General practitioners’ anticipated risk of cancer at referral and their attitude to risk taking and to their role as gatekeeper

Abstract

Objectives

General practitioners have to deal with situations characterized by real but low likelihoods of serious illness. We aimed to investigate variations in general practitioners’ anticipated risk of cancer when referring a patient and associations both with general practitioners’ attitudes to risk taking and with their gatekeeper role.

Methods

In January 2012, all 835 active general practitioners in the county of Aarhus, Denmark, received a questionnaire including the Physician Reaction to Uncertainty scale, The Tolerance for Ambiguity scale, the Physician Risk Attitude scale and a number of single items assessing anticipated risk of cancer when referring a hypothetical 50-year-old patient, use of intuition and perception of their role as a gatekeeper.

Results

A total of 568 (68.0%) practitioners completed and returned the questionnaire. The median anticipated risk of cancer was 30% (inter-quartile range: 15%–50%) and the 5%–95% centiles were 5% and 80%. Increasing tolerance for ambiguity was strongly related to a declining anticipated risk of cancer. None of the other risk attitudes were associated with the anticipated risk of cancer at referral. Increased general practitioners’ age was related to increased anticipated risk of cancer when referring (25% for general practitioners under 45 years to 43% for those 60 years or over) but not statistically important.

Conclusions

General practitioners either overestimate the risk of cancer when referring for suspected cancer or they appear to need to be very sure of the cancer diagnosis before referral. Further, focus on tolerance for ambiguity should be included in the education of general practitioners.

Keywords

anticipated risk of cancer general practice risk taking attitude

Introduction

In cancer diagnosis, general practice forms the important interface between the population and the specialized parts of the health care system.¹ Patients present illness worries and symptoms to the general practitioner (GP) who then must interpret the patient history, combining it with clinical examinations and information from investigations and decide whether to refer the patient for specialist attention or ‘wait and see’.

GPs are specialized in symptom interpretation and diagnosing serious diseases in a low prevalence environment.² The organization of the health care system may affect a GP’s willingness or ability to refer patients for additional investigation.³ Their clinical knowledge, skills, beliefs and attitudes combined with the organization of the health care system determines the threshold for referring patients. Therefore, it is important to understand GPs’ anticipated risk of cancer when referring a patient.^4,5

Referral based on a low anticipated cancer risk may facilitate earlier diagnosis and a better patient evaluation but will lead to many referred patients with a low (3%–8%) positive predictive value (PPV).^4–6 However, a high anticipated cancer risk may mean that patients are referred only when the GP is quite sure that it could be cancer, which may then produce late diagnosis and a poor cancer prognosis.² Thus, a GP’s anticipated risk of cancer when referring to an urgent referral pathway for cancer can be regarded as a combination of a threshold for referring as a gatekeeper and a belief or knowledge about PPVs in general practice.

We aimed to investigate GPs’ anticipated risk of cancer when referring a hypothetical 50-year-old patient for investigation and relate the level of anticipated risk both to the GPs’ risk taking attitudes and to their perception of their role as gatekeeper to specialized care.

Methods

Setting

GPs in Denmark are independent contractors with the regional health authorities. Approximately 40% are single-handed. The patient list size is on average 1550 patients per GP and 99% of citizens are registered with a particular general practice, which they have to consult for medical advice. In 2008, Danish GPs got the possibility to refer patients suspected to have cancer based on specific symptoms to one of 32 fast-track pathways.⁷

Study population

In January 2012, all 835 active GPs in the Central Denmark Region, were invited to participate in a survey on job satisfaction, burnout and working conditions. Active GPs were identified by the Registry of Health Providers, which is managed by the National Board of Health. Non-respondents were sent a reminder after 4 and 13 weeks and GPs were paid 50€ for responding.

Risk attitude scales included in the questionnaire

Three scales were included, all scored on a six-point Likert scale from 1 (strongly disagree) to 6 (strongly agree).

The Physician Reaction to Uncertainty scale^8,9 consists of four subscales: anxiety due to uncertainty (five items), concern about bad outcomes (three items), reluctance to disclose mistakes to physicians (two items) and reluctance to disclose uncertainty to patients (five items). Higher scores on each of the subscales reflect more psychological discomfort in dealing with uncertainty.

The Tolerance for Ambiguity scale¹⁰ includes seven items and higher scores indicate more tolerance of ambiguity.

The Physician Risk Attitude scale is a modified version of the Jackson Personality Inventory¹¹ consisting of six items and higher scores reflect increased risk-seeking behaviour.

