Medical error and hospital deaths

recent editorial by Martin Makary and Michael Daniel in The BMJ (aka British Medical Journal)^[1] has claimed that medical error is the third leading cause of death in the United States. The three top killers, they say, are heart disease, cancer, and medical error, with medical errors killing 251,454 people per year (see table below).

Hospital massacre

This is a big attention-grabbing headline, but is it anywhere close to being true? If medical errors really kill 43% as many people as cancer, patients might reasonably conclude that the safest thing is to stay as far away from hospitals as possible.

Even if no mistakes are ever made, all forms of treatment are inherently dangerous. Every time you take an aspirin there's a slight chance that something will go wrong. Everyone knows when you go to a hospital, there is a chance you will not come back out.

What the patient really wants to know is: what are the chances that a medical error will happen, and more importantly, what is the ‘cure to kill ratio’ for their disease. Patients intuitively perform risk-benefit assessments. They need accurate information.

There were 2,596,993 deaths among US residents for all causes in 2013.^[2] So if Makary and Daniel are right, the average person walking down the street has a 251,454 / 2,596,993 or 9.68% chance of dying from a medical error. If you're admitted for cancer your odds of dying from a medical error would be astronomical.

These numbers are suspiciously high. Makary and Daniel cite a physician-rating website called HealthGrades^[3] which cited an Institute of Medicine (IOM) report estimating that 98,000 deaths due to medical errors occur each year, making it the sixth leading cause of death for hospital patients—a small fraction of the population.

Using the same AHRQ criteria, JAMA estimated the figure to be even lower: 32,591 excess deaths annually.^[4]

The HealthGrades website, however, claimed that 81% of the 323,993 hospital deaths among patients who experienced some sort of adverse event, which they call Patient Safety Indicators or PSIs, were “potentially attributable to patient safety incidents.” This works out to 263,864 deaths per year. It is a pretty slippery number. The weasel word ‘potentially’ lets them use whatever number they want, confident that the news media will omit the qualifier.

The website attributed 60% of these PSIs to three causes: failure to rescue, decubitus ulcer, and post-operative sepsis. Failure to rescue is delayed diagnosis, and is by definition always fatal. Decubitus ulcer, or bedsores, accounts for 41.59% of the patient safety incidents. Thus, your doctor taking too long to make a diagnosis and your lying in a hospital bed too long count as medical errors by HealthGrades's criteria.

We can do a sanity check on these numbers by comparing them with the total number of deaths after hospital admissions. In the UK this number is 14,217,640 admissions and 187,337 in-hospital deaths within 30 days of admission, making the total inpatient death rate 1.318%^[5]. Using Makary and Daniel's figure that 0.71% of patient admissions lead to death due to medical error would imply that 53.9% of all deaths in the hospital are due to malpractice. Something is seriously wrong with these numbers.

Their article didn't present any new data. It was just an editorial pushing for their recommendation that medical error be entered as the cause of death on death certificates. This might sound reasonable, but it's a terrible idea.

For one thing, given the vast number of lawyers in this country, imagine the chaos they would cause should they gain access to these records. At present there are only about 17,000 malpractice lawsuits every year. If medical error were listed as the cause of death, that fact alone would be sufficient grounds for a lawsuit. The hospital might as well put up a big flashing Broadway sign saying “please come and sue us!”

There are practical reasons as well. First, medical error would have to be counted under ‘accidental death’ since, after all, it is not deliberate. Secondly, medical error is never a proximate cause of death. If a person dies of cancer because of a misdiagnosis, the cause of death is still cancer.

Third, it discriminates against hospitals. What about all those deaths that occur in, say, post offices? We'd need a category for postal clerk error, auto body repair shop error, bakery error, and so forth. Do you know how many people are killed each year by those slippery waxed floors in office buildings? Neither do I. We need a category for it.

ICD (International Classification of Disease) coding is a bona fide problem for medical researchers: when a patient has multiple diseases, such as cancer and Alzheimer's disease, there's a space for only a single COD, and cancer usually wins. This causes researchers to waste resources chasing down spurious inverse correlations between cancer and neurodegenerative diseases.

But it is unfair to lump bedsores and delayed diagnoses into the same category as medical errors. Of course, bedsores don't grab as much attention as blaming doctors for killing a quarter million people a year, which may explain why they weren't given prominence.

Mortality statistics are not just compiled for fun. People use them extensively to evaluate the efficacy of treatment and to allocate resources. It's important to avoid exaggerating them for any reason, and it's disappointing to see it in a prominent medical journal. Luckily, creating a phony myth to stampede doctors into doing what you want rarely works. But it's not the doctors you have to watch out for.

1. MA Makary and M Daniel (2016). Medical error—the third leading cause of death in the US. BMJ 2016;353:i2139

2. Xu J, Murphy SL, Kochanek KD, Bastian BA (2016). Deaths: Final Data for 2013. National Vital Statistics Reports 64(2). Hyattsville MD: National Center for Health Statistics. PDF

3. HealthGrades quality study: patient safety in American hospitals. 2004. http://www. providersedge.com/ehdocs/ehr_articles/Patient_Safety_in_American_Hospitals-2004.pdf. Link. See also critique at http://www.ncbi.nlm.nih.gov/pubmed/23372115 Link

4. Zhan C, Miller MR. Excess Length of Stay, Charges, and Mortality Attributable to Medical Injuries During Hospitalization. JAMA. 2003;290(14):1868-74.

5. Freemantle N, Richardson M, Wood J, Ray D, Khosla S, Shahian D, Roche WR, Stephens I, Keogh B, Pagano D (2012). Weekend hospitalization and additional risk of death: An analysis of inpatient data. J Royal Soc Med 105(2), 74–84. Link

Last updated May 19, 2016, 9:41 pm.