To compute the SFR of galaxies from the rest-frame UV it is essential to take into account the obscuration by dust. To do so, one of the most popular methods consists in combining the UV with the emission from the dust itself in the IR. Yet, different studies have derived different estimators, showing that no such hybrid estimator is truly universal. In this paper we aim at understanding and quantifying what physical processes drive the variations between different hybrid estimators. Doing so, we aim at deriving new universal UV+IR hybrid estimators to correct the UV for dust attenuation, taking into account the intrinsic physical properties of galaxies. We use the CIGALE code to model the spatially-resolved FUV to FIR SED of eight nearby star-forming galaxies drawn from the KINGFISH sample. This allows us to determine their local physical properties, and in particular their UV attenuation, average SFR, average specific SFR (sSFR), and their stellar mass. We then examine how hybrid estimators depend on said properties. We find that hybrid UV+IR estimators strongly depend on the stellar mass surface density (in particular at 70 and 100 micron) and on the sSFR (in particular at 24 micron and the TIR). Consequently, the IR scaling coefficients for UV obscuration can vary by almost an order of magnitude. This result contrasts with other groups who found relatively constant coefficients with small deviations. We exploit these variations to construct a new class of hybrid estimators based on observed UV to near-IR colours and near-IR luminosity densities per unit area. We find that they can reliably be extended to entire galaxies. The new estimators provide better estimates of attenuation-corrected UV emission than classical hybrid estimators. Naturally taking into account the variable impact of dust heated by old stellar populations, they constitute a step towards universal estimators.