Family Practice Vol. 21, No. 4, 396-412
Family Practice Vol. 21, No. 4 © Oxford University Press 2004, all rights reserved.
Quality of morbidity coding in general practice computerized medical records: a systematic review
Primary Care Sciences Research Centre, Keele University, Keele, Staffs ST5 5BG, UK
E-mail: k.p.jordan{at}keele.ac.uk
Received 19 March 2003; Revised 27 August 2003; Accepted 10 March 2004.
Jordan K, Porcheret M and Croft P. Quality of morbidity coding in general practice computerized medical records: a systematic review. Family Practice 2004; 21: 396–412.
 |
Abstract |
|---|
 Top Abstract IntroductionMethodsResultsDiscussionReferences |
|---|
Background. Increased use of computers and morbidity coding in primary care delivery and research brings a need for evidence of the quality of general practice medical records.
Objective. Our aim was to assess the quality, in terms of completeness and correctness, of morbidity coding in computerized general practice records through a systematic review.
Methods. Published studies were identified by searches of electronic databases and citations of collected papers. Assessment of each article was made by two independent observers and discrepancies resolved by consensus. Studies were reviewed qualitatively due to their heterogeneity.
Results. Twenty-four studies met the inclusion criteria for the review. There was variation in the methodology and quality of studies, and problems in generalizability. Studies have attempted to assess the completeness and correctness of morbidity registers by reference to a gold standard such as paper notes, prescribing information or diagnostic tests and procedures, each of which has problems. A consistent finding was that quality of recording varied between morbidities. One reason for this may be in distinctiveness of diagnosis (e.g. coding of diabetes tended to be of higher quality than coding of asthma).
Conclusions. This review highlights the problems faced in assessing the completeness and correctness of computerized general practice medical records. However, it also suggests that a high quality of coding can be achieved. The focus should now be on methods to encourage and help practices improve the quality of their coding.
Keywords. Medical records, primary care, systematic review.
|
Introduction |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
Computers in primary care are used for clinical management, administration, research and planning. A 1996 survey of general practices in England showed that 96% were computerized and 81% of these used computers for entering clinical data during consultations.1 A systematic review of published studies on primary care computing concluded that most GPs accepted computers in their working lives, thought they were more accurate than paper notes, gave them better access to records and improved patient care.2
Morbidities can be entered onto the computer using various coding classifications. One of the more common in the UK is the Read Code classification, a hierarchy of morbidity, symptom and process codes which become more specific further down the hierarchy.3 However, the quality of consultation recording and morbidity coding needs to be established. One review of the quality of computer-held patient records mainly assessed studies of hospital records.4 A recent review of how the quality of primary care computerized records (including morbidity, prescription, referral, lifestyle and socio-economic data) has been assessed concluded that there was a lack of standardized methods and that recorded prescription data appeared to have the highest overall quality.5
Our objective was to conduct a review focused on morbidity data in order to assess in detail the quality of recording of morbidity codes and morbidity registers held in computerized primary care records in the UK and to evaluate the different approaches used to assess this quality.
|
Methods |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
We define a morbidity register as a list of people with a certain morbidity code or disease. This does not have to be an actual computerized list, as it may also be established by searching the electronic records for all patients with a particular morbidity code or codes.
Defining quality
The quality of coding is often described by reference to its ‘accuracy’, but defining this term is difficult. The accuracy of morbidity coding is subjective, and can often only be established by reference to a gold standard.
The review by Hogan and Wagner assessed accuracy in terms of the completeness and correctness of records.4 Completeness was defined as equivalent to sensitivity: the proportion of observations ‘about the world’ that were actually recorded. For example, with respect to a morbidity register, are all the subjects with the morbidity included on the register? Correctness was defined as equivalent to the positive predictive value (PPV): the proportion of observations that reflect the ‘true state of the world’. For example, do all subjects on a register actually have that disease? Both are necessary for a database to be accurate. A high level of correctness may be achieved at the expense of failing to record all information, i.e. poor completeness.4 Similarly, a high level of completeness may be obtained at the cost of poor correctness.
