The aim of this study is to reduce the uncertainty associated with determining dose-to-water using an amorphous silicon electronic portal imaging detector (EPID) under reference conditions by identifying and accounting for operational and environmental factors influencing the long-term stability of EPID response. Methods: Measurements of the EPID relative response,corrected for variations in linear accelerator (linac) output, were performed regularly over a period of 12 months. For every acquired image set, measurements of detector supply voltages, internal operating temperature, and ambient environmental conditions were obtained. Pearson r correlation coefficients were then calculated for each pair of variables, a subset of which were fitted using multiple linear regression to develop a predictive model of EPID response. Principal component analysis was performed on the dataset to reveal the internal structure of the data in a way that best accounts for the observed variations. Results: The +5.5 V power supply voltage, internal operating temperature, and the accumulated dose absorbed in EPID were identified as having the greatest influence on the long-term stability of EPID response. By correcting for the combined effect of these variables, the mean difference in linac output as measured by the EPID relative to a reference class chamber improved from -0.46% to 0.23% over the period of the study. Conclusions: This work suggests that the stability of an EPID over a period of a year can be improved by a factor of two by monitoring and accounting for the effects of variations in power supply voltage, internal temperature of the detector, and accumulated absorbed dose.