We want COVID vaccines to succeed and get on with our lives. And, reports of “90%+ effective” have primed the public to expect miracles. But, vaccine acceptability will be contingent on a yet nebulous profile of safety. Therefore, we must proactively evaluate side effect reports and optimize vetted information for dissemination. We are not throwin’ away our shot.
To factcheck vaccine side effect rumors and results in real-time. To deliver vetted information to the public effectively.
We are soliciting interest and funding to rapidly build a lightweight science-as-a-service tool for Jan-Feb deployment. The rumor control website will require scientific expertise, message crafting will require scicomms exerts, and the chatbot will require software developers. Please WhatsApp (+1-919-260-3808) or email [email protected] if interested.
Conventional wisdom about vaccine misinformation has disproportionately focused on the lurid fringe. Scientists and journalists studying viral misinformation suggest a different strategy. Our target audience is the middle-of-the-pack, to prevent progression to the fringe at a specific moment of vulnerability. It is the interaction between undecided populations and anti-vaccination clusters that leads to wild misinformation spread.
We have identified two specific moments to intervene.
- At vaccination: What to expect - Overburdened medical service providers will have to set expectations in person. Even though these expectations are required and repeated disclosures in COVID vaccine RCTs, some patients do not recall having been told what to expect. In higher throughput primary care, this problem will be worse. More tools and resources will be dedicated to this obvious point of message failure as the vaccine is deployed, and those efforts are welcomed.
- Side effect onset: What is normal - Less recognized are the cognitively vulnerable "poor-me" hours as side effects manifest. The onset of side effects is likely to prompt digital sleuthing and reaching out to trusted family and friends. A Google search is predictable, and our efforts are to pre-empt entry into conspiracy-laden networks. Psychology of conspiracy theories reveals that people are likely to turn to conspiracy theories when they are anxious and feel powerless. Therefore, our fundamental proposition is to provide reliable choice-oriented information and comfort suggestions at side effect onset.
Read more about our Misinformation Strategy
Proposed chatbot interaction
Mass vaccinations, common side effects, and ubiquitous mobile phones mean millions will be searching for reliable information on handhelds. In this 1 minute demo video we show how to answer three essential questions regarding side effects that patients find most important.
- Is it vaccine-related?
- How long will it last?
- What should I do now?
The three essential questions were identified from our experience deploying mobile apps, monitoring social media, and most importantly, listening to patients. Platform development is oriented towards SMS because the technology is low threshold and widespread. WhatsApp integration and an embeddable web-based chatbot API are also envisioned. Non-minor and ambiguous events will be referred to clincians and pharmacists, with links for official adverse event reporting.
Jump to Regulatory and Legal Considerations and Comfort Suggestions
Jump to funding options and Project Team
Anonymized user-generated text can be mined for shifts in public sentiment and provide specific real-time targets for countering misinformation. Private text exchanges with a bot may provide different insight than other data sources. Unlike search engine trends and social media monitoring, we will be able to query the time between vaccination and onset of symptoms. Time-to-onset is the single most important element to determining if a side effect was caused by the vaccine. But, the real goal of this project is to provide an avenue for meaningful information (pharmacovigilance is not the primary intent). Data privacy and retention policies will favor public trust over research interests.
From years of deploying patient-centric health apps, we proceed with 4 central insights.
1. The "worried well" need encouragement to get second doses
Offering encouragement to those with minor expected side effects will be essential to overcome hesitancy for regimen completion. Triaging the worried well also aids the healthcare system from being overrun.
2. Syndication speeds uptake
To rise above pandemic media saturation, a data tool must be easily propogated while fulfilling a fundamental public need, as the success of COVID tracker APIs demonstrates.
3. Patients expect instant feedback
As described in this video, patients with side effects expect instant and compassionate feedback.
4. Patients are reluctant to download single-use reporting apps
Based on our experience in 4 countries, mobile apps for side effect reporting have limited uptake. The forthcoming CDC mobile app (V-SAFE) focuses on collecting information, while patients expect to receive information. Referrals to V-SAFE will be made via the chat for non-minor events. In this way, the chatbot fills unmet patient needs and serves as an entry point for official data collection.
Our team has experience in evaluating side effects in real-time from the public. The goal is to evaluate side effect rumors and study results in real-time. We will monitor social and news media, scientific literature, pharmaceutical industry filings, and government reports to identify emerging concerns about vaccine side effects. We will scientifically assess each new side effect ("adverse event") using data and medical knowledge to determine causality. The team will consist of experts in pharmacovigilance, pharmacoepidemiology, pharmacology, and related disciplines.
Our team has experience building and scaling NLP-based apps. We will develop a chatbot allowing users to describe symptoms and learn if they could be vaccine-related. To operationalize, three types of tools are envisioned: rumor curation tables and interface, core chatbot functionality, and real-time NLP layers to respond to user input. Bolded elements represent the anticipated MVP pathway.
Anticipated build components:
- Scientist-facing backend interface to log and ajudicate rumors and results, with Git-oriented off-the-shelf tools for version control
- Master tables for each vaccine with known/suspected/unlikely side effects and comfort suggestions, with syndication capability (REST API, dynamic GitHub table)
- Decision tree for actions
- A blog-style public-facing Rumor Control website (example)
- Real-time NLP tool (integrated with chatbot API/SDK) to interpret user responses and auto-generate text
- To support NLP, a dynamic dictionary to translate vernacular symptoms to medical ontology (allowing comparison to scientific reports)
- Python and Django chatbot for message testing and embedding in third-party websites
- Twilio deployment for SMS text and WhatsApp
- Data export capability for retrospective/static analysis
Uptake by end users will require concerted promotion. Ideally, materials will be provided to patients at the point of vaccination. All vaccination venues, pharmacists, and clinic offices/receptionists are potential partners. Promotion plans will be more fully developed as a team.
