Case fatality rates after admission to hospital with stroke: linked database study
BMJ 2003; 326 doi: https://doi.org/10.1136/bmj.326.7382.193 (Published 25 January 2003) Cite this as: BMJ 2003;326:193All rapid responses
Rapid responses are electronic comments to the editor. They enable our users to debate issues raised in articles published on bmj.com. A rapid response is first posted online. If you need the URL (web address) of an individual response, simply click on the response headline and copy the URL from the browser window. A proportion of responses will, after editing, be published online and in the print journal as letters, which are indexed in PubMed. Rapid responses are not indexed in PubMed and they are not journal articles. The BMJ reserves the right to remove responses which are being wilfully misrepresented as published articles or when it is brought to our attention that a response spreads misinformation.
From March 2022, the word limit for rapid responses will be 600 words not including references and author details. We will no longer post responses that exceed this limit.
The word limit for letters selected from posted responses remains 300 words.
Case-fatality rates after stroke
We agree with the comments made by Michael Hill and by David Liebeskind about contrasting case-fatality rates for intracerebral haemorrhagic and ischaemic stroke. However, there is a problem with reliably differentiating the two, at least when using administrative databases for comparisons over long periods of time and between hospitals, because a high percentage of cases of stroke are not specified as haemorrhagic or ischaemic.
We amplify here our findings for the time-periods and hospitals covered in
our paper. In the first period, 1978-87, 7% of the cases were recorded as haemorrhagic,
9% as ischaemic, and 84% were unspecified. In the second period, 1988-97, 8%
were recorded as haemorrhagic, 10% as ischaemic, and 82% were unspecified. The
table shows case-fatality rates (CFRs) at 30 days after admission. It shows,
for example, that in the second period the CFR at 30 days was 41% for haemorrhagic,
21% for ischaemic, and 34% for unspecified stroke. CFRs varied considerably
by age (see table below; see also the web table accompanying our published paper).
For example, in people under 55 years of age the CFR at 30 days was 32% for
haemorrhagic and 10% for ischaemic stroke (table below).
In our paper we reported that CFRs had fallen between the two time periods
we compared, although we did not give the data separately for the different
types of stroke. As we show in the table, the decline was seen for each type
of stroke and, at least in this respect, the overall decline reported by us
is unlikely to be attributable to just a change in case mix.
We also reported significant differences in CFRs between the teaching and the
non-teaching hospitals. The difference is unlikely to be a consequence of different
case-mix in respect of the types of stroke. The CFRs at 30 days were lower in
the teaching than the non-teaching groups, in each period of time, for each
of the three types of stroke shown.
However, in such comparisons, consideration needs to be given to factors that determine whether a stroke is investigated and/or recorded specifically as haemorrhagic or ischaemic. The level of investigation varies with age (a smaller percentage were of unspecified type in the younger age groups in our data). It has no doubt changed over time with the introduction of new diagnostic methods. The level of investigation and precision of diagnosis is also probably related to case severity, and therefore to outcome, in ways that are difficult to quantify. On the one hand, investigation is more likely to be undertaken in patients whose prognosis may be improved by specific intervention and who may not be in imminent danger of death; on the other hand, post-mortem information about people who die may add to the likelihood that the hospital record specifies the stroke by type. There may not be consistency between hospitals in these respects. The level of either investigation or recording, or both, varied between hospitals in our study: for example, a smaller percentage of strokes were of unspecified type in the teaching than in the non-teaching hospitals. The teaching hospital serves a large, natural catchment population and the majority of its patients admitted as emergencies with stroke are local rather than specialist referrals. There were also some quite large differences between the individual non-teaching hospitals in their level of recording of stroke of specific and unspecified types. We doubt that there is much real difference between the populations served by these hospitals in the percentage of strokes that are haemorrhagic or ischaemic. For all these reasons, using retrospective data for comparisons over long periods of time and between hospitals, we think that analysis of stroke overall, combining types, is probably more secure than analysis of the different types of stroke. However, we agree that information about the individual types is also important.