In this review, we expand these definitions and examine four criteria. (i) The completeness of consultation recording—for each contact a patient has with the GP, is there a morbidity code recorded on the computer? This is an important element of completeness because, if no code is allocated or if the contact goes unrecorded, then completeness of the database is compromised. Further, is each different clinical morbidity consulted about within one contact coded? (ii) The correctness of consultation recording—are the codes given during this contact appropriate? (iii) The completeness of a morbidity register—is everyone included on the register that should be? (iv) The correctness of a morbidity register—should everyone on a register be on that register?
Criterion (i) can normally be evaluated by the percentage of consultations recorded on computer and the percentage of these with a morbidity code. Criteria (ii)–(iv) can be split into external and internal validation.
- Does the patient actually have the morbidity indicated by the code given? This is examined by reference to external sources such as experts in the field of that morbidity examining the patient, and relates to the diagnostic abilities of the GP. This will be referred to as external validity.
- Is the code given accurate based on the evidence that is available within the primary care practice or clinic? This could be evidence obtained during a consultation, or information from external sources supplied to the practice, i.e. is the code the one which would be expected given information such as the history of the patient, hospital letters and test results or prescriptions issued by the GP. This could also relate to whether the code given is correct based on what morbidity the GP thinks the patient has. This will be referred to as internal validity.
In this review, we examine the four criteria above in relation to internal validity. The objective of the review is not to assess the diagnostic ability of GPs, but to assess whether diagnoses in primary care electronic records are a true and complete reflection of the diagnoses given by the practitioner and the information available to the practitioner.
Criteria for inclusion of studies in the review
Inclusion criteria.
- Studies should be based in primary care.
- Studies should assess computerized records or a computerized morbidity register (studies which did not explicitly state that the records were computerized were excluded).
- Studies should be based in the UK. This allows us to assess the quality of recording in the UK without disentangling the effects of different primary care systems.
- Studies should have a stated objective of assessing the quality of these records based on criteria (i)–(iv) above.
Exclusion criteria.
- Studies which solely compare disease prevalence rates with external (to the practice) rates.
- Studies which attempt to validate general practice diagnosis by reference to external sources such as experts in the field of that morbidity reviewing the patient (i.e. external validity).
- Studies which use fictional or simulated patients. These represent another method to investigate external validity.
- Studies comparing patient self-reported consultation with medical records. Here, it is uncertain whether the medical records are validating the self-report, or vice versa.
Databases searched were MEDLINE, Science Citation Index, Social Science Citation Index, Cumulative Index to Nursing and Allied Health Literature (CINAHL), English National Health Care database, the Cochrane Library and the National Research Register. Citations of collected articles were also searched. English language articles up to September 2002 were collected. Keywords used were at three levels, with articles examined for at least one word in its title, abstract or keywords from each level. Level one keywords were: ‘primary care’, ‘general pract*’, ‘family pract*’, level two were ‘morbid*’, ‘computer*’, ‘record*’, ‘electronic’, ‘register’, ‘consult*’, ‘contact*’; and level three were ‘agree*’, ‘valid*’, ‘accura*’, ‘complete*’, ‘correct*’, ‘reliab*’.
Following the search strategy, studies which obviously failed to meet the inclusion/exclusion criteria (based on their abstract) were discarded and the full papers of the remainder read. Studies which fitted the inclusion and exclusion criteria were included in the review.
Two independent observers assessed each included article using a data extraction sheet. Discrepancies between the observers were resolved through consensus. One observer (KJ) is a biostatistician, the other (MP) is a practising GP. Due to the heterogeneity of the studies, no formal meta-analysis or pooling of the data was possible. Studies were reviewed qualitatively.
|
Results |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
Following the search strategy, 344 potential studies were identified. After assessment of their abstracts, 89 papers were read and 24 included in the review. Reasons for exclusion are given in Figure 1.
|
A summary of the methods of the studies in the review are given in Table 1 with their results in Table 2.