All risk attitude scales were translated into Danish using the forward and backward translation procedure suggested by WHO.¹²

Single items included in the questionnaire

Single items were developed by the research team based on previously used items. The items were exposed to semi-structured cognitive interviews with six GPs and tested with 30 GPs from the North Denmark Region. They included

use of intuition: ‘To what degree do you use intuition in your everyday work as a GP, e.g. in relation to referral for investigation, sampling etc.?’

role as gatekeeper: ‘The most important role as gatekeeper is to prevent overuse of secondary health care services’ and ‘The most important role as gatekeeper is to ensure proper medical guidance and referral to secondary health care services.’

anticipated cancer risk at referral: ‘In your judgment, what is the probability that a 50-year-old patient has cancer when you choose to refer the patient to fast track diagnostic services?’

The last question focused on a 50-year-old patient as the incidence of cancer is low (0.5%)¹³ but is important as the risk starts increasing from this age.

Analyses

The anticipated risk of cancer when referring a hypothetical 50-year-old patient for suspected cancer was reported as a percentage between 1 and 100. The reported percentages were divided into five groups: 1%–14%, 15%–24%, 25%–49%, 50%–74% and 75%–100%. The association between anticipated risk of cancer and independent variables was analysed with unadjusted and adjusted linear regression models controlling for age and sex of the GP. The adjusted analyses were corrected for clusters of GPs working in the same practice using robust variance estimates. All risk attitude scales were divided into four bands based on quartiles to allow for non-linear relationships. However, reluctance to disclose mistakes to other physicians was stratified into three bands only, as the 50% and 75% percentiles were identical. The 95% confidence intervals (CI) for coefficients were calculated and p values of 5% or less were considered statistically significant. Data were analysed using STATA 13.

Results

Of the 835 eligible GPs, 570 (68.3%) returned completed questionnaires. Two respondents who reported anticipated cancer risks of 0% and 130% were excluded since we assumed that they had misunderstood the question. Thus, 568 (68.0%) were included in the analyses.

The median anticipated cancer risk was 30% (inter-quartile range 15%–50%) and the 5%–95% centiles were 5% and 80%, respectively. The distribution of GPs in the five categories of anticipated cancer risk is shown in Table 1. Those aged 60 years and older reported a higher (though not a statistically significant difference) anticipated risk of cancer compared to GPs younger than 45 years (median 43% vs. 25%; p = 0.055) (Table 2).

Table 1.

Definition of GPs’ anticipated risk of cancer in a 50-year-old patient referred to fast track cancer diagnostic services.

Categories of anticipated risk of cancer (%)	Characterization of anticipated risk of cancer	N (%)
1–14	Advisable	115 (20.3)
15–24	Fairly reasonable	99 (17.4)
25–49	High	117 (20.6)
50–74	Very high	160 (28.2)
75–100	Unrealistically high	77 (13.6)
Total		568 (100.0)

Table 2.

Median values and interquartile ranges (IQR) of anticipated risk of cancer.

Category	N (%)	Median (IQR)	χ^2,a	p
Sex
Female	267 (47.0)	30 (20–50)
Male	299 (52.6)	30 (15–50)	0.06	0.801
Age
<45 years	120 (21.1)	25 (15–50)	Ref
≥45 and <60 years	304 (53.5)	30 (15–50)	3.22	0.073
≥60 years	128 (22.5)	43 (20–60)	3.68	0.055
Anxiety due to uncertainty (quartiles)
Lowest	204 (35.9)	30 (15–50)	Ref
Second	113 (19.9)	30 (20–50)	0.34	0.557
Third	105 (18.5)	30 (20–50)	0.52	0.470
Highest	139 (24.5)	30 (20–50)	0.31	0.580
Concern about bad outcomes (quartiles)
Lowest	210 (37.0)	30 (15–50)	Ref
Second	125 (22.0)	30 (20–50)	0.02	0.898
Third	108 (19.0)	30 (13–50)	0.06	0.813
Highest	120 (21.1)	30 (20–50)	0.09	0.762
Reluctance to disclose mistakes to physicians (quartiles)
Lowest	162 (28.5)	30 (15–50)	Ref
Second	285 (50.2)	30 (15–50)	0.02	0.901
Third	Omitted
Highest	118 (20.8)	30 (20–56)	0.35	0.552
Reluctance to disclose uncertainty to patients (quartiles)
Lowest	171 (30.1)	30 (15–50)	Ref
Second	146 (25.7)	30 (15–60)	0.78	0.376
Third	125 (22.0)	30 (20–50)	0.55	0.460
Highest	121 (21.3)	30 (15–50)	0.05	0.820
Tolerance for ambiguity (quartiles)
Lowest	153 (26.9)	50 (20–60)	Ref
Second	154 (27.1)	30 (15–50)	1.88	0.170
Third	111 (19.5)	25 (15–50)	3.95	0.047
Highest	134 (23.6)	30 (15–50)	5.52	0.019
Physician Risk Attitude (quartiles)
Lowest	146 (25.7)	30 (20–60)	Ref
Second	148 (26.1)	30 (15–50)	0.35	0.556
Third	144 (25.4)	25 (10–50)	5.55	0.018
Highest	115 (20.3)	30 (20–50)	0.50	0.479
Use of gut feeling
No	267 (47.0)	30 (15–50)
Yes	300 (52.8)	30 (15–50)	0.01	0.929
The most important role as gatekeeper is to prevent overuse of secondary health care
No or neutral	370 (65.1)	30 (15–50)
Yes	191 (33.6)	30 (20–60)	0.35	0.553
The most important role as gatekeeper is to ensure proper medical guidance and referral
No or neutral	45 (7.9)	25 (15–50)
Yes	521 (91.7)	30 (15–50)	1.99	0.158

