Table 3.

Examples of sources that comprise health-related information

Source of Evidence	Strengths	Limitations	Example
Materials studies in the laboratory	Can determine the physical properties of something	May not incorporate various human behavioral and social factors	Studies that show how many different particles face masks can filter out but do not account how the person may wear the mask
Laboratory molecular and cellular studies	Isolate specific mechanisms	Situation may be very different when molecules or cells are viewed in isolation vs. part of the greater system.	Laboratory studies show how well a virus can enter a cell, but the virus has to travel down through the respiratory tract to reach the cell
Laboratory animal studies	May be done when human studies are not yet possible	Results may not necessarily be applicable to humans	Ferret studies that have shown that humidity may affect transmission of respiratory viruses does not necessarily apply to humans
Anecdotes	Can raise questions for further exploration	Typically, does not qualify as scientific evidence; Not verifiable or validated	Stories of a person having an adverse event after a vaccine may not be verifiable
Case studies	Can raise questions for further exploration	Unclear how unusual the findings may be.	Case studies showed that a majority of people infected with SARS-CoV-2 developed some sign of myocarditis but is it actually not that common
Observational cohort studies	Evaluate what’s happening in the real world under real world conditions; Can show trends, patterns, and associations	Cannot demonstrate cause and effect; May overlook many factors and complexities; can oversimplify situations	Observational cohort studies suggested that Vitamin D intake may be associated with lower Covid-19 severity but there were many potential confounding factors such as geographic location, use of other COVID-19 precautions, and ability to social distance
Randomized controlled trial (RCT)	Good at evaluating medications and treatments for an individual	Not as effective at evaluating more complex interventions that are affected by many factors external to an individual; Findings may not hold in the real world	An RCT of an antiviral can show the medication may affect the duration and severity of symptoms; however, an RCT of face mask use may not be able to measure and control for factors that affect efficacy such as where the person moved, how the person wore the mask, and how many others were wearing face masks
Data-driven statistical and machine learning studies	Can help identify trends, patterns, and associations in historic data that may not be obvious; Can help extrapolate these trends; Can help identify areas that need further exploration	Cannot determine cause-and-effect relationships; can oversimplify situations; Often assumes that historical trends will hold; The past does not necessarily predict the future	Studies in early 2020 tried to predict the subsequent course of the pandemic and the resulting mortality, however, conditions changed in the latter half of 2020, and again in 2021 and 2022.
Mechanistic computer modeling studies	Able to pull together, integrate, and synthesize data and evidence from different sources; Attempts to represent the underlying mechanisms; Can run hypothetical scenarios; Can help determine what may happen should circumstances change; May forecast the future; Can help evaluate multiple layered interventions	Can be substantial variation in types and quality of models; The construction of the model may not always be readily understood; Dependent on how well the actual mechanisms are represented; Ability to replicate real-world situations can vary based on the construction of the model and accompanying assumptions	Computer models from 2020 showed the anticipated course of the pandemic and impact of Covid-19 vaccines but did not always account for emergence of new variants