what medical errors add up to

for ten years, every time the country’s cardiologists left for their two big annual meetings, the sickest heart patients admitted to teaching hospitals died less often. not more. less. that finding is real, it was published in a major journal, and its authors could not fully explain it. it sits on top of a bigger argument about how many people die from medical care in the united states each year, where the answers range from about 25,000 to over 400,000 depending on who is counting and what they call a death from error. here is the conference study told straight, the death counts with the fight attached, the medication numbers that are the most solid part of the whole picture, and what a patient can do with any of it.

when the cardiologists left town

every year, on the order of thirteen to nineteen thousand cardiologists and other heart specialists go to the american heart association’s annual meeting, and a similar number to the american college of cardiology’s.^[1] four researchers asked a simple question: what happens to the heart patients admitted to hospital on those days? they pulled every medicare admission for heart attack, heart failure and cardiac arrest from 2002 through 2011, took the ones that landed on meeting dates, and compared them with admissions on the same weekdays in the three weeks before and after. that is about 19,000 heart-failure admissions on meeting days against about 115,000 on the days around them, and 1,564 cardiac arrests against 9,580.^[1]

in major teaching hospitals, the places where the most cardiologists are likely to be away, the high-risk patients did better when the specialists were gone. thirty-day mortality for high-risk heart failure was 17.5 percent on meeting days and 24.8 percent otherwise. for cardiac arrest, 59.1 percent against 69.4.^[1] those are not small gaps. for heart failure it is roughly seven fewer deaths per hundred of the sickest patients, and the confidence intervals do not overlap.

the heart attack result is the clue. among high-risk heart attack patients at teaching hospitals, mortality was the same on meeting days and off (39.2 versus 38.5 percent), but they were much less likely to get a stent: 20.8 percent had a percutaneous coronary intervention during meetings against 28.2 percent otherwise.^[1] fewer procedures, same survival. and the effect had edges that make sense: nothing changed for low-risk patients, nothing changed at non-teaching hospitals, and cardiac deaths did not move during the oncology, gastroenterology or orthopaedics meetings, nor did hip-fracture or GI-bleed deaths move during the cardiology ones.^[1]

the authors gave three explanations and said plainly that their data could not tell them apart. one: the cardiologists who stay home during a conference may be different from the ones who go. two: less intensive care during meetings, whether from reluctance to intervene without back-up or from a change in who is deciding, may mean that the usual procedures done to these patients on ordinary days are, for the sickest, doing harm that outweighs the benefit; they read the stent finding as “potential overuse of PCI in this population.” three: fewer elective procedures and clinic visits may free attention for the high-risk patients who remain. their stated principal limitation was “an inability to establish the mechanism.”^[1] it is observational. it is medicare only. it cannot say who was in the building. and it is one study, though it echoes, as the authors note, reports of falling mortality during physician strikes in israel and elsewhere. what it does not say is that cardiologists kill people. what it says is that for the sickest heart patients, doing less was at least not worse, and might have been better.

the numbers, and the fight over them

the conference study is one clean signal inside a much noisier argument about how many americans die from their care. here are the figures that get quoted, who produced them, what each one actually counts, and what is wrong with it.

the third leading cause of death, and why that is contested

in 2016 martin makary and michael daniel published an analysis in the BMJ arguing that medical error kills 251,454 americans a year, which would rank it third after heart disease and cancer.^[4] their point about the death certificate is sound: the CDC’s cause-of-death list is built from ICD codes, and “causes of death not associated with an ICD code, such as human and system factors, are not captured.”^[4] the number went everywhere.