Michael Goldacre
Stephen Roberts
Table: Age-specific case-fatality rates by type of stroke, time period
1978-87 and 1988-97, and age-group: admissions (A), deaths (D), case-fatality
rates at 30 days (CFR) and their 95% confidence intervals
(a) |
Haemorrhagic stroke*
|
||||
1978-87 | |||||
age group |
A |
D |
CFR |
(95% CI) |
|
<55 |
111 |
48.9 |
42.4 |
-55.4 |
|
55-64 |
200 |
122 |
61.0 |
54.2 |
-67.8 |
65-74 |
221 |
140 |
63.3 |
57.0 |
-69.7 |
75-84 |
185 |
131 |
70.8 |
64.3 |
-77.4 |
85+ |
33 |
26 |
78.8 |
64.8 |
-92.7 |
Total ‡ |
866 |
530 |
61.2 |
58.0 |
-64.4 |
1988-97 | |||||
age group |
A |
D |
CFR |
(95% CI) |
|
<55 |
382 |
123 |
32.2 |
27.5 |
-36.9 |
55-64 |
302 |
123 |
40.7 |
35.2 |
-46.3 |
65-74 |
443 |
195 |
44.0 |
39.4 |
-48.6 |
75-84 |
423 |
184 |
43.5 |
38.8 |
-48.2 |
85+ |
139 |
74 |
53.2 |
44.9 |
-61.5 |
Total ‡ |
1689 |
699 |
41.4 |
39.0 |
-43.7 |
(b) |
Ischaemic stroke*
|
||||
1978-87 | |||||
age group |
A |
D |
CFR |
(95% CI) |
|
<55 |
182 |
17 |
9.3 |
5.1 |
-13.6 |
55-64 |
225 |
44 |
19.6 |
14.4 |
-24.7 |
65-74 |
342 |
111 |
32.5 |
27.5 |
-37.4 |
75-84 |
347 |
152 |
43.8 |
38.6 |
-49.0 |
85+ |
121 |
72 |
59.5 |
50.8 |
-68.3 |
Total ‡ |
1217 |
396 |
32.5 |
29.9 |
-35.2 |
1988-97 | |||||
age group |
A |
D |
CFR |
(95% CI) |
|
<55 |
288 |
30 |
10.4 |
6.9 |
-13.9 |
55-64 |
375 |
48 |
12.8 |
9.4 |
-16.2 |
65-74 |
606 |
105 |
17.3 |
14.3 |
-20.3 |
75-84 |
632 |
179 |
28.3 |
24.8 |
-31.8 |
85+ |
243 |
89 |
36.6 |
30.6 |
-42.7 |
Total ‡ |
2144 |
451 |
21.0 |
19.3 |
-22.8 |
(c) |
Unspecified stroke*
|
||||
1978-87 | |||||
age group |
A |
D |
CFR |
(95% CI) |
|
<55 |
595 |
105 |
17.6 |
14.6 |
-20.7 |
55-64 |
1359 |
367 |
27.0 |
24.6 |
-29.4 |
65-74 |
3186 |
1134 |
35.6 |
33.9 |
-37.3 |
75-84 |
4110 |
1676 |
40.8 |
39.3 |
-42.3 |
85+ |
1441 |
692 |
48.0 |
45.4 |
-50.6 |
Total ‡ |
10691 |
3974 |
37.2 |
36.3 |
-38.1 |
1988-97 | |||||
age group |
A |
D |
CFR |
(95% CI) |
|
<55 |
679 |
74 |
10.9 |
8.6 |
-13.2 |
55-64 |
1592 |
317 |
19.9 |
18.0 |
-21.9 |
65-74 |
4188 |
1162 |
27.7 |
26.4 |
-29.1 |
75-84 |
7218 |
2695 |
37.3 |
36.2 |
-38.5 |
85+ |
3662 |
1614 |
44.1 |
42.5 |
-45.7 |
Total ‡ |
17339 |
5862 |
33.8 |
33.1 |
-34.5 |
* The ICD-8 codes and terms used were: haemorrhagic stroke (431), occlusive or thrombotic or embolic stroke (432-434) and ‘acute but ill-defined’ stroke (436). The ICD-9 codes and terms used were: haemorrhagic stroke (431-432), occlusive stroke (433-434) and ‘acute but ill-defined’ stroke (436). The ICD-10 codes and terms used were: haemorrhagic stroke (I61-I62), cerebral infarction (I63) and ‘acute but ill-defined’ stroke (I64).
‡ The age of 81 patients in 1978-87, and 53 in 1988-97, were not known.
Competing interests:
None declared
Competing interests: 227
EDITOR - Stroke is a clinical syndrome characterized by rapidly
developing clinical signs of focal (or global) disturbance lasting 24
hours or longer or leading to death with no apparent cause other than of
vascular origin. The definition of stroke incorporates hemorrhagic and
ischemic lesions, with additional subtypes in both categories. The
incidence and mortality rates vary widely across stroke subtypes, likely
reflecting underlying differences in etiology and pathophysiology.
