Search tips
Search criteria 


Logo of brjgenpracRCGP homepageJ R Coll Gen Pract at PubMed CentralBJGP at RCGPBJGP at RCGP
Br J Gen Pract. 2007 January 1; 57(534): 7–14.
PMCID: PMC2032694

Measuring disease prevalence: a comparison of musculoskeletal disease using four general practice consultation databases

Kelvin Jordan, PhD, Senior Lecturer in Biostatistics
Primary Care Musculoskeletal Research Centre, Keele University, Keele
Alexandra M Clarke, MSc, Research Assistant and Deborah PM Symmons, MD, MFPH, FRCP, Professor of Rheumatology and Musculoskeletal Epidemiology
Arthritis Research Campaign Epidemiology Unit, University of Manchester, Manchester
Douglas Fleming, OBE, FRCGP, FMedSci, Director
RCGP Birmingham Research Unit, Birmingham
Mark Porcheret, FRCGP, GP Research Fellow, Umesh T Kadam, MRCGP, PhD, GP Research Fellow, and Peter Croft, MD, Professor of Primary Care Epidemiology



Primary care consultation data are an important source of information on morbidity prevalence. It is not known how reliable such figures are.


To compare annual consultation prevalence estimates for musculoskeletal conditions derived from four general practice consultation databases.

Design of study

Retrospective study of general practice consultation records.


Three national general practice consultation databases: i) Fourth Morbidity Statistics from General Practice (MSGP4, 1991/92), ii) Royal College of General Practitioners Weekly Returns Service (RCGP WRS, 2001), and iii) General Practice Research Database (GPRD, 1991 and 2001); and one regional database (Consultations in Primary Care Archive, 2001).


Age-sex standardised persons consulting annual prevalence rates for musculoskeletal conditions overall, rheumatoid arthritis, osteoarthritis and arthralgia were derived for patients aged 15 years and over.


GPRD prevalence of any musculoskeletal condition, rheumatoid arthritis and osteoarthritis was lower than that of the other databases. This is likely to be due to GPs not needing to record every consultation made for a chronic condition. MSGP4 gave the highest prevalence for osteoarthritis but low prevalence of arthralgia which reflects encouragement for GPs to use diagnostic rather than symptom codes.


Considerable variation exists in consultation prevalence estimates for musculoskeletal conditions. Researchers and health service planners should be aware that estimates of disease occurrence based on consultation will be influenced by choice of database. This is likely to be true for other chronic diseases and where alternative symptom labels exist for a disease. RCGP WRS may give the most reliable prevalence figures for musculoskeletal and other chronic diseases.

Keywords: consultation and referral, databases, musculoskeletal disease, prevalence


Estimating the proportion of a population with a particular disease (the prevalence) provides the basis for the public health functions of needs assessment (how many people may require care for the particular disease), monitoring of changes in disease occurrence over time, and investigating differences in disease frequency between geographical areas and sociodemographic groups. As a defined number of people are registered with a general practice for the provision of health care in the UK system, prevalence estimates using self-reported data from national surveys in the UK can be used to calculate the likely numbers of people with a particular disease in the practice. However, continual recording of the reasons for consultation (a longstanding feature of British general practice) provides an alternative method of directly estimating morbidity prevalence in general practice populations and of providing prevalence estimates for symptoms and conditions which have not been studied in population surveys. The increasing computerisation of practice records has enhanced these opportunities. There are, however, a number of drawbacks to using consultation data to estimate morbidity prevalence in this way. First, not all morbidity will be presented to the GP — consultation data will always be about ‘morbidity for which primary health care is sought’ and the extent to which it reflects the true population prevalence of a condition will vary with the condition. Second, longstanding morbidity, for which active health care is not regularly needed or sought, may appear at some point in consultation records, but not necessarily in the records for the time period chosen to estimate prevalence figures. Conditions which are episodic, or that require frequent consultation, may occur more than once in any single individual's record during such a time period. The measure of occurrence will thus be affected by the length of the time period chosen for scrutiny and whether repeat consultations are included. Third, the consultation entry will reflect the individual GP's choice of label or diagnosis and of what is considered important in complex consultations — in the absence of any standardisation, this will vary considerably between practitioners and across different morbidities. Fourth, the quality of recording of the reason for consultation varies between practices and also between morbidities.1

