  |
|
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Journal of the New Zealand Medical Association,
31-March-2006, Vol 119 No 1231
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
Representative case series from New Zealand public hospital
admissions in 1998—III: adverse events and death
Robin Briant, John Buchanan, Roy Lay-Yee, Peter Davis
The New Zealand Quality of Healthcare Study (NZQHS)1
examined 6579 medical records using a two-stage retrospective review applied to
a representative sample of hospital admissions for the calendar year 1998. The
sample was drawn by systematic list selection (after exclusion of specialist
institutions) from 13 public hospitals providing acute care and over 100 beds.
The main aim of the Study was to quantify the adverse outcomes of healthcare
management in the New Zealand public hospital system.
NZQHS found that 12.9% of public hospital admissions were
associated with an adverse event1 (within the range of the 16.6% and 10.8% found
in similar studies from Australia2 and the United Kingdom3 respectively).
Subsequently, there has been increased awareness of the incidence and outcomes
of adverse events. The Institute of Medicine Report4 extrapolated from the
Harvard Medical Practice Study (HMPS)5 and the Utah and Colorado Study (UTCOS),6
and estimated the numbers of deaths in the United States from adverse healthcare
events. McDonald et al7 have challenged the reliability of such extrapolations
and Hayward and Hofer,8 in a careful examination of deaths attributed to adverse
events and sub-optimal care, have argued that the degree of preventability is
probably grossly over-estimated.
This paper is the third in a series on clinical aspects of
adverse events following those examining medication-9 and surgery-related10
cases. We aim to document in detail the deaths that occurred in patients
admitted to New Zealand public hospitals in 1998 for whom an adverse event was
identified. Because of the way the study instrument is constructed, the
concurrence of an adverse event and death did not necessarily mean that the two
were related. The paper seeks, through the careful re-analysis of those adverse
events where death was the outcome, to determine the extent of this
relationship.
MethodsNZQHS’s data collection method has been reported
elsewhere.1 It involved structured implicit review of the medical record, while
seeking evidence of harm to patients from healthcare management.
An adverse event (AE) was defined as:
The degree of causation was scored on a scale
of 1–6, where 1 is no relation of the injury to healthcare management (and
therefore by definition not an AE), 2 is slight association, 6 is definite
association, and 3 and 4 are on either side of
50:50.
Preventability of the event, scored on the same scale,
was defined as an error in healthcare management due to failure to follow
accepted practice at an individual or system level. Patient impact was measured
by disability defined as temporary or permanent disability (respectively lasting
less or more than 1 year) or death. Attributable or added bed-days refer to
those extra days associated with an AE that were spent in the study hospital
during one or more admissions.
Nurse screeners and medical reviewers wrote clinical
summaries of each case as part of the main study. For the purpose of this
re-analysis of deaths, two of the authors (RB and JB) independently studied
those summaries and the details of injury, with the aim of discriminating those
cases where there was a clear link between the AE and death (from those cases
where death was not the result of the identified AE).
Attention was paid to the medical reviewer’s
scoring of the degree of healthcare causation in the production of injury, but
the authors had to make their own determination of the degree of causation of
death. A simple categorisation was developed and all cases were placed in one or
other category. The authors (RB and JB) then met, and those instances where
there was not initial agreement on category, were resolved by discussion and
agreement reached.
The categories of relationship established were:
It was not possible to
use the same process to determine preventability of a death, as access to only
the clinical summaries did not provide the authors with sufficient information.
The high degree of preventability of an AE cannot be extended to draw a
conclusion on the preventability of the death. However, we have made a
sub-analysis of the cases where an AE had both high causation and high
preventability scores (4–6), and was detected during the sampled
admission; this sub-group matches most closely those patients included in the
American studies5,6 from which extrapolations to total deaths due to medical
error have been made.
Life tables, by age and gender, for the New Zealand
population 2000–0211 were consulted to determine the years of life lost by
those patients whose deaths were caused by an AE.
Estimated death rates and 95% confidence intervals have
been adjusted for the sample design.