Estimates of stroke mortality that fail to consider these inherent
characteristics are misleading. Statistical analyses of stroke incidence
and mortality are critically dependent on the definition of stroke.
Roberts et al report case fatality rates after hospital admission for
"stroke" based on ICD-9 (International Classification of Diseases, 9th
revision) codes.1 This method of coding for stroke diagnosis is known to
be inaccurate due to lack of specificity, and analyses should be adjusted
for published positive predictive values of each ICD-9 code. The authors
exclude subarachnoid hemorrhage (ICD-9 430), yet they include other and
unspecified intracranial hemorrhage (ICD-9 432). Ischemic strokes (ICD-9
433, 434, 436) are then lumped with intracerebral hemorrhage (ICD-9 431).
This lumping leads to a gross exaggeration of ischemic stroke mortality
figures. The authors report unadjusted case fatality for in-hospital
stroke at 29.7% between 1988-1997 in the former Oxford health region of
England.1 In-hospital mortality for ischemic stroke admissions in the
United States during 1997 has been estimated at 5.3%, compared to 30.9%
for intracerebral hemorrhage.2 The mortality figures reported by Roberts
et al are likely skewed by the elevated in-hospital mortality associated
with intracerebral hemorrhage. The in-hospital stroke case fatality rate
in the database analyzed by Roberts et al appears inflated due to their
selected definition of stroke. Determination of stroke mortality rates is
meaningless, unless the definition of stroke is reasonably specific or
representative of a particular clinical syndrome.
1. Roberts SE, Goldacre MJ. Case fatality rates after admission to
hospital with stroke: linked database study. Bmj 2003;326(7382):193-4.
2. Liebeskind DS, Sayre JW. Decreasing incidence of stroke in the United
States: 1996-2000 [abstract]. Stroke 2003;34(1):287.
Competing interests:
None declared
Competing interests: No competing interests
With more aggressive efforts being made to improve the poor prognosis
in patients with acute strokes(1) it may be of value to consider the
distinction between the putative bnefits of treatment as opposed to
prevention as considered for myocardial infarctions (2).
In short resuscitative efforts and therapeutic revascularisations are
of no benefit to brain that is irrversibly damaged and might be largely
responsible for the secondary brain dysfunction and injury that develops
in these patients and those with head injuries. From a pathophysiologic
perspective there would appear to be much more to be gained from the much
earlier identification and elective reversal of metabolically significant
occlusive vascular lesions in asymptomatic patients. The challenge is to
indentify these lesions, vascular anatomy having no absolutely correlation
with metabolic status in those tissue beds that have been appropriately
investigated.
1. Case fatality rates after admission to hospital with stroke:
linked database study
Stephen E Roberts and Michael J Goldacre
BMJ 2003; 326: 193-194
2.Fiddian-Green RG, rapid reponse to: Impact of changing diagnostic
criteria on incidence, management, and outcome of acute myocardial
infarction: retrospective cohort study
J P Pell, E Simpson, J C Rodger, A Finlayson, D Clark, J Anderson, and A C
H Pell
BMJ 2003; 326: 134-135
Competing interests:
None declared
Competing interests: No competing interests
To the editor:
The short report by Roberts et al[1] on stroke mortality in-hospital
and at one year post-stroke raises several issues. The case-mix ratio,
presence or absence of on-site neurosurgical services and stroke unit care
should be emphasized as important and easily measured explanatory
covariates. Unfortunately, the most important predictor of 30-day
mortality is stroke severity at baseline and this variable is not captured
using administrative data. Nevertheless, intracerebral haemorrhage (ICH)
is accurately coded in hospital discharge abstracts (ICD-9 code 431).
Therefore, grouping the cases by diagnosis - ICH vs. ischemic stroke (ICD-
9 codes 433, 434, 436) - is valid for ICH. Although ICH makes up only
about 10% of all stroke, ICH has a higher acute mortality than ischemic
stroke. In Toronto, Canada, the 30-day mortality rate at a neurosurgical
centre for primary intracerebral haemorrhage was 27.4%[2]. Therefore,
centres that receive ICH patients may have higher aggregate mortality
rates and adjustment for case-mix would be relevant.
While it remains unclear whether acute neurosurgical intervention
improves overall long-term outcome in ICH, it may reduce early
mortality[3]. At experienced centres, aggressive interventional care
shows promise in improving both morbidity and mortality[4]. One surrogate
for neurosurgical availability is teaching hospital status and
observational evidence that differences in mortality exist by hospital
type, after adjustment for case-mix is worth exploring.