the following year, in the BMJ’s own patient-safety journal, two of the field’s senior researchers took it apart.^[5] the paper “presented no new data”; it took “the arithmetic average of four estimates,” one of them from a for-profit ratings company. the peer-reviewed studies underneath were designed to count harms, not deaths, and they contained almost no deaths to count: the north carolina study had 14, two government reports had 12 and 3, another had 9. “any extrapolation that generalises from so few deaths (14 or fewer) to so many (200 000–400 000) surely warrants substantial scepticism.” when studies are built to do the one thing that matters here, to look at each inpatient death and ask whether it was avoidable, they find 1 to 3 percent, and the largest found 3.6 percent. about 700,000 people die in US hospitals a year. at 3.6 percent, “something like 25 000 deaths might be averted each year by eliminating medical error—a far cry from 251 454.”^[5]

the same critique applies, more gently, to the 1999 institute of medicine range of 44,000 to 98,000 that started all this.^[2] the upper figure comes from a 1984 review of 30,121 new york hospital records in which 3.7 percent of admissions had an adverse event and 13.6 percent of those events “led to death.”^[3] an event that leads to death and an error that causes one are not the same thing, and the reviewers were not asked to separate them. shojania and dixon-woods note that even one of the original investigators expressed disquiet about the figures.^[5]

what the cleanest recent count found

the best current answer to “how often does hospital care hurt people” is the 2023 SafeCare study: 2,809 randomly chosen admissions to 11 massachusetts hospitals in 2018, records screened by trained nurses and every flagged event adjudicated by physicians.^[6] at least one adverse event turned up in 23.6 percent of admissions. of 978 events, 22.7 percent were judged preventable and 32.3 percent serious or worse. a preventable event happened in 6.8 percent of admissions; a serious preventable one in 1.0 percent. adverse drug events were the most common category, 39.0 percent of everything found, ahead of surgery (30.4), nursing care such as falls and pressure injuries (15.0) and infections (11.9).^[6]

and the deaths: seven in the sample, one judged preventable.^[6] that is one preventable death in 2,809 admissions. scale that the way the big estimates are scaled and you get a number nearer the critics’ 25,000 than the 251,454. the study was paid for by the malpractice insurer of the harvard hospitals, which is worth knowing and does not change the arithmetic.

two older studies fill in the shape. in ten north carolina hospitals from 2002 to 2007, harms ran at 25.1 per 100 admissions and did not fall over the six years, a decade after the institute of medicine report was supposed to change everything.^[7] and when three hospitals had the same patients checked three ways, the voluntary incident reports and the government’s patient safety indicators “missed 90 percent of the adverse events” that a systematic chart-trigger review found.^[8] the hospital’s own count is not the count.

medication errors

drugs are where the numbers are most solid, because a wrong drug, wrong dose or missed interaction leaves a trace in the record. the institute of medicine’s preventing medication errors report put it at “at least 1.5 million preventable adverse drug events” a year: 380,000 to 450,000 in hospitals, 800,000 in long-term care, 530,000 among outpatient medicare patients alone, and called each of those likely an underestimate. its other line is the one to remember: “on average, a hospital patient is subject to at least one medication error per day.”^[9]

the hospital figure traces to a 1995 study of 4,031 admissions at two boston hospitals: 6.5 adverse drug events per 100 admissions, 28 percent of them preventable, 42 percent of the serious and life-threatening ones preventable, and the errors made mostly at ordering (56 percent) and administration (34 percent).^[10] outside hospital, among 30,397 person-years of older medicare patients in one practice, adverse drug events ran at 50.1 per 1,000 person-years, 27.6 percent preventable, with the preventable ones concentrated at prescribing and monitoring.^[11]

the sharpest recent figures come from emergency departments. about 99,600 americans over 65 are admitted from an emergency department for an adverse drug event each year; 65.7 percent of those are unintentional overdoses, and two-thirds involve four ordinary drug classes, warfarin, insulin, antiplatelets and oral diabetes drugs, while the drugs officially labelled high-risk account for 1.2 percent.^[12] across all ages there are about 4 emergency visits for adverse drug events per 1,000 people a year, and 27.3 percent of those end in admission.^[13]

the psychiatric angle, where it has a number: among children and teenagers aged 6 to 19, antipsychotics were the second most common drug class behind an emergency visit for a drug reaction, at 4.5 percent, after antibiotics.^[13] in a state psychiatric hospital, a review team went through 31 charts and the dispensing records and found 2,194 medication errors, 58 percent rated high risk of harm; the staff’s own reporting system had logged nine.^[14] the FDA’s own medication-errors page notes that reporting to it is voluntary, defines an error as a preventable event that “may or may not result in an adverse event,” and lists among its interventions approving a new brand name for the antidepressant vortioxetine to stop confusion with an antiplatelet drug.^[15]

