This study investigated radiation exposures from nuclear medicine patients by systematically comparing the dose rates calculated using various source models, ranging from simplified point, line, and cylinder sources to high-quality anthropomorphic phantoms. Three widely used radionuclides, (99m)Tc, (18)F, and I(131), were considered in these source models with uniform or organ-dependent distributions. Conducting Monte Carlo simulations with anthropomorphic phantoms is a realistic but time-consuming approach. The point source model is simple but too conservative, overestimating dose rates by approximately a factor of 2 at a distance of 30 cm and by 30-40% at 1 m. Both the line and cylinder source models provided improved estimates, reducing the overestimation of dose rates to 10-20% at distances of interest. The line source model was comparable to the cylinder source model because of the offset of two competing effects (i.e., attenuation and buildup) caused by the source volume. The influence of various photon energies and cylinder sizes on the result of compensating errors was examined to evaluate the effective range of the line source model. The line source model, which is relatively easy to implement and predicts slightly conservative dose rates, is considered the most practical method for calculating dose rates near radioactive patients. An application of the line source model to 51 post-thyroidectomy patients in Taiwan was demonstrated. The consistency between calculations and measurements was satisfactory after considering the room-scattering effect.