Informatics 98may6

Journal of Informatics in Primary Care 1998 (May):14-18


Papers


Characteristics of practices contributing to the MediPlus database and the implications for its use in epidemiological research

RA Lawrenson a MRCGP. FAFPHM. MD.   Senior Lecturer in Public Health
G Coles b BSc.   Research Assistant
K Walton c MA. MBBS. MRCP.   General Practitioner
RDT Farmer b MRCGP. FFPHM. PhD.   Professor

a European Institute of Health and Medical Sciences, University of Surrey, Stirling Road, Surrey Research Park, Guildford, Surrey, GU2 5RF. Tel 01483 302239

b Department of Epidemiology and Public Health, Imperial College School of Medicine, Chelsea and Westminster Hospital, 369 Fulham Road, London SW10 9NH.

c 7 Leighton Avenue, Pinner, Middlesex, HA5 3BW.

Correspondence to Dr. Lawrenson: email: r.lawrenson@surrey.ac.uk

 

Introduction

Increasingly general practitioners in the United Kingdom are using computers to help manage their practices. These computers are used to generate prescriptions, to keep patient records and to help fundholders track the referral of patients to hospital. Over 80% of practices in the UK are computerised1, and about 8% are believed to be paperless, that is, the practices use their computer record as the primary source of information about their patients2. Combining the anonymised records from a number of practices allows researchers to use the collected information for medical research. There are over 30 different software systems designed for use by general practitioners that are currently available. Only two of these systems, the VAMP and Meditel systems, have been used to produce commercially available databases. The General Practice Research Database (GPRD), owned and managed by the Department of Health, is based on the VAMP system, uses OXMIS codes and takes data from over 400 practices. The MediPlus database is based on information from 151 practices that use the Meditel system. It has information on 1.8 million patients from practices throughout the United Kingdom and Northern Ireland. All the diagnostic records, prescriptions and notes are coded into Read codes, the coding system adopted by the Department of Health for general practice. Doctors using Meditel System 5 can attach free text to the appropriate code; for reasons of confidentiality this information is not held on the MediPlus database.

Examples of studies that can be carried out using general practice databases include epidemiological studies of the prevalence and incidence of different diseases3, studies looking at the utilisation of health services4 and studies on pharmaco-epidemiolgy5. However if these databases are to be used for research then there are a number of issues that need to be investigated.

The three main issues include :

Generalisability – are the patients and doctors that contribute to the database representative of all patients and doctors?

Validity of the data – what is the likelihood that a patient with a diagnosis recorded on the database truly has that disease?

Completeness of the recording – for example, if a patient has received a prescription which is recorded elsewhere what is the likelihood that this will also be recorded on the computer record?

How we deal with these issues to ensure that studies using general practice databases are both valid and generalisable requires research into the development and maintenance of the databases. To a certain extent the more complete a database the less generalisable the data, that is, the sorts of doctors who religiously record every piece of data on the computer are not typical of general practitioners as a whole. However this does not necessarily mean that the patients they see are different from patients in general or the doctor’s diagnostic skills are better or worse than doctors whose recording is less complete. The degree of variation in these factors needs to be measured.

Practices comprise one or more doctors who work together as a partnership. Practices may also have assistant doctors, trainees, nursing, clerical and other staff using the computer system. Within a practice it is not unusual to find that the doctors differ in the way they use the computer. Many practices have at least one 'computer phobic' whose recording of data will be less complete than other members of the practice who are more comfortable with the technology.

Over 95% of the population are registered with a General Practitioner and individuals can only be registered with one doctor at any given time6. Thus general practice databases are population-based and include patients of all ages. This means that GP data is able to supply the denominator for incidence and prevalence studies in a way that hospital data cannot. The General Practitioner is the main point of access to the National Health Service. Consequently their records contain considerable information on health services utilisation, both of primary and secondary care. Discharge summaries from hospitals and other specialist centres are sent to the GP as a matter of routine and important clinical information can be entered onto the computer (at present usually manually).

