constructs that bear upon the future adoption and use of this system among the general public.

Our overall receptivity assessment (Table II) offered positive, above average scores (>4.00) for each of the three aspects of potential adoption of Mo-Buzz (response efficacy, self efficacy and intention-to-use), suggesting a definitive need for such a system. The strong feedback on response efficacy alludes to a prevailing public sentiment about the value of a pro-active, interactive system that can protect people from dengue by forewarning them about dengue outbreaks, and enables real-time communication with health authorities. The positive response on self-efficacy is testament to the latent power of mobile phones as tools of public health intervention delivery in Sri Lanka. Finally, the strong intention-to-use scores provide evidence-based encouragement for us to consider launching a public version of the Mo-Buzz system.

Our demographic analyses revealed findings that strengthen extant claims in the literature. For instance, the youngest age group in our study reported highest levels of dengue-related perceived severity and susceptibility. A similar pattern was observed in at least two other studies [16, 29], one noting that older age was associated with lower perceived severity of breast, colon and ovarian cancer; and the other reporting that younger women perceived themselves at greater risk for HIV infection than their older counterparts. One possible explanation for this finding is that younger people in Colombo, equipped with better education and media exposure, might be more aware about the prevailing dengue situation and the possible threats posed to them by dengue. This rationale dovetails into our other finding which reveal that participants who were better educated tended to report stronger perceptions related to health beliefs (severity, susceptibility), and reported greater response efficacy and self-efficacy pertaining to the use and effectiveness of Mo-Buzz.

From the analyses by income, low threat perceptions about dengue among higher-income respondents lend credence to PHI’s views about prevalent attitudes among this group. These views were gleaned through in-depth interviews conducted by our research team during a qualitative needs assessment in 2013 (findings from which are being reported by the authors in a separate paper under preparation). During this exercise, PHIs informed us that households in up-market residential areas in Colombo were often uncooperative in terms of sharing health information and did not encourage the PHI's entry into their compounds or houses for inspections. Based on this survey’s findings, the underlying reasons for this apparently lie in an underestimation of the gravity of the dengue situation among this group, and the thought of themselves being at low risk for being infected with dengue. It is important for public health authorities to devise strategies to educate these higher income groups on the fact that their relatively manicured surroundings do not make them immune from dengue: an argument supported by two facts. First, the CMC has earmarked Cinnamon Gardens—an area located in the popular Colombo 07 zipcode and boasting high real estate prices—as a dengue hotspot [30]. Second, a popular and wealthy Bollywood film director living in a plush household located in one of Mumbai’s most upmarket real estate localities lost his life to dengue in the year 2012 [31].

The analyses by ethnicity offer new contributions to current understanding of dengue-related health beliefs among the Sri Lankan population. The high levels of perceived severity but low levels of perceived susceptibility among the Sinhalese group suggests that while this group might be cognizant of the overall threat of dengue (at the city or country level), their perception of the proximity of this threat to themselves might be much lower—perceptions that can potentially shape low motivation to practice protective behaviors. From the four ethnic groups we studied, the Indian Tamil group stood out for consistently low perceptions related to severity and susceptibility (compared to the other groups), and only marginally higher perceptions related to response efficacy and intention-to-use Mo-Buzz.

The programmatic implications of our findings pertain to the need for an awareness campaign about Mo-Buzz prior to the launch of the application for the general public. As such, our analyses provide a useful segmentation framework to inform such a campaign. For instance, it is incumbent among such a campaign to be designed in a manner that is aimed at creating greater awareness about the severity and the level of threat of dengue. These message components need to be especially disseminated to groups that are low literate, older, belong to Indian Tamil ethnicity or are higher income. Specific messages emphasizing the purpose of Mo-Buzz in combating dengue might help enhance the perceived utility (or response efficacy) of the system among Indian Tamil and Muslim groups, 31–40 year olds and those with lesser education. Campaign sessions demonstrating the usability of the Mo-Buzz application need to be targeted at Indian Tamil and Muslim groups, older populations (41 years of age or more), low literate and high income groups. Possibly requiring user training sessions, user manuals, and demonstration videos, these messages will be aimed at explaining the functionalities and features of Mo-Buzz in a simple and clear manner, so as to enhance the confidence of such groups to operate the application on their own.

Study limitations

The seemingly counter-intuitive findings where higher educated and higher income groups report similar perception strengths can be explained in terms of sampling limitations. For example, the higher educated groups in our sample mainly comprised university students who were earning very little (hence, low income). However, the high income respondents in our sample were relatively senior in age and employed with such institutions as banks, and were not as highly educated. We expect to incorporate these sampling nuances while utilizing the findings from our study for devising an audience segmentation strategy towards the pre-launch Mo-Buzz awareness campaign. The other main limitation of this study pertains to the convenience sampling strategy which mitigates the generalizability of our findings to the rest of the Colombo population.

Conclusions

Conceptual advances in the field of mobile health and the ubiquitous penetration of smartphones have allowed us to develop a one-of-its-kind crowdsourcing-based dengue prevention intervention in a low-resource, developing country settings such as Sri Lanka. As cutting-edge as the proposition sounds, this paper reinforces the value in the traditional approach of conducting formative research as opposed to a technologically deterministic approach of parachute interventions (directly implementing a health technology innovation without prior research). Our theory-based receptivity assessment not only provided valuable insights to our intervention design but also contributed to our conceptual understanding of health beliefs surrounding dengue in Sri Lanka. While we aim to utilize these findings to inform a pre-launch campaign and conduct subsequent evaluation studies, future investigations can build upon our evidence—both in terms of conducting more representative theoretical inquiries and designing innovative technological interventions – to alleviate the dengue burden in Sri Lanka and the rest of the South Asian region.

Funding

This research is supported by the National Research Foundation, Prime Minister's Office, Singapore under its International Research Centres in Singapore Funding Initiative and administered by the Interactive Digital Media Programme Office.

Conflict of interest statement

None declared.