Mortality following emergency admission for fractured neck of femur at west hertfordshire hospitals nhs trust

Supporting your commitment to excellence
Mortality following
emergency admission for
fractured neck of femur at
an English NHS Trust
Dr Rod Jones (ACMA, CGMA)
Statistical Advisor
Healthcare Analysis & Forecasting
[email protected]
01276 21061
For further articles please go to: www.hcaf.biz
Page 1 of 10
Supporting your commitment to excellence
The aim of this study is to determine if decease following an emergency admission for
a fractured neck of femur is higher than expected at the NHS Trust.
This report was commissioned due to the claim by the Dr Foster website that the age
standardised mortality was 13% higher than the national average.
The conclusions in this report are the independent views of Healthcare Analysis &
Forecasting.
1. Conclusions

1. Fractured neck of femur is a traumatic injury with a high incidence of ensuing 2. Approximately 1 in 10 patients die within 30 days of admission although this 3. Comparison of mortality between hospitals is not a direct measure of the quality of care since mortality is dependant on patient specific factors such as: exact type of the fracture, suitability for anaesthesia and associated co-morbidities such as respiratory infection, previous heart failure, dementia, malnutrition, alcoholism, etc. 4. Given the very small number of patients who actually die the claim for a real 13% higher mortality at the NHS Trust is not supported by statistical analysis. 5. The NHS Trust is within the expected range around the average in which 50% of hospitals will lie due to random variation in the condition of the arriving patients.
3. Introduction
Falls resulting in a fractured neck of femur are a significant cause of death1. The reason for the high mortality following a fractured neck of femur is partly to do with the relatively high average age of the patients and partly to do with the traumatic nature of this particular type of fracture and the associated major surgery required to correct the fracture. The risk of a fall is significantly increased by a sedentary lifestyle (such as confinement in a nursing home or due to general ill health)2. Some 80% of UK admissions are for women and decease within one year has been reported to lie in the range of 20 to 35% of admissions3 although for those over 85 years of age this risk has increased to 46% or higher (24% dying in hospital)4. In the UK some 73% of admissions are aged 69 and over5, however, for the NHS Trust this 1 Holden,J. et al (1998) British Journal of General Practice, 48 (432):1409-1412 2 Galgali,G. et al (1998) New Zealand Medical Journal, 111(1058), 7-10 3 Goldacre,M.J. et al (2002) British MedicaL Journal, 325 4 Jennings,A.G. and de Boer,P. (1999) Injury, 30(3), 169-172. 5 DOH data from HES (2001/02) Page 2 of 10
Supporting your commitment to excellence

proportion varies from 78% to 86% (00/01 to 02/03), i.e. a significantly higher
proportion of older persons.
The risk of mortality is higher for men although the risk of such a fracture and hence
the number of admissions is lower due to the lower incidence of osteoporosis in men6.
In the UK the proportion of male admissions averages 28% while for the NHS Trust
this average varies from 19% to 26% (00/01 to 02/03), i.e. significantly fewer male
admissions than national average.
Before launching into an analysis of the data it is always helpful to understand the
nature of the problem. In this instance mortality is not a direct measure of the quality
of care delivered by a hospital. This is due to the fact that the physical condition of the
patient has an enormous influence on the outcome. Hence it would be a cause for
concern if a relatively younger patient in good health and with an uncomplicated
fracture were to die. On the other hand a patient with a history of alcoholism, hence
malnutrition and poor kidney function would have a high expectation of mortality.
The point of relevance is that mortality statistics used within the UK (and the Dr
Foster website) although having some adjustment for age do not take any other factors
into account relating to the general state of the patient. Given that we are dealing with
relatively small numbers the claim for a higher mortality rate needs to be understood
in its correct context. Indeed we need to understand if the quoted figure of 13% higher
mortality is merely a statistical artefact. In other words is the mortality in the range
expected given the influence of patient condition on the final outcome?
To answer these questions we will explore the range of patient specific risk factors;
look at the indicators of hospital performance which could contribute to the outcome
and finally explore the limitations of the statistical methods used to report relative
mortality rates.
4. Patient Specific Risk Factors

