Journal of Public Health Advance Access originally published online on July 13, 2006
Journal of Public Health 2006 28(3):283-287; doi:10.1093/pubmed/fdl044
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Ethnicity recording in general practice computer systems
P. Kumarapeli, Research Assistant1
R. Stepaniuk, DataNet Project Lead2
S. de Lusignan, Senior Lecturer1
R. Williams, DataNet Lead Clinician2
G. Rowlands, Professor3
1 Primary Care Informatics, Division of Community Health Sciences, St. George’s—University of London, London SW17 0RE, UK
2 Primary Care Service Improvement, Lambeth PCT, 1 Lower Marsh, London SE17NT, UK
3 Institute of Primary Care and Public Health, Faculty of Health and Social Care, London South Bank University, 103 Borough Road, London SE1 0AA, UK
Address correspondence to S. de Lusignan, E-mail: slusigna{at}sgul.ac.uk
Background Ethnicity data in general practice (GP) computerized medical records can be utilized to audit equity in health care.
Methods We evaluated a patient profiling project targeted to improve ethnicity recording.
Results Data extracted from 16 practices showed an increase in ethnicity recording from <1% before the intervention to 48% after. Recorded codes could be mapped onto the basic national statistics six-category ethnicity classification headings, and their proportions were similar to the 2001 census values.
Conclusion Recording of data using multiple coding hierarchies has reduced the utility of data as clinically important ethnic subgroups cannot be identified. Practitioners should be encouraged to use the single recommended ethnicity coding hierarchy.
Keywords: computers, demography, ethnic groups, ethnicity, general practice, medical records systems computerized, primary care, vocabulary controlled
 |
Introduction |
|---|
 TOP Introduction MethodResultsDiscussionConclusionAcknowledgementsReferences |
|---|
Ethnicity can be linked with ill health and utilization of health care, yet it is poorly recorded in medical records.1 For example, the uptake of certain preventive services is reduced in some ethnic groups,2 patterns of presentation3 and treatment can also be ethnic specific,4 and different ethnic groups may also respond differently to medication.5 UK general practices (GPs) are almost universally computerized, and facilities exist in GP computer systems to record ethnicity information. Despite its potential utility, we have found <0.3% of patients have their ethnicity recorded.6
Comprehensive ethnicity data in computerized patient medical records would enable health services to audit any differences in disease prevalence and any differential access to and delivery of services. A south London primary care organization implemented an intervention, the Individual Patient Registration Profile Project (IPRP), to improve the recording of ethnicity and other patient characteristics. Patients were asked to self-ascribe their ethnic group they thought to best describe them. These data were recorded in their computerized medical records in a structured (Read coded) format (Box 1). The 16 practices that volunteered for the project were provided with a data entry template to streamline data entry and standardize the Read codes used to record ethnicity. We carried out this evaluation to report the effectiveness of the intervention.
| Box 1 Read codes and the ethnicity hierarchies There are several versions of the Read codes. All these practices used the 5-byte version of Read code version 2. Read codes have a hierarchical structure organized like a family tree. Ethnicity codes appear in different parts of the hierarchy. For example ‘134... Country of origin’ codes appear in Chapter 1—History/Symptoms, ‘9S... and Ethnic group (Census)’ appear in Chapter 9—Administration. When a new classification is added, an additional hierarchy is added in the same chapter. Three different ethnicity coding hierarchies exist within Chapter 9—Administration of the Read codes; ‘9S... Ethnic group (Census)’, ‘9i... Ethnic category (2001 census)’ and ‘9T... Ethnicity and other related nationality data’. Further information about coding systems and where the read codes fit in: http://www.pcel.info/pdf/sdel_coding.pdf. For information about Read ethnicity hierarchies: http://www.pcel.info/ethnicity.
|
|
Method |
|---|
|
TOPIntroductionMethodResultsDiscussionConclusionAcknowledgementsReferences |
|---|
We developed the audit criteria to assess the project’s effect on ethnicity recording levels. We wanted to explore the prevalence of ethnicity data recording and the proportion that could be mapped onto the 16 ethnic categories recommended by the NHS and used in the 2001 census.7 We also wished to see what proportion could be mapped onto the simpler five main ethnic classification headings developed by National Statistics.8
We defined the data set to extract demographic details and ethnicity data. The two principal groups of ethnicity codes are the ‘9i... Ethnic category—2001 census’ (currently recommended by the NHS as it maps onto the 2001 census data. It has 16 principal categories and 86 ‘child codes’, the subdivisions which allows comprehensive description of ethnic subgroups) and ‘9S... ethnic group’ codes. Other available hierarchies are the ‘9T... Ethnicity related national data’, ‘134.. Country of Origin codes’ and the ‘226.. On examination ethnic group’. We also looked for the local codes created within individual practices. EMIS, the commonest brand of GP clinical computer system used in England, has some codes unique to that system and allows users to create their own codes. We collectively refer to these as ‘EMIS local codes’.