How this fits in

Primary care consultation data are an important source of information on the patterns and occurrence of morbidity. Large national consultation databases are popular sources of such data, and should give the most reliable prevalence figures because of the training given to practices in recording morbidity although it is not known how reliable such figures are. Using musculoskeletal conditions as an example, this paper shows the considerable variation in persons consulting prevalence figures derived from different databases. Instructions given to healthcare professionals in the recording of morbidity strongly influence the recorded prevalence and researchers and health service planners need to be aware of these differences.

Despite all these concerns, the idea of using ‘persons consulting prevalence’ (the proportion of people on the practice register who consult at least once with a particular morbidity during a specified course of time such as 1-year) as a measure of true population prevalence is a popular one. As a result, national primary care consultation databases are used increasingly as sources of information on morbidity occurrence. Although such routinely collected primary care consultation data is a valuable resource for researchers and health service planners, the limitations of such data need to be understood to ensure it is used appropriately and that healthcare decisions based on it are reliable.2,3

Consultation prevalence figures in the UK have been based on the Morbidity Statistics from General Practice national studies (the most recent being the fourth, MSGP4, in 1991–1992),4 the General Practice Research Database (GPRD)59 and the Royal College of General Practitioners Weekly Returns Service (RCGP WRS).10 The assumption is that these large national databases, using practices trained in the electronic recording of morbidity, will give the most reliable prevalence figures. However, few studies have compared prevalence rates derived from these different national databases.11,12

Musculoskeletal conditions are a common reason for consultation. Fifteen per cent of the population in MSGP4 had a recorded musculoskeletal problem during the study year and one in five of all recorded consultations was a musculoskeletal problem.4 The umbrella term of musculoskeletal conditions includes both chronic conditions, which carry a specific diagnosis label such as osteoarthritis and rheumatoid arthritis, and also symptom-type labels such as ‘joint pain’ and ‘soft tissue disorder’ which may be episodic or acute in nature. Consultation prevalence figures for musculoskeletal conditions have previously been gathered from MSGP4.13

As the first step in updating the estimates of consultation incidence and prevalence rates for musculoskeletal disease, and to examine their patterns over time, the aim of this study was to assess the consistency of consultation prevalence rates from four general practice databases. Specifically, the objective was to compare the annual persons consulting prevalence rates for musculoskeletal conditions overall, and for rheumatoid arthritis, osteoarthritis and arthralgia in these databases.


The databases

The data sources were: i) MSGP4; ii) the RCGP WRS; iii) GPRD and iv) the Consultations in Primary Care Archive (CiPCA). Three (MSGP4, RCGP WRS, GPRD) are national UK databases of computerised general practice consultation records. The fourth (CiPCA) is a database of consultations from practices within North Staffordshire, UK.

i) MSGP4 was the fourth (and last) National Survey of Morbidity in General Practice conducted by the RCGP, the UK Office of Population and Census Surveys, and the Department of Health and coordinated by the RCGP Birmingham Research Unit.4 The survey took place between 1 September 1991 and 31 August 1992 and included all consultations during that period at 60 general practices in England and Wales. Practice staff were trained in morbidity data recording prior to the start of the survey. Doctors and nurses were asked to provide a diagnostic morbidity code for each problem consulted with during a contact.4 In particular, they were specifically encouraged to record a diagnosis at each contact rather than symptoms (for example, to record asthma rather than wheeze).

ii) The RCGP WRS database, also coordinated by the RCGP Birmingham Research Unit, holds morbidity data from consultations routinely collected from participating practices.10 For this study, annual persons consulting rates are derived for the year 2001 (the likely year of the next MSGP survey if it had been conducted) based on the 38 practices in England and Wales able to submit the requisite annual data. Over half of these had contributed to the 1991–1992 MSGP4 survey. Doctors and nurses are asked to enter a predominant symptom code only if no clear diagnostic code can be allocated with reasonable probability.10 The morbidity encountered at each consultation is required to be entered.

