Cause and Effect in Health Science

We have all heard the mantra “correlation does not equal causation.” This helps to ensure that causation is not assigned where it should not be assigned. But that begs the question, what does establish causation?

Causation is notoriously difficult to establish, but that does not mean that it is difficult to recognize. While there are numerous methods used to establish causation, the most common approach is to focus on the three necessary criteria most often used in research studies. 

Correlation Matters

While correlation is not sufficient to establish causation, it is one of the requirements. Without a relationship between the two variables, causation cannot be established. This is why many research studies are designed using methodology which identifies and establishes relationships between two variables (aka correlation), a practice which leads to confusion among those who are not familiar with research methods. When the exposure is present, the outcome should occur. When the exposure is removed, the outcome should cease to occur. This should predictably take place time and time again.

For example, to establish a relationship between inhalation of an essential oil and the reduction of anxiety, the reduction should take place regularly when the oil is inhaled and that reduction should not occur when the oil has not been inhaled. If we wanted to establish a correlation between dietary supplement intake and lead levels within the body, we may conduct a wide range of research study designs which are developed with the goal of ensuring that a relationship exists.

However, while correlation is a necessary component to establish causation, it is not sufficient to establish causation in and of itself. There are a couple of reasons for this. Correlation does not on its own verify that the outcome is caused by the exposure, rather than causing the exposure. Sure, these two variables occur together, but can we be sure that this one is the causative factor? Correlation itself also fails to verify that some other additional variable did not cause the outcome. So while the two factors should have a documented relationship, that is only the first step towards establishing causation.


Timing is Everything 

The second criterion is that the outcome must occur after the potential causative factor. If we wanted to establish that eating ice cream sundaes causes weight gain, the weight gain must have started after the daily ice cream sundae habit. Or if we wanted to say that an essential oil reduced anxiety, the reduction has to begin after the use of the essential oil.

This one typically seems and feels pretty obvious. Of course the outcome would have to occur after the exposure, but there are many claims regarding causation that get this basic step wrong.

Some factors that can lead to false assumptions about a timeline include the natural history of a disease and the pharmacological actions of a substance. Let’s expound on those a bit. Natural history of a disease tells us how long it takes from exposure to the onset of symptoms, how long a disease typically lasts, and what to expect from recovery. That first element, the incubation period (in the case of infectious disease) or the latency phase (for environmental exposures) can cause confusion in the timeline.

We know, for example, that the incubation period for the flu is 24-72 hours. So an exposure would have to take place within that timeframe to be causative. If you had lunch with a friend who has the flu and became sick an hour later, it’s likely that you were the microbial exposure at lunch, and not the other way around. This is the case even though it appears that the lunch occurred before the onset of the flu.

Similarly, active substances such as herbs and essential oils do not always exhibit immediate actions in the body. There is often a lapse of time between the onset of the protocol and when results can be expected.

Timing still does not fully establish causation. To confirm causation, all three criteria must be met.

Eliminate Spurious Correlations

In the words of Sherlock Holmes, “once you eliminate the impossible, whatever remains, no matter how improbable, must be the truth.” To establish causation, the relationship identified between the two variables cannot be explained by the presence of any third variable. Many relationships exist in which neither factor is causal. These are known as spurious correlations. Researchers spend a great deal of time employing advanced statistical techniques and study designs to isolate the variables from any alternate explanations. This ensures that causation is truly due to the exposure or variable in question.

Researchers have to answer the questions: Could this outcome simply be a fluke? Would this outcome have occurred otherwise?  Is there some other variable that could be causing this outcome? Is there an interaction between variables that makes this relationship appear to be causal when it is not? Is this factor the only explanation for the outcome?

We account for other causal factors in both the design of a research study and the analysis of the dataset. There are many issues a researcher must consider during the process. These range from validity issues with the measurement instrument to interactions between variables during the analysis process. This requires investing a substantial amount of time into the development of the research design as evaluating these issues requires gathering the right kind of information from the right people. In fact, the design process is the longest phase of a study, for this reason.

This is where things become difficult when you are looking for a single research study to establish causation. No single research study can accomplish all of the tasks above. Researchers and health educators therefore cannot point to a single study as evidence of a causal factor. These conclusions must be dependent upon the entire body of scientific literature as a whole.

Complex Methods

To ensure that all of the three simplified criteria are met, many researchers use the Bradford Hill criteria, which include nine total criteria which rule out the potential for other possible explanations. There are many more advanced methods of establishing causation, but all procedures ultimately reflect these three key criteria. The way in which researchers achieve this goal is complex and requires multiple different studies and methods.


Get Updates

Sign up for Franklin Health Research Foundation news & updates to get the latest cutting edge research.