4.1.1. Type of Fracture

Risk of mortality is higher for trochanteric type fractures7.
4.1.2. Bronchopneumonia

Surgery for fractured neck of femur on patients with bronchopneumonia results in
only a 10% survival rate. Due to this enormous risk of decease it has been suggested
that early surgical intervention should be denied8.
4.1.4. Suitability for Anaesthesia

The operation to repair a fractured neck of femur is major surgery with prolonged
exposure to anaesthesia. Using the American Society of Anaesthetists (ASA) scoring
6 Kazar,G. et al (1997) Orvosi Hetilap, 138(50), 3173-3177 7 Kazar,G. et al (1997) Orvosi Hetilap, 138(50), 3173-3177 8 Ryder,S.A. et al (2001) Injury, 32(4), 295-297 Page 3 of 10
Supporting your commitment to excellence

system (1 = fit for anaesthesia through to 5 = totally unfit) one study showed that
patients with an ASA score of 4 or 5 had only a 10% chance of survival. It was
considered that early surgical intervention should be denied to such patients9. Another
study looking at one and three year mortality showed a 1.7-times higher risk for ASA
3,4 or 5 patients10.
4.1.4. Previous Heart Failure

A previous heart failure increases the risk of decease following surgery however on
its own it was not deemed to be sufficient to warrant the withholding of surgery11.

4.1.5. Nutritional Status

Patients classified as being malnourished are twice as likely to die within one year
after admission to hospital for any reason12. A similar risk has been identified for
patients with chronic body mass depletion (malnutrition or other causes)13. This risk
also applies to patients admitted for fractured neck of femur14.
4.1.6. Physical Activity

Physical activity is known to reduce the likelihood of all cause mortality15. The lack
of physical activity is a causative factor in the increased risk of fractured neck of
femur16.
4.1.7. Mental State

Patients with senile dementia were reported to have 3-times higher mortality at four
months or one year after surgery for fractured neck of femur. Patients with senile
dementia also had a higher risk of being far less mobile after the operation. For this
group of patients the less invasive method of internal fixation is recommended rather
than hemiarthroplasty17. However, patients with a prior cerebrovascular accident did
not appear to have a higher risk of decease (except that they tended to have an ASA
rating of 3 or 4)18.
4.1.8. Secondary Infection by Salmonella

The fact that a patient has had a fall of sufficient violence to cause a fractured neck of
femur and the resulting internal trauma caused by the fractured bones is sufficient to
increase the risk of Salmonella infection with consequent multi-organ failure19.
9 Ryder,S.A. et al (2001) Injury, 32(4), 295-297 10 Hamlet,W.P. et al (1997) American Journal of Orthopaedics, 26(9), 621-627. 11 Ryder,S.A. et al (2001) Injury, 32(4), 295-297 12 Persson,M.D. et al (2002) Journal of the American Geriatrics Society, 50(12): 1996-2002. 13 Liu,L. et al (2002) Journals of Gerontology Series A, 57(11), M741-746 14 Ryder,S.A. et al (2001) Injury, 32(4), 295-297 15 Crespo,C.J. et al (2002) Annals of Epidemiology, 12(8), 543-552. 16 Galgali,G. et al (1998) New Zealand Medical Journal, 111(1058), 7-10 17 vanDortmont,L.M. et al (2000) Injury, 31(5), 327-331 18 Youm,T. et al (2000) Journal of Orthopaedic Trauma, 14(5), 329-334 19 Mofredj,A. et al (2001) Intensive Care Medicine, 27(5), 950-951 Page 4 of 10
Supporting your commitment to excellence

4.1.9. Pulmonary Fat Embolism

Some patients are at risk of death due to the onset of fat embolism in the lungs and
kidneys leading to hypouresis, pulmonary edema & hypoxia and cardiac arrest20.
4.1.10. Patients with Cancer

Patients with significantly advanced malignancy had poor outcomes following
surgery and should not automatically be referred for surgery21.
4.1.11. Alcoholism, Drug Addiction & Smoking

Patients suffering from alcoholism, drug addictions or long term smoking all have a
higher risk of decease.
5. Hospital Specific Risk & Quality Factors
While the risk of decease may be a possible gross indicator of ‘quality’ the two
measures, namely, risk of decease and quality of care should perhaps be considered as
separate elements to the full picture.

