Good test-retest reproducibility for an instrument to capture self-reported melanoma risk factors

Abstract

Objective: To examine the test-retest reliability and validity of self-reported items capturing phenotypic characteristics and sun exposure measures in the baseline survey instrument used for a prospective study of skin cancer and melanoma.

Study design and setting: Repeatability/validity study conducted among 114 participants randomly selected from the cohort to complete the survey instrument a second time and to undergo a physician skin examination. We calculated intraclass correlation coefficients (ICCs) and kappa (κ) statistics as measures of agreement for continuous and categorical measures, respectively.

Results: Measures of phenotypic characteristics showed moderate-to-high agreement (e.g., eye color, κ=0.87; 95% confidence interval [CI]: 0.80, 0.94). Measures of sun exposure had slightly lower estimates of agreement. The repeatability of items relating to medical and family history of skin cancer was high (e.g., the number of skin cancers removed surgically, κ(w)=0.79; 95% CI: 0.71, 0.88). Physician counts of nevi correlated well with categorical measures of self-reported nevus density at the age of 21 years but correlated only modestly with absolute nevus counts conducted by participants (ICC, 0.38; 95% CI: 0.19, 0.54).

Conclusion: Our survey instrument demonstrated fair-to-good test-retest reliability for most self-reported risk factors for melanoma, indicating the suitability of these items for developing risk prediction tools in the future.