Do patients' faces influence General Practitioners' cancer suspicions? A test of automatic processing of sociodemographic information

Abstract

Background: Delayed cancer diagnosis leads to poorer patient outcomes. During short consultations, General Practitioners (GPs) make quick decisions about likelihood of cancer. Patients' facial cues are processed rapidly and may influence diagnosis.

Aim: To investigate whether patients' facial characteristics influence immediate perception of cancer risk by GPs.

Design and setting: Web-based binary forced choice experiment with GPs from Northeast Scotland.

Method: GPs were presented with a series of pairs of face prototypes and asked to quickly select the patient more likely to have cancer. Faces were modified with respect to age, gender, and ethnicity. Choices were analysed using Chi-squared goodness-of-fit statistics with Bonferroni corrections.

Results: Eighty-two GPs participated. GPs were significantly more likely to suspect cancer in older patients. Gender influenced GP cancer suspicion, but this was modified by age: the male face was chosen as more likely to have cancer than the female face for young (72% of GPs;95% CI 61.0-87.0) and middle-aged faces (65.9%; 95% CI 54.7-75.5); but 63.4% (95% CI 52.2-73.3) decided the older female was more likely to have cancer than the older male (p = 0.015). GPs were significantly more likely to suspect cancer in the young Caucasian male (65.9% (95% CI 54.7, 75.5)) compared to the young Asian male (p = 0.004).

Conclusion: GPs' first impressions about cancer risk are influenced by patient age, gender, and ethnicity. Tackling GP cognitive biases could be a promising way of reducing cancer diagnostic delays, particularly for younger patients.

Fig 1. Example of choice set presented to GPs.

In this example the comparison is between a middle-aged and a young woman, faces differ in age.

Fig 2. The influence of gender and age on GP choice of face more likely to have cancer.

Fig 3. The percentages of GPs choosing age, gender, and ethnicity categories.

In the ethnicity comparison (Fig 3), 65.9% (95% CI 54.7–75.5) of GPs decided the young Caucasian male was more likely to have cancer than the young South Asian male (p = 0.004). There was no significant difference in the proportion of GPs choosing between the young Caucasian female and the young South Asian female (54.9% (95% CI 43.8–65.5) chose the young Caucasian female (p = 0.377)).