“…Among the claims-only studies, there was very little reported use of either a previously published algorithm (n ¼ 8, 12.9%) [20,27,37,48,49,58,60,61] or a validated algorithm (n ¼ 4, 6.5%) [48,58,60,61]. Almost half (n ¼ 31, 48.4%) of claims-only studies relied solely on diagnosis codes to select their study sample [7,9,[12][13][14][15]19,20,25,[27][28][29]31,32,34,47,49,50,56,59,60,63,69,70], with 35.5% (n ¼ 11) of these studies using a single claim with a cancer diagnosis as a prerequisite for study entry [15,29,32,34,36,37,46,50,63,69,70]. Only 4.8% of the articles discussed the implications of their selection criteria on the study findings…”