The Digital Health Lab has been analyzing how existing misinformation risk frameworks apply in the public health context.

The practice of risk management is inherently one of tradeoffs: a process of trying to find the ideal balance of competing priorities. In public health, the risk frameworks that are set in place are significant because the stakes of human health and wellbeing are high. Though triaging risk in health contexts has been in place for millenia, the concepts of "risk assessment" and "risk management" are relatively new, with formal adoption having emerged in the last 4 to 5 decades as the complexity of risk increased and scientific knowledge advanced and became more nuanced.

Today, public health risk frameworks are utilized for a wide range of applications: allocating resources, triaging medical care, and setting policy on exposure levels to contaminants. In hospital emergency departments, for example, triaging is a process in which a patient enters the hospital, they are assessed for specific criteria that would determine the severity of the condition, and that assessment is then used to prioritize how quickly a patient is seen by a health care provider. The criteria for this type of triaging can include a person’s overall diagnosis, their pulse, their respiratory rate, and the presence of bleeding, among other factors. This information is used to prioritize the urgency of responding to particular patients relative to other patients also waiting for care. It is also important to note that, in hospitals, this triaging process is dynamic, so as a patient’s situation or vulnerabilities change, the prioritization of their care might change as well.

This is an example of a risk framework being used in the context of decision-making. Not all risk frameworks, however, are used for decision-making; some are used for analysis, some for monitoring, and others for communication to the public about risk. Each of these elements of risk management are crucial pieces for public health and mitigating harm. Unfortunately, however, there are often great inconsistencies across the approaches used to evaluate risk. These inconsistencies and gaps can result in poor use of resources, lacking health communication, and, as a result, negative health outcomes.

These inconsistencies apply to the realm of health misinformation as well. Attempts to address health misinformation online by platforms, academics, and policymakers have been manifold, yet are accompanied by a range of challenges. There is limited, if any, standardization of important terms that determine content moderation priorities; harm, for example, is a term that serves as a backbone of many content moderation systems." Content is triaged for review based on the estimated level of harm it may cause. But without standardization, the term is vague and open to interpretation. Without such standardization, platforms are at higher risk of yielding both false positives (dangerous for over-moderation) and false negatives (dangerous in the context of harmful information circulating).

Public health risk frameworks serve as a guide, helping researchers and practitioners make some of the most difficult decisions. Risks are identified and labelled in order to create standardization and proactively mitigate health harms. Such frameworks are used for a range of purposes including determining which public health program to carry out based on the population most in need, setting climate policy based on the areas at highest risk for natural disasters, and triaging acute medical care. Risk frameworks are set by a range of entities including international bodies, governmental organizations, industrial groups, or private companies to establish common standards. Significant benefit could be gained from the adoption of such public health risk frameworks by social media companies for the application of triaging and addressing health misinformation online.

Over the last 18 months, Meedan Digital Health Lab’s team of health experts have been reviewing and collating the most relevant risk frameworks for monitoring and moderating misinformation, summarizing how they may be beneficial when applied to addressing specifically health-related misinformation in order to highlight some of the strengths, benefits and tradeoffs of each when it is used to prioritize health misinformation response.

Based on some of the gaps in existing approaches, the Digital Health Lab team developed an additional risk framework that is based on both potential harms and negative impacts for users exposed to health misinformation online. This framework accounts for some of the challenges of health misinformation response that we uncovered in a process evaluation of our Health Desk project, which includes limited ability to access scientific experts in timing that aligns with journalistic and fact-checking turnaround times, and the challenge of knowing what kind of expertise is required for different public health topic areas. This risk framework is based on standardized criteria that account for: a) the type of query and b) the expertise needed to answer a particular public health question.

Fact-checking organizations, journalists, and social media companies have identified the challenge of finding the right approach to triaging health misinformation content and determining if and how experts should be integrated into fact-checks. Frameworks like this one can assist in streamlining risk prioritization of health misinformation as well as lead to more targeted and manageable expert outreach if and when needed. This framework can be adapted to a range of public health contexts and topics, and we are always excited to collaborate with other groups interested in expanding on the framework or developing relevant case studies together.

Below is an example of the Query Risk & Response Criteria framework, developed in collaboration with Health Desk Public Health Research Lead Christin Gilmer, as applied to COVID-19 misinformation, oriented towards our question-fielding process with Health Desk:

Response Level of Risk Responder Criteria Comments High Responded to by an infectious disease expert Reference to severe symptoms (with and without shortness of breath) including blood oxygen level of 95% or below; BP below 90/60; seizures; etc. (not necessarily meeting all the COVID-19 testing requirements/high-risk groups)

Questions about active traveling from a high-risk region to a non-high-risk region (differs by day

Questions about autoimmune conditions and their risk profiles

Questions about interpretation of study findings

Questions about pre-published studies with links

Questions submitted by elected officials

Questions submitted by international NGOs

Questions about clinical conditions, presentations, or risk factors

Questions about biological responses to coronaviruses

Questions about diagnostics for COVID-19 (RT-PCRs, CTs, chest x-rays, etc.) Regarding clinical knowledge and recommendations

Interprets latest study information into understandable language

Establishes and conveys baseline recommendations for major groups Response Level of Risk Responder Criteria Comments High Requires validation from an infectious disease expert Questions regarding hygiene practices by groups or businesses

Presentation of severe symptoms

Questions submitted by local NGO staff members

Questions about quarantine procedures and timelines following potential and confirmed COVID-19 illnesses

Questions about weather patterns, geography, and coronaviruses/upper and lower respiratory diseases

Clinical questions following exposure to patient with presumptive COVID-19 diagnosis

Questions about respirators, ARDS, oxygen treatments, etc.

