Article Text

Download PDFPDF

Cohort study
Predictive values of GPs’ suspicion of serious disease are high enough to warrant subsequent investigation
  1. Knut Holtedahl
  1. Department of Community Medicine, UiT The Arctic University of Norway, Tromsø, Norway
  1. Correspondence to : Professor Knut Holtedahl, Department of Community Medicine, UiT The Arctic University of Norway, N-9037 Tromsø, Norway; knut.holtedahl{at}

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Commentary on: OpenUrlAbstract/FREE Full Text


Although general practitioners (GPs) act as gatekeepers to further investigation into patients’ medical issues, much is unknown about what factors cause them to suspect serious disease and to what extent such suspicions correspond with final diagnoses, especially in the case of cancer. The cohort study conducted by Hjertholm et al attempts to answer these questions, while also examining what actions are taken by GPs when serious disease is suspected.


GPs in a region of Denmark were invited to participate in a 1-day registration of data concerning consulting patients, including a final question, “Are you left with the slightest suspicion of cancer or another serious disease?” Of the invited GPs, 404 (46%) participated, with some over-representation of female and younger GPs. After exclusions, 4518 patients participated, of whom 4% lacked an answer to the question about suspicion; these were included as ‘no suspicion’ patients. A unique personal identification number was linked to two national health registers as well as to Statistics Denmark, with sociodemographic data. General linear models were used to analyse associations between patient characteristics and suspicion of serious disease, as well as the use of healthcare services in secondary and primary care. Cox proportional hazard models were used to calculate the risk of new serious disease occurring within 2 and 6 months after the consultation. Adjustments were made for various patient characteristics.


GPs suspected serious disease in 256 (5.7%) patients. Suspicion increased with higher age and when symptoms were related to the digestive system, blood or blood-forming organs and female genital organs. In 42 patients GP suspicions were confirmed, with cancer revealed in eight patients. Two hundred and seventy-nine patients had a subsequent serious diagnosis not suspected after the consultation; the authors found that for 22% of these patients diagnosis was related to the reason for initial primary care encounter. The rate of a serious diagnosis within 2 months was threefold among patients suspected of serious disease relative to those not suspected (HR=2.98 (95% CI 1.93 to 4.62)). Positive predictive value of GPs’ suspicion was 9.8% for a serious diagnosis within 2 months and 16.4% within 6 months. Referrals and overall use of healthcare services increased when there was suspicion.


This study highlights the complexity of diagnostic thinking regarding cancer and other serious disease. The authors avoided a Hawthorne effect commonly present in studies focusing on cancer, because the question was part of a larger survey not specifically focused on serious disease. However, the seemingly high number of patients produced only eight patients suspected to have cancer, which limits the precision of calculated parameters, at least in cancer sub-group analyses. It also remains vague how the authors defined and distinguished ‘serious’ versus ‘non-serious’ incident diagnoses.

The number of patients with a subsequent cancer diagnosis who escaped suspicion was 50–60. This suggests that physician suspicion, while predictive, has low sensitivity. This leaves us with many unanswered questions. We have relatively good data about the importance of alarm symptoms and we know that ‘low-risk but not no-risk’ symptoms play a lesser but not negligible role. But we also know little about the importance of what the GP learns during the clinical examination or from the results of simple tests and imaging.

We do know that it matters that GPs use simple tools (eg, medical history, clinical examination, supplementary tests) well, while we also know that GP's suspicions are accurate to the extent that they should be taken seriously when patients are referred.1 More qualitative data may increase our understanding of what the numbers fail to explain; interpersonal awareness (ie, being alert to changes in the patient's appearance or behaviour) is important.2 Cognitive factors are also important, as is the current interest in physicians’ gut feelings.3–5

The possibility of linking with high-quality national registers proved its value in this study. High numbers of patients will be needed in similar studies, but although individual written consent was not required in Denmark, it may be required in other settings, providing researchers with an administrative hurdle to overcome. Hospitals commonly inform their patients that anonymised patient data may be used for research purposes, asking them to opt out if they do not want their data used. Could this be a possibility for primary care practices in research networks?


View Abstract


  • Competing interests None.

  • Provenance and peer review Commissioned; internally peer reviewed.