iii) The GPRD is a computerised database of general practice consultation data within the UK.14 Practices are required to reach a set standard in recording before contributing data and a series of recording guidelines are set down. Morbidity codes only have to be entered under three circumstances: at first diagnosis, if the event results in a change of treatment, or if another significant event, such as a referral, occurs. A chronic condition only needs to be recorded on the first occasion a new drug is prescribed.11,14 Symptom codes are allowed if seen as appropriate.11 The correctness of recorded morbidity codes assigned to patients in the GPRD has been examined for several conditions by comparison with external information such as hospital information. These conditions include psychoses,15 venous thromboembolism,16 bowel disease17 and pressure ulcers.8 For this study, data for 2001 was used when 227 practices contributed data for the whole year. To allow comparison with the MSGP4, prevalence figures were also calculated for 1991 (157 practices).

iv) CiPCA is a database of consultations from nine general practices in North Staffordshire, UK. Data for 2001 was again used. The practices are part of the Keele GP Research Partnership and, as such, have regular cycles of training, assessment and feedback in the quality of their computerised morbidity coding. Current estimates show that 93% of doctor contacts at these practices are given a morbidity code.18 Doctors and nurses are requested to enter at least one morbidity code for every contact. While encouraged to use diagnostic codes, symptom codes may be used until a diagnosis is made.

Consultations are typically recorded in the UK using the Read Code morbidity coding system. Read Codes are a hierarchy of morbidity, symptom and process codes which become more specific further down the hierarchy.19 In MSGP4 and RCGP WRS, these Read Codes have been mapped to ICD9 (International Classification of Disease revision 9) codes for presentation of prevalence and incidence of disease.4,10 However, data from the MSGP4 were analysed here based on the original Read Code format.

Statistical analysis

Annual persons consulting prevalence rates were calculated per 10 000 registered persons aged ≥15 years. To be counted as a prevalent case, patients had to have one or more recorded consultations during the 12-month period with the specified Read Code. Patients who consulted more than once with that code were only counted once. Prevalence rates for any musculoskeletal condition were calculated first. The specified codes were all codes within Chapter N ‘Musculoskeletal and connective tissue diseases’ of the Read Code system (equivalent to ICD9 codes 710–739). Then, prevalence rates for three chosen morbidities were derived. Rheumatoid arthritis was chosen to represent a specific disease category. Read codes starting N04 (‘Rheumatoid arthritis and other inflammatory polyarthropathy’), equivalent to ICD9 codes beginning 714, were identified. Osteoarthritis and arthralgia were chosen to represent the contrast between disease and symptom-type labels for a similar condition. The Read codes for osteoarthritis were those starting N05 (‘Osteoarthritis and allied disorders’), equivalent to ICD9 codes beginning 715. The Read codes for arthralgia were those starting N09 (‘Other and unspecified joint disorders’) and ICD9 codes beginning 719. Prevalence rates were calculated for adults (aged 15 years and over) from the raw data for the MSGP4, GPRD and CiPCA. Due to the method of collating data within the RCGP WRS, this was not possible for this database and prevalence rates were extracted or calculated from the annual prevalence report.10 Fifteen was the minimum age chosen because of the age groupings used in the RCGP WRS report.

The databases have slightly different definitions of the denominator population. The denominator population for MSGP4 was derived in terms of person years, so a person registered for half the year would add half a person year to the denominator population. In the GPRD, only patients registered throughout the 12-month period could contribute to the numerator or denominator. The denominator populations for CiPCA and RCGP WRS were the registered mid-year or end of year populations depending on practice. Persons consulting rates per 10 000 patients were produced for each database by sex and by age group. To take into account any difference in the age–sex structure of the populations covered by the databases, an overall age–sex standardised prevalence rate (with 95% confidence interval [CI]) was determined. The standard population used was that of England and Wales in 2001 (Office of National Statistics).