|
|
Methodologies of the studies
Three studies were single practice evaluations6–8 and eight studies evaluated the VAMP database system used for the General Practice Research Database (GPRD).9–16 Two other studies assessed practices contributing towards a multipractice database; the MediPlus database17 and the GPASS system in Scotland.18 The number of practices for the remaining 11 studies ranged from two to 25 (median 7). After VAMP, the most common computer system used was EMIS (seven studies).6,8,19–23 No more than two studies used any other system. Only six studies were published before 1995.7,9–11,24,25
All the VAMP studies used OXMIS for morbidity coding. Ten studies had practices stated to be using Read Codes,6,8,17–19,21–23,26,27 whilst two studies used ICD9.11,25
The paper information in the practice was the most commonly used gold standard (nine studies).6,8,11,18–22,27 This could include, for example, hospital letters as well as notes about the consultation kept in these manual notes. Prescription information (paper or electronic) on drugs relating to a specific disease was used as the sole gold standard for at least one quality assessment by eight studies.6,13,15,21–24,28 Combinations of disease-specific information stored electronically,6,12 information including drug records and hospital letters in paper format18 or information stored either electronically or in paper notes21,26,29 were also used. For example, Hassey et al. used drugs and, where appropriate, other diagnostic tests stored electronically to validate 15 diagnoses.6 Pringle et al. used a combination of paper notes, computerized prescription data and related diagnoses stored on computer.21 They also videotaped consultations to assess correctness of consultation coding compared with independent coding. Several GPRD studies sent questionnaires to the patients' GPs asking for information about symptoms or reason for diagnosis12,15,16 or to ask for hospital correspondence.9,10,14 One study used hospital discharge information maintained by the Health Authority.25
Several studies only included the highest performing practices in terms of computer use, or those which had undergone extensive training in the use of computers and coding.6,18,21,23 To be included in the GPRD also requires a period of training. Problems were sometimes encountered enrolling practices into the study, due either to the practice lacking the information required (e.g. a morbidity register) or to the practice refusing to enter the study. For example, in one study, 31 out of 38 practices approached were unable to supply the required data.20 In another, only nine out of 48 practices contacted supplied useable information for the study.28
Quality of consultation recording
Eight studies looked at the completeness of consultation recording.6,7,11,17,20–22,27 This was in terms of either the extent to which consultations were being recorded on the computer6,11,20,22 or the extent to which morbidity codes were being allocated to recorded consultations.7,17,21,22,27 The former were all compared with paper notes, except one study which used an appointment book.6
The extent of electronic consultation recording was generally high (>90%), except in one study where only three-quarters of consultations were on computer.11 In this study, a quarter of consultations were not in the paper notes. In other studies, 
10% of consultations were on computer but not in the paper notes.20,22
In terms of morbidity codes allocated to consultations, there was high variability between the practices studied, even within the separate studies, ranging from 67 to 99%.7,21,22,27 This may have improved since these studies were published (1992–1996). de Lusignan et al. suggested that feedback improved the detail of coding but, in general, in their study, 80% of problems had Read Codes down to level 3 or lower (i.e. were more specific).17
GPs in one study suggested that mental and psychological problems were often not recorded due to their difficulty of coding and that chronic problems were also under-recorded.22 In another study, doctors were more likely to fail to record consultations with older females who were more frequent consulters.20 Discrepancies were considered to be due to unavailability of a computer (e.g. for a home visit), lack of motivation or forgetfulness. One study suggested that simple requests to GPs to record diagnoses, with or without feedback of previous results, could greatly improve the completeness.7
The only study which assessed consultation content as a measure of comprehensiveness of the codes given for the contact did so by reference to videotapes of consultations.21 The authors suggested that items missed were not of clinical importance. Diagnoses not apparent from observation of the videotape were sometimes coded on the computer.
