Discussion document: comments please
1.8.98
html version at http://www.irnham.demon.co.uk/morbid1.htm
Information requirements for PCG's
Aims: PCG's represent a major change in the organisation of UK general
practice. It groups together practices on a geographical basis, and for the
first time IT advanced (whatever that means) practices will be working with
relative novices. We are currently joining the Sedgemoor (Somerset) PCG, and
have a GP who currently has never had a computer - a rare breed in the South
West. This document discusses the need for morbidity information, and seeks
to provide a structure for this recording so that we are all doing things
the same way. This conformity is essential when we start using and
amalgamating the data to influence contracting - mistakes will be made if
the data is not of a reasonably high standard (and I will argue that
validating this data is the only way to ensure the quality remains high).
Background: I am a GP in Minehead Somerset. Our practice has been
computerised since 1989 (AMC - EMIS since 1992) and providing morbidity data
for the RCGP "Weekly Returns Service" since 1991. We partook in the 4th
National morbidity survey (4thNMS) in 1991-92 linked to the 1991 census and
are currently providing data to the weekly returns service and the "Somerset
Morbidity Project" - one of the CHDGP sites
(http://www.nott.ac.uk/gp/chdgp ). The Somerset Project has been collecting
and amalgamating annonymised morbidity information from all GP/patient and
nurse/patient face-to-face consultations for 4+ years. It has 12
participating GP practices (70,000+ patients) matched for age/sex
urban/rural big/small practices; to be similar to the population of Somerset
as a whole.
Likely PCG morbidity data requirements: With the dissemination of "Audit"
all UK GP's are familiar with the data collection stage, and the value of
"disease registers" for asthma etc. It is a small step from being able to
identify groups of patients with a few conditions to being able to identify
groups of patients with ANY medical condition. The Somerset Project has
shown the value of having a representative morbidity database; and uses this
database for asking questions such as "how many patients do we have with
angina, and how many CABG operations do we need to purchase?" - on the basis
of the morbidity presenting to general practice rather than last years
historical purchasing. No-one knows exactly what morbidity information will
be required next "contracting round" - which diseases will be looked at
next. There are seasonal differences in many conditions - so data collection
may need to occur for 6 months+ (arguably 2yrs + to account for annual
variation). We need to be urging as many GP's as possible to collect and
record "morbidity data" from all consultations - and this needs to begin
ASAP!
Models for such data collection: Several similar models exist. Most recently
is the CHDGP project, but unfortunately this has felt unable to insist on
the discipline of recording information at EVERY consultation, and has not
insisted on validation. The one thing I know a lot about is "busy GP
surgeries" and unless these disciplines are forced upon them, data on
Monday mornings, Friday afternoons and visits is likely to be very
incomplete! "Paperless practices" have an advantage here, but need to code
each morbidity presenting, ie Diabetes, AF, CCF and hypertension, which is
tiresome and perhaps impractical?
Data recording itself
The Read codes provide both problems and answers. We code and episode type
each episode of morbidity (problem title) in each consultation
(face-to-face, telephone, OOH) and trawl all correspondence to keep our
records up-to-date. Finding the appropriate Read Code is straightforward
about 60% of the time. Other times there are several similar codes,
sometimes pages of them!
Ideally the PCG should circulate "suggested codes" for conditions (e.g..
glue ear) where there are several "correct" codes in different hierarchies,
and the "computer lead GP" sets up synonyms so his partners can find
preferred codes easily. We use "LBP" to find preferred back pain codes and
"OTALGIA" to find the earache codes. As many "problems" as possible have
protocols or templates which automatically put in the correct read code.
Diabetes, asthma and contraception are run entirely by protocols.
Partners are encouraged to use the same code during a single episode, to
avoid double counting.
See http://www.irnham.demon.co.uk/morbidit.htm
How can we get this data out of GP systems? It is possible that PCG's
circulate request for information to their practices - "please return
age/sex distribution for your patients presenting with angina between 1.4.97
and 1.4.98". The practices can then run individual computer (or manual)
searches and return data for central amalgamation at PCG HQ! Much better
would be for PCG's to invest time and effort to use HQL search engines such
as MIQUEST, construct a single search and circulate (Email?) practices.
There are many GP system compatibility issues , but Somerset is in the
enviable position that the majority of GP's have MIQUEST compatible systems.
Perhaps PCG's should insist that its practices invest in systems that can
provide this information in this way? I much prefer the model with patient
data being on individual GP systems (confidentiality!) rather than having
one massive PCG computer system - but this may be the answer for some areas.
How can we ensure that the data is valid? The only way I can see to do this
is to validate the data. Some things can be done by computer systems - the
4thNMS study used a program that checked for "males with menorrhagia" or
patients with rare diseases such as polio (usually polio read code entered
instead of polio immunisation). Somerset Morbidity Project employs a
research assistant ([log in to unmask] ), who
regularly visits practices (quarterly) to check 95% of consultations have
appropriate "Read Code Problem titles", and to re-enthuse practices with
"roadshows" showing how the data is being used. The project also has a very
active "User group" where a GP from each practice attend to relay
difficulties in coding, sharing computer templates and protocols, support
individual research projects, and help the "slower" practices keep up to
speed (and stop some racing ahead??). The project pays "Weekly Returns Rate"
for validated data - 40p per patient/year based on average list sizes, and
this would stop if data is too dirty. This pays for some receptionist/typist
hours to trawl letters and visits and check that as much as possible
morbidity data gets on the computer - with a little left over to convince my
partners to continue?
I personally believe that to provide good data, each PCG should pay
practices a similar amount, and employ a data manager to validate etc.
Comments please!!
Huw Thomas
GP Minehead
[log in to unmask]
http://www.irnham.demon.co.uk
Internet Links Page
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|