Validation and validity of diagnoses in the General Practice Research Database: a systematic review

By E. Herrett, S. Thomas, W. Schoonen, L. Smeeth, A. Hall. British Journal of Clinical Pharmacology, 2010

1,148 citations recorded by the source corpus.

AIMS To investigate the range of methods used to validate diagnoses in the General Practice Research Database (GPRD), to summarize findings and to assess the quality of these validations. METHODS A systematic literature review was performed by searching PubMed and Embase for publications using GPRD data published between 1987 and April 2008. Additional publications were identified from conference proceedings, back issues of relevant journals, bibliographies of retrieved publications and relevant websites. Publications that reported attempts to validate disease diagnoses recorded in the GPRD were included. RESULTS We identified 212 publications, often validating more than one diagnosis. In total, 357 validations investigating 183 different diagnoses met our inclusion criteria. Of these, 303 (85%) utilized data from outside the GPRD to validate diagnoses. The remainder utilized only data recorded in the database. The median proportion of cases with a confirmed diagnosis was 89% (range 24–100%). Details of validation methods and results were often incomplete. CONCLUSIONS A number of methods have been used to assess validity. Overall, estimates of validity were high. However, the quality of reporting of the validations was often inadequate to permit a clear interpretation. Not all methods provided a quantitative estimate of validity and most methods considered only the positive predictive value of a set of diagnostic codes in a highly selected group of cases. We make recommendations for methodology and reporting to strengthen further the use of the GPRD in research.

Subfields: Family Practice, General Health Professions, Statistics and Probability

View Paper: https://www.semanticscholar.org/paper/d33ce574ae7602385f1861c2fe77ce6db2b11099

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In General Health Professions

In Statistics and Probability

Citations in the corpus

The REporting of studies Conducted using Observational Routinely-collected health Data (RECORD) Statement (2015, background): “Two recent systematic reviews also document poor reporting of studies undertaken to validate data from routine data sources [5,6], which can obscure sources of bias, hamper efforts to undertake meta-analyses, and lead to erroneous conclusions.”
Data Resource Profile: Clinical Practice Research Datalink (CPRD) (2015, background): “However, reporting of validation studies was often too poor to permit a clear interpretation, and the majority of studies focused on positive predictive value rather than sensitivity or specificity.(39) The quality of primary care data is variable because data are entered by GPs during routine consu”
Extracting Diagnoses and Investigation Results from Unstructured Text in Electronic Health Records by Semi-Supervised Machine Learning (2012, methodology): “It has been widely used for research on drug safety and clinical epidemiology [17].”
Antidepressant use and risk of adverse outcomes in older people: population based cohort study (2011, result): “We did not specifically validate our outcomes as part of this study. However, reviews of validation studies in other UK primary care databases have shown overall high levels of validity, and we would expect similar levels of validity in QResearch.”
Comparison of Random Forest and Parametric Imputation Models for Imputing Missing Data Using MICE: A CALIBER Study (2014, neutral): “CALIBER is a database of linked routinely collected electronic health records from England (16), comprising data from primary care (Clinical Practice Research Datalink) (18), hospital admissions (19), the national registry of acute coronary syndromes (20), and the national death registry.”
Rising burden of gout in the UK but continuing suboptimal management: a nationwide population study (2014, background): “A recent systematic review supported a high validity of recorded diagnoses in CPRD, with a median proportion of cases with a confirmed diagnosis of 89% for 183 different conditions.[22]”