A further analysis calculated the percentage of all musculoskeletal consultations that were for rheumatoid arthritis, osteoarthritis and arthralgia.


The study populations were similar in terms of the percentage of the adult population who were female, and similar to that of England and Wales as a whole (Table 1). However, RCGP WRS and MSGP4 adult populations were younger than that of England and Wales as a whole, while the CiPCA adult population were older.

Table 1
Age and sex of the study populations (aged 15 years and over).

Table 2 shows the age-sex standardised persons consulting prevalence rates and age-specific rates for musculoskeletal conditions, rheumatoid arthritis, osteoarthritis and arthralgia. A consistent feature for all morbidities was that the GPRD 1991 prevalence rates were far lower than those derived from the other databases.

Table 2
Persons consulting prevalence rates per 10 000 people aged 15 years and over of musculoskeletal conditions, rheumatoid arthritis, osteoarthritis and arthralgia.

Musculoskeletal conditions

The persons consulting rates of musculoskeletal conditions followed a similar trend for all of the databases (Figure 1). Prevalence increased steadily over the age groups. CiPCA, RCGP WRS and MSGP4 showed similar prevalences for each age group. Overall annual prevalence of musculoskeletal conditions for adults aged ≥15 years was around 2000 persons per 10 000 based on these three databases; however, the GPRD 2001 prevalence rate was only around two-thirds of this figure. The prevalence rate in males was around 75% of that in females (range across databases 74–77%).

Figure 1
Age-specific persons consulting prevalence rate of musculoskeletal conditions per 10 000 people aged ≥15 years.

Rheumatoid arthritis

MSGP4 and RCGP WRS gave the highest adult prevalence rates for rheumatoid arthritis at around 44 per 10 000, CIPCA rates were similar, and all three gave rates about twice those of GPRD 2001 (Supplementary Figure 1). Prevalence rates for females were consistently about twice as high as for males.


Similar overall persons consulting prevalence rates for osteoarthritis were derived from the CiPCA and RCGP WRS databases (230–280 per 10 000 people aged ≥15 years). The overall MSGP4 prevalence was over 50% higher than the prevalences from these databases. By contrast, GPRD 2001 prevalence was two-thirds that of CiPCA and RCGP WRS, and less than half that of MSGP4. The prevalence rate of osteoarthritis increased sharply after age group 25–44 years (Figure 2). Prevalence for males was around 60% that of females (range across databases 58–62%).

Figure 2
Age-specific persons consulting prevalence rate of osteoarthritis per 10 000 people aged ≥15 years.


Similar adult persons consulting prevalence rates for arthralgia were derived from each of the CiPCA, RCGP WRS and GPRD 2001 databases, 400 per 10 000 people aged 15 and over. The MSGP4 prevalence rate was only two-thirds of that from the other databases. There was a sharp increase in prevalence rate in all four databases for the age group 45–64 years compared to the younger age groups (Figure 3). The prevalence in males was around 70% of that in females (range across databases 66–76%).

Figure 3
Age-specific persons consulting prevalence rate of arthralgia per 10 000 people aged ≥15 years.

Percentage of all musculoskeletal consultations that are rheumatoid arthritis, osteoarthritis and arthralgia

In MSGP4 one-fifth of all Chapter N consultations in persons aged ≥15 years were recorded as osteoarthritis (Table 3), increasing with age to 40% of all musculoskeletal consultations in the ≥75 years age group; this contrasted with only one-tenth recorded as arthralgia in the same database. The percentages were reversed for GPRD 2001, which had 22% of Chapter N consultations recorded as arthralgia and 9% as osteoarthritis. The percentages for CiPCA (arthralgia 15%, osteoarthritis 10%) and RCGP WRS (arthralgia 18%, osteoarthritis 12%) were similar to each other. The GPRD 1991 figures differed from the other databases (arthralgia 13%, osteoarthritis 4%).