The only intervention that reduces mortality for ischemic stroke (and
probably also for ICH) is stroke unit care. The publication of randomized
trial evidence for and development and implementation of stroke units have
occurred over the time period of the current study. Evaluation of 30-day
mortality by hospital stratified by presence or absence of a stroke unit
could provide important observational evidence of the “real-life”
effectiveness of stroke units. Such observations are important for the
development of appropriate health policy.
1. Roberts SE, Goldacre MJ. Case fatality rates after admission to
hospital for stroke: linked database study. BMJ 2003; 326: 193-194
2. Hill MD, Silver FL, Tu JV, Austin P. Rate of recurrence of stroke
after primary intracerebral hemorrhage. Stroke 2000; 31: 123-127
3. Qureshi AI, Tuhrim S, Broderick J, Batjer HH, Hondo H., Hanley D.F.
Spontaneous Intracerebral Hemorrhage. The New England Journal of Medicine
2001;344:1450-1460.
4. Naff NJ, Carhuapoma JR, Williams MA, Bhardwaj A, Ulatowski JA,
Bederson J, Bullock R, Schmutzhard E, Pfausler B, Keyl PM, Tuhrim S,
Hanley DF. Treatment of intraventricular hemorrhage with urokinase:
effects on 30-day survival. Stroke. 2000:31:841-847.
Competing interests:
None declared
Competing interests: No competing interests
Rapid Fall in Hospital Case-Fatality may be due to Decline in Stroke Severity
Dear Sir
Roberts and Goldacre [1] found a fall in in-hospital case fatality
rates for stroke from 35.4% to 29.7% over an average 10 year period. They
speculate that much of this fall could be due to changes in case-mix, and
particularly in stroke severity. We have used our hospital stroke
register to investigate this possibility.
We kept a prospective register of all acute stroke admissions to our
hospital between January 1997 and December 2001. Diagnosis was confirmed
by a stroke physician. Stroke severity was estimated from the Scandinavian
Stroke Scale score on admission (SSS>44/58 = ‘mild’; <_19 _="_" severe.="severe." _5-year="_5-year" trends="trends" were="were" assessed="assessed" using="using" rank="rank" correlation="correlation" kendall.="kendall." p="p"/> During this period 2199 patients were admitted with acute stroke. The
total number of annual admissions, age (median 74 yrs) and sex
distributions (50% male), the proportion with previous stroke or TIA (31%)
and length of in-patient stay (median14 days) did not change
significantly. The proportion of mild strokes increased from 46% in 1997
to 54% in 2001, while severe strokes decreased from 13% to 9% (p<_0.01 for="for" trend="trend" in="in" both="both" cases.="cases." between="between" _1997="_1997" and="and" _2001="_2001" case="case" fatality="fatality" fell="fell" from="from" _28="_28" to="to" _16="_16" p0.001="p0.001" the="the" proportion="proportion" discharged="discharged" with="with" barthel="barthel" index15="index15" _20="_20" _23="_23" p="p" though="though" these="these" trends="trends" were="were" no="no" longer="longer" significant="significant" after="after" adjusting="adjusting" initial="initial" stroke="stroke" severity="severity" treating="treating" sss="sss" score="score" as="as" a="a" continuous="continuous" variable.="variable."/> Although our sample was relatively small, our data were collected
prospectively according to standardised criteria [2], and our results are
not subject to the limitations of retrospectively-coded hospital activity
data [3]. The year-on-year trends were remarkably consistent and amounted
to a fall in case fatality rate of over 40% in 5 years, apparently due to
a decline in stroke severity. If our findings are replicated in other
units, the changes in the pattern of stroke disease in British hospitals
may be even more rapid than Roberts and Goldacre’s results indicate.
1. Roberts SE, Goldacre MJ. Case fatality rates after admission to
hospital for stroke: linked database study. BMJ 2003; 326: 193-194
2. Barer D, Ellul J, on behalf of the European Stroke Database
Collaboration. A "basic stroke register" for studying variations in
casemix and outcome of stroke care in different European centres.
Cerebrovasc Dis 1996; 6(S2):55
3. Barer D, Ellul J, Watkins C. Correcting outcome data for case mix in
stroke medicine. BMJ 1996; 313:1005-6 (letter)
Competing interests:
None declared
Competing interests: No competing interests