Any "atomic state" except the ground state have a finite width alone from the spontaneous emission contribution of QED. What you calculate as "atomic states" is an approximation, where you take into account the static contributions of electromagnetic interactions in the Coulomb gauge only. That's not the whole truth, because the electromagnetic field is a dynamical entity too and has to be quantized. Since the em. field as a spin-1 field is necessarily a boson field, as follows from the fundamental assumptions of local QFT demanding a stable ground state, there's also spontaneous emission in addition to induced emission and absorption which also is described in semi-classical electromagnetics, where the em. field is kept classical.
The possibility of spontaneous emission, i.e., the transition of an excited "atomic state" to a lower one under emission of one (or sometimes more than one) photon(s) has a finite probability (except if selection rules, i.e., conservation laws forbid some specific transition), the "atomic states" are not true energy eigenstates anymore as soon as you take into account the "radiation corrections" with a quantized em. field. This leads to the conclusion that in reality these "atomic states" are (quite narrow) resonances of finite width and the corresponding finite lifetime. It's also called "natural linewidth".