Note: Missing data varied from 1 (0.2%) to 16 (2.8%) by variable.

K-sample test.

An increasing score on tolerance for ambiguity was statistically significantly associated with lower levels of anticipated cancer risk at referral (Table 3). GPs who scored in the lowest quartile of tolerance for ambiguity reported a median anticipated risk of cancer on 50%, whereas GPs who scored in the third and highest quartiles of tolerance for ambiguity reported a median risk on 25% and 30%, respectively.

Table 3.

Linear regression analyses (adjusted for sex and age of GP and for GPs working in the same practice).

Category	Coefficient	95% CI	p
Anxiety due to uncertainty (quartiles)
Lowest	Ref
Second	0.10	−0.22 to 0.49	0.245
Third	0.02	−0.34 to 0.30	0.898
Highest	0.12	−0.18 to 0.43	0.424
Concern about bad outcomes (quartiles)
Lowest	Ref
Second	0.18	−0.12 to 0.48	0.236
Third	−0.02	−0.34 to 0.30	0.898
Highest	0.09	−0.24 to 0.43	0.590
Reluctance to disclose mistakes to physicians (quartiles)
Lowest	Ref
Second	−0.03	−0.31 to 0.25	0.821
Third	Omitted
Highest	0.18	−0.16 to 0.52	0.306
Reluctance to disclose uncertainty to patients (quartiles)
Lowest	Ref
Second	0.13	−0.18 to 0.44	0.419
Third	0.23	−0.07 to 0.54	0.130
Highest	0.08	−0.25 to 0.41	0.631
Tolerance for ambiguity (quartiles)
Lowest	Ref
Second	−0.27	−0.59 to 0.04	0.089
Third	−0.32	−0.66 to 0.01	0.057
Highest	−0.38	−0.70 to −0.06	0.022
Physician risk attitude (quartiles)
Lowest	Ref
Second	−0.13	−0.44 to 0.19	0.431
Third	−0.37	−0.69 to −0.06	0.021
Highest	−0.07	−0.38 to 0.24	0.648
Use of intuition
No	Ref
Yes	0.07	−0.16 to 0.30	0.545
The most important role as gatekeeper is to prevent overuse of secondary health care services
No or neutral	Ref
Yes	0.14	−0.11 to 0.38	0.272
The most important role as gatekeeper is to ensure proper medical guidance and referral
No or neutral	Ref
Yes	0.23	−0.15 to 0.62	0.232

Note: Dependent variable was anticipated cancer risk at referral.

GPs who scored in the third quartile of propensity to risk taking reported a lower anticipated risk of cancer at referral compared to GPs in the lowest quartile (adjusted coefficient −0.37, 95% CI −0.69 to −0.06). The median anticipated risk of cancer for GPs in the third quartile of propensity to risk taking was 25% compared to 30% in GPs who scored in the lowest quartile.

None of the other risk attitude scales or the perceived role as gatekeeper was associated with the anticipated cancer risk.

Discussion

Main findings

This study revealed great diversity in anticipated cancer risk when referring a hypothetical 50-year-old patient. The 5%–95% centiles of anticipated cancer risk were 5% and 80%, respectively, a 16-fold variation. More than a quarter of GPs reported 50%–74% risk of cancer and nearly 14% reported an even higher risk, a risk level which the authors consider unrealistic. Increasing age was associated with an increased anticipated cancer risk though did not reach statistical significance.

