Can social media encourage diabetes self-screenings? A randomized controlled trial with Indonesian Facebook users

Non-communicable diseases (NCDs) are a leading cause of death globally, with diabetes posing a significant challenge, particularly in middle-income countries (MICs) like Indonesia. While screening for NCDs is crucial for early intervention and prevention, encouraging participation among individuals without obvious symptoms remains difficult, especially in MICs with low NCD awareness. The World Health Organization (WHO) advocates for mass media campaigns, but these typically focus on traditional media. This study explores the potential of social media, specifically Facebook, as a cost-effective tool for promoting diabetes self-screening in Indonesia, a country with a high diabetes burden and extensive Facebook usage. This research addresses a gap in the literature by evaluating the effectiveness of a social media campaign targeting a disease with low public awareness in an MIC context. It also differs from existing studies by measuring actual offline behavior change following online engagement and conducting a cost-effectiveness analysis that considers the cost per diagnosed case, not just the cost per person reached.

Existing literature demonstrates the potential of social media campaigns to impact various health behaviors in high-income countries. However, evidence regarding their effectiveness in MICs with low disease awareness is limited. Studies have explored social media's role in improving vaccination rates, addressing COVID-19 infections, promoting healthy eating habits, and encouraging physical activity. These campaigns largely focused on providing information, lacking the integration of concrete actions beyond online engagement like clicks or likes. This study extends the existing research by incorporating an actual screening activity and measuring offline behavior changes through a follow-up survey, improving the assessment of the campaign’s real-world impact. The cost-effectiveness aspect, usually overlooked, is also a key focus of this research.

A three-week Facebook campaign titled "Ada Gula, Ada Diabetes" targeted Indonesian Facebook users in Jakarta and Yogyakarta. Five different ad designs were randomly assigned to users above age 35, using Facebook's A/B testing function: two loss-framed ads ("diabetes consequences" and "shock"), and three neutrally framed ads ("family," "religion," and "geography"). Ads directed users to a website containing a diabetes risk screening questionnaire adapted for the Indonesian population. The questionnaire, based on existing tools like the American Diabetes Association's test and the FINDRISC, calculated a risk score (0–16). Participants received a risk assessment and recommendations, including advice to seek professional screening if high-risk. Six weeks post-campaign, a follow-up survey was sent to participants who provided their email addresses to assess their compliance with the recommendations. Data analysis involved logistic regression models to evaluate the effectiveness of different ad designs in generating link clicks and conversions (completed questionnaires), controlling for age, gender, and region. A cost-effectiveness analysis projected the impact of scaling the campaign to the entire island of Java.

The campaign reached 286,776 individuals, generating 758,977 impressions and 5274 link clicks (1.84% click rate), exceeding rates observed in similar studies. 1469 individuals completed the screening questionnaire (cost of US$0.75 per person). The "diabetes consequences" and "shock" ads significantly outperformed other ads in terms of both link clicks and conversions. The "consequences" ad yielded a 23% higher click-through rate compared to the reference ad, while the conversion rate increased by 57%. Gender differences in ad responsiveness were observed, with women showing a stronger preference for loss-framed ads. The follow-up survey (n=53) showed that 34% of high-risk respondents who were unaware of their condition planned to schedule a professional screening. Extrapolating this to the main sample, an estimated 250 individuals at high risk complied with the recommendation. A cost-effectiveness analysis projecting a one-year campaign across Java estimated the diagnosis of almost 170,000 previously undetected diabetes cases at a cost of approximately US$9 per newly diagnosed person. This was considerably less expensive than other screening strategies.

The study demonstrates the effectiveness of a Facebook-based campaign in promoting diabetes self-screening in Indonesia. Loss-framed ads proved particularly effective, aligning with theoretical predictions. The campaign's high reach and low cost suggest significant potential for cost-effective NCD screening in resource-constrained settings. Gender differences highlight the need for tailored messaging strategies. Although the campaign showed promise in reaching high-risk individuals, the self-reported compliance with professional screening needs further validation. The findings underscore social media's potential for public health interventions in MICs, although scalability considerations remain.

This study demonstrates the feasibility and effectiveness of using Facebook to promote diabetes self-screening in Indonesia. Loss-framed advertisements proved most effective, achieving high reach and engagement at a low cost. Future research should focus on validating self-reported compliance, exploring the long-term effects of loss-framed messaging, and investigating potential negative psychological consequences. The findings advocate for the integration of social media into public health strategies for NCDs in MICs.

The study's reliance on self-reported compliance with professional screening constitutes a limitation, potentially overestimating true compliance rates. The limited sample size in the follow-up survey may restrict the generalizability of the findings. Additionally, the potential long-term psychological impact of loss-framed messaging requires further investigation. The generalizability is limited to specific demographics and settings.