ResultsIn total, 6579 admissions were screened according to set
criteria by nurses at the first stage for indications of an AE, with 4119
progressing to second-stage review by physicians using a structured protocol. Of
these admissions, 850 were judged to be AEs (12.9% of admissions) that occurred
or were detected during the sampled admission, and where there was any evidence
of healthcare management causation (scores 2–6). Death was the outcome for
38 of those admissions (4.5% of AEs).1
In Table 1, the results of the assessment of the causal
status of AEs are outlined, together with key patient characteristics.
Twenty-four (63%) of the 38 deaths were judged attributable to the AE, either
definitely or probably, with 14 considered not related to the reported AE.
Table 1. Causal status of adverse event associated with
death, other AEs, all AEs, and all reviewed admissions: characteristics of
patients
AE=Adverse event: occurred or
was detected during the sampled admission, and any evidence of healthcare
management causation (scores 2–6).
Of the deaths definitely or probably associated with the AE,
the average age was 75.6 years (Group A – Definite) and 57.9 years (Group
B – Probable) respectively; this compared with the average age of 51.5
years for all patients suffering AEs. In the combined Groups A and B, exactly
half were males, 62.5% were aged under 70 years, 12.5% were Maori, and
two-thirds were from high-deprivation areas of residence. This profile was very
similar to that for all AEs.
Key characteristics of the events are outlined in Table 2.
For the combined AE-attributable group, the AE causation score was over 50% for
nearly all in the group, and the AE preventability score was over 50% for
exactly half of them. There were 9.6 mean added bed-days for Group A and 20.6
for Group B, compared with the 9.3 days overall for all AEs.
(View Tables 2,5,6-8 in the PDF
version)
Comorbidity was slightly below average for Group A and well
above for Group B. The patients in Group A (where death was definitely related
to the AE) lost an average of 11.8 years of life, while those in Group B (where
the AE probably caused death) lost an average of 25.0 years of life.
The preventability of deaths associated with AEs is
addressed in Table 3. The data were also analysed using the same criteria as
UTCOS6 to identify highly preventable AEs; that is, AEs that were detected
during the sampled admission and where both the causation and preventability
scores were 4 or more.
Applying these criteria, 267 AEs were judged to be highly
preventable (4.1% of admissions), of which 15 were associated with death as the
outcome (5.6% of highly preventable AEs) (Table
3).
We judged that 9 (60.0%) of the 15 deaths were causally
related to the AE, either definitely or probably. Thus, with this calculation, 9
(3.4%) of 267 highly preventable AEs caused death.
Table 4 provides a summary of the distribution of deaths,
together with rates of occurrence per 1000 admissions. For all 6579 reviewed
admissions in NZQHS, the overall death rate (at discharge) was 1.8%.
The mortality rate was 0.22% (15/6579) where death (either
at or after discharge) was the outcome of a highly preventable AE. After
re-analysis, the adjusted rate was 0.13% (9/6579), equating to a rate of 1.3
patients per thousand admitted to New Zealand public hospitals
Table 3. Causal status of AEs associated with death, by
preventability of AE
AE=Adverse event; was detected
during the sampled admission, and high healthcare management causation (scores
4-6); *Preventability scores 2–6;†Preventability scores
4–6.
Table 4. Summary of death rates by causal status and
preventability of AE
AE=Adverse event; was detected
during the sampled admission, and high healthcare management causation (scores
4-6);
*All
deaths in this table are those associated with AEs (at discharge or some time
after discharge), except those for all reviewed admissions (at discharge only);
†Preventability scores 2–6; ‡Preventability scores 4–6;
§All rates have a common denominator (n=6579); rates and 95% confidence
intervals have been adjusted for the sample design.
Group A (definite
relationship between AE and death)—There were 11 patients in this
group (Table 5). Five of these deaths followed surgery—two
cholecystectomies and three operations for carcinoma of the colon (one rendered
urgent by colon perforation during diagnostic colonoscopy).
There were two fatal infections: a
Clostridium difficile colitis case
following antibiotic for exacerbation of COPD, and an under-treated
staphylococcal bloodstream infection associated with a pacing wire.