the other direction

three things have to be kept apart, and the headlines never do. an adverse event is an injury from care, error or not. an error is a mistake, harm or not; most cause none, which is why the FDA’s definition says so.^[15] and a death from error is a death that would not otherwise have happened, which is a judgement, made after the fact, by a reviewer reading a chart. the large numbers on this page are almost all the first thing scaled up; the small ones are the third thing counted one at a time. the 1.5 million is events, not deaths. the 251,454 is deaths, but it was built out of events.^[5][9]

the denominator matters as much. the medicare-based estimates were applied to all US admissions, a population in which the single most common reason for admission is having a baby, an event medicare patients rarely have.^[5] patients who develop a hospital-acquired complication are also the patients who were most likely to die anyway; in one analysis the critics cite, those who got a C. difficile infection had a baseline risk of death of 8.0 percent against 1.8 percent for those who did not. “many patients die with, rather than of, these conditions.”^[5]

and the conference study cuts the other way too. it does not show that hospitals are dangerous; the same paper found no effect at non-teaching hospitals and none for low-risk patients, which is most patients. it shows that for a small, very sick group, a particular kind of intensity may not be helping. its authors called their interpretation consistent with “less is more” findings elsewhere in intensive care, and they were careful to say “may.”^[1] the strongest honest summary of everything above is: care hurts people far more often than the reporting systems admit, drugs are the most common way, and nobody has a defensible national count of deaths.

what to do with it

the numbers that are most solid, the drug numbers, are also the ones a patient can act on. the questions below are AHRQ’s where they say so; the rest are mine, with the reason attached. none of them involves changing what you take.

• every time a new drug is added: what does this interact with, out of everything else i take, including supplements? (AHRQ’s own wording is “is this medicine safe to take with other medicines or dietary supplements i am taking?”)^[16]
• keep one written list of everything you take, prescription and not, and bring it, or the bottles, to every visit. AHRQ calls it brown-bagging. the 2,194-versus-9 study is what happens when the record is the only list^[16][14]
• ask: is this dose adjusted for my age and my kidneys? no guideline on this page says it in those words; the reason to ask is that two-thirds of emergency hospitalisations for drug harm in people over 65 were unintentional overdoses, mostly of ordinary drugs^[12]
• at the pharmacy: is this the medicine my doctor prescribed? name confusion is common enough that the FDA has approved renaming drugs over it, including an antidepressant^[16][15]
• at discharge: ask for the plan in writing, and specifically which of the medicines you were taking before the stay you should keep taking, and which you should not. AHRQ says this may prevent an unexpected return trip^[16]
• do not stop anything on the strength of this page. AHRQ: talk with your doctor or pharmacist before you stop taking any medicine that was prescribed^[16]