Thus far this is a collaboration between the University of North Carolina at Chapel Hill and Rhizomik Tech, LLC. We are open to other collaborators. Funding could be directed at the university (scientific strategy) or the small business (tech strategy).
Traditional Development
We are soliciting funding and suggestions to make this vision a reality. We would like to develop a rapid MVP by January 2021. Using standard development process we will need a software developer ($25-$43k), purchased computing resources ($30k), and a scientific project manager ($40k). Ideally, 2 research assistants ($21k each) and more programmer resources would be available, either funded, crowdsourced, or in-kind. These resources are anticipated to cover the initial build and 6 months of moderate uptake.
Crowdsourced Development
Capitalizing on boredom during a socially distanced holiday and altruistic impulse, an alternate approach is to rely on crowdsourcing for much of the devleopment as an open source project. For example, scientific tasks could be distributed to separate teams: searching for rumors on TikTok, monitoring and extracting results from scientific reports, message testing, creating the vernacular-to-medical language dictionary, assesing causality and writing blog posts, analyzing user-generated text. Similarly, once an architecture has been established, tech development tasks could be crowdsourced too. However, for this model to be successful, we still need a lead software designer and a lead scientific project manager to oversee task deployment. And, we would need to excite and motivate hundreds to yield a core dozen who would do most of the work. Crowdfunding could also be envisioned. Both crowdsorucing and crowdfunding would be heavily reliant on institutional support for promotion.
Scaling Costs
In the event of high volume uptake (more than 100,000 sessions), text messaging via Twilio can be scaled with volume discounts, but would require additional funding. Each chatbot session costs about $0.12 (assuming 16 send/receive messages) in the United States, and is available for paralell deployment in dozens of other countries.
A recent headline in Science summarizes the need.
Professional societies and industry experts have warned of this challege for months. Misinformation and humor are already spreading on TikTok about COVID vaccines.
- USA Today, Dec 2, 2020: Will there be side effects from a COVID-19 vaccine? When can you get it? We answer your vaccine questions
- CNN, Dec 2, 2020: What it feels like to get an mRNA coronavirus vaccine
- Washington Post, Nov 17, 2020: The next big disinformation fight is coming – over coronavirus vaccines
- WebMD, Nov 24, 2020: Experts: COVID Vaccine May Cause Side Effects
- BBC News, Nov 15, 2020: Vaccine rumours debunked: Microchips, 'altered DNA' and more
- NBC News, Nov 12, 2020: Covid-19 vaccines may have potentially unpleasant side effects
"The actual injection felt, at first, just like a flu shot, which is basically just a little pinch in the side of your arm," Batalvi said. "Once I left the hospital, that evening, the stiffness got a little bit worse. It was definitely manageable, but you kind of don't really feel like moving your arm too far above your shoulder. But the side effects are pretty localized. I mean, it's just in the muscle in your arm. And that's about it. It doesn't really affect anything else and you feel fine." That was after the first dose. But the second dose was different. "I actually had some pretty significant symptoms after I got the second dose. Once I got the second dose, I was fine while I was in the hospital. But that evening was rough. I mean, I developed a low-grade fever, and fatigue and chills," Batalvi said. He said he was out for that day and evening, but he "felt ready to go by the next morning." He said he called the study doctors to let them know about his symptoms. They weren't alarmed and told him he shouldn't be either.
CNN - Dec 2, 2020
This summer, computational biologist Luke Hutchison volunteered for a trial of Moderna's COVID-19 vaccine. But after the second injection, his arm swelled up to the size of a “goose egg,” Hutchison says. He can't be sure he got the vaccine and not a placebo, but within a few hours, Hutchison, who was healthy and 43, was beset by bone and muscle aches and a 38.9°C fever. “I started shaking. I had cold and hot rushes,” he says. “I was sitting by the phone all night long thinking: ‘Should I call 911?’” Hutchison's symptoms resolved after 12 hours. But, he says, “Nobody prepared me for the severity of this.”
Science - Nov 27, 2020
Despite the extensive information I had on the research process and vaccine, on a personal level I did not get the message that I should anticipate a reactogenic response. I was scared when I saw that I had a fever, and my gut reaction after months of scrutinizing myself for all the possible COVID-19 symptoms was: Do I have COVID-19? I texted a few friends about my experience, and their response was the same: “Wait, does this mean you have COVID-19? Are you contagious?” I assured them I did not and was not, but every physician and nurse in the US needs to be prepared to have a conversation about adverse effects with patients. I can already see the wrong message about the COVID-19 vaccine going viral.
JAMA Internal Medicine - December 7, 2020
We will not be providing treatment recommendations to users. Instead, we are delivering vetted side effect information in a format that is meaningful, for example augmenting with how long symptoms might last or age-specific profiles. For the most common, easily managed, minor side effects like chills and soreness, providing general comfort suggestions (ice, arm movement exercises, standard OTC pain relievers, etc.) is entirely reasonable. We may not have to create this content ourselves, since news media are already making these suggestions. To mitigate risks we can focus on non-therapeutic comfort suggestions at launch. Providing information that patients expect is inherent to the value proposition. The general posture is to refer anything ambiguous or complicated to clinicians/pharmacists, as with PoisonHelp.org. We recognize that pharma and government authorities have traditionally shied away from comfort suggestions, but the exceptional conditions of the pandemic call for patient-centric strategies that are bolder. Read more about regulatory and legal considerations.
- Project team
- Patient perspective video we compiled on from advocates discussing COVID vaccine adverse event reporting
- Misinformation and humor about COVID vaccines
- Example rumor dispelling
- Regulatory and legal considerations
fin.