Another study suggested that 87% of 1090 records checked had the appropriate Read Codes when compared with paper notes, although little information is given on the methodology for assessing this ‘appropriateness’.27
Morbidity registers
Heart disease registers, including angina, myocardial infarction (MI) and ischaemic heart disease (IHD), have been assessed most often (10 studies). These have been by comparison with prescribed medication (typically nitrates),21–23,28 combinations of information (including drugs and hospital letters) in paper notes,18 medication and procedures stored electronically,6 tests and procedures19,26,29 and hospital discharge information collected by the Health Authority.25
Completeness of heart disease registers appears poor. Seventy-two percent of patients with validated coronary heart disease (CHD) based on related information held in paper notes and computer records were on the registers of 18 practices (one practice's register was not computerized).29 One study reported that only two out of six practices had >60% of those with angina medication with an angina code,22 and another study reported that a search using the IHD Read Code (G3) identified only 47% of probable IHD patients.19 Forty-three percent of subjects identified as having left hospital following a MI were coded at four practices.25 Hassey et al. in their single practice study noted IHD as a high priority area for identifying undiagnosed cases (an estimated 24 in a population of 10 000), although 96% of their suspected cases, based on medication and procedures, were identified by Read Code.6 Whitelaw et al. detected a median 60% of suspected angina cases and 80% of MI cases in the electronic records of 41 practices.18 However, two studies comparing the prevalence of CHD diagnostic coding with that of nitrate medication at the same practices reported higher prevalences for the diagnosis.23,28 This may suggest that estimation of the prevalence of CHD is improved when based on diagnosis, e.g. not all patients may be on medication, but may also suggest that some patients are incorrectly placed on the register.
Correctness of heart disease coding appears slightly better. Seventy percent of patients (ranging from 47 to 92% between practices) with the morbidity code had appropriate medication in one study.21 This figure improves to 83–100%6,18,19,29 when the comparison included tests and procedures, although Connolly et al.26 could only validate 53% of cases with a CHD Read Code but no nitrate prescription. Moher et al. found that surgery contact was the main predictor of being on the CHD register [odds ratio (OR) 2.1]. Other significant predictors were repeat prescriptions (OR 1.6), MI diagnosis (OR 1.5) and revascularization procedure (OR 1.5). Diagnosis of angina, year of diagnosis, age and gender were not significant predictors.29
Seven studies examined completeness of diabetes registers, and three of these also looked at correctness.6,18,21 One study compared medication and diagnostic code prevalence.23 Gold standards were medication22–24, combinations of related information (electronic and/or non-electronic)6,18,21 and hospital discharge information.25 Completeness of diabetes registers appears high, with consensus of 90% or more of diabetics identified in practice. The exception to this used a comparison with hospital discharge data (72%) and was the oldest study reviewed (data collected 1982–1984).25 Correctness was agreed to be near to maximum.
Asthma and COPD registers have been investigated in seven studies, four of which examined completeness6,18,22,24 and three of which examined correctness.6,18,21 These were in relation to medication6,21,22,24 and paper information including medication.18 Two studies compared medication and diagnostic prevalence.13,23 Quality of the asthma registers was variable: compared with medication, five out of six practices coded 70% of patients on asthma medication,22 but only 58% were coded in Coulter's older study.24 Asthma was again highlighted by Hassey et al.6 as a high priority area for identifying undiagnosed cases. They estimated 68 unrecorded cases in a population of 10 000. Pringle et al. found only 54% of asthma-coded cases in four practices with the appropriate medication.21 Comparison of medication and diagnostic prevalences gave variable results.13,23
Completeness of epilepsy coding was moderate (generally 40–70%) when compared with medication22,24 and hospital discharge information,25 but higher in two other studies (>90%) using medication6 or combinations of relevant paper-held evidence18 as the gold standard. These also found high correctness.