questions people ask

sources

1. Jena AB, Prasad V, Goldman DP, Romley J. Mortality and treatment patterns among patients hospitalized with acute cardiovascular conditions during dates of national cardiology meetings. JAMA Internal Medicine 2015;175(2):237-244. Medicare fee-for-service beneficiaries 65 and over, 2002 to 2011, admitted with heart attack, heart failure or cardiac arrest during the dates of the American Heart Association and American College of Cardiology annual meetings, compared with “identical nonmeeting days in the 3 weeks before and after” (heart failure: 19,282 meeting-day vs 114,591 nonmeeting-day admissions; cardiac arrest 1,564 vs 9,580; heart attack 8,570 vs 57,471). In teaching hospitals, adjusted 30-day mortality among high-risk patients was lower on meeting dates: “heart failure, 17.5% [95% CI, 13.7%-21.2%] vs 24.8% [95% CI, 22.9%-26.6%]; P < .001; cardiac arrest, 59.1% [95% CI, 51.4%-66.8%] vs 69.4% [95% CI, 66.2%-72.6%]; P = .01.” High-risk heart attack: mortality “similar” (39.2% vs 38.5%; P = .86) but PCI rates “lower during meetings (20.8% vs 28.2%; P = .02).” “No mortality or utilization differences existed for low-risk patients in teaching hospitals or high- or low-risk patients in nonteaching hospitals.” Other specialties’ meetings had no effect on cardiac deaths. Full text (PMC4314435) gives the authors’ three explanations and their “principal limitation… an inability to establish the mechanism.” PMID 25531231. https://doi.org/10.1001/jamainternmed.2014.6781
2. Institute of Medicine. To Err Is Human: Building a Safer Health System. National Academies Press, 2000 (the catalogue page dates it 2000; it is usually cited as 1999). From the report summary on the National Academies’ own site: “When extrapolated to the over 33.6 million admissions to U.S. hospitals in 1997, the results of the study in Colorado and Utah imply that at least 44,000 Americans die each year as a result of medical errors. The results of the New York Study suggest the number may be as high as 98,000.” Also: “Medication errors alone, occurring either in or out of the hospital, are estimated to account for over 7,000 deaths annually.” The National Academies’ catalogue page carries the one-line version: “as many as 98,000 people die in any given year from medical errors that occur in hospitals.” https://doi.org/10.17226/9728
3. Brennan TA, Leape LL, Laird NM, et al. Incidence of adverse events and negligence in hospitalized patients: results of the Harvard Medical Practice Study I. New England Journal of Medicine 1991;324:370-376. The New York study behind the 98,000 figure: 30,121 records from 51 hospitals in 1984. “Adverse events occurred in 3.7 percent of the hospitalizations… and 27.6 percent of the adverse events were due to negligence… 13.6 percent led to death.” Weighted to the state’s 2,671,863 discharges: “98,609 adverse events and 27,179 adverse events involving negligence.” PMID 1987460. https://doi.org/10.1056/NEJM199102073240604
4. Makary MA, Daniel M. Medical error: the third leading cause of death in the US. BMJ 2016;353:i2139. An analysis piece, not a study; its estimate is 251,454 deaths a year in US hospitals. The open portion of the text (read in an archived copy) makes the argument for why the CDC list misses error: “a major limitation of the death certificate is that it relies on assigning an International Classification of Disease (ICD) code to the cause of death. As a result, causes of death not associated with an ICD code, such as human and system factors, are not captured,” and “We focus on preventable lethal events to highlight the scale of potential for improvement.” The rest is paywalled; the 251,454 figure on this page is taken from the critique below, which quotes it. PMID 27143499. https://doi.org/10.1136/bmj.i2139
5. Shojania KG, Dixon-Woods M. Estimating deaths due to medical error: the ongoing controversy and why it matters. BMJ Quality & Safety 2017;26:423-428. Full text read. “Makary and Daniel attribute an even higher toll to medical error: 251 454 deaths in US hospitals per year, making, they say, medical error the third-leading cause of death in the USA… We propose that the new estimate is very likely to be wrong.” “The new paper is not a study… The authors simply took the arithmetic average of four estimates,” one from HealthGrades, “a for-profit company that markets quality and safety ratings.” The source studies “typically involve very small numbers of deaths”: 14 in North Carolina (0.6% of patients), 3 and 12 in two government reports, 9 in another; “Any extrapolation that generalises from so few deaths (14 or fewer) to so many (200 000–400 000) surely warrants substantial scepticism.” Studies designed to judge preventability of inpatient deaths find “proportions under 5%, typically in the range of 1%–3%”; the largest, 3.6%. “Around 700 000 deaths occur in US hospitals annually… If… 3.6% is the more correct rate, then it means something like 25 000 deaths might be averted each year by eliminating medical error—a far cry from 251 454.” Also: HealthGrades 389,576; Leapfrog 206,021; “many patients die with, rather than of, these conditions.” PMID 27733444. https://doi.org/10.1136/bmjqs-2016-006144
