Fig 1.
Flowchart of patients in the derivation and validation cohort with an overview of the reasons for and number of excluded patients and number of implausible and missing values.
BMI, body mass index; CRC, colorectal cancer; EQUIP-3, Endoscopic Quality Improvement Program-3 study; IBD, inflammatory bowel disease; N, number of patients.
Table 1.
Provider and patient characteristics for the derivation and validation cohort and the adenoma detection rate per patient subgroup.
Table 2.
Univariable odds ratios of possible risk factors for the detection of ≥1 adenoma within the derivation and validation cohort.
Table 3.
Prediction modela for the detection of ≥1 adenoma per patient based on multivariable logistic regression within the derivation cohort.
Fig 2.
Calibration plot of the model performance A) within the derivation cohort after internal validation (figure at the left), and B) within the validation cohort after logistic re-calibration (figure at the right).
Grouped observations are grouped per decile of patients. The observed proportion of patients with ≥1 adenoma detected is displayed on the y-axis, and the predicted proportion of patients with ≥1 adenoma on the x-axis. C-statistic, concordance statistic.
Fig 3.
Observed adenoma detection rates (including 95% Wilson confidence intervals) versus predicted proportions of patients with ≥1 adenoma per physician within the derivation cohort.
The physicians are ranked in ascending order of adenoma detection rate. The numbers on the x-axis denote the number of patients per physician.
Fig 4.
A. Observed adenoma detection rates (including 95% Wilson confidence intervals) versus predicted proportions of patients with ≥1 adenoma per center within A) the derivation cohort (figure at the left) and B) the validation cohort (figure at the right).
The centers are ranked in ascending order of adenoma detection rate. The numbers on the x-axis denote the number of patients per center.
Fig 5.
Observed adenoma detection rates (including 95% Wilson confidence intervals) versus predicted proportions of patients with ≥1 adenoma per physician within the validation cohort.
The physicians are ranked in ascending order of adenoma detection rate. The numbers on the x-axis denote the number of patients per physician.