5.1. Hospital Factors Increasing the Risk of Decease
5.1.1. Delay to Surgical Intervention

There is clear evidence to show that survival is improved for those patients receiving
surgical intervention on the day of admission compared to those receiving
intervention on the second day22. The risk of decease was reported to be 1.7-times
higher23 in one study and 4.5-times higher in another24.
The delay experienced by a patient before surgery is made up from two components:
Suitability for immediate surgery
Section 4 has already detailed a number of instances where medical opinion is that immediate surgery should be withheld until the patient is fit enough. Size of the department
Patients arriving with fractured neck of femur can arrive outside normal working hours and over the weekend. Their treatment usually requires a dedicated theatre (and associated dedicated theatre staff) to avoid cancellation of the normal non-emergency surgery. The immediate availability of an anaesthetist is also required. Smaller departments are not able to offer this level of service and hence will experience a delay to surgery particularly when a patient is admitted over the weekend. 20 Mori,K. et al (1999) Japanese Journal of Anaesthesiology, 48(4), 416-418 21 McNamara,P. et al (1997) Age & Ageing, 26(6), 471-474 22 Smektala,R. et al (2000) Zentralblatt fur Chirurgie, 125(9), 744-749 23 Beringer,T.R. et al (1996) Ulster Medical Journal, 65 (1), 32-38 24 Hamlet, W.P. et al (1997) American Journal of Orthopaedics, 26(9), 621-627 Page 5 of 10
Supporting your commitment to excellence

Hence analysis of data from German hospitals showed that the larger departments
able to specialise in trauma had a delay before operation which was on average 0.5 to
0.7 days shorter than the smaller departments.
International benchmarks are not readily available, however, in 1989 the Royal
Victoria Hospital (Belfast) was reporting that 57% of patients received surgery on the
day of admission25. Of all the factors this is the one most likely to be behind a
persistent higher mortality rate at particular NHS sites.
The NHS Trust delivers Orthopaedic services from two hospital sites, namely, A and
B. Volume of admissions per annum is around 200 at A with slightly more at B.
.
5.1.2. Treatment for blood clotting

The risk of fatal pulmonary embolism following surgery for fractured neck of femur
is known to be relatively high. One study showed that treatment with heparin reduced
the risk of death but that the risk of embolism extended well beyond discharge from
hospital26. Another study showed that aspirin had the potential to prevent 4 fatal
pulmonary emboli per 1,000 patients; however, this was balanced against an increased
risk of aspirin-induced bleeding requiring post operative transfusion (6 per 1,000
patients)27.
5.1.3. Preparation of the cement used to fix the bones

The following is one example of a recent development in medical technology which
may make a significant impact on future hospital practice. In this German study the
authors postulated that the high level of circulatory and pulmonary problems
following repair of fractured neck of femur may be due to air embolism during
polymerisation of the methylmethacrylate used to cement the bones. To prevent this
they proposed that the cement should be mixed under vacuum. In a small trial with 72
patients (36 in each group) the group with cement mixed under vacuum had a 6-times
reduction in the incidence of pulmonary embolism and a 4-times reduction in
complications from poor arterial oxygenation. Mortality was reduced from 13.8% in
the group with normal mixing of cement down to 2.8% where cement was mixed
under vacuum28.
The above finding needs to be validated by further studies; however, it does
demonstrate how a breakthrough in practice can lead to a dramatic reduction in
mortality.
5.2. Hospital Factors Leading to Quality of Care

Our knowledge of the factors leading to higher quality of care is developing over
time. Hence it has taken many years of controlled trials to establish which factors lead
25 Beringer,T.R. et al (1996) Ulster Medical Journal, 65(1), 32-38 26 Frostick,S.P. (2000) Haemostasis, 30 Suppl 2, 82-87 27 Anon (2000) Lancet, 355(9212), 1295-1302 28 Leidinger,W. et al (2002) Unfallchirung, 105(8), 675-679 Page 6 of 10
Supporting your commitment to excellence

to a high ‘quality’of care. This is because as many of the patient specific factors as
possible need to be ruled out to establish the true cause of the outcome.
5.2.1. Choice of Procedure

