The novel coronavirus has killed nearly three million people worldwide, more than 560,000 of them in the United States. Those numbers alone, it seems, might persuade every person who works in healthcare to leap at the opportunity of vaccination.
The truth is more complicated: a survey of roughly 3,500 healthcare workers at Yale Medicine and Yale New Haven Hospital, conducted at the time of FDA approval of the Pfizer-BioNTech vaccine, found that one in six were reluctant to get vaccinated.
“Health care leaders may assume that [healthcare workers] would have little hesitancy to take a vaccine,” writes Vineet Kumar, associate professor at Yale SOM, who conducted the survey with two colleagues at the Yale School of Medicine. “In reality, many [of them] may balk for a host of reasons.”
In their survey, the Yale team found that close to 15% were hesitant to take a vaccine that was readily available. To understand the underlying motivations behind the hesitancy, Kumar used text analytics to mine the open-ended responses. What they found, was a wide range of concerns that varied greatly, from the lack of medium- and long-term follow up to a desire for more information on particular groups, such as pregnant women and minorities. Close to 11% of hesitant respondents said that nothing would make them feel more comfortable with the vaccine.
The question, of course, is what to then do with this information. Kumar and his colleagues began by utilizing a sentiment analysis method typically employed to capture customer feedback, taking the top 15 reasons people expressed reluctance and analyzing the overall sentiment of the words they used: when talking about long-term risks, for example, did respondents use predominantly negative words and phrases? Positive? Was the wording neutral? “People expressing themes for which we see more positive sentiments might be persuadable,” the researchers hypothesized. “Whereas those with highly negative sentiments might be less so.”
Once they knew which groups might be more or less persuadable, the researchers devised hypothetical interventions to increase vaccine uptake. Considering again people with long-term safety concerns, an employer or the state could communicate to this group the latest trial results and provide regular updates on vaccine safety.
Kumar and his colleagues suggest that microscopically mapping concerns this way can help policymakers avoid a misunderstanding of the motives behind the data that might lead to an investment in the “wrong” types of messaging and interventions. The authors take as an example those who are reluctant to get vaccinated because they want to see others get the vaccine first. What, precisely does that mean? Well, respondent A could say this because she wants others who are at higher risk to get the vaccine first; alternatively, respondent B could say this because she first wants to observe others getting the vaccine in case there are unexpected side effects.
While respondent A might respond positively to seeing that high risk individuals have been prioritized and have adequate supplies, respondent B may respond more positively to messaging around positive outcomes or lack of side effects for people similar to them.
Ultimately, though, Kumar and his colleagues note that the specific findings of this study are of less relevance than the broad contours of the analysis—its future potential in the world of medicine. The method, both easy and illuminating, is what they highlight. “The design, administration, and analysis of this survey was completed within 1 week,” they write. Also worth noting, it is relatively inexpensive. Subsequent experimentation to test different interventions or messages and determine their potential resonance with different target populations can also be done relatively quickly. And yet, we find that this type of research, while common in marketing and insights functions, may be underutilized in the health systems.
“The rich insights provided by this approach demonstrate the potential for health systems to learn from consumer marketing firms that routinely apply such survey methods for customer service improvement, as well as unstructured text analysis to learn about performance issues in service industries.”