There are several approaches to the issues of completeness and validity:

  • Does the information on the database represent the clinical history of the patient accurately? For instance, if the database says 100 patients have had a myocardial infarction (MI) what proportion of these patients will truly have had an MI?
  • Alternatively does the information on the database represent the primary (written) medical record of the patient? If 100 patients have had an MI and this is known to their general practitioner what proportion are recorded on the database?
  • Does the information on the database represent the information on the practice computer database accurately? Sometimes the software that converts the general practitioner records on System 5 will lose information (particularly free text) when it is anonymised and converted into the MediPlus database. (This is done to ensure there is no transfer of information that might identify a patient, such as free text of a family name.)

The answer to some of these questions can be found by asking general practitioners how they use their systems. This can lead to the generation of some hypotheses that can then be further explored using different methodologies. Thus the aim of this study was to ascertain the characteristics of the practices contributing to the MediPlus database and to find out how these practices used the AAH Meditel system.

 

Methodology

The MediPlus database is collated by Intercontinental Medical Statistics (IMS). At the time of the study, the MediPlus database contained patient information from 151 practices. A questionnaire was developed in conjunction with general practitioner advisers to IMS and was trialled in ten general practices. This questionnaire was then sent to the remaining practices. Five practices were visited as part of a further validation study and in these practices a number of general practitioners and administrative staff were interviewed in order to ensure that there was consistency across the practices in the responses.

The questionnaire had 23 sections. The intention was that it should be answered by one person on behalf of the practice as a whole (although in some circumstances different people filled in different sections). Firstly there were questions about the make up of the practices such as the number of partners, whether the practice was fundholding and whether it was a training practice. There were then some questions about how long the practice had used the Meditel system, what training they had received and what links there were between the system and other agencies. These were followed by detailed questions about how the practices used the system as opposed to written records. Furthermore, practices were asked to indicate where they thought information on patients might be missing from the computer record. This included questions about home visits and special clinics. As the MediPlus system does not provide de-registration data, identification of patients who have died can vary from practice to practice. The opportunity was therefore taken to obtain information about the recording of deaths on the system.

 

Results

At the update in April 1997, 151 practices were contributing to the MediPlus database. Because some practices have since left the panel, and others have joined, questionnaires were posted only to 138 of the contributing practices. Replies were received from 129 practices (94% response rate). The number of partners in each of the practices contributing to the database is shown in Table 1.

Table 1. The number of partners in practices within the MediPlus database

Number of Partners

Full time partners - number of practices

Part time partners - number of practices

0

0

10

1

26

39

2

20

32

3

25

3

4

26

2

5

15

1

6

10

 

7

2

 

8

3

 

9

0

 

10

1

 

The number of practices that were fundholding was 89/129 (69%). Sixty-two of the practices (48%) indicated they were training practices. The distribution by year of the introduction of the Meditel system into the practices is indicated in Table 2. Only 9 practices contributing to the database have been using the system for less than six years. Sixty-seven out of 129 (52%) had had some training from Meditel in the last two years. In addition all practices had initial training on the system when it was first installed.

Table 2. The number of practices that began recording data on the Meditel system by year

Year practice began collecting data

Number of practices

1986

3

1987

16

1988

29

1989

31

1990

23

1991

10

1992

5

1993 - 1995

4

Information not provided

8

One hundred and eighteen out of 129 practices (91%) were linked to the local health authority for registration purposes. Ninety practices out of 129 (70%) were also linked for the purpose of recording one or more item of service payments. However only 9/129 (7%) were linked to hospital laboratories.

Fifty-four out of 129 practices (41%) said that the computer record was their primary record whilst 18/129 (14%) indicated that they put equal reliance on the computer and the written record. Fifty-five out of 129 (43%) indicated that despite the fact that they are recording all key diagnoses and prescriptions on the computer they still considered the written records as the primary source of data. (Two practices did not answer this question.) Thirty-six of out 129 (28%) said they rarely or only occasionally consulted the written records. These could be considered the paperless practices.

