M Shahid, Jiayi Song, Zaiba Ali, Hafiz Muhammad Naveed, Serhat Yuksel, Hasan Dincer
Background: The role of social media in in shaping the public perception of health crises is growing, yet we know little about the contribution of platforms such as YouTube in shaping knowledge regarding child undernutrition. We aimed to explore the digital discussion of and sentiment around YouTube content on child undernutrition in Pakistan, where at least 38% of children are stunted (as per global rankings). Methods: We compiled a dataset of 1847 videos and 42 963 comments YouTube users that were uploaded between January 2010 and March 2024. We used Latent Dirichlet Allocation (LDA) topic modelling to analyse this data and identify narrative frames present in the content. We also used the Valence Aware Dictionary for sEntiment Reasoning (VADER) to analyses sentiment in the comments. Results: The analysis reveals six dominant narrative frames, where 'policy failure' made up 32% of the discourse, while just 5% addressed 'solution-focused actions'. The VADER sentiment analysis supports this with a compound score of -0.72, suggesting an overall negative and highlighting considerable public concern about institutional responses. The data further suggests specific engagement patterns that point to a clear break between crisis-oriented and solution-oriented information. Conclusions: Algorithms on digital platforms amplify the narratives that serve as digital determinants of health, shaping public perceptions and reinforcing public distrust in institutional policy responses. It will not be enough to simply 'get the message right' if we seek to create effective health communication - we must, instead, be prepared to actively challenge harmful discourses, champion fair and locally relevant narratives, and engage for algorithmic transparency and public health-centred content curation from these platforms. These results provide an important evidence base for the reframing of nutrition communication initiatives in Pakistan and beyond.