There are three possible interventions for fractured neck of femur, namely, internal
fixation, uncemented hemiarthroplasty or cemented total hip arthroplasty gives the
best outcomes.
In this respect one study of nearly 300 over 65 year olds over a period of 13 years
showed no statistical difference in mortality between the three procedures, however,
in terms of the need for longer-term revisional surgery, complications or patient
mobility and discomfort total hip arthroplasty gave the best results (in spite of higher
earlier complications, increased blood loss and longer time for surgery)29. This
general conclusion has been supported by other recent studies and reviews30,31,32,33.
One review has suggested that internal fixation is a cause for higher longer-term
mortality34
5.2.2. Post-operative Rehabilitation

For the surviving patients the ‘quality’ of the post operative rehabilitation is probably
the most influential aspect of a return to a normal quality of life (walking ability and
associated painfulness)35.
6.
Measuring Increased Risk

6.1. Death in Hospital
Death within hospital could be considered to be a valid measure of the risk of decease.
However the medical literature is quite clear regarding the fact that the risk of decease
is elevated for any patient undergoing surgery for fractured neck of femur with
increased levels of in-hospital and ex-hospital 1, 3 & 5 year mortality, i.e. mortality
must be measured at an identical point in time to give comparable results.
Decease while in hospital is therefore not an equally comparable measure for the
simple reason that the time spent in hospital depends greatly on the supporting
infrastructure around a hospital. Hence some hospitals discharge patients very early
for convalescence in a nearby community hospital or with a supporting hospital-at-
home scheme. Those hospitals with supporting infrastructure will therefore have
correspondingly lower numbers of deceases. As a general comment the NHS Trust
does not have access to early discharge schemes for its orthopaedic patients and
would therefore be expected to have a higher number of in-hospital deaths due to the
simple fact that the patients stay for a longer period of time.
29 Kumar,K.J. & Marsh, G (2000) Injury, 31(10), 793-797 30 Khan,R.J. et al (2002) Injury, 33(1),13-17 31 Rogmark,C. et al (2002) Acta Orthopaedica Scandinavica, 73(6), 605-610 32 Cree,M. et al (2002) Canadian Journal of Surgery, 45(4), 248-254. 33 Broos,P.L. (1999) Acta Chirurgica Belgica, 99(4), 190-194 34 Hudson,J.I. et al. (1998) Clinical Orthopaedics & Related Research, 1998 March (348), 59-66 35 Kazar,G. et al. (1997) Orvosi Hetilap, 138(5), 3173-3177 Page 7 of 10
Supporting your commitment to excellence

6.1. Death within 30 Days of Admission

The choice of death (from all causes) within 30 days of admission used within the UK
appears to be a somewhat arbitrary cut off point, however, the main aim is to
introduce an element of consistency in order to compare like with like. The inclusion
of death from all causes has been defended in a recent study on the grounds that
mention of a previous fractured neck of femur is often omitted from the death
certificate36.
Hence the UK measure of death within 30 days is as good a measure as any other of
the in-hospital and short-term ex-hospital deaths. Given the short timescale between
admission and the 30 day cut off point the inclusion of deaths from all causes appears
to be justified.
5.2. Is a one year snap-shot a good choice for such an indicator?

