More than 90% of sepsis deaths are unpreventable, study concludes
"Nihilist" is one of the harshest insults that can be lobbed at a physician. Even while one knows intellectually that every patient can't be saved, it's considered odious to openly acknowledge that actuarial reality. Accepting the truth that some patients will inevitably die despite our best efforts is seen by some as the threshold of a slippery slope that leads to defeatism, decreased care quality, and death panels. It's better -- some would say essential -- to eschew probabilism and believe every life is savable, with ideal care.
The history of medicine generally supports such optimism. For myocardial infarction, traumatic injuries, HIV and innumerable other infectious diseases, fatalism yielded to medicine's triumphant progress, as improved treatments now routinely commute what were once death sentences for millions of people.
Sepsis, too, has had improved survival over the past two decades. Academics, government regulators, guideline-writers and patient-advocacy organizations argue that with more funding, research, and enforced care implementation, sepsis could soon join that pantheon of well-controlled, if not defeated diseases.
But a new study in JAMA throws a wrench in the optimistic narrative, concluding that for the vast majority of patients who die with sepsis, their deaths are most likely inevitable even with today's best treatments.
Authors reviewed the medical records of 568 patients who died after admission to six hospitals (at Brigham and Women's, Barnes-Jewish, and Duke). About half (n=300) had sepsis (possibly or definitely), which was by far the most common cause of death, and was considered to have directly caused the deaths of 198.
Could these patients have been saved? Among the 300 who died with sepsis, 121 were considered to have hospice-qualifying conditions -- metastatic cancer, dementia, strokes with severe disability, etc. Most of the rest also had a high burden of chronic disease (cancer, heart disease, dementia, etc). Average age was 71 years.
After detailed review of the medical records, only 3.7% of patients' deaths from sepsis were considered moderately or definitely preventable, based on apparent deviations from reasonable care. Reviewers agreed that 88% of the sepsis deaths were unpreventable, although with only moderate agreement using the Likert scale.
An editorialist pointed out that 23% of the patients did receive suboptimal care, most often delays in antibiotics. Further, the rate of suboptimal sepsis care at these excellent tertiary referral hospitals was lower than has been observed at other centers, so the study may underestimate the number of preventable sepsis deaths through improved care nationally.
The study suffers from the limitation of subjectivity and potential bias among the reviewers, who might tend to underestimate suboptimal care -- or be unable to detect it by medical record review.
Those kinds of biases can be eliminated by using "dumb" tools like the IHI's (requiring only low-skilled data clerks, or bots) to classify deaths as preventable. For their part, those tools lead to laughable overestimates by systematically blaming inevitable deaths on minor care lapses with no plausible means of causation. (These tools' flawed output are the source of the "100,000 deaths from medical errors" falsehood that gets repeated breathlessly and credulously by major news outlets.)
Years back, I performed a review of several dozen patients who died on an internal medicine service, over a year's time at a large, busy urban hospital, using methods like these investigators'. As in this study, most of the patients died from sepsis, and most had arrived to the hospital with metastatic cancer, extreme age or frailty, dementia, irreversible aspiration syndromes, end-stage congestive heart failure, etc. My bias was strongly tilted toward finding errors, because "quality improvement" was the rubric providing me with protected time (and part of my paycheck) to continue performing the reviews. Unfortunately for my career prospects in QI, there were no deaths that could be blamed on medical errors.
Using the IHI's tool, on the other hand, would have classified many of the deaths in aged, demented septic patients as due to trivial errors, like a dose of insulin being given late. I had started the project using their tool, but looked at the results and thought, this is ridiculous, I can't sign my name to this, completing the review using my own trained judgment.
To protect our ethics and integrity, we must give our best efforts every time to every patient with sepsis. To avoid despair, we must do so with acceptance that despite those efforts, many will not survive. Holding those two imperatives in our minds and hearts simultaneously is the central cognitive and spiritual challenge of critical care medicine.