In our study, patients who take any symptoms less seriously than others have an almost 2-fold higher prevalence of cancer (OR 1.85, 95%CI: 1.29, 2.64) than those who take their symptoms with the same seriousness as others, and a more than a 3-fold higher prevalence of cancer (OR = 3.28, 95%CI: 2.02, 5.33) than those who take their symptoms more seriously than others. Even if adjustment is made for a wide range of other predictors of colorectal cancer, the effects are still almost 1.5 and 2-fold respectively. An increase in prevalence is apparent in all age groups, in males and females and also in people with, and without, symptoms. The assessment of how seriously people took their symptoms was based on a question which is reliable and easy to answer which only a very small percent of patients did not complete.
Strengths of our study are the use of a reliable question to assess how seriously people reported they took symptoms and testing predictive validity in a large high quality study of an important clinical disease. The question is asked in a general way and is not restricted to any particular set of symptoms or disease. High quality of the predictive validity study was assured by several features. The data were collected prospectively with the question completed before disease was identified, all patients underwent colonoscopy with a caecal intubation rate of 98% and all lesions were examined for pathology.
The weakness of our study is that we investigated only one condition – colorectal cancer – so we cannot be certain that the results would be as striking for other clinical outcomes. That can be readily assessed by researchers adding the question we used to their studies.
We are not aware of any previous tool that encapsulates how seriously patients take their symptom in a single question. It is possible that the question relates to stoicism, on which there is some research. Murray et al. showed that the Liverpool Stoicisim Scale has adequate reliability. However, this requires a 20-item questionnaire. They found that stoicism was associated with lower reported quality of life though they did not explore whether serious clinical diagnoses are more common in stoical people. More generally, they point out that there is very little research on stoicism . Miller has shown that individual differences in health–seeking behaviour and health status are influenced by whether people typically scan for threat-level information (high monitors) or ignore threat-relevant information (low monitors) . High monitors, who are assessed by their doctors as having less severe medical problems, complain about these as much as low monitors. However, measurement of whether people are high or low monitors is by completion of 32 responses to 4 imaginary, stress-invoking scenarios . These measures differ from ours in concept, in that they have been developed for use in people already labelled as having a disease and in their complexity.
We have shown that our simple single question can predict disease, prior to diagnosis, and irrespective of the presence of symptoms. Our single question does not measure or evaluate the underlying constructs and mechanisms that may explain why a person takes their symptoms as seriously as they do; the question effectively provides an assessment of how seriously a person takes their symptoms, and links this to patient outcomes. We are not aware of any other studies that have used a single question in this way.
Our results are based on a clinical population – all patients had been referred for specialist consultation and subsequently underwent colonoscopy, although the question was answered prior to this. The presence of bowel symptoms is a common reason for seeking medical attention . In our study 6 of the symptoms were reported as present by more than 30% of patients. Our cancer detection rate is 1.9%, suggesting it is not a high-risk referred population. Although the population is not highly selected, we think that our finding is not a feature of a general population, but rather reflects who seeks medical help.
One way of explaining our findings is by considering a hypothetical scenario loosely based on our results. Imagine that the colorectal cancer prevalence is about 3% in people going for colonoscopy who report taking symptoms less seriously than others; one can represent this as 3 cancers in 100 people. It seems likely that people who take their symptoms more seriously than others access health services for minor symptoms that do not reflect organic pathology. If they are also referred for colonoscopy, this might add an extra 100 people to the pool of patients investigated, without adding any more cancers, so that the cancer prevalence is now only 1.5%.
Another way of explaining our findings is that people who take their symptoms more seriously present for colonoscopy earlier or more frequently than those who take their symptoms less seriously, allowing polyps to be detected and removed preventing their progression to colorectal cancer.
One might expect the effect on outcomes of how seriously people take their symptoms to disappear once the effect of symptom perception is added to a model containing predictive bowel symptoms. The fact that the gradient of risk for our symptom perception question is diluted but still evident suggests that the effect must be partly mediated by factors outside those measured which are important in people’s health-seeking behaviour.