The completeness and correctness of glaucoma coding has been found to be >80%6,18,21 whilst the prevalence of glaucoma based on diagnostic coding was greater than for that based on medication and procedures at two practices.23
The completeness of thyroid conditions ranged from 42 to 84% across practices in the two oldest studies.24,25 Hassey et al.'s study found high correctness and completeness of 82% for hypothyroidism and 98% for hyperthyroidism.6
The evidence for hypertension is mixed. Hassey et al. found high completeness (98%) and correctness for hypertension when compared with medication.6 However, Whitelaw et al. found poor completeness in their study (median across 41 Scottish practices of 43%).18 One other study found correctness and completeness for hypertension coding of 70% in a random sample of 100 elderly patients in one practice compared with paper notes, but gave little information on methodology.8
The completeness and correctness of gout has been found to be good.6,18 However, the results for cancer have been variable. Hassey et al. in a triangulation with drugs and other treatments found high completeness and correctness for breast and prostate cancer, although both had <60 cases.6 Whitelaw et al., however, found poor completeness for breast tumour (median across practices of 57%).18 Mant and Tulloch also found poor completeness (52%) for cancer compared with hospital discharge information in their older study.25
Whitelaw et al. found poor completeness for depression (47%).18 A study of GPRD London practices assessed diagnoses of schizophrenia, affective psychoses and other psychoses compared with information contained in paper notes.11 Psychosis diagnosis appeared to be 74% complete based on just 50 cases identified through a random search of paper notes. Correctness was high for 149 affective and other psychosis computer-coded patients, although slightly lower for 102 schizophrenia-coded patients (64–89%) depending on strictness of criteria for diagnosis. However, >90% of the schizophrenia-coded patients were deemed to have a non-organic psychosis.
Hassey et al. found good completeness and correctness for rheumatoid arthritis in their practice,6 but completeness was poorer across 35 Scottish practices (median 67%).18
Other conditions have been assessed in the GPRD studies. Lawrenson et al. explored cases of venous thromboembolism (VTE), defined by morbidity code with evidence of treatment. They compared this with hospital investigation information forwarded to the research team by the GP. VTE was supported in 83% of the 169 cases.14 In a further study, a correctness of 85% of bowel disease diagnosis was found out of 157 cases based on a valid reason for diagnosis (e.g. intestinal biopsy, gastroenterology consultation, surgery). This information was obtained by a questionnaire to the GP asking how the diagnosis was made. However, only around a half of almost 7000 inflammatory bowel disease patients had a relevant prescription.15 One study suggested high correctness and completeness of pressure ulcer morbidity coding based on information from a questionnaire completed by the GP;16 another study suggested correct diagnosis of 60% in anorexia nervosa and bulimia coded cases when experts assessed information on the practice computer record and a GP questionnaire on symptoms and treatment.12 However, both studies suffered from small sample sizes.
Finally, two other GPRD studies suggested that 90% of patients prescribed non-steroidal anti-inflammatory drugs (NSAIDs) and with a hospital consultant's letter forwarded by their GP had a diagnosis on computer that matched the diagnosis in the letter.9,10
|
Discussion |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
The quality of morbidity coding appears variable. Conditions with clear diagnostic features such as diabetes have higher quality recording than conditions with more subjective criteria such as asthma. It is difficult to ascertain an improvement in quality over time. Although, two studies in the 1980s looking at chronic disease registration in Oxford practices found poor completeness of registers,24,25 quality has tended to be mixed in studies performed both in the 1990s and since the start of 2000. Hassey's more recent study (2001) in a trained practice found reasonably high completeness and correctness but suggested a number of morbidities where some cases may not have the relevant morbidity code recorded.6
The GPRD studies have shown reasonable correctness and completeness of morbidity registers (although poorer for some diseases such as anorexia nervosa and bulimia). Of the other multipractice studies, the studies on the GPASS group in Scotland,18 the MediPlus database17 and practices in Northern Ireland26 all suggest there is room for improvement in the areas assessed, but showed variation across practices. Most studies were of a reasonable size, although for rarer diseases the number of cases were small.
The completeness of consultation recording was generally high. However, although most contacts with a GP may have a computer-recorded morbidity code, this does not mean that all problems addressed during a multiproblem encounter are recorded. Only one study21 addressed this issue, which requires further research. It may be that only new problems or problems regarded as the most important are recorded, leaving ongoing or minor problems unrecorded.
