Completeness and diagnostic validity of recording acute myocardial infarction events in primary care, hospital care, disease registry, and national mortality records: cohort study
BMJ 2013; 346 doi: https://doi.org/10.1136/bmj.f2350 (Published 21 May 2013) Cite this as: BMJ 2013;346:f2350
All rapid responses
Rapid responses are electronic comments to the editor. They enable our users to debate issues raised in articles published on bmj.com. A rapid response is first posted online. If you need the URL (web address) of an individual response, simply click on the response headline and copy the URL from the browser window. A proportion of responses will, after editing, be published online and in the print journal as letters, which are indexed in PubMed. Rapid responses are not indexed in PubMed and they are not journal articles. The BMJ reserves the right to remove responses which are being wilfully misrepresented as published articles or when it is brought to our attention that a response spreads misinformation.
From March 2022, the word limit for rapid responses will be 600 words not including references and author details. We will no longer post responses that exceed this limit.
The word limit for letters selected from posted responses remains 300 words.
We read with interest the article by Herrett and colleagues, reporting that a primary care database alone captured approximately 75% of all cases of acute myocardial infarction (Herrett, Figure 3). This could have implications for every study that has used primary care databases to derive, or validate, clinical prediction rules for cardiovascular disease.
For example, a study using The Health Improvement Network (THIN) database to evaluate cardiovascular risk equations reported that Framingham equations overestimated cardiovascular risk by 23% [1]. The Framingham Investigators used a comprehensive standardised ascertainment method for cardiovascular outcomes [2] whereas THIN relies on similar data to the CPRD database studied by Herrett. The apparent discrepancy in accuracy between Framingham and cardiovascular events in this THIN study appears remarkably similar to the discrepancy in ascertainment reported by Herrett et al., albeit for MI alone rather than total cardiovascular risk. Subsequent external validations using THIN also reported that Framingham over estimated risk [3,4].
We suggest that the choice of risk score for use in clinical practice should ideally be based on head to head validation studies that have used linked methods, similar to those of Herrett, to establish event rates.
1. Collins GS, Altman DG. An independent external validation and evaluation of QRISK cardiovascular risk prediction: A prospective open cohort study. BMJ (Online). 2009;339(7713):144-7.
2. Shurtleff D: Some characteristics related to the incidence of cardiovascular disease and death: Framingham study, 16-year follow-up, in Kannel WB, Gordon T (eds): The Framingham Study: An Epidemiological Investigation of Cardiovascular Disease, section 26. US Government Printing Office No. 0-414-297, Washington, D.C.,1971
3. Collins GS, Altman DG. Predicting the 10 year risk of cardiovascular disease in the united kingdom: Independent and external validation of an updated version of QRISK2. BMJ (Online). 2012;345(7867).
4. Collins GS, Altman DG. An independent and external validation of QRISK2 cardiovascular disease risk score: A prospective open cohort study. BMJ (Online). 2010;340(7758):1231.
Competing interests: Richard Stevens is a member of the Independent Scientific Advisory Committee for MHRA database research. Both authors have used primary care databases in their research including CPRD, THIN and QRESEARCH.
Re: Completeness and diagnostic validity of recording acute myocardial infarction events in primary care, hospital care, disease registry, and national mortality records: cohort study
We thank Stevens and McManus for pointing out that using unlinked primary care databases may underestimate the absolute risk of myocardial infarction as they fail to capture all events. Non-differential under-recording may not cause bias in studies solely focussing on relative effects, but for predicting absolute risks or rates, using a single data source whether from primary or secondary care settings, may lead to biased estimates. Ideally all events should be recorded definitively in a single electronic health record, but as this does not currently happen in the NHS or other health systems, we recommend the use of linked data sources, such as those available through the Clinical Practice Research Datalink (CPRD), to overcome under-recording in individual sources.
In the CALIBER programme [1] we are developing prognostic models for patients with coronary disease in a linked dataset, using multiple data sources for outcome ascertainment. We are investigating the use of free text entered by doctors as an additional source of diagnostic information [2], and we are part of the new network of four UK e-Health Informatics Research Centres [3] which will make use of linked datasets available through the CPRD, facilitate further linkages and enable greater use of electronic health records for research.
1. Denaxas S, George J, Herrett E, Shah A, Kalra D, Hingorani AD, et al. Data resource profile: Cardiovascular Disease Research using Linked Bespoke Studies and Electronic Records (CALIBER). Int J Epidemiol. 2012;41:1625-38. doi: 10.1093/ije/dys188
2. Shah AD, Martinez C, Hemingway H. The freetext matching algorithm: a computer program to extract diagnoses and causes of death from unstructured text in electronic health records. BMC Med Inform Decis Mak. 2012;12:88. doi: 10.1186/1472-6947-12-88
3. http://www.mrc.ac.uk/Ourresearch/ResearchInitiatives/E-HealthInformaticsResearch/index.htm
Competing interests: No competing interests