Questions about international health system comparison and treatment protocols

Questions about food deliveries/pick-ups and infection potential Knowledge of clinical practices

Interpretation of latest CDC and state/national guidelines

Ability to translate newest information into risk provisions for populations Response Level of Risk Responder Criteria Comments Medium Responded by someone with public health expertise Questions regarding contact/exposure with someone who has tested positive for COVID-19 (confirmed and unconfirmed cases)

Questions regarding disinfectants related to COVID-19 (CDC guidelines updated daily)

Questions related to plausible but disputed misinformation (i.e. administration of vitamin C and its potential impacts; zoonotic transmission routes; vaccine development, etc.)

Questions about statistical models

Questions about infection modeling and projections

Questions about viral load

Questions about supply chain and logistics

Questions about testing kit shortages General knowledge of national and international health systems and protocols

Ability to navigate established scientific journals and Ministry/Department of Health studies

Knowledge of regression analysis and public health terminology Response Level of Risk Responder Criteria Comments Medium Requires validation from someone with public health expertise Questions regarding state and regional health restrictions and guidelines

Questions regarding infection prevention on an individual basis

Questions about public health terminology (i.e. case-fatality rate; R0; DALYs; etc.)

Questions regarding public transportation

Questions about over-the-counter treatment recommendations

Questions about breaking news stories related to COVID-19

Questions about conflicting information (i.e. self-isolating vs. testing; new study contradicts prior study’s findings; etc.)

Questions about asymptomatic transmission

Questions about viral shedding (general)

Questions about pediatric patterns and presentations of COVID-19

Questions about needs to leave/end self-isolation/quarantine

Questions about regional, national, and international travel Knowledge of current at-home treatment standards and recommendations

Ability to translate CDC guidelines for infection prevention in professional environments

General understanding of public health prevention theories Response Level of Risk Responder Criteria Comments Low Responded to by an experience fact-checker Questions regarding transmission methods

Questions regarding high-risk groups (major categories of hypertension; cancer; diabetes; cardiovascular disease; chronic respiratory disease)

Questions regarding CDC testing protocols

Question about COVID-19’s origins and spread

Questions about testing locations and procedures (including requirements)

Questions regarding outbreak locations

Questions about school closures

Questions about incubation period (general)

Questions about social distancing/isolating/quarantining

Questions regarding pregnant women (limited case studies, so pre-populated results will suffice)

Questions regarding asthma as a risk (limited case studies, so pre-populated results will suffice)

Questions regarding false, widely disproven information about COVID-19

Questions regarding age risks

Questions regarding global and local incidence rates, mortality rates, and recoveries

Questions regarding CDC guidelines for testing

Questions regarding general presentation of COVID-19-related symptoms (i.e. shortness of breath, fever, cough, fatigue)

Questions regarding differentiation of allergies vs. colds vs. influenzas

Questions regarding Department of Health/Ministry of Health/local emergency contact information or recommendations

Questions regarding infection prevention General ability to use popular terminology to complete research on pre-populated topics and guidance

Knowledge of national and international resources for schools, food, emergency services

Ability to quickly respond to popular, frequently asked topics with developing information

Tags
COVID-19
Footnotes
  1. Online conversations are heavily influenced by news coverage, like the 2022 Supreme Court decision on abortion. The relationship is less clear between big breaking news and specific increases in online misinformation.
  2. The tweets analyzed were a random sample qualitatively coded as “misinformation” or “not misinformation” by two qualitative coders trained in public health and internet studies.
  3. This method used Twitter’s historical search API
  4. The peak was a significant outlier compared to days before it using Grubbs' test for outliers for Chemical Abortion (p<0.2 for the decision; p<0.003 for the leak) and Herbal Abortion (p<0.001 for the decision and leak).
  5. All our searches were case insensitive and could match substrings; so, “revers” matches “reverse”, “reversal”, etc.
References
Authors
Words by

Jenna Sherman, MPH, is a Program Manager for Meedan’s Digital Health Lab. Her work has focused on digital health challenges across information access, maternal incarceration, and discrimination in AI. She has her MPH from the Harvard T.H. Chan School of Public Health in Social and Behavioral Sciences.

Nat Gyenes, MPH, leads Meedan’s Digital Health Lab. She received her masters in public health from the Harvard T. H. Chan School of Public Health, with a focus on equitable access to health information and human rights. She is a lecturer at Harvard University on the topic of health, digital media and human rights.

Megan runs Meedan’s Health Desk initiative as Senior Program Manager. She has worked for news outlets in Canada and the US, and holds a Peabody Award for her work on Netflix’s Patriot Act series. She has a Master of Science from the Columbia Journalism School.

Jenna Sherman
Nat Gyenes
Megan Marrelli
Words by
Organization
Published on
September 16, 2021
April 20, 2022