Two patients died before their diagnosis was established,
one in-hospital patient with high probability of thromboembolism not adequately
investigated or treated, and one community patient with high risk for ischaemic
heart disease where chest pain was not acted upon and an untreated myocardial
infarction progressed to cardiogenic shock.
One patient with many comorbidities and an abdominal aortic
aneurysm (AAA) waited 3 months on an urgent tertiary centre vascular waiting
list; then when rupture occurred, waited many hours for surgery at the
peripheral hospital and died before the operation began.
One frail resthome resident fell fracturing the neck of
femur (falls in rest homes or other health care facility were classed as AEs)
and died of acute haematemesis 3 weeks after the fracture and its
fixation.
Group B (probable/possible
relationship between AE and death)—This group comprised 11 patients
(Table 6). Several patients are included in this group rather than Group A
because (although there was a close time relationship between a procedure and
death) the cause of death was not clarified in the case records. For two of
these patients, a very long lag occurred between relevant healthcare procedure
and death.
In this group, there were four deaths related to surgery and
two deaths related to diagnostic procedures. There was one probable infection in
the remaining lung of a patient after pneumonectomy, and one probable myocardial
infarction after partial hepatectomy in another patient.
A patient with carcinoma of the colon experienced major
postoperative complications, and another patient had persistent subacute bowel
obstruction and died 6 months after complicated pelvic surgery. There was one
fatal myocardial infarction after bronchoscopy. One patient with end-stage renal
failure on dialysis and many comorbidities had a liver biopsy to evaluate
abnormal enzymes and died a few days later with a hepatic subcapsular
haemorrhage, ascites, and pleural effusions.
There were two cases of non-diagnosis and one of late
diagnosis. A patient with typical cardiac pain was discharged early from a
hospital emergency department after a single normal electrocardiogram (ECG),
returned later with a completed myocardial infarction too late to thrombolyse,
and died after 2 months of severe cardiac failure.
One patient died suddenly probably from pulmonary embolism
from a brachial venous thrombosis after arm compression in a community drug
coma. One had melanoma, mistaken for a benign lesion, so referral for excision
was not made urgent, and 5 months elapsed before the diagnosis and excision; the
patient presented with cerebral metastases 1 year after surgery and died soon
after.
There were two infective complications in complex cases, one
a young person with severe asthma requiring repeated ICU admissions and one with
aspiration pneumonia. Two patients had very late deaths; one gastric bypass
operation for obesity in 1968 was probably responsible for cirrhosis and hepatic
failure in 1998, and a woman who had cervical biopsy for carcinoma
in situ in 1973 had no follow-up smears
and died of invasive carcinoma of cervix in
1998.
Group C (unlikely
relationship between AE and death)—There were eight patients in
this group (Table 7). Although these patients suffered an AE, either it was
judged that there was no connection between AE and death, or the AE did no more
than hasten inevitable and imminent death.
Five of the patients, aged 83–91 years, were described
as frail and they suffered a range of terminal events where healthcare causation
may have played a small part in death—a fall in a rest home, a fall from a
hospital bed, a delay in inserting a gastrostomy leading to aspiration
pneumonia, delay in sorting out a complex presentation, and complications of
cardiovascular medication. There was one patient with advanced malignancy, one
with terminal muscular dystrophy; and one with myelodysplasia and pneumonia for
whom appropriate healthcare actions may have hastened imminent death.
Group D (no relationship
between AE and death)—There were six patients in this group (Table
8) where the authors could not identify any connection between healthcare
management and death and, in some instances, could not confirm that an AE had
occurred.
One demented patient with atrial fibrillation had warfarin
stopped (in consultation with family) after risk-benefit assessment; she died 6
months later of a cerebral embolism.
One acutely-admitted patient died before he could have upper
gastrointestinal endoscopy—but death was due to small bowel infarction,
and there would have been no benefit from the endoscopy.
