Cell-free (CF) massive multiple-input multiple-output network and its low-complexity user-centric (UC) alternative rely on accurate channel knowledge for combining and precoding to achieve their claimed performance gains. When serving vehicular users with high mobility, the non-negligible channel aging effect will inevitably degrade system performance, and accurate performance evaluation is essential to system design. In this paper, we investigate the aging uplink (UL) spectral efficiency (SE) in the UC CF networks with non-isotropic scattering conditions. We adopt the von Mises distribution to model the angle-of-departure (AoD), resulting in an analytically tractable channel autocorrelation function that allows us to analyze the time-varying properties of the channel for an arbitrary AoD spread and the user’s moving direction. We derive a closed-form signal-to-interference-and-noise ratio expression with large-scale fading decoding (LSFD) for an achievable UL SE. The simulated results in a 3GPP-recommended vehicular network scenario show that there is an optimal subframe length to achieve the maximum area average SE. The LSFD cooperative strategy of the UC CF network significantly increases the optimal subframe length in the non-isotropic scattering environment, which reduces the pilot overhead and improves the spectrum utilization eventually.