Accurately predicting aging of lithium-ion batteries would help to prolong their lifespan, but remains a challenge owing to the complexity and interrelation of different aging mechanisms. As a result, aging prediction often relies on empirical or data-driven approaches, which obtain their performance from analyzing large datasets. However, these datasets are expensive to generate and the models are agnostic of the underlying physics and thus difficult to extrapolate to new conditions. In this article, a physical model is used to predict capacity fade caused by solid-electrolyte interphase (SEI) growth in 62 automotive cells, aged with 28 different protocols. Three protocols parametrize the time, current and temperature dependence of the model, the state of charge dependence results from the anode's open circuit voltage curve. The model validation with the remaining 25 protocols shows a high predictivity with a root-mean squared error of 1.28%. A case study with the so-validated model shows that the operating window, i.e. maximum and minimum state of charge, has the largest impact on SEI growth, while the influence of the applied current is almost negligible. Thereby the presented model is a promising approach to better understand, quantify and predict aging of lithium-ion batteries.