The temperature dependence of the spin susceptibility \ensuremath{\chi}(T) is studied in a weakly disordered itinerant-fermion system close to a magnetic instability. The paramagnon model is used with a Hubbard-type contact repulsion among opposite spins, with the Stoner enhancement of the pure system (1-\ifmmode \bar{I}\else \={I}\fi{}${)}^{\mathrm{\ensuremath{-}}1}$\ensuremath{\gg}1. The result is shown to be different from the usual one obtained in the case of a screened Coulomb interaction: The 2-phDP (two--particle-hole-diffusion-propagator) diagrams, which cancel altogether in the Coulomb interaction case, are shown to give, here, for the contact interaction, a finite contribution which is of the same order in (${\ensuremath{\varepsilon}}_{F}$\ensuremath{\tau}${)}^{\mathrm{\ensuremath{-}}1}$ as the 3-phDP and 4-phDP diagrams where \ensuremath{\tau} is the lifetime due to disorder, and ${\ensuremath{\varepsilon}}_{F}$ the Fermi energy. Instead of a unique temperature range T\ensuremath{\tau}\ensuremath{\ll}1 in the Coulomb case, here one has to distinguish two ranges:When T\ensuremath{\tau}\ensuremath{\ll}(1-\ifmmode \bar{I}\else \={I}\fi{}), the usual 3- and 4-phDP diagrams dominate and one recovers \ensuremath{\chi}(T)\ensuremath{\approxeq} \ensuremath{\surd}T\ensuremath{\tau} /(1-\ifmmode \bar{I}\else \={I}\fi{}${)}^{2}$ as was first announced by Al'tshuler and Aronov; but when (1-\ifmmode \bar{I}\else \={I}\fi{})\ensuremath{\ll}1, the 2-phDP diagrams dominate, yielding \ensuremath{\chi}(T)\ensuremath{\sim}(T\ensuremath{\tau}${)}^{3/2}$/(1-\ifmmode \bar{I}\else \={I}\fi{}${)}^{5/2}$, which is non-negligible near the magnetic instability. At T=0, the 2-phDP diagrams definitely dominate as they enhance the effective interaction (I${\ifmmode\bar\else\textasciimacron\fi{}}_{\mathrm{eff}}$-\ifmmode \bar{I}\else \={I}\fi{})\ensuremath{\approxeq}(${\ensuremath{\varepsilon}}_{F}$\ensuremath{\tau}${)}^{\mathrm{\ensuremath{-}}2}$(1 -\ifmmode \bar{I}\else \={I}\fi{}${)}^{\mathrm{\ensuremath{-}}1/2}$ and let the system be closer to magnetism, while the 3- and 4-phDP diagrams play a minor role at 0 K. The present study accounts only for phDP processes.Effects due to ppDP's (particle-particle diffusion propagators) should also be studied within the same framework to incorporate the contributions of the 2-ppDP diagrams. Finally, the latest developments using renormalization-group analysis of Finkel'stein and of Castellani et al. with a screened Coulomb interaction ought to be modified to account for the contact-interaction case where spin constraints yield the noncancellation of the 2-phDP---and also most likely of the 2-ppDP---diagrams.