By Nana Dadzie Ghansah
In the times we live in, one may ask, “Compared to the general population, what is the risk a frontline healthcare worker (HCW) has of contracting COVID-19?”
If one calculated this risk, it would be a ratio:
Risk of contracting COVID-19 in the general population: Risk of contracting COVID-19 as a frontline worker.
The risk in the group with very low risk is usually 1. That of the high-risk group is then calculated.
This ratio is called the “hazards ratio” (HR) and is a part of the branch of statistics called “Survival Analysis.” This branch entails a set of statistical methods used to investigate the time and probability it takes for an event of interest to occur (the hazard function) as well as the time and probability of the event not occurring (survivor function). In this case, the event would be “Contracting COVID-19.” Since we are interested in whether it occurs, we will structure a study that would allow us to calculate the hazards ratio.
Besides stating the hazard ratio, a value called the “Confidence Interval — CI” is also stated. The hazard ratio is often stated this way: Adjusted Hazard Ratio 11·61, 95% CI 10·93–12·33. A CI of 95% means that in 95% of the cohort, the calculated hazard ratios were between 10.93 and 12.33. The CI is meant to estimate the spread of the hazard ratio in the general population, and 95% means the sample approximates the general population.
The best study for such a question is a prospective cohort study. In such a study, participants are grouped according to their risk for the event e.g., those at risk of seeing the event occur and those with a much lesser risk of seeing the event. They are followed over a set period, after which the number of events is compared to get the hazards ration.
To get the hazard ratio, when only one risk factor (univariate) is involved, the statistical methods used are Kaplan-Meier curves and log-rank tests. However, when the results are affected by several factors (multivariate), the Cox proportional hazards regression analysis is used. (Note that “regression” here means “teasing out.” Regression analysis allows one to tease out the rear factors that impact an event to of the many).