Background: Information about the Li6(d,α)He4 reaction rates of astrophysical interest can be obtained by extrapolating direct data to lower energies, or by indirect methods. The indirect Trojan horse method, as well as various R-matrix and polynomial fits to direct data, estimate electron screening energies much larger than the adiabatic limit. Calculations that include the subthreshold resonance estimate smaller screening energies. Purpose: Obtain the Li6(d,α)4He reaction R-matrix parameters and the bare astrophysical S factor for energies relevant to the stellar plasmas by fitting R-matrix formulas for the subthreshold resonances to the S-factor data above 60 keV. Methods: The bare S factor is calculated using the single- and the two-level R-matrix formulas for the closest to the threshold 0+ and 2+ subthreshold states at 22.2,20.2, and 20.1 MeV. The electron screening potential Ue is then obtained by fitting it as a single parameter to the low-energy data. The calculations are also done by fitting Ue simultaneously with other parameters. Results: The low-energy S factor is dominated by the 2+ subthreshold resonance at 22.2 MeV. The influence of the other two subthreshold states is small. The resultant electron screening is smaller than the adiabatic value. The fits that neglect the electron screening above 60 keV produce a significantly smaller electron screening potential. The calculations show a large ambiguity associated with a choice of the initial channel radius. Conclusions: The R-matrix fits do not show a significantly larger Ue than predicted by the atomic physics models. The R-matrix best fit provides Ue=149.5 eV and Sb(0)=21.7 MeV b.