The use of the Read codes within the Meditel system is the basis of the MediPlus database and is a key advantage for epidemiological research. One hundred and thirteen practices out of 128 (88%) said they had a policy to encourage the consistent use of Read codes. If no Read code was available then practices will usually pick the best available code or use free text. Consistent collection and coding of data, particularly in common conditions like asthma and diabetes can be encouraged by the use of specially written protocols – known in the Meditel system as SOPHIEs ("Screening Of Patient Health in an Interactive Environment"). Only 27/129 (21%) of practices said they did not use SOPHIEs whilst 41/128 (32%) create their own.

Letters from outpatients and hospital discharge summaries are routinely filed in the paper records, but 55/129 (43%) of practices indicated that less than 50% of hospital discharge summaries were entered on their computer. Comments made on the questionnaire, supported by the interviews held in the 5 practices, suggested that laboratory and x-ray results judged to be significant were "almost always" entered but 36/129 of practices (28%) recorded less than 50% of routine laboratory test results on the computer. The fact that the test has been ordered is entered onto the database about 50% of the time (59/122 practices that answered this question said they recorded the ordering of laboratory tests more than 90% of the time).

One hundred and five of the 129 practices said they routinely recorded the date of death of a patient and 84/129 said they recorded the cause of death. However 25/129 (19%) said they always or usually recorded this using free text (which does not appear on the MediPlus database for reasons of data confidentiality noted earlier).

Almost all practices recorded major features of patients’ past medical history on the computer. This included significant diagnoses (95%), chronic diseases (94%), family history (82%), major surgery (87%), and allergies (84%). Other features that practices commented they usually recorded included past cervical smears and vaccinations. Eighty-four of 129 practices (65%) said over 95% of consultations will generate a computer record and over 90% of practices said at least 70% of consultations will generate a record. Two practices suggested that 45% or more consultations will not generate a computer record. Twenty-two out of 129 practices (17%) indicated that if there was no computer record then there would be no written record either.

Some prescriptions are not generated by the computer system and may not be entered. Thus 26% of practices indicated that private prescriptions may not be entered, 50% of practices said that prescriptions written on home visits may not be entered, and 11% of practices said controlled drugs would not typically be entered.

Ninety-one out of 129 practices (71%) use a deputising service for home visits out of hours. Of these 53/91 (58%) said the visit would be recorded on the computer whilst 7/91 (8%) said no record of the visit would be kept by the practice

 

Discussion

This survey indicates the likely validity and completeness of the MediPlus database. We have no reason to believe that this database is any less reliable as a source of general practice activity than any other UK general practice database. Indeed the quality control checks and continual feedback from the general practitioner advisors would suggest that the computer records in the contributing practices are more complete than most computerised practices. The fact that over 40% of practices considered the computer to be the primary record and 28% only occasionally or rarely called for the written notes suggest that a considerable proportion of the practices are essentially paperless. This is as expected a much higher proportion than the 8–10% quoted nationally2. It seems likely that information from these practices will more accurately reflect the care of patients because the recording of data will be more complete than data from those practices that still consider the written notes to be the primary record. Whether the activities in the paperless practices are substantially different from those in practices that rely on the written record is open to debate. The generalisability of the data may also be compromised by the fact that the practices contributing to the database are larger, more likely to be fundholding, and are more likely to be training practices than practices in general. Again the extent to which these characteristics affect the activities and outcomes of care can be debated. Some differences in the prescribing of fundholding practices have been noted7, although a study of repeat prescribing showed no difference between fundholders and other practices on the database8.

All the practices in England were linked to the local health authority for Registration purposes and the majority were linked for some item of service payments. Whilst these administrative links are extremely useful the clinical links with hospitals are less well developed and at the time of the study were only found in a few practices. The speed of uptake will depend to a great extent on the hospital Trusts. Scotland and Northern Ireland are only beginning to pilot Registration Links. Again the speed at which this happens is likely to be determined by the Health Boards rather than being initiated by the practices.