The higher the tolerance for ambiguity, the lower the anticipated risk of cancer. GPs with the highest levels of tolerance for ambiguity were approximately half as sure that the referred patient had cancer as were GPs with the lowest level of tolerance for ambiguity. This indicates that being able to accept uncertainty also may affect the GP’s limit for referral. GPs’ perception of their role as a gatekeeper was not associated with anticipated risk of cancer.

Comparison with existing literature

The considerable diversity in anticipated cancer risk observed fits in with literature showing variation in levels of prescriptions of medicine, ordering of investigations and requests for follow up,¹⁴ with GP practice style being stable over time.¹⁵ One study in cancer referral rate by general practice found a six-fold variation, after excluding the highest and lowest 10% of practices⁶ which suggested significant differences in referral thresholds.

There is no accepted referral rate for suspected cancer.⁶ In this study, anticipated risks of cancer from 1%–100% were reported, and insofar as anticipated risks of cancer correlate with actual referral rates, both ends of the spectrum will probably reflect poor quality care or gaps in the knowledge about PPVs of alarm symptoms of cancer in general practice. The considerable variation in anticipated risk of cancer at referral, taken together with the observed variation in clinical activity, points to the need of evidence-based decision aids and encouragement of GPs to consider whether they tolerate too little or too much risk.²

On the one hand, it may appear counter-intuitive that increased tolerance for ambiguity was associated with a lowered anticipated cancer risk when referring patients. However, this could reflect that GPs with high tolerance for ambiguity feel less discomfort with uncertainty as they disclosed a low PPV of their cancer suspicion. Most cancer alarm symptoms have low PPV (<5%)¹⁶ and even though a definite level of PPV for triggering referral has not been established, it has been suggested that symptoms or symptom patterns with PPV levels around 1% should prompt investigation.¹⁷ With this in mind, a self-assessed risk of cancer above 25% when referring to diagnostic services as reported by the majority of GPs in this study could reflect either acceptance of too much risk (high risk of sending home patients having a high risk of cancer) and/or a lack of knowledge concerning the low PPVs of most cancer alarm symptoms.

The revealed tendency that older GPs report higher anticipated cancer risk than younger may be explained by their many years of experience which may have given them a feeling of high levels of success when it comes to their ability to differentiate serious symptoms from trivialities. However, a number of studies have not documented a relationship between the age of GPs and their referral rates.^18–20 It could also be that older GPs were educated at a time when making a definite diagnosis was deemed to be high-quality care. In contrast, today’s education in family medicine might give younger GPs a more realistic impression of the diagnostic process of cancer in general practice. This indeed stresses the importance of continuing medical education on cancer diagnosis including working with ambiguity.

In one study of US primary care doctors, increased scores on the subscales ‘Anxiety due to uncertainty’ and ‘Reluctance to disclose uncertainty to patients’ were found to be associated with increased patient charges.²¹ Differences in levels of risk attitudes appear not to be the explanation of the incongruent findings as the GPs in our study scored slightly lower on anxiety due to uncertainty than the US doctors (mean scores: 13.7 and 14.6, respectively), but higher on reluctance to disclose uncertainty to patients (mean scores: 14.9 and 10.7, respectively). Perhaps cultural differences, such as an increased risk of litigation in the United States, can explain why these subscales were associated with the outcome in the US study but not in our study. Moreover, the cited US study focused on actual GP behaviour whereas the present study focused on anticipated risk. This may also explain the incongruent findings.

GPs who scored in the third quartile of propensity to risk taking reported a lower anticipated cancer risk at referral compared to GPs in the lowest quartile. However, there was no trend as GPs with the highest propensity of risk taking did not differ from those GPs with the lowest propensity of risk taking concerning anticipated cancer risk. Therefore, we consider the finding to be incidental.

Strengths and weaknesses of the study

The high response rate and use of validated scales for assessment of attitude to risk are among the strengths of the present study. The risk attitude scales were translated for the purpose of the present study and pilot-tested in a small group of Danish GPs, but the limited use of the scales outside North America introduces the risk of possible cultural biases. Since no validated scale for assessing referral threshold exists, we had to develop a question for capturing this phenomenon for the purpose of this study. The validity of the developed item has only been investigated by means of cognitive interviews with six GPs, and the lack of formal testing of validity and reliability of the item is a weakness of the study. As referral threshold was assessed by a single item, the risk of poor precision cannot be excluded. Lastly, since actual referral rates for the individual GP are not registered in Denmark, the association between this and anticipated risk when referring a hypothetical patient could not be determined. Examination of the relationship between risk tolerance, anticipated risk and actual urgent referral rate would be relevant in future studies.

