Encrypted login | home

Program Information

An Empirical Model for Predicting Qualified Medical Physicist Staffing Level From Equipment and Procedures

T Li

T Li1*, Y Sheng2 , M Mills3 , Y Yu4 , (1) Thomas Jefferson University, Philadelphia, PA, (2) Duke University, Durham, NC, (3) James Graham Brown Cancer Center, Louisville, KY, (4) Thomas Jefferson University, Philadelphia, PA


SU-I-GPD-P-3 (Sunday, July 30, 2017) 3:00 PM - 6:00 PM Room: Exhibit Hall

Purpose: To provide an empirical model that predicts staffing requirement for qualified medical physicist (QMP) for small to mid-sized clinics.

Methods: 116 institution’s clinic setup information and QMP staffing levels (FTEs) reported in the 2016 AAMD and 2014 Abt IV surveys were used in this study. Clinic setup information were broken down into 11 numeric factors including equipment and procedures. A stepwise linear regression method with 10-fold cross validation was then applied to 116 training data to build a preliminary model which links staffing level to clinic setup information. The model was later adjusted to balance prediction errors between clinics with low and high number of FTEs. The prediction accuracy of the final model was evaluated using survey data.

Results: Because 90% of survey responses had QMP ≤ 5 the scope of this prediction was focused on institutions with ≤ 5 QMPs. The model used only five parameters, and their relative FTE factors are: 0.75 FTE per LINAC, 0.55 FTE per unit of tomotherapy/cyberknife/gammaknife, 0.2 FTE per unit of HDR system/simulator/TPS/image registration system, 0.003 FTE per SRS/SBRT/TBI/TSE patient, and 0.001 FTE per brachytherapy patient.When validated using the training data, for clinics with QMP ≤ 5 average prediction error is 0.4±1.0 FTE. For institutions with QMP > 5 in the survey, the prediction accuracy decreased to -4.1±5.58.When rounded to integer level, the difference between predicted and actual staffing level is 0.4±1.0 FTE. Prediction agreed with actual staffing level in 53 (50%) out of the 106 institutions with QMP ≤ 5, and agreed within 1 FTE for 80 institutions (75%).

Conclusion: An empirical model was developed to predict QMP staffing levels with acceptable accuracy up to 5 QMPs.

Contact Email: