"Repeatability of Repeatability": the stability of self-reported melanoma risk factors in two independent samples

To determine the test-retest repeatability of a self-completed survey with items capturing skin cancer risk factors.

We invited 238 randomly selected participants of the QSkin II cohort to complete the baseline survey a second time. Responses were compared using kappa statistics and intraclass correlation coefficients to quantify agreement for categorical and continuous variables, respectively. We compared the performance of key items with that observed in an earlier repeatability study using the same survey instrument in an independent cohort.

Measures of phenotypic characteristics had moderate to almost-perfect test-retest repeatability (e.g. eye colour weighted kappa (κ_w ) = 0.87, 95% confidence interval [CI]: 0.81, 0.92). Items measuring sun exposure showed lower agreement (κ_w range 0.36-0.54) compared with phenotypic characteristics (κ_w range 0.59-0.87). Items relating to treatment of skin cancers demonstrated almost-perfect test-retest repeatability (e.g. excisions for skin cancers κ_w 0.85, 95%CI: 0.80, 0.89). In aggregate, the repeatability of key items was very similar across the two independent repeatability samples.

Fair to almost-perfect repeatability for self-reported skin cancer risk factors was robust across independent and temporally distant cohorts. Implications for public health: These self-assessed risk factors for skin cancer are repeatable and suitable for use in clinical practice and research.