We developed the data extraction queries for Morbidity Information Query and Export Syntax (MIQUEST)—a Department-of-Health-sponsored data extraction tool designed to extract data from the different brands of GP computer systems. We aggregated and processed these data using our in-house method which ensures traceability.9 Responses were exported into Statistical Package for Social Services (SPSS) to produce a single flat file for analysis. We derived categorical variables to represent the different ethnicity hierarchies. We translated the ‘9i...’ codes using the published guidelines10 and used our own mapping process for the other hierarchies (mapping table available at http://www.pcel.info/ethnicity). We first mapped these data onto the 16 category classifications and then to the five headings National Statistics classification.
|
Results |
|---|
|
TOPIntroductionMethodResultsDiscussionConclusionAcknowledgementsReferences |
|---|
The combined study population was 117 158. This represented 47.9% (117 158/244 834) of the borough’s population. Baseline recording of ethnicity data was poor; <1% of the practice population had ethnicity codes recorded. The median level of ethnicity recording after the study was 46.85% (interquartile range—IQR 12.85%); minimum and maximum levels were 14.01 and 74.77%, respectively. Thirteen practices achieved more than 40% ethnicity recording and five achieved more than 50%. Four hundred and eighty three (0.9%) were recorded as ethnicity not given.
Ethnicity recording increased with age, but superimposed on this were high levels of recording for young adults: 63.25 and 51.07% for 21–30 and 31–40 age bands, respectively. More codes were recorded for females than males; the medians were 57.15 (IQR 3.9%) and 46.03% (IQR 7.6%), respectively. Patients over 40 years had 46.74% (17 709/37 888) ethnicity recorded. For patients aged over 65 years, this was 54.94% (6349/11 556).
Ethnicity recording was primarily carried out using ‘9S...’ (68.37%) and ‘9i...’ codes (28.18%) (Table 1). The ‘226... On examination’ Read codes could not be mapped because of the use of non-specific ‘O/E Ethnic group’ code. We did not return any data in the ‘134..’ hierarchy, because of a technical problem with the queries.
|
The commonest recorded ethnicity was in the ‘White’ ONS ethnic category (60.88%, 34 013/55 871); this represents 29.03% (34 013/117 158) of the study population. All but three practices had a majority of ‘White’ ethnicity data; these three practices had a majority of ‘Black or Black British’ category. Overall, ‘Black or Black British’ was recorded for 22.99% (12 844/55 871). ‘Mixed’, ‘Asian or Asian British’ each account for about 3% of the ethnicity codes. People of ‘Chinese or other’ ethnic group were more likely to have their ethnicity recorded using the 9i hierarchy and local EMIS codes.
The mixed use of hierarchies and the use of non-specific codes made it difficult to identify individuals who might be in specific high-risk groups. For example, we identified 542 south Asian women, out of a total of 920 women with Asian ethnic codes. We could differentiate south Asians because both the ‘9S...’ and ‘9i...’ hierarchies allow them to be separately coded. A total of 2016 men were identified as Afro-Caribbean from 5917 ‘black or black British’ codes; only 237 were coded using ‘9i...’.
Despite the inter-practice variation in the rate of ethnicity recording, nearly all the practices have used the full range of hierarchies available and proportions of people in each ethnic category was not statistically different from the 2001 census findings (Fig. 1).
|
|
Discussion |
|---|
|
TOPIntroductionMethodResultsDiscussionConclusionAcknowledgementsReferences |
|---|
Main finding of this study
The intervention improved ethnicity recording levels although the completeness of these data was variable. Over 95% of the data could be mapped onto the five category ONS classification, and the proportion of each ethnic category was not statistically dissimilar from 2001 census values. Coding ethnic group into the less specific and relatively crude ethnic groups decreases the ability to identify high-risk ethnic subgroups that might benefit from being targeted for specific health interventions. However, a system that allows data recording from multiple detailed hierarchies is unhelpful because of the difficulty in mapping between them. A single system for coding ethnicity is needed if local records are to have consistency and comparability to national census data. We believe that 9i... hierarchy would best meet this requirement.
What is already known on this topic
Ethnicity data quality is generally poor in primary care. NHS organizations have only recently received proper guidance for ethnicity recording,7 recommending the 16 ethnic categories used in the 2001 census. Ethnicity recording has been added to the list of financially incentivized quality targets for GPs to improve data quality, though it allows use of the ‘9i...’ or ‘9s...’ hierarchy. However, the older codes remained within the coding systems, making it all too easy for primary care clinicians to use these inadvertently.