Completeness and correctness of data entry may rely on the enthusiasm of practices and of individual GPs. GPs may have personal preferences for certain codes which may not always be appropriate. Morbidity coding is subjective and relies on the characteristics and idiosyncrasies of individual GPs.30 Many of the studies reported here looked at practices with explicit interests in recording information electronically or with a substantial amount of training in morbidity coding. Several multipractice studies had to discard practices from their study which were unable to provide suitable data. This biases the sample in favour of the better recorders, leading to a higher quality of recording than that which would be achieved by examining all practices. The majority of studies were also based in one localized area (exceptions include the studies based on the GPRD), which makes generalization difficult.
Evaluation of the accuracy of electronic records is enhanced by use of common coding systems. One question is at what level of a coding system should validation be demonstrable. For example, recording diagnoses only at diagnostic Chapter heading (level 1 of the Read Code) may improve the accuracy of recorded data but is too general for most research or clinical purposes. It could be argued that the insistence on use of a coding system loses the richness of the material contained in consultations, which may be better expressed in free text descriptions. GPs may feel pressured into using the codes available even if they are inappropriate for that patient.
A variety of gold standards have been used, and completeness and correctness can only be inferred in relation to the quality of the gold standard(s) used. A number of studies used the terms ‘sensitivity’ and ‘positive predictive value’. In diagnostic testing, these terms are used to imply whether patients really do have the disease. To avoid this implication, we have used the terms ‘completeness’ and ‘correctness’, as we have only examined internal validity, e.g. where one source of information (e.g. prescription data or paper notes) has been used to justify the existence of a morbidity code.31 We have not examined external validity which relates more to the normal interpretations of sensitivity and PPV.
Information in paper notes and medication were most commonly used to validate morbidity coding. Paper notes have to be scrutinized by hand, preferably by two researchers, to improve the reliability and completeness of the paper trawl. A disadvantage is the assumption that the paper notes are accurate. Entries in paper notes may be missing or illegible, and the notes may be poorly organized, which means that searching for information can be difficult. Increases in the number of paperless practices will also mean that, increasingly, only other electronic data can be used as a gold standard. Problems of comparing diagnosis with medication, which may be in paper notes or stored electronically, is that the same medication may be prescribed for many conditions, or diagnoses may be old and patients may no longer need, and therefore have prescribed, a medication. This may be one of the reasons asthma had poorer quality of recording. It also assumes that the medication information is complete and correct. Studies can only include morbidities where a medication is specific to that disease and results cannot be confounded by patients obtaining the medication, e.g. over the counter. Whilst prescriptions, hospital tests and procedures are likely to be the best internal validation for morbidities (as recommended previously5), this limits the number that can be validated.
Our review has been more specific in focus than that of Thiru et al.5 who reviewed methods of assessing the quality of all data contained in electronic records, using any gold standard and including non-UK studies. This has allowed us to give a specific assessment of the quality of computerized morbidity recording in primary care in the UK and to critique methods for assessing this quality.
The decision to include only UK studies in this review will have reduced the number of studies. However, the difference in the structure of primary care services and coding and computer systems between nations makes it difficult to combine studies from different countries. Estimates based on our search suggest that 6–10 non-UK studies would have been included although they are unlikely to have uncovered any different methodologies. We have further excluded studies which have compared patient's self-report of disease with that contained in computerized medical records. This would be likely to entail a systematic review of its own, and the objective of such studies may be different. It is difficult to know quite which is being validated, the self-report or the medical records. Studies comparing prevalence rates with external data (from national sources, other practices or at different time points) were also not accepted into the review. Illness prevalences from different localities or at different time points are subject to a range of influences other than the quality of coding, notably variations in prevalence, since certain morbidities or morbidity codes may become more fashionable over time or criteria for a diagnosis may change.32 We also did not search for grey literature and other unpublished studies. It is possible these may have added to our review.