Table 3
Percentage of all musculoskeletal consultations that are rheumatoid arthritis, osteoarthritis or arthralgia by age group.

The percentages of all musculoskeletal consultations recorded as rheumatoid arthritis were around 4% for CiPCA, RCGP WRS and MSGP4, twice that of GPRD 2001.


Summary of main findings

The annual persons consulting rates for musculoskeletal conditions were compared across three national general practice consultation databases and one regional database. Trends across the age and sex groups were similar for all databases. However, the actual prevalence of disease consultation varied considerably.

MSGP4 gave the highest persons consulting prevalence rates for osteoarthritis, but lower prevalence rates for the less specific diagnosis of arthralgia. This is most likely explained by differences in recording practice, with increased use of diagnostic codes in MSGP4 relative to symptom-type codes. The use of arthralgia and osteoarthritis as alternatives is most likely to occur in patients over the age of 50 years and it is in the 45–64 years age group where the divergence between databases in prevalence rates becomes apparent.

The GPRD 2001 data underestimated prevalence rates for musculoskeletal conditions in general, and for the specific disease categories of rheumatoid arthritis and osteoarthritis. This is likely to reflect the fact that, in the GPRD, not every consultation for a chronic problem has to be coded, and that there are probably a number of ‘invisible’ people who are consulting with chronic disease during the year but are not recorded as doing so. By contrast, the recording of arthralgia in GPRD 2001 was consistent with that from CiPCA and RCGP WRS. The low prevalence rate of rheumatoid arthritis in the GPRD may give most cause for concern as this is a specific disease with a definite diagnosis.

The persons consulting prevalence rates for two minor acute morbidities (chicken pox and hayfever) have been shown before to be consistent between GPRD and MSGP4.11 However, rates for two chronic conditions (diabetes and asthma) were lower when based on the GPRD. A further study compared the GPRD and MSGP4 and showed reasonably similar rates for most of 11 respiratory diagnoses.12 Both of these studies suggested differences between the two databases were due to the difference in recording instructions given to contributing GPs.

MSGP4 has been used as the gold standard to assess the face validity of prevalence rates derived from the morbidity records of general practices in order to assess quality of recording.18,2022 However, this study suggests that comparisons of rates to MSGP4 may not be helpful because differences in prevalence rates may be due to differences in recording practice rather than under or over-recording of morbidity.

Persons consulting prevalence rates for musculoskeletal conditions and osteoarthritis were reported to have increased between MSGP3 (1981–1982) and MSGP4, but had fallen for rheumatoid arthritis.4 The GPRD has shown increased prevalence rates of disease over time for several conditions.7,9,23 However, these results suggest that identifying trends in prevalence of musculoskeletal disease based on changes in consultation rates either using the two GPRD databases or comparing MSGP4 with GPRD 2001 is not a reliable reflection of true changes in consultation prevalence.

The RCGP WRS and CiPCA databases gave comparable prevalence rates. RCGP WRS is a national database, consisting of practices with training and long experience in coding morbidity, and, as such, appears to represent the most accurate current morbidity patterns based on data collected as patients consult. RCGP WRS may also allow investigation of trends over time from 1994 when the database was first computerised. CiPCA is limited by its local focus, based entirely in North Staffordshire. However, the results shown here suggest that CiPCA is a good local epidemiological resource and that if effort is put in to ensure good quality data collection, then local databases can give comparable prevalence rates for all but the rarer morbidities to larger national databases.

Limitations of the study

There are general problems in determining consultation prevalence. The derived prevalence estimates in all databases are likely to be underestimates of actual consultation prevalence as every problem discussed during a single encounter may not be coded and many people with musculoskeletal morbidity may not consult their GP with the problem.