Conclusions

There was considerable variation in anticipated risk of cancer when referring a hypothetical 50-year-old patient. Increasing tolerance for ambiguity was related to a declining anticipated cancer risk. GPs have to deal with situations in which there is a real but low likelihood of serious illness, and the results of the present study point to the need for less variation and more realistic levels of anticipated cancer risk when referring patients suspected for cancer.

Footnotes

Acknowledgements

We would like to thank Christina Maar Andersen for her help during data collection.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

Centre for Cancer Diagnosis in Primary care (CaP) is funded by the Novo Nordisk Foundation and the Danish Cancer Society. The project was further supported financially by the Committee of Quality and Supplementary Training (KEU) in Central Denmark Region.

Research ethics

The study is conducted in accordance with ethical guidelines. According to Danish law, questionnaire surveys do not require an ethical approval.

References

Hamilton

. Cancer diagnosis in primary care. Br J Gen Pract 2010; 60: 121–128.

Buntinx

Mant

Van den Bruel

. Dealing with low-incidence serious diseases in general practice. Br J Gen Pract 2011; 61: 43–46.

Vedsted

Olesen

. Are the serious problems in cancer survival partly rooted in gatekeeper principles? An ecologic study. Br J Gen Pract 2011; 61: e508–e512.

Hamilton

Lancashire

Sharp

. The risk of colorectal cancer with symptoms at different ages and between the sexes: a case-control study. BMC Med 2009; 7: 17–17.

Jones

Latinovic

Charlton

. Alarm symptoms in early diagnosis of cancer in primary care: cohort study using general practice research database. BMJ 2007; 334: 1040–1040.

Baughan

Keatings

O'Neill

. Urgent suspected cancer referrals from general practice: audit of compliance with guidelines and referral outcomes. Br J Gen Pract 2011; 61: e700–e706.

Probst

Hussain

Andersen

. Cancer patient pathways in Denmark as a joint effort between bureaucrats, health professionals and politicians – a national Danish project. Health Policy 2012; 105: 65–70.

Gerrity

DeVellis

Earp

. Physicians' reactions to uncertainty in patient care. A new measure and new insights. Med Care 1990; 28: 724–736.

Gerrity

White

Devellis

. Physicians' reactions to uncertainty: refining the constructs and scales. Motiv Emotion 1995; 19: 175–191.

10.

Geller

Tambor

Chase

. Measuring physicians' tolerance for ambiguity and its relationship to their reported practices regarding genetic testing. Med Care 1993; 31: 989–1001.

11.

Pearson

Goldman

Orav

. Triage decisions for emergency department patients with chest pain: do physicians' risk attitudes make the difference? J Gen Intern Med 1995; 10: 557–564.

12.

WHO. Process of translation and adaptation of instruments, www.who.int/substance_abuse/research_tools/translation/en/ (2015, accessed 4 August 2015).

13.

Engholm

Ferlay

Christensen

. NORDCAN – a Nordic tool for cancer information, planning, quality control and research. Acta Oncol 2010; 49: 725–736.

14.

Davis

Gribben

Lay-Yee

. How much variation in clinical activity is there between general practitioners? A multi-level analysis of decision-making in primary care. J Health Serv Res Policy 2002; 7: 202–208.

15.

Davis

Gribben

Scott

. Do physician practice styles persist over time? Continuities in patterns of clinical decision-making among general practitioners. J Health Serv Res Policy 2000; 5: 200–207.

16.

Shapley

Mansell

Jordan

. Positive predictive values of ≥5% in primary care for cancer: systematic review. Br J Gen Pract 2010; 60: e366–e377.

17.

Astin

Griffin

Neal

. The diagnostic value of symptoms for colorectal cancer in primary care: a systematic review. Br J Gen Pract 2011; 61: e231–e243.

18.

Verstappen

ter Riet

Dubois

. Variation in test ordering behaviour of GPs: professional or context-related factors? Fam Pract 2004; 21: 387–395.

19.

Verstappen

ter Riet

van der Weijden

. Variation in requests for imaging investigations by general practitioners: a multilevel analysis. J Health Serv Res Policy 2005; 10: 25–30.

20.

Kushnir

Greenberg

Madjar

. Is burnout associated with referral rates among primary care physicians in community clinics? Fam Pract 2014; 31: 44–50.

21.

Allison

Kiefe

Cook

. The association of physician attitudes about uncertainty and risk taking with resource use in a medicare HMO. Med Decis Making 1998; 18: 320–329.