Short answer:
One cannot determine whether a reference frame is inertial. It is simply not possible.
This situation is avoided in General Relativity (or any of its extensions), since in GR the physics does not depend on the reference frame (any reference frame), and the concept of inertial frame is not necessary.
Of course one can determine an inertial frame which is good enough for a given system, but one cannot determine, for example, an inertial frame to describe the motion of galaxies in the whole universe.
Long answer:
As pointed out by Andrea Di Biagio, one can reason by exclusion. If a particle moves on a straight line and no force is applied, the reference frame is inertial. Since there is a finite number of known forces, one can in principle rule out each of them and reduce to a situation where no force is applied to the particle. There are at least two weak points in this reasoning:
1) The list of known forces in modern physics is: gravitational, electromagnetic, weak and strong interactions. No one knows if this list is complete of course, therefore in principle reasoning by exclusion cannot work.
2) In classical mechanics, two distant bodies can interact over a long distance through the electromagnetic or gravitational field. Therefore to exclude, say, electromagnetic and gravitational forces, one should in principle know the charge and mass distribution of the entire universe.
Consider the following situation. In a hypothetical laboratory in deep space, far away from galaxies and other visible mass densities, a neutron is observed to travel in a certain reference frame. Is this reference inertial? What if a very massive structure is placed just beyond the portion of universe observable by the laboratory? Will this mass distribution exert a force on the neutron in the laboratory?
As one can see, the concept of inertial frame is very problematic, and the paradoxes which arise from this concept are not curable in the framework of classical (Newton) mechanics.