Study design

A cross-sectional retrospective observational review of comments made by the public to specific social media recruitment posts (n = 2) was undertaken. This secondary data approach was chosen to capture the public perceptions as it allowed for unfiltered opinions. It also avoided social desirability biases such as the Hawthorne effect where participants may change their behaviour, or censor their opinions, if they are aware they are being studied [22,23]. The data collection and analysis method complied with the terms and conditions of the Facebook data source.

Data collection

The comments were manually extracted on a Microsoft Excel spreadsheet from two Facebook posts advertising a research opportunity for a study on psychological support for care staff during the pandemic. Data were stored on the Northumbria University secure server and only the research team had access to it. Post 1 (see Fig 1) was a sponsored post from Northumbria University’s official Facebook page and set to reach adults in the UK. The recruitment advert in this post was hosted on the university’s Facebook page and included university branding and reference to the funding organisation. Post 2 (see Fig 2) was advertised from a Facebook page created by the author (MS) but not affiliated with Northumbria University, which advertised research relating to the impact of COVID-19 on mental health and quality of life. Both posts were live at the same time between June and October 2021. MS had author access and therefore an opportunity to reply to questions on the Facebook Post 2, while this was not the case with the Facebook Post 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. University sponsored recruitment post.

https://doi.org/10.1371/journal.pone.0303609.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Facebook group recruitment post.

https://doi.org/10.1371/journal.pone.0303609.g002

Ethical considerations

Ethical approval was granted from the Faculty of Health and Life Sciences at Northumbria University REF:32662NU and consent was waived.

Analysis

Comments were qualitatively analysed using the 6-steps of thematic analysis defined by Braun and Clarke [24]. All members of the team familiarised themselves with the data. During phase two and three, comments were independently coded by MS, CM and MB with initial codes inductively generated and then grouped into themes via collaborative discussion. Subsequently, phase four and five followed, where all the research team met and reviewed if the initial themes were supported by the code extracts and the overall data set. Themes were defined and named with the final names of the key themes and sub-themes agreed by consensus. A report (phase 6) on the findings was agreed by all team members.

The quantitative analysis included exploration of word usage. Several key word categories were identified from the findings of the thematic analysis. Use of these word categories were counted and recorded. Singular and plural pronoun use were quantified to gain insight into perceptions of personal responsibility in the context of the pandemic, displacement of responsibility and locus of control. Table 1 displays word categories and word bank. Finally, the number of words used in each comment was recorded.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Word categories.

https://doi.org/10.1371/journal.pone.0303609.t001

A series of t-tests were conducted to compare use of each word category and comment length between posts 1 and 2. Correlations were conducted to explore relationships between the word categories in each post.

Themes

Three themes were identified from the data: i) support, ii) mistrust, and iii) blame.

Word count analysis

Table 2 displays descriptive statistics and t-test results for the word count analysis.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. T-test analysis for word counts.

https://doi.org/10.1371/journal.pone.0303609.t002

Comments on post 2 (M = 1.33, SD = 2.14) featured significantly more first-person singular pronouns than the comments on post 1 (M = 0.20, SD = 0.65), t(197) = -5.60, p < .001.

Post 2 comments also included significantly more negative emotion words (Post 2 M = 0.45, SD = 0.73, Post 1 M = 0.20, SD = 0.44), t(197) = -3.03, p < .001, words associated with support (Post 2 M = 0.21, SD = 0.45, Post 1 M = 0.06, SD = 0.24), t(197) = -3.13, p < .001, and words associated with choice (Post 2 M = 0.32, SD = 0.64, Post 1 M = 0.09, SD = 0.34) t(197) = -3.31, p < .001, than comments on post 1.

Correlations.

Table 3 displays correlations between word categories in posts 1 and 2.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Word category correlations.

https://doi.org/10.1371/journal.pone.0303609.t003

Use of positive emotion words was positively correlated with use of first-person singular pronouns in both posts (Post 1 r = .19, p = .03, Post 2 r = .31, p = .01), first-person plural pronouns in post 2 (r = .31, p = .01), and second-person pronouns in post 1 (r = .22, p < .001).

Use of negative emotion words was positively correlated with use of first-person singular pronouns (r = .36, p < .001), first-person plural pronouns (r = .35, p < .001), and third-person plural pronouns (r = .36, p < .001) in post 2. Negative emotion word use also positively correlated as follows: with words associated with support (r = .27, p = .03); the COVID-19 pandemic (r = .38, p < .001) in post 2; and words associated with vaccines in post 1 (r = .18, p = .04).

