LATE last month, Harold Shipman was convicted of killing 15 of his elderly
patients, and is thought to have murdered more than a hundred others. How did
Shipman, a family doctor in a small town on the outskirts of Manchester, get
away with his crimes for so many years?
It is the worst example of serial killing in British history, prompting the
government to announce a public inquiry and a rash of calls for new ways of
keeping tabs on doctors and their practices鈥攏ot least from within the
medical profession. No sooner had Shipman begun his life sentence than the
British Medical Association (BMA) demanded the introduction of computer
technology which could alert the authorities to such anomalies as the abnormally
high death rate among Shipman鈥檚 patients.
鈥淗ere we are, with technology in most parts of our lives, but there isn鈥檛 the
ability to actually select out trends with the sort of technology we should
have,鈥� says Ian Bogle, chairman of the BMA.
Advertisement
In fact, some NHS experts say the technology has been around for years, and
is neither horribly expensive to install nor especially complex to use. Indeed,
geographical information systems (GIS) software has already been used to
investigate another notorious medical scandal surrounding implausibly high death
rates. Basically, GIS is software that allows geographical data to be stored and
manipulated in a way that makes spotting trends very easy. Technologically it
doesn鈥檛 represent a huge advance.
But according to GIS experts, this technology has yet to be used
significantly by senior health officials in any developed nation鈥攍et alone
exploited to its full potential. But there are signs that in Britain, at least,
the authorities are waking up to its potential. According to Alistair Calder,
who manages a GIS at Britain鈥檚 Office for National Statistics (ONS), 鈥渋f there
is a desire to stop future Dr Shipmans we would have to tell the DoH they should
get this sort of technology鈥�.
Identifying clusters of death has certainly led to major advances in public
health over the last 150 years. In 1854, a London doctor called John Snow had a
hunch about a major cholera outbreak which had already claimed 500 lives. Simply
by plotting the location of victims on a street map and correlating this with
the source of their water supply, Snow was able to pinpoint one contaminated
water pump. Snow removed the handle, and halted the epidemic.
Environmental hazard
Amazingly, almost 150 years later, clusters of deaths caused by disease,
environmental hazard or murderous doctors are still most likely to be spotted by
anecdote and pins on maps. For while doctors are compelled to report outbreaks
of many communicable diseases to the Public Health Laboratory Service, which
advises the Department of Health (DoH), a wealth of other potential insights go
begging for lack of systematic data collection.
鈥淲ith good, up-to-date data, then it would be possible to use computer
technology to regularly look at areas and see if and when excess deaths occur,鈥�
says GIS expert Lars Jarup at St Mary鈥檚 Hospital, London. As a member of the
Small Area Health Statistics Unit (SAHSU), Jarup speaks from experience. SAHSU
was set up by the DoH in 1987 in the wake of concern about links between nuclear
installations and childhood leukaemia. Over the years, the unit has been called
in to investigate many clusters of death and disease, using sophisticated
statistical methods to explore their true significance.
The basic idea, however, is simple: the cluster is identified on a digital
map of Britain, which stores both relevant medical information and standard
geographical information. Then the rate of illness within a radius of 2
kilometres is compared with that in the surrounding areas between 2 and 7.5
kilometres from the centre of the cluster. Statistical tests then show whether
the cluster is down to chance or something more malign.
While highlighting disease clusters may cause concern, the findings are often
reassuring. For example, the Department of Public Health Medicine at Barking and
Havering Health Authority, Essex, recently asked SAHSU to investigate complaints
about a suspected cluster of respiratory illnesses around two local factories
making chemicals for the paint industry. Comparing the rate of death and
hospital admissions from lung disease within 2 kilometres of the factories with
those for the greater area stored on a digital map, the SAHSU team was able to
allay fears about the effect of the factories.
Sometimes, however, SAHSU finds hard statistical support for mere anecdote
and suspicion. Last autumn, the unit submitted key evidence to a government
inquiry set up following 29 deaths between 1984 and 1995 among babies with a
rare heart condition who had been operated on by surgeons at the Bristol Royal
Infirmary.
By comparing mortality rates for the child heart surgery at the Bristol
hospital with those for the rest of Britain, the SAHSU team found that it had
the worst mortality rate for these operations in England鈥攁t around 63 per
cent, almost four times the national average.
No differences emerged that could plausibly explain such a rate. The
conclusion had to be that some of the surgeons performing the operations were
not up to the job. Yet as with the Shipman case, the results of their
actions initially came to light only through the suspicions of individuals. All
the heavy-duty statistical analysis came later.
So could GIS and analysis of patient information really have prevented it? In
theory, says Jarup, but only if health records are accurate and up-to-date.
Unfortunately, he points out that the reverse is often true, citing one example:
鈥淭he latest mortality figures available to us are for 1997, which is pretty
辞濒诲.鈥�
It is a concern shared by Ralph Smith, GIS manager with the NHS West Midlands
Cancer Intelligence Unit, and member of the Association for Geographic
Information. 鈥淚nformation from GPs鈥� active records is very poor,鈥� says Smith.
鈥淚鈥檝e used information from clinics where 25 per cent of the postcodes have
turned out to be wrong.鈥� What is needed, says Smith, is a well thought out GIS
strategy for Britain鈥檚 health service, which focuses on both the technology and
on the quality of information. But what we are likely to get, he suspects, is
just a knee-jerk response to the Shipman case, with already overworked GPs just
getting more forms to fill in. 鈥淭he fact is that the powers that be in the NHS
haven鈥檛 paid GIS any attention at all,鈥� says Smith. 鈥淚t鈥檚 a scandal that not
more use is made of it.鈥�
However, there are some signs that GIS may finally have its day. The ONS,
which collects statistics on most things except agriculture, is now very
interested in the technology, spurred on by keenness in the Cabinet Office to
use such software to track areas of extreme deprivation.
Calder told 快猫短视频 that the ONS might be using a system which
could detect abnormal death clusters 鈥渨ithin two to three years鈥�. His optimism
comes from the availability within the past 18 months of new software that
allows grid references to be applied to addresses.
But he is also concerned that the standards of medical record-taking in
Britain will have to be vastly improved for the system to work. And somebody
would have to be in charge. 鈥淭here鈥檚 no point in doing this unless you give
someone responsibility. We would have to be commissioned to provide the warning
to the appropriate government department,鈥� he argues. So far, no one is
monitoring deaths using GIS, let alone keeping track of the bureaucracy needed
to run it.
Above all, GIS experts such as Jarup, Smith and Calder believe this
technology has the power to do far more than catch the one-in-a-million GP who
murders. It would provide new insights into public health that could save vast
sums in the long run. Jarup estimates the basic cost of setting up the system in
the ten regional centres of the health service in England and Wales would be no
more than a few million pounds. And that he says, 鈥渨ould be money well spent鈥�.