Two patients had falls at home and died within hours; one of
an acute subdural haematoma before he could be taken to theatre, and one had an
extensive myocardial infarction earlier which probably caused the fall. The
remaining two patients died 1 and 5 months after surgery, but there was no
relationship between the surgery and the death.
DiscussionBecause of the way the study instrument functions, one
cannot necessarily conclude that the death is causally linked to the identified
AE. Therefore, we have reassessed carefully all aspects of the information
provided and made a determination on the causal relationship in each case.
Key
findings—The numbers are very small, and so the information is
indicative only. Our re-analysis showed that 14 of 38 deaths were found to have
a very tenuous link, or none at all, with healthcare causation. This makes the
other reported deaths from similar AE studies at least slightly suspect and
probably an overestimate of the fatal impact of AEs.
Based on the American definition of an AE, we concluded that
19/659 AEs (2.9%) led to death. In our studied population of 6579 patient
admissions, the death rate due to healthcare management was 0.28% (19/6579), or
2.8 healthcare-related deaths per 1000 patients admitted to New Zealand public
hospitals; the overall death rate was 1.8% (118/6579). Furthermore, three-fifths
of highly preventable AEs where death was the outcome (9/15) directly caused the
deaths of 1.3 per 1000 New Zealanders admitted to hospital.
Strengths and limitations
of this study—As far as we are
aware, this is the first in-depth
examination of deaths from an AE study using retrospective chart review, and it
calls into question the estimations of death rates drawn from other similar AE
studies. The main utility of studies of AEs and their outcome is to provide a
basis for understanding how we can improve the healthcare system and reduce the
amount of disability and the number of deaths from medical misadventure.
To this end, the question of the preventability of AEs is
crucial, but alas it is also the most difficult aspect of the study. Thus the
reviewer must ‘read between the lines’ as well as carefully study
the written record of the patient’s hospitalisation to understand the
process of care and determine if there were errors or omissions in management,
meaning that another standard of care might have prevented the mishap.
In this analysis, the primary sources of information were
not available and the secondary ones insufficiently detailed in the process of
care, so preventability of the death itself could not be assessed.
Interpretations and
implications—The New Zealand Quality of Healthcare Study (NZQHS)
found that AEs were associated with 12.9% of admissions, and 4.5% of AEs
resulted in death. The Harvard Medical Practice Study (HMPS)5 produced a rate of
AE of 3.7%; and of these AEs, 13.6% were associated with death. The Utah and
Colorado Study (UTCOS)6 showed death rates of 6.6% of all AEs reported and 8.8%
of all negligent AEs identified. The Quality of Australian Health Care Study2
showed a death rate of 4.9%; 69.6% of these deaths were judged to be highly
preventable and patients who died had 8.2 added bed-days compared with all AEs.
In these studies, and ours, data acquisition was carried out
in a comparable manner (although there were some methodological differences), so
these are the studies with which we can most accurately compare our
results.
None of these previous reports have included critical
analysis of the deaths, so raw numbers have been used to make extrapolations of
the impact of adverse healthcare and death. Extrapolating from the results of
the HMPS to the whole of New York State for 1984, the conclusion was that 13451
people would have died, at least in part from an AE. The mortality rate, where
death was the outcome of a highly preventable AE, extrapolated from UTCOS4,6 was
0.13% or 1.3 per 1000 admissions; this compares with the NZQHS rate of 0.22%
(15/6579) or 2.2 per 1000 admissions.
Controversies and
directions—We have attempted to estimate the number of years of
life lost by the patients where we identified an association between the AE and
death. For this estimate, we used the expected life span of each person.
Group A patients were much older than those patients
suffering AEs overall (75.6 versus 51.5 years), and many had significant
pathology or physiological compromise. Without taking their presenting disease
or comorbidities into account, they lost an average of 11.8 years of life.
Group B patients are distinguished by being younger (57.9
years) than all others who died, although they were still older than the average
person suffering an AE; their average loss of life was 25.0 years.
However, this assessment over-estimates the number of years
lost by this particular group of patients; we are applying normal
life-expectancy tables to a group of patients, most of whom already had serious
disease along with comorbidities12 which would have markedly reduced their years
of life available. We have been unable to find a method that would provide an
appropriate comparison of life expectancy for our cases.