Use of words associated with vaccines positively correlated with first-person singular pronouns in both posts (post 1 r = .32, p < .001, post 2 r = .44, p < .001), and first-person plural pronouns (r = .33, p < .001), and second-person pronouns (r = .47, p = .03) in post 2. Vaccine words also positively correlated with words associated with choice in post 2 (r = .48, p < .001) and the COVID-19 pandemic in post 1 (r = .44, p < .001).

Words associated with support positively correlated with first-person singular pronouns (r = .29, p = .02) in post 2. Choice words were positively correlated with first-person singular pronouns (r = .26, p = .04) and second-person pronouns (r = .30, p = .01) in post 2. Words associated with the COVID-19 pandemic were also positively correlated with first-person singular pronouns (r = .18, p = .04) in post 1.

Support

Our findings underscore the crucial role of support, encompassing both psychological and logistical assistance, in mitigating the adverse effects of the pandemic on care home staff’s well-being. The Health Belief Model provides a valuable framework for interpreting these findings, particularly regarding perceptions of susceptibility, severity, and perceived benefits of seeking support. Perceived susceptibility to negative outcomes, such as mental health challenges, may motivate individuals to seek support services. However, our study suggests that care home staff received little psychological support during the pandemic, aligning with previous research demonstrating low uptake of well-being interventions among HCWs [11,25]. This perceived lack of support may exacerbate feelings of vulnerability and contribute to increased depression, anxiety, and insomnia [26–28]. The positive relationship between words associated with support and negative emotive words in our word count analysis further underscores the importance of addressing support deficits to alleviate psychological distress among care home staff. Clearly, care home staff’s physiological and safety needs would require to be met before they could participate in research. Meeting these fundamental needs are important considerations when recruiting to care home studies. Researchers must ensure physical comfort is obtained from adequate breaks, minimal participant burden and creating an environment for psychological safety i.e., confidentiality etc. Importantly, ensuring care home staff have an opportunity to participate in research has important ethical principle i.e. fairness and equity of opportunity.

Mistrust

Mistrust was a prevalent theme throughout the data, particularly in relation to the effectiveness and risks of the COVID-19 vaccine, authorities, and researchers. Public and care home staff mistrust is important as it can affect all levels of trust in societal and social systems [29]. For example, mistrust in authorities can also become mistrust in researchers. Trust depends on people’s previous experiences, with negative experiences being more powerful at building mistrust, than positive experiences building trust [30]. Therefore, resources need to be directed towards creating lots of positive experiences with care home workers to build trust, and subsequently increase their likelihood of engaging with research projects. Additionally, how trust is perceived is dependent upon individuals’ perceptions of what is fair [31]. Many respondents in this study felt that it was unfair to enforce vaccinations for care home staff, as this removed individual autonomy. Some posts even suggested that care home staff should be vaccinated in retribution for allowing elderly residents to be vaccinated. To understand care home workers values there is a need for researchers to build relationships with them.

Additionally, there were intense views around the efficacy and safety of the COVID-19 vaccine with some reporting serious adverse medication events from people known to them. Of course, these could be extreme views by those who have had recent negative experiences, otherwise known as proximity fallacy. These posts contained a discussion about vaccine and related choices and consequences, despite no mention of vaccine in the recruitment post. However, the timing of the recruitment post coincided with public health and government discussions and activities around the vaccine in general and the compulsory vaccination for HCWs, placing the responsibility to protect vulnerable people (care home community) on HCWs (individual level). Thus, it is not surprising to see vaccine and choice discussion on the posts.

Post 2 comments featured significantly more choice words and first-person singular pronouns than post 1. In addition, post 2 had stronger associations between vaccine words and choice words or positive emotion or singular pronouns in comparison to post 1. It is possible to assume that the people commenting on post 2 felt they had more choice or more choice should be available to people regarding the vaccine, but also could indicate a higher internal locus of control where they have a personal responsibility around limiting the transmission of the disease. This fits with the theory of locus of control which looks at the extent to which people believe they control their lives (internal locus) or if others control their life, such as the influence and power of government or other forces (external locus) [32]. This could be particularly important in the context of the global pandemic and the mental health of the public and HCWs. Having a higher external locus can exacerbate negative feelings, feelings of threat and having depressive thoughts in comparison to having a higher internal locus where people feel more in control and that their actions matter [33,34]. What this means for barriers to recruitment is that those that feel they have less control over their life are more likely to have mistrust towards government and researchers and are therefore less likely to participate in research. Thus, as mentioned before, doing work that can increase trust in researchers by building relationships with care homes, can change the perception HCWs have towards researchers.

