Why is a physician working in the Achham district of Nepal worried when he finds that a patient tested positive for HIV, but a physician working in Humla district is worried but also skeptical? Why do we generally not prescribe high dose IV Vitamin C + Thiamine + Hydrocortisone when the combination has shown to provide a substantial mortality benefit in sepsis? Why do we encourage a patient, very rightly so, to get flu shots every year?
When making decisions, we think, we use our knowledge, weigh pros and cons, and make a choice. The variables, whose salient feature is that we barely think of them, are biases and heuristics. We are influenced by various medical journals we read, colleagues we work with, and even movies and advertisements we watch. Another, sometimes lethal feature of these decision influencers is that their influence is inversely proportional to the time we have to make a choice. This becomes relevant in ED, where split-second decisions are the norm.
So how do we make decisions that are backed more by studies and less by our implicit biases? How do we compare two tests that measure the same variable or two vaccines that work against the same infectious agent? There comes the role of statistics. Every physician, especially those making life-saving decisions in a fraction of seconds, should have an intuitive understanding of medical statistics. This will help us make decisions that are backed by our best understanding and understand our limitations.
Achham district of Nepal has the highest prevalence of HIV/AIDS in the country. When the disease’s prevalence is high, the chance that your patient has the disease given the positive result is high. This is the Positive Predictive Value (PPV). The same physician would want to re-run the test on asymptomatic patients if the test was negative. That is because, given the high prevalence, the Negative Predictive Value (NPV) of the test is low. One would also worry about the sensitivity and specificity of the test in question. Although these are properties intrinsic to the test and do not change with the prevalence of a disease in a population, their knowledge adds to the confidence with which we can prescribe a test to a patient.
One way of thinking about sensitivity is: among 100 diseased patients, how many will the test identify? You would want your screening test to have very high sensitivity so that you do not miss any diseased person. Specificity can be thought of as: among 100 healthy patients, how many will the test identify as negative for the disease? If a highly specific test tells you that a patient has a disease, chances are – he does. So the worried physician of Achham district probably used a very sensitive test and followed it with a highly specific test to confirm before talking to the patient about the result.
We encourage all patients to get the flu vaccine every year because of something called the Number Needed to Treat (NNT). It is the number of patients you need to treat to prevent one additional bad outcome e.g., severe flu, death, etc. Every 12 – 37 flu shots prevent one healthy adult from influenza when the vaccine is well-matched. That means the NNT of the flu vaccine is 12 to 37. [1]
The combination of high dose IV vitamin C + Thiamine + Hydrocortisone had shown to provide a substantial mortality benefit in a small retrospective study in 2016. We generally do not prescribe this in sepsis because we do not have a large RCT that supports the claim yet. The GRADE working group suggests a system for grading the quality of evidence. [2] When we say that evidence is graded 1A or 3B, we are commenting on the type, quality, and the number of studies that back the claim. Familiarizing ourselves with the grading system and hierarchy of evidence can be a good start in the world of evidence-based medicine.
References
- Kolber MR, Lau D, Eurich D, Korownyk C. Effectiveness of the trivalent influenza vaccine. Can Fam Physician. 2014;60(1):50.
- Petrisor B, Bhandari M. The hierarchy of evidence: Levels and grades of recommendation. Indian J Orthop. 2007;41(1):11-15. doi:10.4103/0019-5413.30519