Characterizing the transmissibility of cancer.

Abstract

e13076 Background: Although cancer screening has a remarkable impact on worldwide morbidity and mortality, patients are often non-compliant with screening recommendations. Our ability to understand and manage the seeds of non-compliance can reap substantial public health benefits. The Twitter social network provides a unique perspective on public sentiment regarding cancer screening modalities. Methods: We utilized a Naive Bayes algorithm to analyze over 30,000 tweets related to cancer screening via colonoscopy, pap smear or mammography. Tweets were classified as having positive, negative, or neutral sentiment. For each screening modality, we correlated tweet sentiment with demographic information as well as markers of social network visibility (i.e. Klout score). We also identified the most commonly occurring words among positive, negative and neutral tweets. Finally we used re-tweet rate to study the contagion of sentiment. Results: For mammography and pap smear-related tweets, sentiment was more likely to be positive than negative. In contrast, more colonoscopy-related tweets carried negative sentiment than positive sentiment. Contagion of sentiment paralleled the spread of infectious diseases with a ‘re-tweet’ rate that was close to the critical value of R0 = 1. Sentiment toward colonoscopy, mammoraphy, and pap smear did not correlate with Klout score. However, there was significant heterogeneity in re-tweet rate, beyond what would be expected by chance. Positive and negative tweets among each screening modality tended to share lexical features. Positive tweets showed a strong resonance with the benefit of early detection. Fear and pain were the principal lexical features seen in negative tweets. Conclusions: Twitter data provides a quantitative framework for how to improve the dissemination of positive sentiment, both in terms of how cancer screening is discussed with patients and how public health messages can be disseminated via social media.