The Institute of Medicine Report4 extrapolating from the
UTCOS6 study of AEs, claimed that at least 44,000 Americans (1.3 per 1000
admissions) die each year as a result of medical error; this would make medical
misadventure the eighth most important cause of death in the United States.
McDonald et al7 questioned the use of this methodology to
make such extrapolations, describing the review methodology as observational and
not appropriate for defining cause and effect. They concluded that the estimates
of rates of death from medical errors—which were based on the methods of
review used in our study—were misleading and exaggerated. We agree with
his view. These cautions were countered by Leape13 in the same issue of
JAMA, where he claimed that the numbers
had never been seriously called into question before. In fact, there has never
been a critical analysis of the causal association of death to AE in the studies
reported.
Hayward and Hofer8 set out to answer the question of whether
the implicit review tool gave an overestimate of the frequency of preventable
deaths. They were particularly concerned to identify how much the death would
have been preventable in the absence of an AE. Their 14 trained and carefully
controlled reviewers conducted 383 reviews of 111 deaths (excluding expected
deaths) sampled from VA hospitals in 1995–6. Even in this situation, there
was poor inter-reviewer reliability as to preventability. The final conclusion
was that only 6% of those dying would have likely survived to discharge if given
optimal care, and only 0.5% of those 111 would have been expected to survive in
good cognitive health for 3 months or more. Hayward and Hofer believe that,
while the implicit review tool probably underestimates the occurrence of AEs, it
almost certainly overestimates their consequences. Our conclusion is in
concordance with his view.
In New Zealand, Tobias and Turley14 have estimated (on the
basis of NZQHS findings) that about 1500 deaths per year are due to in-hospital
AEs of any degree of preventability, hence making AEs the 11th most important
cause of death; however, our re-analysis would reduce that estimated rate by
perhaps a third.
ConclusionOf the total reviewed population of 6579, 118 patients died;
and of 850 patients suffering an AE, 38 died. Based on the American AE
definition, there were 30 deaths among 659 AEs. However, only 19 of these 30
deaths (19/6579 or 2.8 per 1000 of the reviewed population) were causally
related to the AEs; 63% of the raw number reported.
Where the AE itself was considered to be highly preventable,
3.4% (9/267) or 1.3 per 1000 reviewed admissions were causally related to death.
It is likely that a similar re-analysis of the deaths in previous studies would
reduce the number (from which extrapolations of the fatal consequences of
healthcare have been made) by about a third according to our estimates.
Moreover, the retrospective review process (used to determine AE rate) may
underestimate the frequency of AEs, but conversely may overestimate their
consequences.
Author information:
Robin Briant, Formerly Clinical Director (NZQHS), Centre for Health Services
Research and Policy, School of Population Health, Faculty of Medical and Health
Sciences; John Buchanan, Associate Professor, Epidemiology and Biostatistics
Section, School of Population Health, Faculty of Medical and Health Sciences;
Roy Lay-Yee, Research Fellow, Department of Sociology, Faculty of Arts; Peter
Davis, Professor, Department of Sociology, Faculty of Arts; University of
Auckland, Auckland
Acknowledgements:
This study was funded by the Health Research Council of New Zealand. We
are also very grateful to the 13 participating New Zealand hospitals;
Dr David Richmond (Chair) and members of the study’s Advisory and
Monitoring Committee; Professor Stephan Schug; Professor Alastair Scott; Sandra
Johnson; Wendy Bingley; our medical review and data processing teams; and
hospital records staff.
Correspondence:
Associate Professor John Buchanan, Epidemiology and Biostatistics Section,
School of Population Health, Faculty of Medical and Health Sciences, University
of Auckland, Private Bag 92019, Auckland. Fax: (09) 373 7503; email: j.buchanan@auckland.ac.nz
References:
|
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| | Current
issue | Search journal |
Archived issues | Classifieds
| Hotline (free ads) Subscribe | Contribute | Advertise | Contact Us | Copyright | Other Journals |
|