This paper has reviewed the quality of morbidity coding in general practice electronic records. As this review has highlighted, it does appear possible to assess and verify the quality of coding of certain morbidities, and some of these have been shown to be generally well coded; others have poorer levels of coding. Training of practices, as shown in the GPRD studies and in the study by Hassey et al.,6 can lead to a reasonable quality of coding. As practices increasingly use computers to record consultations and other medical information, there is a need to ensure that there is a high level of completeness and correctness of not just morbidity codes relevant to the consultation, but also information from external sources such as hospital letters. The focus should be now be on methods to encourage and improve the quality of this coding in general practice.
|
References |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
1 RCGP. General Practice Computerisation. London: Royal College of General Practitioners; 2000.
2 Mitchell E, Sullivan F. A descriptive feast but an evaluative famine: systematic review of published articles on primary care computing during 1980–97. Br Med J 2001; 322: 279–282.
3 NHS Information Authority. The Clinical Terms Version 3 (The Read Codes). Birmingham: NHS Information Authority; 2000.
4 Hogan WR, Wagner MM. Accuracy of data in computer-based patient records. J Am Med Inform Assoc 1997; 4: 342–355.
5 Thiru K, Hassey A, Sullivan F. Systematic review of scope and quality of electronic patient record data in primary care. Br Med J 2003; 326: 1070–1072.
6 Hassey A, Gerrett D, Wilson A. A survey of validity and utility of electronic patient generated records in a general practice. Br Med J 2001; 322: 1401–1405.
7 Gilliland AEW, Mills KA, Steele K. General practitioner records on computer—handle with care. Fam Pract 1992; 9: 441–450.
8 Hooker RC, Cowap N, Newson R, Freeman GK. Better by half: hypertension in the elderly and the ‘rule of halves’: a primary care audit of the clinical computer record as a springboard to improving care. Fam Pract 1999; 16: 123–128.
9 Jick H, Jick SS, Derby LE. Validation of information recorded on a general practitioner based computerised data resource in the United Kingdom. Br Med J 1991; 302: 766–768.
10 Jick H, Terris BZ, Derby LE, Jick SS. Further validation of information recorded on a general practitioner based computerized data resource in the United Kingdom. Pharmacoepidemiol Drug Saf 1992; 1: 347–349.[CrossRef]
11 Nazareth I, King M, Haines A, Rangel L, Myers S. Accuracy of diagnosis of psychosis on general practice computer system. Br Med J 1993; 307: 32–34.
12 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; 169: 705–712.
13 Hansell A, Hollowell J, Nichols T, McNiece R, Strachan D. 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.
14 Lawrenson R, Todd JC, Leydon GM, Williams TJ, Farmer RD. Validation of the diagnosis of venous thromboembolism in general practice database studies. Br J Clin Pharmacol 2000; 49: 591–596.[CrossRef][Web of Science][Medline]
15 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: 211–218.[CrossRef][Web of Science][Medline]
16 Margolis DJ, Bilker W, Knauss J, Baumgarten M, Strom BL. The incidence and prevalence of pressure ulcers among elderly patients in general medical practice. Ann Epidemiol 2002; 12: 321–325.[CrossRef][Web of Science][Medline]
17 de Lusignan S, Stephens PN, Adal N, Majeed A. Does feedback improve the quality of computerized medical records in primary care? J Am Med Inform Assoc 2002; 9: 395–401.
18 Whitelaw FG, Nevin SL, Milne RM, Taylor RJ, Taylor MW, Watt AH. Completeness and accuracy of morbidity and repeat prescribing records held on general practice computers in Scotland. Br J Gen Pract 1996; 46: 181–186.[Web of Science][Medline]
19 Gray J, Majeed A, Kerry S, Rowlands G. Identifying patients with ischaemic heart disease in general practice: cross sectional study of paper and computerised medical records. Br Med J 2000; 321: 548–550.
20 Neal RD, Heywood PL, Morley S. Real world data—retrieval and validation of consultation data from four general practices. Fam Pract 1996; 13: 455–461.
21 Pringle M, Ward P, Chilvers C. Assessment of the completeness and accuracy of computer medical records in four practices committed to recording data on computer. Br J Gen Pract 1995; 45: 537–541.[Web of Science][Medline]
22 Scobie S, Basnett I, McCartney P. Can general practice data be used for needs assessment and health care planning in an inner-London district? J Public Health Med 1995; 17: 475–483.
23 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 two-practice pilot study. Health Informatics J 1999; 5: 224–232.