What this study adds
This study highlights the need to have a limited list of ethnicity codes in primary care. Whilst templates, as used in this study, limit the choices to recommended codes, all the hierarchies are visible on the standard coding screens. Only an informatician or clinician, with a special interest in ethnicity coding, would realize the subtleties of the different Read code hierarchies for ethnicity and know to select a ‘9i...’ code from the list presented.
Educational programmes, data recording initiatives and limiting GP computer systems to display only a preferred list of codes all have a role to play in improving the quality of data in this domain. Further research is needed to explore how ethnicity might best be recorded to enable health interventions tailored to individual ethnic categories to be provided and monitored. We find the decision to include both ‘9i...’ and ‘9s...’ hierarchies in the quality targets for GPs hard to understand. Research is also needed to understand the justification for using two hierarchies that do not precisely map.
Limitations of this study
We could only access the structured data within the GP computer databases. It is possible that more specific data could have been recorded within the narrative (free-text) record.
|
Conclusion |
|---|
|
TOPIntroductionMethodResultsDiscussionConclusionAcknowledgementsReferences |
|---|
This intervention has achieved higher levels of ethnicity recording. Its method could be replicated to improve ethnicity data quality. However, although the data recorded could almost all be grouped into crude ethnic categories, it remains a challenge to improve data quality so that it can be utilized effectively for providing health care. In the absence of any obvious benefit of any of the other hierarchies, we would recommend the use of the ‘9i...’ hierarchy for ethnicity recording in UK practice. ‘9i...’ codes are the most comprehensive and most readily map to 2001 UK census data. Researchers and health service managers need to recognize the complexity associated with multiple coding hierarchies for ethnic data and adopt strategies to reduce the variability in their use.
|
Acknowledgements |
|---|
|
TOPIntroductionMethodResultsDiscussionConclusionAcknowledgementsReferences |
|---|
We thank the participating practices and members of the Datanet team. We also thank Nigel Hague for developing the MIQUEST queries and Neil Dhoul for supporting the data collection.
|
References |
|---|
|
TOPIntroductionMethodResultsDiscussionConclusionAcknowledgementsReferences |
|---|
- London Health Observatory. Indications of public health in the English regions. no. 4: Ethn Health. http://www.lho.org.uk/viewResource.aspx?id=9841 (5 January 2006, date last accessed).
- Griffiths C, Cooke S, Toon P. Registration health checks: inverse care in the inner city? Br J Gen Pract 1994;44:201–4.[Web of Science][Medline]
- Boyd AE, Murad S, O’shea S et al. Ethnic differences in stage of presentation of adults newly diagnosed with HIV-1 infection in south London. HIV Med 2005;6:59–65.[CrossRef][Web of Science][Medline]
- Hull SA, Aquino P, Cotter S. Explaining variation in antidepressant prescribing rates in east London: a cross sectional study. Fam Pract 2005;22:37–42.
[Abstract/Free Full Text] - Blann A, Hewitt J, Siddiqui F et al. Racial background is a determinant of average warfarin dose required to maintain the INR between 2.0 and 3.0. Br J Haematol 1999;107:207–9.[CrossRef][Web of Science][Medline]
- London health Observatory. Ethnic Health Intelligent Program (EHIP)—Ethnicity coding 2005. http://www.lho.org.uk/Health_Inequalities/EHIP/EthnicityCoding.aspx (12 January 2006, last accessed).
- National Health Services Information Authority. Data Set Change Notice 21/2000 CDS, HES and Workforce:Ethnic Data and DSCN 2000. http://www.connectingforhealth.nhs.uk/dscn/dscn2000/212000.pdf (5 January 2006, last accessed).
- National Statistics.Ethnic group statistics: A guide for the collection and classification of ethnicity data. Neighbourhood Statistic 2003. http://www.statistics.gov.uk/about/ethnic_group_statistics/downloads/ethnic_group_ statistics.pdf (22 February 2006, date last accessed).
- van Vlymen J, de Lusignan S, Hague N et al. Ensuring the quality of aggregated general practice data: lessons from the Primary Care Data Quality Programme (PCDQ). Stud Health Technol Inform 2005;116:1010–5.[Medline]
- Department of Health. A practical guide to ethnic monitoring in the NHS and social care Annex D: Detailed breakdown of the ONS 2001 census codes for ethnic group. Health and Social Care Information Centre. http://www.dh.gov.uk/assetRoot/04/11/70/09/04117009.pdf. (12 January, 2006 date last accessed).
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||