Implications for future research and clinical practice

Primary care consultation data has many potential uses.2 Local databases, for example, have potential value in measuring the local burden of disease and healthcare utilisation, evaluating trends over time, and allowing comparison with national data. Primary care trusts and health planners are already using data collected from their local practices, and they are making increasing use of national databases in the same way that they use other national sources of health data (for example, the Census, health surveys) to construct estimates of need and provision. Unlike the survey and Census data, however, the meaning and validity of consultation data have not been well rehearsed in the public health and policy arenas. The instigation of the Quality and Outcomes Framework (QOF)24 may mean that prevalences of the included morbidities in particular are likely to be used more extensively by primary care trusts and regional planners for health service planning and that the electronic patient record will become a major resource for describing the occurrence of these specified conditions. However, they will not be directly comparable to those derived from the consultation databases examined in this paper as prevalences reported from the QOF are lifetime prevalences and not annual persons consulting prevalence rates. Although most of the included morbidities have objective diagnoses (such as diabetes) and may not be affected by the coding practice of GPs, others, such as asthma, may well be. Variations in regional prevalence may, thus, continue to reflect variations in diagnosis and labelling practice as well as sociodemographic characteristics of the populations. Furthermore, many conditions, including musculoskeletal morbidity, are not included in the QOF, and the quality and reliability of routine recording of these may be slower to improve.

The current study has found considerable variation in estimated prevalence rates for musculoskeletal conditions, and researchers and health service planners should be aware that prevalence rates of consultation for disease will be influenced by the database used to generate them, as well as the methods used to calculate the prevalence rates. All databases using routine consultation coding will reflect GP variation in the application of diagnostic criteria and in the use of labels. Databases cannot ‘enforce’ diagnostic criteria since this does not reflect clinical practice. However, they can influence coding practice, and the RCGP WRS format of encouraging diagnostic coding, but allowing symptom codes if a clear diagnosis cannot be made with reasonable probability, appears to lead to more consistency. Other influences, such as the QOF, may stimulate further uniformity in the future.

The differential effect of coding requirements within the MSGP4 and GPRD is likely to extend to conditions other than musculoskeletal conditions that are also chronic in nature or that have alternative symptom-type labels. Due to the problems of interpretation of prevalence rates derived from MSGP4 and the GPRD, the RCGP WRS may give the most reliable prevalence figures for musculoskeletal disease, and this may be true for other chronic diseases also.

Supplementary Material

Supplementary Figure 1:


The authors thank Helen Ogden for the organisation of the CiPCA database, the Informatics team in the Primary Care Musculoskeletal Research Centre for collecting the data and all the general practices who contribute to CiPCA. We are also grateful for the contribution to the RCGP WRS of the collaborating sentinel practices.


Supplementary information

Additional information accompanies this article at

Funding body

CiPCA is funded by the North Staffordshire Primary Care Research Consortium and the National Coordinating Centre for Research Capacity Development (NCCRCD). Use of the GPRD is funded by a Medical Research Council equipment grant. The RCGP WRS is funded by the Department of Health

Ethics committee

Ethics approval for CiPCA was gained from North Staffordshire Research Ethics Committee (LREC Project 03/04)

Competing interests

Kelvin Jordan, Mark Porcheret, Umesh T Kadam and Peter Croft all work in the Primary Care Musculoskeletal Research Centre, Keele University, which is responsible for the collection and maintenance of the CiPCA database. Douglas Fleming is Director of the RCGP Birmingham Research Unit which coordinates the RCGP WRS database.


