At the end of part two of this series, two charts on deaths in the United States (reproduced below) were presented and I asked the question “What do they tell you?”
These charts present data (statistics) from two sources for the year 2013, which is not indicated on the graphics. The value for heart disease is 611,105 and the value for cancer is 584,881, both from the U.S. Center for Disease Control and Prevention. Not shown is the statistic for respiratory disease deaths of 149,205, same source. The third value shown was developed by Martin Makary, MD and MPH (master of public health) and Michael Daniel, MD of the Johns Hopkins University School of Medicine. They attributed 251,454 deaths to medical error. Assuming correctness, medical errors would be the third leading cause of deaths in the U.S. in 2013.
According to Makary, in 1949 the U.S. adopted the use of the International Classification of Diseases, the tool used for epidemiology, health management and clerical purposes. Medical error is not a cause of death in that classification system. Managed by the World Health Organization, its uses, among others, include observing reimbursements including billing codes to tally death causes and resource allocations trends. In 1948, morbidity was included. According to Makary, “at that time, it was under recognized that diagnostic errors, medical mistakes, and the absence of safety nets could result in someone’s death…, medical errors were unintentionally excluded from health statistics.”
Return to the charts. The top chart shows the full range of the statistics from zero to 600,000 and shows that medical error is significant. The second chart truncates the measurement axis, thereby making medical error appear less significant, at least visually. And if you wanted to show respiratory disease in the second chart, its value would fall below the implied starting height of 200,000. The chart does not say that is the starting point, but the perception implies that heart disease and cancer are many times more significant than medical error.
Now look at the data (statistics) from a different perspective, adding a few more categories. We will use percentage of the U.S. population that died from various causes in 2013 (the “official” U.S. population in 2013 was 316,200,000). Then you can figure your odds, adjusting for age dependence (correlation) for heart disease and cancer deaths and so forth.
Cause Percent of U.S. Population
Heart Disease 0.19327%
Medical Error 0.07952%
Respiratory Disease 0.04719%
Guns, Including 0.013014%
Guns, no Suicide 0.003454%
Gun Suicides 0.006697%
Auto Fatalities 0.010403%
Is that really helpful in assessing the problem of where to collectively invest in potential solutions? No, it is not. The numbers are not in the comprehension domain of most people, and most includes me. Can I comprehend a percentage to five or six decimal places? What is the meaningful difference between 0.012345 percent and 0.012346 percent? Do a crude calculation and the odds of being killed by a gun are 1 in 28,212. That’s a little better, but still hard to grasp other than it is not imminent. How much of an improvement would it be of the odds went to 1 in 28,213? For auto fatalities, the odds are 1 in 9,613. But I still drive, after all many factors affect your personal odds just like age has a factor in heart disease death.
So how do we make decisions when presented the statistics? More importantly, how do decision makers (such as government officials, elected and unelected) make decisions affecting large numbers of people? Perhaps the graphic mode is easier to understand, but is the graphic unbiased? In the charts shown above, which convey the “true” information, are there subtle hints carried in the graphic structure? Is the glass half empty or half full? Are statistics (and probabilities therewith associated) for investing, making money, providing governance, managing the budget, making choices, or something else?