The Plural of Anecdote is Not Data
Anecdote - a short account of a particular incident or event, especially of an interesting or amusing nature.
Data - a series of observations, measurements, or facts; information
Data is useful. Data represents information from either all relevant cases or a representative sample of all the relative cases. Inferences made from properly acquired data lead to scientific discoveries. Data is scientific. Unfortunately, data tends to be dry. It tends to be boring. We generally do not connect emotionally to 'data'.
Anecdotes are stories. People have been telling stories for as long as there has been speech. The ability to tell stories likely led to tremendous survival advantages for those early human groups that achieved it. Our minds seem adapted to dealing with relatively small groups of 150 people or less. Language and storytelling played an important role in maintaining the groups' wellbeing. A tribe whose members could relate their individual experiences about the location of food or predators would be expected to survive in greater numbers. It is speculated that stories in the form of gossip may have given rise to important survival advantages.
Anecdotes and Data
Large Numbers Trumped by Small Stories
The problems of our society are no longer limited to sample sizes of 150 or less. The problems that medical science tackles are often subtle. Their effects can only be teased out by examining very large numbers of cases. The benefits of blood pressure and cholesterol control may only become apparent when comparing groups of thousands. This means that a doctor may have to treat many people before preventing one catastrophe (see Number Needed to Treat on the Statistics and Risk page). For instance, in a large meta-analysis on the statin cholesterol drugs, it was found that only 27 people at risk needed to be treated to prevent one adverse cardiovascular event. For a doctor who sees about a hundred patients a week, realizing this fact can prevent many disasters.
But it will not be obvious unless one understands the data. The problem with prevention is that when it is done well, nothing noticeable happens.
The same study showed that the number needed to harm one patient with statin therapy was 197. "Treating 1000 patients with a statin would prevent 37 cardiovascular events, and 5 AEs (adverse events) would be observed." The majority of these rare events were minor, such as reversible muscle aches. Serious adverse reactions, such as actual muscle damage, were vanishingly rare. In fact, the number needed to harm with respect to seriously high muscle enzymes was 3,400. The number needed to harm with respect to active muscle destruction was 7,428. So, you only need to treat 27 risky patients to prevent a serious consequence like a heart attack, but you have to treat 7,428 patients to seriously damage one patient's muscles.
We empathize with a good storyteller because our ancestors passed on this trait as a survival skill. Hence, it literally is natural for us to become emotionally moved by a good story.
Now, consider a patient (let's call him 'Joe Alarmist') who is placed on a statin. He had heard on T.V. that statins can cause muscle damage. He is reluctant to take a statin even though his cholesterol is very high and he has multiple risks for heart disease. After 2 weeks on the medication, he feels that his back pain is worse than usual. He then notes that his legs ache after working. Now, according to the data, there is about a 1 in 200 chance that the statin is actually causally related to the perceived increase in muscle pain. But the reporter on T.V. was quite convincing. Joe stops the medication. At a family gathering, he tells the story about how terrible that statin was. "I would never go on that drug again." Joe's family is emotionally connected to him. They make an emotional connection to his story. Being family members, many of them are also at risk for heart disease. It is likely that one day, a doctor will prescribe a statin for one of them. It would then be unlikely that the doctor's cold 'data' would trump Joe's anecdote. (see the Availability Heuristic)
Doctors must understand this. We understand the numbers. We must also understand our patients.
Experience Based Medicine
There have been many proposed 'treatments' that have not shown benefit. For instance, arthroscopic surgery for osteoarthritis of the knee has been practiced for decades. It seemed plausible that it should work. Doctors who perform the procedure receive positive feedback from patients often enough to enforce their belief in the procedure. They hear anecdotes from patients whose pain is a bit better after the operation. However, the data from randomized, controlled trials comparing 'sham' surgery to real surgery failed to show a significant difference in outcomes. In other words, it doesn't work. Does this mean that orthopedic surgeons have abandoned arthroscopic surgery for osteoarthritis of the knee? No. Although the data is acknowledged, efforts are made to rescue the falsified theory.
Similarly, antidepressants are commonly prescribed for stressed-out patients who seem depressed with relatively low depression scores. Placebo controlled trials have falsified the notion that antidepressants are actually doing any good for these patients. Yet, antidepressants are prescribed at alarming rates. Why is this?
We learned in The "Placebo Effect" page that many conditions have symptoms that wax and wane. Other conditions are self-limited and get better despite treatment. Professionals who treat these conditions hear the anecdotes told to them by their patients who happen to feel better after the treatments. Confirmation bias kicks in. Cognitive dissonance prevents them from accepting the real data. To these practitioners, these anecdotes become their data.
As Mark Crislip, M.D. is fond of saying, the three most dangerous words in medicine are, "In my experience...".
We are barraged with anecdotes from 'satisfied customers' on television, radio, newspaper and internet advertisements. Any claim about any product can be supported by anecdotes. Consider the hypothetical 'Product X'.
"Product X cured my joint pain!" exclaims one happy customer. "I have never felt so good!" states another.
Whether Product X does anything at all is irrelevant. If enough people are treated with Product X, at least some people will feel better purely by statistical chance. Others may only claim to feel better due to the secondary gain they receive from the attention of the T.V. camera. The point is, if you want to sell Product X, all you need to do is interview the 'responders' and place their anecdotes in the commercial.
Many people are persuaded by anecdotes in commercials. Such 'data' can be worse than worthless. People who buy Product X may do so in lieu of effective treatments. By making such testimonials public, the marketers exploit the availability heuristic and power of the anecdote.
Patients are people and so are doctors. We are all persuaded on some level by anecdotes because we are a social species. Stories from our fellow tribesmen touch us emotionally, and our emotions often trump our rationality. Doctors who recognize this can avoid some of the pitfalls that occur when conveying scientific conclusions to their patients. Most will not hear the data. Many will listen if the doctor communicates facts like a good storyteller.
John Byrne, M.D.
Sometimes, the anecdote comes from an authority figure. This technique exploits the Appeal to Authority fallacy as well. The authority figure may be a celebrity who is paid to promote Product X. Why this should be of any value is a mystery. Sometimes the authority is an actual doctor in a white coat with a stethoscope. The doctor may actually be sincere with his anecdotes because of his "experience".
References and Links
"Grooming, Gossip, and the Evolution of ... - Harvard University Press." 2010.
"Robin Dunbar: We can only ever have 150 friends at ... - The Guardian." 2010.
"Statins for Heart Disease Prevention (With Known Heart ... - The NNT." 2011.
Silva, Matthew A et al. "Statin-related adverse events: a meta-analysis." Clinical therapeutics 28.1 (2006): 26.