Practice within the UK as promoted by the DOH appears to be out of line with
international studies on mortality following any hospital intervention. It would seem
that the pressure for short-term measures via hospital star ratings has confused many
of the issues.
International practice tends to recognise the need for higher volumes of data and
hence studies within a single hospital or between groups of hospitals are usually an
aggregate of 6 or more years of data37,38. The study by Jarman et al (1999)39 of deaths
from all causes in English hospitals used three years data across 183 hospitals – a very
large data set indeed. This latter study indicated that the most significant factor in all
deaths within hospital was the number of doctors per 100 beds and the number of
GP’s per head of population. This explains why the Dr Foster website gives
information on the number of doctors per 100 beds as a measure of ‘quality’.
The next figure gives data for all English hospital Trusts for death following a
fractured hip. The age and sex standardised rates have been calculated on two
successive years. It is immediately apparent that in most instances the score from one
year to the next is controlled by random variation rather than an genuine measure of
‘quality’.
In this respect one Trust had a score of 18,897 per 100,000 deaths per admission in
1998/99 (the highest in the UK) yet in the following year had a score of only 3,893
(the lowest in the UK)! Such massive variation is simply not possible – rather it is an
artefact of the method used to score and hence rank hospitals.
The random variation caused by the patient specific risk factors in what are small
number events is overwhelming the attempt to make any meaningful measure of
‘quality’.
36 Goldacre,M.J. et al (2002) British Medica Journal, 325 (7369), 868-869 37 Broos,P.L. (1999) Acta Chirurgica Belgica, 99(4), 190-194 38 Smektala,R. et al (1999) Chirurg., 70(11), 1330-1338 39 Jarman,B. et al (1999) British Medical Journal, 318, 1515-1520 Page 8 of 10
Supporting your commitment to excellence
Age and sex standardised mortality rates for English hospital
0
/0
9 12000
9
9
1 10000
in
re
o 8000
Score in 1998/99

5.3. Is a league table approach helpful?

A league table approach is one where hospitals are ranked from so-called best to
worst. This is similar to that used on the Dr Foster website. Hence at first sight a score
of 113 for mortality from fractured neck of femur at the NHS Trust would lead to the
conclusion that any patient had a 13% higher risk of decease at this hospital Trust.
Fortunately this is not the case and a recent publication in the British Medical Journal
has suggested that an approach based on Industrial Control Charts gives a much more
reliable indication of the relative risk of decease40. This approach gives proper
recognition to the role of random variation due to the multitude of patient specific risk
factors.
A similar approach will be employed to demonstrate that the NHS Trust has a normal
level of decease from fractured neck of femur.
Age and Sex Standardised Mortality Rates

Due to the fact that the patient specific factors discussed above are not routinely
recorded for every patient in the data used by the Department of Health the minimum
level of adjustment is therefore made to account for the effect of the age and sex of
the patient – since these pieces of information about the patient are standard data
items.
While this may seem a relatively unambiguous way of achieving some degree of
standardisation there are still pitfalls for the unwary. These pitfalls arise from the use
of what are called standard 5 year age bands up to the age of 85 and a final age band
for those over 85 years. Given that the risk of decease is far greater for anyone above
this age the use of a single age band is extremely unsatisfactory. Even the use of 5
40 Adab,P et al (2002) British Medical Journal, 324, 95-98
Page 9 of 10
Supporting your commitment to excellence

year age bands can be questioned on the grounds that a couple of patients more than
average at the upper end of the age band could be sufficient to skew the results. The
availability of Pentium processors in even the most basic desktop computer means
that single year standardisation should be the basic method.
5.5. Is the National Average a Valid Benchmark?

The comparison between a single hospital and the national average assumes that the
national average is a reliable benchmark.
However data for fractured hip shows that the so-called national average varied from
9,898 deaths per 100,000 admissions in one year to 10,331 the next, i.e. the national
average is itself showing considerable movement.
As can be seen even the so-called national average is itself uncertain. This is because
the risk of decease is influenced by the multitude of patient specific risk factors. Even
at a national level the volume of patients is still too small to even out the random
fluctuation caused by such patient specific factors.
Hence it should be obvious that if the national average is uncertain then the
calculation of a standardised mortality rate for a single hospital will be an even less
certain indicator of ‘quality’.
How do we account for random variation?

Given the fact that data for UK hospitals is not adjusted for the patient specific risk
factors how do we discern if one hospital with an age and sex standardised mortality
rate of 87 is any different from another hospital with a rate of 113.
After all, as a potential patient we are looking for real assurance that the local hospital
is at least as good as the national average. We can also certainly do without the
unnecessary doubt cast over the relative performance of our local hospital should a
score of 113 prove to measure little more than a score of 87, i.e. a measure of random
variation rather than any real ‘quality’ factors.
Dr Rod Jones provides consultancy and analysis for healthcare organisations. He can
be contacted at [email protected]
Page 10 of 10

Source: http://www.hcaf.biz/Hospital%20Efficiency/Mortality_FNOF.pdf