24 Coulter A, Brown S, Daniels A. Computer held chronic disease registers in general practice: a validation study. J Epidemiol Community Health 1989; 43: 25–28.
25 Mant D, Tulloch A. Completeness of chronic disease registration in general practice. Br Med J 1987; 294: 223–224.
26 Connolly P, Cupples ME, Cuene-Grandidier H, Johnston D, Passmore P. The importance of validating the diagnosis of coronary heart disease when measuring secondary prevention: a cross-sectional study in general practice. Pharmacoepidemiol Drug Saf 2002; 11: 311–317.[CrossRef][Web of Science][Medline]
27 Pearson N, O'Brien J, Thomas H, Ewings P, Gallier L, Bussey A. Collecting morbidity data in general practice: the Somerset morbidity project. Br Med J 1996; 312: 1517–1520.
28 Bottomley A. Methodology for assessing the prevalence of angina in primary care using practice based information in northern England. J Epidemiol Community Health 1997; 51: 87–89.
29 Moher M, Yudkin P, Turner R, Schofield T, Mant D. An assessment of morbidity registers for coronary heart disease in primary care. Br J Gen Pract 2000; 50: 706–709.[Web of Science][Medline]
30 Fleming DM. The measurement of morbidity in general practice. J Epidemiol Community Health 1991; 45: 180–183.
31 Newcombe RG, Altman DG, Bryant TN. Electronic patient records in general practice. Methods of evaluation of electronic patient records entail dangers. Br Med J 2001; 323: 1184–1185.
32 Anderson JE. Reliability of morbidity data in family practice. J Fam Pract 1980; 10: 677–683.[Web of Science][Medline]
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  J.-K. Soler, I. Okkes, M. Wood, and H. Lamberts The coming of age of ICPC: celebrating the 21st birthday of the International Classification of Primary Care Fam. Pract., August 1, 2008; 25(4): 312 - 317. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. Majeed, J. Car, and A. Sheikh Accuracy and completeness of electronic patient records in primary care Fam. Pract., August 1, 2008; 25(4): 213 - 214. [Full Text] [PDF]
|
|
|
|
|
|
E. P Klein Woolthuis, W. J. de Grauw, W. H. van Gerwen, H. J. van den Hoogen, E. H van de Lisdonk, J. F. Metsemakers, and C. van Weel Identifying people at risk for undiagnosed type 2 diabetes using the GP's electronic medical record Fam. Pract., June 14, 2007; (2007) cmm018v2. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. J Calvert, R. J McManus, and N. Freemantle The management of people with type 2 diabetes with hypoglycaemic agents in primary care: retrospective cohort study Fam. Pract., June 1, 2007; 24(3): 224 - 229. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
H. Brouwer, P. Bindels, and H. Weert Data quality improvement in general practice Fam. Pract., October 1, 2006; 23(5): 529 - 536. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. de Lusignan and C. van Weel The use of routinely collected computer data for research in primary care: opportunities and challenges Fam. Pract., April 1, 2006; 23(2): 253 - 263. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. G. Deckers, W J. Paget, F. G Schellevis, and D. M Fleming European primary care surveillance networks: their structure and operation Fam. Pract., April 1, 2006; 23(2): 151 - 158. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. Downs, S. Turner, M. Bryans, J. Wilcock, J. Keady, E. Levin, R. O'Carroll, K. Howie, and S. Iliffe Effectiveness of educational interventions in improving detection and management of dementia in primary care: cluster randomised controlled study. BMJ, March 25, 2006; 332(7543): 692 - 696. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 K. Jordan, M. Porcheret, U. T. Kadam, and P. Croft The use of general practice consultation databases in rheumatology research Rheumatology, February 1, 2006; 45(2): 126 - 128. [Full Text] [PDF]
|
|
|
|
 |
|
 H. Buurma, P. A. De Smet, M. Kruijtbosch, and A. C. Egberts Disease and Intolerability Documentation in Electronic Patient Records Ann. Pharmacother., October 1, 2005; 39(10): 1640 - 1646. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||