Professional Documents
Culture Documents
The 20th-century statistician and quality scholar HMPS in a sample of 2809 admissions in 11
W. Edwards Deming proposed the “14 Points for Massachusetts hospitals in 2018 — 34 years af-
Top Leaders” — a checklist of management ter the New York hospital records were sampled
principles for executives who wish to nurture in the HMPS. As in the HMPS, trained nurses
improvement in complex systems. First on his reviewed the records to identify admissions that
list was “constancy of purpose for improve- included possible adverse events, and trained
ment.”1 In Deming’s view, when leaders slacken physicians then validated the findings; an at-
their visible commitment to a goal, progress tempt was also made to distinguish “prevent-
slows or stalls. able” adverse events from other adverse events.
For a period of time after the publication of However, unlike in the HMPS, a trigger tool was
the 2000 report by the Institute of Medicine used to help flag suspicious records, and all the
(now known as the National Academy of Medi- records were scrutinized for certain types of
cine) titled, “To Err Is Human: Building a Safer harm that were not examined in the HMPS, such
Health System,”2 improving patient safety was a as diagnostic errors and failure to treat decom-
priority in U.S. health care. That report rested on pensating patients.
a strong foundation of empirical research — The authors’ findings are disturbing. At least
with the f lagship study being the Harvard one adverse event was identified in 23.6% of the
Medical Practice Study (HMPS) — and progress admissions, and 9.0% of the admissions includ-
in safety science (largely in other industries), ed an adverse event that was rated as serious
such as studies of human error, cognition, team (i.e., caused harm that resulted in substantial in-
dynamics, and resilience. The authors of the re- tervention or prolonged recovery), life-threaten-
port famously estimated the number of lives lost ing, or fatal. Overall, 22.7% of the adverse events
to the consequences of errors in health care to were judged to be preventable. The types of ad-
be 44,000 to 98,000 per year in hospitals alone, verse events that were identified were consistent
thereby flagging shortcomings in patient safety with those previously reported in the literature:
as a public health threat as large as motor ve- 39.0% were adverse drug events (defined as inju-
hicle accidents and breast cancer. ries resulting from drugs that were taken),
Subsequently, a decade-long burst of strategic 30.4% were events related to a surgical or other
activities to improve patient safety occurred; procedure, 15.0% were events related to general
these activities included national campaigns, patient care (such as falls), and 11.9% were
research studies, training programs for patient hospital-acquired (health care–associated) infec-
safety officers, and changes in Medicare pay- tions. The mean length of stay for admissions
ment. Measuring the results of these initiatives during which an adverse event occurred was
proved difficult because they were confounded more than twice as long as that for admissions
by deficient data and a lack of standardized without an adverse event: 9.3 days as compared
methods, but there were inklings of progress with 4.2 days.
against specific threats, such as some hospital- A direct comparison of the quantitative find-
acquired infections. Despite such progress, a ings of this study with those of the HMPS is
firm answer to even the most basic question has tempting but is not warranted. Bates and col-
been elusive: Has the nation made progress or leagues used search methods that were guaran-
not since “To Err Is Human: Building a Safer teed to identify more injuries than the methods
Health System” rang the alarm? used in the HMPS, and their definition of an
In this issue of the Journal, Bates and col- adverse event was broader. However, the inci-
leagues3 have attempted to estimate progress in dence of adverse events that were identified
patient safety by replicating the methods of the (23.6%) does not suggest dramatic progress.
On the contrary, these findings suggest that the models, staff burnout, and decarbonization, to
safety movement has, at best, stalled. name a few. They may not welcome the duty to
The 2022 National Steering Committee for push patient safety back to strategic promi-
Patient Safety,4 as well as the authors of a na- nence. Nevertheless, “first do no harm” remains
tional action plan for patient safety sponsored by a sacred obligation for all in health care, and
the Agency for Healthcare Research and Qual- success requires “constancy of purpose for im-
ity,5 reached the same conclusion. The President’s provement.” Without renewed board and execu-
Council of Advisors on Science and Technology tive leadership and accountability for safety and
has been preparing recommendations for the without concerted, persistent investment in and
President to reignite an effective safety move- monitoring of change, a summary study 34
ment. This effort could hardly be timelier. years from now may again look all too familiar,
The study by Bates and colleagues illustrates with millions upon millions of patients, fami-
several of the difficulties in tracking safety. lies, and health care staff paying the price for
First, event rates are highly sensitive to the inaction.
method of review. Essentially, the harder one Disclosure forms provided by the author are available with the
looks for hazards and patient injuries, the more full text of this editorial at NEJM.org.
one finds. (Voluntary reporting is so unreliable
as to be nearly worthless in the calculation of From the Institute for Healthcare Improvement, Boston.