1. Jordan K, Porcheret M, Croft P. Quality of morbidity coding in general practice computerized medical records: a systematic review. Fam Pract. 2004;21(4):396–412. [PubMed]
2. Majeed A. Sources, uses, strengths and limitations of data collected in primary care in England. Health Stat Q. 2004;21:5–14. [PubMed]
3. De Lusignan S, van Weel C. The use of routinely collected computer data for research in primary care: opportunities and challenges. Fam Pract. 2005;23(2):253–263. [PubMed]
4. McCormick A, Fleming D, Charlton J. Morbidity Statistics from General Practice: Fourth national study 1991–1992. London: HMSO; 1995.
5. Wallace H, Shorvon S, Tallis R. Age-specific incidence and prevalence rates of treated epilepsy in an unselected population of 2,052,922 and age-specific fertility rates of women with epilepsy. Lancet. 1998;352(9145):1970–1973. [PubMed]
6. Lawrenson R, Wyndaele JJ, Vlachonikolis I, et al. A UK general practice database study of prevalence and mortality of people with neural tube defects. Clin Rehabil. 2000;14(6):627–630. [PubMed]
7. Majeed A, Moser K, Carroll K. Trends in the prevalence and management of atrial fibrillation in general practice in England and Wales, 1994–1998: analysis of data from the general practice research database. Heart. 2001;86:284–288. [PMC free article] [PubMed]
8. Margolis DJ, Bilker W, Knauss J, et al. The incidence and prevalence of pressure ulcers among elderly patients in general medical practice. Ann Epidemiol. 2002;12(5):321–325. [PubMed]
9. Soriano JB, Kiri VA, Maier WC, Strachan D. Increasing prevalence of asthma in UK primary care during the 1990s. Int J Tuberc Lung Dis. 2003;7(5):415–421. [PubMed]
10. RCGP Birmingham Research Unit. Weekly Returns Service Annual Prevalence Report 2001–2003. Birmingham: 2004.
11. Hollowell J. The General Practice Research Database: quality of morbidity data. Popul Trends. 1997;87:36–40. [PubMed]
12. Hansell A, Hollowell J, Nichols T, et al. Use of the General Practice Research Database (GPRD) for respiratory epidemiology: a comparison with the 4th Morbidity Survey in General Practice (MSGP4) Thorax. 1999;54:413–419. [PMC free article] [PubMed]
13. Symmons D, Asten P, McNally R, Webb R. Healthcare needs assessment for musculoskeletal diseases: The first step — estimating the number of incident and prevalent cases. 2nd edn. Chesterfield: The Arthritis Research Campaign; 2002.
14. Lawson DH, Sherman V, Hollowell J. The General Practice Research Database. Q J Med. 1998;91:445–452.
15. Nazareth I, King M, Haines A, et al. Accuracy of diagnosis of psychosis on general practice computer system. BMJ. 1993;307(6895):32–34. [PMC free article] [PubMed]
16. Lawrenson R, Todd JC, Leydon GM, et al. Validation of the diagnosis of venous thromboembolism in general practice database studies. Br J Clin Pharmacol. 2000;49(6):591–596. [PMC free article] [PubMed]
17. Lewis JD, Brensinger C, Bilker WB, Strom BL. Validity and completeness of the General Practice Research Database for studies of inflammatory bowel disease. Pharmacoepidemiol Drug Saf. 2002;11(3):211–218. [PubMed]
18. Porcheret M, Hughes R, Evans D, et al. Data quality of general practice electronic health records: the impact of a program of assessments, feedback, and training. J Am Med Inform Assoc. 2004;11(1):78–86. [PMC free article] [PubMed]
19. NHS Information Authority. The Clinical Terms version 3 (The Read Codes) Birmingham: NHS Information Authority; 2000.
20. Pearson N, O'Brien J, Thomas H, et al. Collecting morbidity data in general practice: the Somerset morbidity project. BMJ. 1996;312(7045):1517–1520. [PMC free article] [PubMed]
21. Thiru K, de Lusignan S, Hague N. Have the completeness and accuracy of computer medical records in general practice improved in the last five years? The report of a twopractice pilot study. Health Informatics J. 1999;5:224–232.
22. Hassey A, Gerrett D, Wilson A. A survey of validity and utility of electronic patient generated records in a general practice. BMJ. 2001;322(7299):1401–1405. [PMC free article] [PubMed]
23. Moser K, Majeed A. Prevalence of treated chronic diseases in general practice in England and Wales — trends over time and variations by the ONS area classification. Health Statistics Q. 1999;2:25–32.
24. NHS Health and Social Care Information Centre. National Quality and Outcomes Framework Statistics for England 2004/05. Statistical Bulletin 2005/02/HSCIC.

Articles from The British Journal of General Practice are provided here courtesy of Royal College of General Practitioners