Contact| Forum

Research Outputs - Work Packages

Work Package 4

Research Topic
Dr Olga Kostopoulou

Diagnostic errors in primary care

Lead: Health and Social Care Research, King's College London, UK.

Contact: Dr Olga Kostopoulou

Description of work


Little empirical research has been done on the causes of Diagnostic Errors in Primary Care. An Australian analysis of 142 self-reported incident errors suggest that at least 30% were system errors (communication failure, non-follow up) and at least 30% were knowledge errors (diagnosis not considered or inappropriately ruled out), and the remainder were a miscellany including patient deliberately withholding information and equipment failure. If these findings are generalisable to other settings, then the two major areas of primary care errors that may be amenable to interventions are system errors and knowledge errors. This has been confirmed by Dovey and colleagues in a pilot study in the USA and by Makeham and colleagues who reported on the Australian arm of an international study. Esmail and colleagues in Manchester were involved in the international study reported by Makeham and have found similar results from work in the UK. All these authors agree on the role of system errors and knowledge failures in the genesis of primary care errors though the percentages that each type of error contributes to the overall rate of error in primary care varies. Knowledge areas are the domain where diagnosis probably plays the greatest role.

A review on the frequency and nature of medical error in primary care, noted that studies reported wide differences in rates of errors in primary care, varying from 5 to 80 per 100 000 consultations15. Errors related to diagnosis were consistently the most common category, varying from 26 to 78% of identified errors. Errors associated with diagnosis, either delayed or missed, were most likely to result in major harm to the patient or precipitate hospital admission. Work that Esmail and colleagues have completed on the analysis of medico-legal databases confirms the preponderance of errors associated with diagnosis in the primary care setting. We propose a series of linked pieces of work which will between them increase our understanding of diagnostic error in primary care.

Data from both major UK medical defence societies show that the most common claim in primary care is of delay in diagnosis or misdiagnosis (63-66%). Nevertheless, little empirical research has been done on the causes of diagnostic errors in primary care. Evidence from retrospective reviews of litigation cases in UK primary care, and US ambulatory care, as well as from a multi-method US study in internal medicine suggests that diagnostic errors can be caused by both cognitive and system factors. Researchers typically address either the cognitive or the system causes of diagnostic errors.

There is a long tradition of studying how clinicians reason, make decisions, and solve diagnostic problems, spanning more than 3 decades. Psychologists have used a variety of methodologies (for a review see Harries & Kostopoulou), such as:

• Process tracing: general term that characterises a variety of methods developed to track the reasoning process as it unfolds over time, leading to a decision, for example, ‘think aloud’ and ‘active information search’.

• Clinical Judgement Analysis: uses large numbers of patient vignettes and statistically models (via regression analysis) how patient characteristics influence clinical judgements. Carefully controlled, between-subject experiments have investigated the influence of a number of cognitive biases in clinical reasoning.

Clinical judgement analysis and experiments often reduce the ecological validity of the situations under study for the sake of controllability. Process tracing techniques preserve realism but intervene on the reasoning process and can therefore change it (‘reactivity’). More naturalistic methodologies have developed as part of the ‘Naturalistic Decision Making’ (NDM) paradigm and are mainly qualitative:

observations, interviews, video recordings of performance, critical incident analysis, etc. They too have raised methodological concerns, especially, the numerous uncontrolled variables under study and the subjectivity in interpretation. On the other hand, they are more likely to capture system factors of diagnostic errors than any of the traditional 3 approaches above. NDM approaches have not been used to study decision making in primary care and have mainly been used in hospital settings, especially anaesthesia, surgery, and intensive care.

Since the publication of the Harvard Medical Practice study, two methodologies have been extensively used to study patient safety (not limited to diagnostic error): retrospective record review and incident reporting. Neither however is a good way of identifying causes of diagnostic errors. In fact, incident reporting suffers from very low rates of diagnostic errors being reported – ranging from 0.5% to 3.9% in primary care (increasing to 14% to include diagnostic errors by pharmacists and hospital doctors).

In terms of studying the quality of actual clinical practice and therefore the potential for diagnostic (and other types of) error, carefully constructed patient vignettes were shown to be better tools than record review and compared well to the gold standard, i.e. standardised patients (trained actors who presented unannounced to physicians' clinics). Kostopoulou provides methodological recommendations for the study of diagnostic reasoning and error. Nevertheless, the hiatus between methods for studying the cognitive causes and those focussing on the system causes of diagnostic errors is ever-present. Researchers rarely cross this boundary and rarely exchange information and ideas. Since clinicians work within multi-layered systems, research should take account of all aspects of the interaction that can produce errors: clinical cognition, clinical environment, and – a certainly neglected aspect: the types and features of the diagnostic problems encountered by primary care doctors.

In addition to the important role of cognition in diagnosis, there is a growing interest in advanced information and communication technologies (ICT) in patient safety. The eHealth programme has funded a study of the role of ICT in patient safety (ICT2005/175-173232, Impact of ICT on Patient Safety and Risk Management in Healthcare). Although the potential for ICT in this area is considerable, we are particularly interested in the interface between cognition, diagnosis and information management at the point of consultation between a clinician and patient. A recent paper has shown how current electronic health records are deficient in terms of information presentation and alerting functions. There is a need to promote co-ordination between researchers in disparate areas such as data mining, computerised decision support and the electronic health record (eHR), in order to focus on how to improve the role of the eHR in supporting diagnosis. Better presentation of existing records, structured capture of clinical information during the consultation, integration of patient-related data with epidemiological and diagnostic data held elsewhere, coupled with the presentation of salient and cognitively appropriate alerts are all areas of importance to this work package.

We propose a series of linked pieces of work which aim to disseminate knowledge and increase our understanding of diagnostic error in primary care, and provide recommendations for best research practice. Outputs from this work will inform the development of a coherent research agenda at a pan-European level.

Working Group Synopsis (PDF)

Tools and Guidance notes

  • Report on methods of identifying diagnostic errors in primary care

Journal papers


Conference proceedings

  • Gandhi, T. K., A. Kachalia, et al. (2006). "Missed and delayed diagnoses in the ambulatory setting: A study of closed malpractice claims." Annals of Internal Medicine 145(7): 488-496.
  • Graber, M. L., N. Franklin, et al. (2005). "Diagnostic Errors in Internal Medicine." Arch Intern Med 165: 14931499.
  • Harries, C. and O. Kostopoulou (2005). Psychological approaches to measuring and modelling clinical decision-making. Handbook of Health Research Methods: Investigation, Measurement and Analysis. A. Bowling and S. Ebrahim. Maidenhead, Birkshire, Open University Press: 331-361.
  • Kostopoulou, O. (2009). Diagnostic errors: psychological theories and research implications. Health Care Errors and Patient Safety. B. Hurwitz and A. Sheikh. Oxford, Wiley-Blackwell: 97-111.
  • Phillips, R. L., Jr., L. A. Bartholomew, et al. (2004). "Learning from malpractice claims about negligent, adverse events in primary care in the United States." Quality and Safety in Health Care 13(2): 121-126.
  • Sandars, J. and A. Esmail (2003). "The frequency and nature of medical error in primary care: understanding the diversity across studies." Family Practice 20(3): 231-236.
  • Silk, N. (2000). What went wrong in 1000 negligence claims. Health Care Risk Report: 13-16.

LINNEAUS deliverable(s)

 Report on methods of identifying diagnostic errors in primary care

Other LINNEAUS reports