6. Bates DW, Levine DM, Salmasian H, et al. The safety of inpatient health care. New England Journal of Medicine 2023;388:142-153. The SafeCare study: a random sample of 2,809 admissions to 11 Massachusetts hospitals in 2018, trigger method plus record review. “We identified at least one adverse event in 23.6%. Among 978 adverse events, 222 (22.7%) were judged to be preventable and 316 (32.3%) had a severity level of serious… or higher. A preventable adverse event occurred in 191 (6.8%) of all admissions, and a preventable adverse event with a severity level of serious or higher occurred in 29 (1.0%). There were seven deaths, one of which was deemed to be preventable. Adverse drug events were the most common adverse events (accounting for 39.0% of all events), followed by surgical or other procedural events (30.4%), patient-care events… (15.0%), and health care-associated infections (11.9%).” Funded by the Harvard hospitals’ malpractice insurer. PMID 36630622. https://doi.org/10.1056/NEJMsa2206117
7. Landrigan CP, Parry GJ, Bones CB, et al. Temporal trends in rates of patient harm resulting from medical care. New England Journal of Medicine 2010;363:2124-2134. Ten North Carolina hospitals, 2,341 admissions, 2002 to 2007, Global Trigger Tool. “Internal reviewers identified 588 harms (25.1 harms per 100 admissions)”; “no significant changes in the overall rate of harms per 1000 patient-days (reduction factor, 0.99 per year…) or the rate of preventable harms.” “Harms remain common, with little evidence of widespread improvement.” This is the study whose 14 deaths Shojania and Dixon-Woods say were extrapolated to hundreds of thousands. PMID 21105794. https://doi.org/10.1056/NEJMsa1004404
8. Classen DC, Resar R, Griffin F, et al. “Global trigger tool” shows that adverse events in hospitals may be ten times greater than previously measured. Health Affairs 2011;30(4):581-589. Three detection methods on the same patients at three hospitals. Voluntary reporting and AHRQ’s Patient Safety Indicators “fared very poorly compared to other methods and missed 90 percent of the adverse events. The Institute for Healthcare Improvement’s Global Trigger Tool found at least ten times more confirmed, serious events than these other methods. Overall, adverse events occurred in one-third of hospital admissions.” PMID 21471476. https://doi.org/10.1377/hlthaff.2011.0190
9. Institute of Medicine. Preventing Medication Errors. National Academies Press, 2007 (the catalogue page dates it 2007; it is usually cited as 2006). From the report summary on the National Academies’ own site: “The committee estimates that at least 1.5 million preventable ADEs occur each year in the United States: Hospital care—Classen and colleagues (1997) projected 380,000 preventable ADEs occurring annually, and Bates and colleagues (1995b) 450,000… Long-term care—Gurwitz and colleagues (2005) projected 800,000 preventable ADEs… Ambulatory care—Among outpatient Medicare patients alone, Gurwitz and colleagues (2003) projected 530,000 preventable ADEs.” Also: “the committee conservatively estimates that on average, a hospital patient is subject to at least one medication error per day,” and “Assuming conservatively an annual incidence of 400,000 in-hospital preventable ADEs, each incurring extra hospital costs of $5,857, yields an annual cost of… $3.5 billion in 2006 dollars.” https://doi.org/10.17226/11623
10. Bates DW, Cullen DJ, Laird N, et al. Incidence of adverse drug events and potential adverse drug events: implications for prevention. JAMA 1995;274:29-34. All 4,031 adult admissions to 11 units in two tertiary hospitals over 6 months. “247 ADEs and 194 potential ADEs were identified. Extrapolated event rates were 6.5 ADEs and 5.5 potential ADEs per 100 nonobstetrical admissions… Of all ADEs, 1% were fatal (none preventable), 12% life-threatening, 30% serious, and 57% significant. Twenty-eight percent were judged preventable. Of the life-threatening and serious ADEs, 42% were preventable… Errors resulting in preventable ADEs occurred most often at the stages of ordering (56%) and administration (34%).” PMID 7791255. https://pubmed.ncbi.nlm.nih.gov/7791255/
