Primary care clinicians' performance in the diagnosis of actinic keratoses and skin cancer

If skin cancer screening is to become widely adopted, increasing its accessibility would be best achieved by incorporating screening examinations into routine patient care by primary care clinicians (PCCs). Because PCCs' proficiency for detecting skin cancer has not been well described, we compared their ability to diagnose actinic keratoses (AKs) and malignancies using dermatologists' examinations as a pragmatic reference standard. One hundred-ninety patients presenting to a general internal medicine or dermatology clinic at a Veterans Affairs Medical Center were independently evaluated by one dermatologist and one PCC. The skin above the waistline was evaluated, and each examiner was asked to identify AKs and malignant skin lesions based on the single most likely diagnosis. Since counting individual AKs was prohibitive, AKs were described as a site of skin with a single lesion or two or more lesions and were, therefore, recorded as single or multiple AKs. A malignant lesion was defined as one requiring a biopsy to provide a definitive diagnosis. Interobserver agreement for malignancy recognition was moderate between the two specialties with the patient as the unit of analysis (kappa = 0.46, 95% CI - 0.32 - 0.61). With the lesion as the unit of analysis, the kappa statistic was 0.30 (95% CI = 0.23 - 0.36). The kappas for AKs were similar to malignancies when the patient was the unit of analysis, but decreased for individual lesion agreement. The PCCs' examinations for malignancies were highly specific with both the patient (specificity = 88%, 95% CI = 81% - 92%) and the lesion (specificity = 95%, 95% CI = 93% - 96%) as the unit of analysis. Correspondingly, the positive likelihood ratios were good (LR+ = 4.6,95% CI = 2.8 - 7.7 and LR+ = 6.8, 95% CI = 4.6 - 10.0, respectively) indicating that the PCCs were proficient at "ruling in" disease. The PCCs evaluations lacked sensitivity (sensitivity = 57%, 95% CI = 43% - 68% and sensitivity = 37%, 95% CI = 28% - 46%, respectively), and the negative likelihood ratios (LR- = 0.50, 95% CI = 0.36 0.64 and LR- = 0.67, 95% CI = 0.57 - 0.76, respectively) were not helpful for discriminating disease presence. Consequently, the PCCs examinations were not as predictive for "ruling out" the presence of malignancies. PCCs require improved diagnostic skills if their skin examinations are to be used effectively for skin cancer screening.

Duke Scholars

Author Russell P. Hall III Dermatology

Author David Lee Simel Medicine, General Internal Medicine