Methods: Tumblr posts were identified using keywords for SI, depression, and autism between January 2- 7, 2020. All posts exhibiting SI and depression were separated into two categories-those users who had self-identified as autistic and those who had not. 298,498 unique posts were analyzed. The sentiment analysis tool VADER was used to identify the affect and latent Dirichlet allocation (LDA) was used to find common themes in the posts.

Results: There was a statistically significant difference between expressions of depression and SI between the general population and those self-identified as autistic. The autism group had a slightly more positive affect than the general population (0.04, -0.12, p<0.001). When looking at the most common words unique by group, the autism group expressed more social sentiment than did the general population. The LDA also found that themes about mental illness were more common in the autism group and themes about physical symptoms were more common in the general population.

Conclusions: Suicidal ideation among autistic individuals is not well understood. Because expressions of depression and SI are different in this community, traditional screening for suicide risk may not be as accurate. This study helps shed light on new ways to approach suicide prevention and outreach in the autistic community. With this new perspective, researchers and public health practitioners better understand how autistic individuals write about depression and SI and can use this information to improve health outcomes. Additionally, Tumblr may provide another platform for reaching autistic individuals with suicide prevention efforts.