We collected Twitter posts between March and July 2020 that included these COVID-19–related hashtags: #COVID, #Sars-Cov, and #COVID19. We then subsampled posts that used ageist language as well as language that may indicate discussion on ageist discourse. To subsample Twitter posts that used ageist language, we identified a list of ageist language keywords aggregated from existing articles. In addition, research team members who are experts in ageism also brainstormed for keywords. We then searched these keywords on the Twitter website. Using the first 20 posts we retrieved, we assessed whether the posts referred to older adults. We then narrowed down the list of keywords, discarding those that resulted in irrelevant posts. The final keywords were “old people,” “elderly,” “older adult,” “ageist,” and “ageism.” Other keywords, such as “senior,” for instance, retrieved too many irrelevant posts, such as those related to “high school seniors.” Thus, we excluded tweets that used such terms from the analysis.

With the collected data, we used two methods to address the two RQs. First, we developed a codebook for a qualitative content analysis based on an existing survey instrument that measures ageist perception or ageist content. Several instruments exist that measure ageism: the Ambivalent Ageism Scale [27], the Ageism Survey [28], the Fraboni Scale of Ageism (FSA) [29], and the Competence and Warmth Scale [30]. Among these scales, we used the FSA, given its high citation, comprehensiveness, and inclusivity of ageist concepts when compared to the other surveys. We thematized the FSA survey questions into three categories: (1) the perception that older individuals cannot make good decisions, (2) the perception that older individuals are a burden on society, and (3) the devaluing of older individuals’ lives. We then applied these themes as codes for ageist content. We also included a theme to code antiageist content that criticizes ageist messages.

We randomly selected 150 tweets from the Twitter data. We randomly divided these tweets into 3 sets of 50 tweets and distributed the data to 3 coders. Using the codebook, 3 individuals coded 50 randomly selected tweets and 1 coder coded all 150 tweets. Given the implicit nature of the Twitter data (eg, posts can indicate sarcasm), it was challenging to establish high interrater reliability. Instead, we coded a common set of 20 tweets as a group, and using the agreed-upon codebook, we individually coded the assigned tweets. We then negotiated any disagreements.

For the descriptive quantitative analysis, we first counted the total posts that contained the keywords mentioned above. We also counted frequencies of ageist versus antiageist posts from the qualitative findings to understand the relationship between ageist content and terminology used to denote older adults. Specifically, we wanted to understand whether posts that included ageist terms were intended to be ageist and whether posts that included appropriate terms were nonageist.

All data reviewed are publicly available and were collected through the Twitter API (application programming interface) and deidentified. Therefore, no ethical review or approval was deemed necessary.

As shown in Table 1, out of 60.32 million tweets, the term “elderly” occurred in 32,700 tweets (0.05%), “old people” occurred in 26,230 tweets (0.04%), and “older adult” occurred in 11,328 tweets (0.02%).

Next, we investigated whether each term referring to older adults in tweets promoted or spoke against ageism. Table 2 describes the frequency results.

We found that the use of ageist terms was prevalent on Twitter. Twitter users also used ageist terms to criticize ageist messages, showing a lack of understanding of the appropriate use of terms when referring to older adults. Highly hostile ageist content against older adults came from tweets that denoted older adults with derogatory terms.

Ageism has been heightened during the COVID-19 pandemic, and researchers have highlighted the hostile messages being propagated through social media against older adults and their role in society [31-34]. Accordingly, this was an opportune time for researchers to assess what language is being used and how this language is associated with ageist messages.

The terms “old people” and “elderly” were more prevalent than “older adult” in tweets. This finding shows a continued use of ageist terms despite the efforts of various organizations to raise awareness about not “othering” older adults. In addition, tweets with the term “elderly” showed a high percentage of antiageist content, which also suggests a lack of awareness of recommended terminology and of the derogatory nature of the term “elderly” as deemed by the AARP.

Our paper uniquely contributes to the field of aging using social media data and qualitative and quantitative methodologies to assess the public’s use of common ageist terms. This study has a few limitations. First, the terms we used to assess ageist term use were not exhaustive. Such limitation came from the fact that the excluded terms were used in multiple ways depending on the context (eg, “senior” referring to senior housing rather than to older adults). Thus, we limited our use of the terms for analysis to those that were explicitly referring to older adults. Although using exhaustive terms would have generated more complete results, we were able to address our RQs with the terms we identified. Second, Twitter content can often convey sarcasm, which can result in interpretive errors. We engaged all coauthors of this manuscript to discuss and converge on the final interpretation. Lastly, the qualitative analysis was limited to a subset of the Twitter data set due to logistical feasibility (eg, time and resources needed for manual coding). However, this is a common practice in follow-up qualitative research to give richer nuance and context to quantitative results.

We suggest several future directions for this study. First, we can develop a training data set based on our qualitative research results to automate the identification of ageist terms on Twitter and perform a larger-scale study on the RQs. Second, a repeated follow-up study with post–COVID-19 Twitter data will help us understand how term usage has evolved over time. Third, using social network analysis, we can identify how influential Twitter users use ageist terms and how impactful their tweets are among their followers over nonageist content. Lastly, future work should examine how often ageist content is challenged and critiqued by antiageist responses. Public policies, organizations, and the technology industry should develop creative solutions to detect harmful content and educate the public on appropriate terminology use and how to change harmful perceptions of older adults.

The COVID-19 pandemic and older adults’ vulnerability to the disease have perpetuated hostile ageist discourse on social media. This is an opportune time to understand the prevalence and use of ageist language and discuss the ways forward. From examining tweets related to COVID-19, we were able to uncover the prevalence of the ageist terms used and the contexts in which these terms were used. The findings showed a continued lack of understanding among the public on the appropriate use of terms that refer to older adults. This paper emphasizes the need to put more effort into eradicating the perpetuation of ageist messages that challenge healthy aging.