Findings relating to mistrust of researchers and perceptions of government sponsorship highlight the need for better collaboration and sharing of power between researchers and HCWs. All stages of the research process would benefit from co-design, particularly preparation of recruitment and data collection materials to ensure they appeal to and are suitable for the population. Researchers should be aware of the potential impact of including institutional and funders’ information on recruitment posts. Discussion of mistrust towards researchers was most prominent on post Facebook 1. Given the clear need for more research to be conducted with HCWs, it is important that as researchers we work to foster relationships with this population–research champions to promote collaboration with care and nursing homes. This also highlights the importance of researcher responses to comments–where recruitment posts attract some negative comments this may ‘taint’ the post and put off people who might otherwise have considered taking part. Researchers engaging with the comments to provide and clarify information about the study (as would happen face to face) might help to limit this and increase the credibility of the post.

Blame

Finally, a preponderance of negative emotion words was apparent and attribution of blame in relation to care home staff was evident within our findings. Applying the Health Belief Model, individuals’ perceptions of blame may reflect their beliefs about the causes and controllability of adverse outcomes. There was a larger range of negative emotion words than positive emotion words. Negative emotion words were more likely to be higher valency e.g., depressed, trauma, suffering, than the positive emotion words which were typically lower valence e.g., funny, glad, happy. This suggests negative emotions experienced and expressed were more intense and warrants the need for development and accessibility of public health interventions that can address these intense negative emotions. Use of negative emotions words in the comments on post 1 were typically used in reference to others e.g., care home residents being lonely, whereas in post 2 they were more often used in reference to care home staff being burnt out and exhausted. This further demonstrates the dominant external locus of control among the participants, but also highlights the general worry about the vulnerable and the underappreciated members of the public and dissatisfaction about how the challenges of the pandemic were handled.

Strengths and limitations

It is important to acknowledge the strengths and limitations of using social media such as Facebook data as a research source. One of the strengths is that is a quick and easy way of accessing public’s views on specific topics without needing an explicit consent since the information is in public domain [35]. Further, as comments were generated naturalistically, we can assume that opinions expressed are uncensored and not subject to the demand characteristics and social desirability bias often observed within the context of experimental designs. However, this method of data collection also presents notable limitations. Firstly, demographic information could not be collected for those who contributed to the comments. It was unclear from some posts whether these were a lay public perspective or specifically experiences of care home workers. While it can be inferred that contributors to the comments on post 2 were more likely to be HCWs due to the content of the comments and the target audience of the page, similar inferences cannot be made for post 1. Secondly, difficulties arose with determining which were genuine opinions and experiences and which comments were posted to be deliberately controversial or confrontational (particularly in the context of a sensitive and potentially divisive topic such as the COVID-19 pandemic). As such, all comments were extracted and treated as valid data points, however this may mean that excessively negative, anti-vaccine or anti-pandemic opinions were over-represented in the data.

Also, word count analysis does not consider the semantics of the word in the context of the sentence [36]. Many of the positive emotive words were used to convey sarcasm or irony. Contributors to social media discussions may represent a self-selected group with particular views, potentially skewing the data towards certain sentiments and limiting the generalisability of the findings [37]. Finally, we want to acknowledge that the method used is insufficient to address aim 1 entirely and future qualitative studies should be undertaken to explore this further.

The negative media portrayal of care homes during the COVID-19 pandemic can be attributed to several interconnected factors [38]. Heightened mortality rates within care homes, often resulting from the virulent nature of the virus, have become focal points in media coverage, creating a perception of inadequacy in crisis management. Reports on insufficient protective measures, including shortages of essential equipment, and staffing challenges, such as shortages and overwork, contribute to the negative image [39]. Communication issues, both within care homes and between stakeholders, are highlighted, further impacting the perception of care home efficacy. Public and governmental responses, including policies and financial support, are scrutinized, adding another layer to the narrative. Stigmatization of care homes as COVID-19 hotspots and potential sensationalism by the media contribute to a nuanced and often negative portrayal. This complex interplay of factors underscores the multifaceted nature of the media’s influence on public perception during the pandemic.