A time-adaptive reduced order model (ROM) is developed for the electrochemical model for lithium-ion cells derived by Doyle, Fuller, and Newman (DFN) [M. Doyle, T. F. Fuller and J. Newman, J. Electrochem. Soc., 140 1526 (1993)]. The main advantage of a time-adaptive strategy is that it does not require a set of full order model simulations to be generated beforehand and, thus, it is the most cost-effective alternative when no databases are available. However, the reduction of this electrochemical problem exhibits special features that require ad hoc solutions, preventing the application of generic strategies. This complexity is carefully analysed, focusing on mode selection, treatment of non-linearities and error estimation. Despite of all this analysis being done for a pseudo-two-dimensional DFN model, we show that such complexity is intrinsic to the physics of the electrochemical problem, making the analysis applicable to a pseudo-four-dimensional DFN model, where results prove that the benefits of a reduction in the number of degrees of freedom are more self-evident. The efficiency, robustness and accuracy of our method are remarkable, as shown by the macroscopic (cell voltage) and internal (variable distributions) results obtained from the simulation of two different electrochemical cells under several charge/discharge C-rates.