11. Gurwitz JH, Field TS, Harrold LR, et al. Incidence and preventability of adverse drug events among older persons in the ambulatory setting. JAMA 2003;289:1107-1116. 30,397 person-years of Medicare enrollees in one multispecialty practice. “1523 identified adverse drug events, of which 27.6% (421) were considered preventable. The overall rate of adverse drug events was 50.1 per 1000 person-years, with a rate of 13.8 preventable adverse drug events per 1000 person-years.” 38.0% serious, life-threatening or fatal. Errors at “prescribing (n = 246, 58.4%) and monitoring (n = 256, 60.8%)”; “neuropsychiatric (8.6%)” among the most common types of preventable event. PMID 12622580. https://doi.org/10.1001/jama.289.9.1107
12. Budnitz DS, Lovegrove MC, Shehab N, Richards CL. Emergency hospitalizations for adverse drug events in older Americans. New England Journal of Medicine 2011;365:2002-2012. National surveillance, 2007 to 2009. “An estimated 99,628 emergency hospitalizations (95% confidence interval [CI], 55,531 to 143,724) for adverse drug events in U.S. adults 65 years of age or older each year… Nearly two thirds of hospitalizations were due to unintentional overdoses (65.7%)… Four medications or medication classes were implicated alone or in combination in 67.0%… of hospitalizations: warfarin (33.3%), insulins (13.9%), oral antiplatelet agents (13.3%), and oral hypoglycemic agents (10.7%). High-risk medications were implicated in only 1.2%.” PMID 22111719. https://doi.org/10.1056/NEJMsa1103053
13. Shehab N, Lovegrove MC, Geller AI, et al. US emergency department visits for outpatient adverse drug events, 2013-2014. JAMA 2016;316:2115-2125. 42,585 cases from 58 emergency departments. “An estimated 4.0 (95% CI, 3.1-5.0) ED visits for adverse drug events occurred per 1000 individuals annually in 2013 and 2014 and 27.3%… resulted in hospitalization.” Among ages 6 to 19, antibiotics were the most common class implicated (31.8%), “followed by antipsychotics (4.5%; 95% CI, 3.3%-5.6%).” Among adults 65 and older, anticoagulants, diabetes agents and opioids accounted for 59.9%. PMID 27893129. https://doi.org/10.1001/jama.2016.16201
14. Grasso BC, Genest R, Jordan CW, Bates DW. Use of chart and record reviews to detect medication errors in a state psychiatric hospital. Psychiatric Services 2003;54(5):677-681. “In the 31 charts retrospectively reviewed and the dispensing events concurrently reviewed, the team detected a total of 2,194 medication errors, whereas a total of nine errors were self-reported for the same patient group. Administration errors accounted for more than half of the total (66 percent), followed by transcription errors (23 percent), prescription errors (11 percent)… Nineteen percent of errors were rated as having a low risk of harm, 23 percent as having a moderate risk, and 58 percent as having a high risk.” PMID 12719497. https://doi.org/10.1176/appi.ps.54.5.677
15. US Food and Drug Administration. Medication errors related to CDER-regulated drug products (content current as of 01/11/2024). Uses the NCC MERP definition: “any preventable event that may cause or lead to inappropriate medication use or patient harm while the medication is in the control of a healthcare provider, patient, or consumer.” “A medication error may or may not result in an adverse event.” “The reporting of medication errors to FDA’s Adverse Event Reporting System (FAERS) is voluntary in the United States.” The page lists the safety communications it has issued for name and dose confusion, including the brand-name change of the antidepressant vortioxetine from Brintellix “to avoid confusion with antiplatelet drug Brilinta (ticagrelor).” No count of reports received is given on the page. https://www.fda.gov/drugs/drug-safety-and-availability/medication-errors-related-cder-regulated-drug-products
16. Agency for Healthcare Research and Quality. 20 Tips to Help Prevent Medical Errors (patient fact sheet; read in an archived copy of AHRQ’s page). “Make sure that all of your doctors know about every medicine you are taking. This includes prescription and over-the-counter medicines and dietary supplements, such as vitamins and herbs.” “Bring all of your medicines and supplements to your doctor visits.” Ask: “Is this medicine safe to take with other medicines or dietary supplements I am taking?” “When you pick up your medicine from the pharmacy, ask: Is this the medicine that my doctor prescribed?” “When you are being discharged from the hospital, ask your doctor to explain the treatment plan you will follow at home… It is important to know whether or not you should keep taking the medicines you were taking before your hospital stay.” AHRQ’s companion page, Be More Engaged in Your Healthcare: “Talk with your doctor or pharmacist before you stop taking any medicines that your doctor prescribed.” https://www.ahrq.gov/patients-consumers/care-planning/errors/20tips/index.html

related: the overdose the prescription was part of · st john’s wort drug interactions · what works that costs nothing

adam