The use of the nationally recognised coding system, the Read codes, which have been cross-mapped to ICD codes9, is one of the advantages of the Meditel system, and makes the use of these codes in the MediPlus database relatively straightforward for epidemiological research. The other major advantage that the Meditel system has is that all records – notes, prescriptions, etc., can be linked under a problem heading. One disadvantage of the Read coding system is that for some diseases such as depression there are a number of options for a code that can be used. It is therefore pleasing to see that practices have recognised the need for consistency in the use of codes and that 88% have a policy for consistent use. This is also reflected in the number of practices using SOPHIEs.

The key advantage of this general practice database is that it links prescriptions to key clinical diagnoses, and to the age and sex of patients. This is ideal for pharmaco-epidemiology studies and for conducting health services research. It appears that laboratory results, X-rays and some information from hospital discharge summaries are less well recorded in some practices. This means that for epidemiological purposes, whilst there is sufficient data to inform many studies, this information is incomplete. On the other hand the potential for missing prescriptions on the database seems small with only some practices admitting that some home visits and private prescriptions may not be recorded – these being a tiny percentage of the whole. As part of the IMS quality checks the completeness of prescription records is compared with data from the Prescription Pricing Authority (PPA). At the time of writing this showed that 96–97% of prescriptions that reach the PPA are entered on the computer record.

Whilst 105/129 practices said they recorded the death of patients, we have found from comparisons with national data that less than the expected number of deaths appear on the MediPlus database. This can best be rectified by the database including the registration details on all patients including the registration codes for reason for leaving.

A small number of consultations do not result in a record. It has been suggested that counting prescriptions and consultations patterns are a useful way of monitoring the success of treatment for various conditions: it would be expected that after successful treatment consultations should reduce. There is likely to be some variation in consultation rates between practices and part of this is likely to be due to differences in the use of the computer. These differences would need to be adjusted for in any study involving consultation or prescription rates.

 

Conclusions

  • The MediPlus database is a valuable tool for epidemiological research.
  • The accurate and reliable recording of prescription data linked to diagnostic codes, the age and sex of patients is its major strength.
  • The use of the Read Code system for recording data makes analysis relatively straightforward.
  • Useful data in the form of laboratory results, X-rays and hospital notes are also available but their recording will be less complete.

 

References

  1. NHS Management Executive. Computerisation in GP practices 1993 survey. Department of Health, Leeds, 1993
  2. Purves IN. The paperless general practice. Br Med J 1996; 312:1112–1113
  3. Lawrenson R, Tyrer F, Pierce M. Mortality rates in diabetic patients from a community based population compared to local and age and sex matched controls. Diabetic Medicine 1997
  4. Turnbull S, Ward A, Treasure J, Jick H, Derby L. The demand for eating disorder care. An epidemiological study using the general practice research database. Br J Psychiatry 1996 (Dec); 169(6):705–712
  5. Farmer RD, Lawrenson RA, Thompson CR, Kennedy JG, Hambleton IR. Population based study of risk of venous thromboembolism associated with various oral contraceptives. Lancet Jan 1997; 349:83–88
  6. Office of Population Censuses and Surveys. General Practice Morbidity Statistics; Fourth national study 1991-2. OPCS, 1992
  7. Wilson RP, Hatcher J, Barton S, Walley T. Influences of practice characteristics on prescribing in fundholding and non-fundholding general practices: an observational study. Br Med J 1996; 313:595–599
  8. Harris CM, Dajda R. The scale of repeat prescribing. Br J Gen Pract 1996; 46:649–653
  9. Saint Yves IF. The Read Clinical Classification. Health Bulletin Edinburgh 1992; 50:422–427

Refbacks

  • There are currently no refbacks.


This is an open access journal, which means that all content is freely available without charge to the user or their institution. Users are allowed to read, download, copy, distribute, print, search, or link to the full texts of the articles in this journal starting from Volume 21 without asking prior permission from the publisher or the author. This is in accordance with the BOAI definition of open accessFor permission regarding papers published in previous volumes, please contact us.

Privacy statement: The names and email addresses entered in this journal site will be used exclusively for the stated purposes of this journal and will not be made available for any other purpose or to any other party.

Online ISSN 2058-4563 - Print ISSN 2058-4555. Published by BCS, The Chartered Institute for IT