Pneumocystis jirovecii pneumonia (PCP) is an opportunistic infection with airborne transmission and remains a major cause of respiratory illness among immunocompromised individuals. In recent years, several outbreaks of PCP, occurring mostly in kidney transplant recipients, have been reported. Currently, multilocus sequence typing (MLST) performed on clinical samples is considered to be the gold standard for epidemiological investigations of nosocomial clusters of PCP. However, until now, no MLST consensus scheme has emerged. The aim of this study was to evaluate the discriminatory power of eight distinct loci previously used for the molecular typing of P. jirovecii (internal transcribed spacer 1 [ITS1], cytochrome b [CYB], mitochondrial rRNA gene [mt26S], large subunit of the rRNA gene [26S], superoxide dismutase [SOD], β-tubulin [β-TUB], dihydropteroate synthase [DHPS], and dihydrofolate reductase [DHFR]) using a cohort of 33 epidemiologically unrelated patients having respiratory samples that were positive for P. jirovecii and who were admitted to our hospital between 2006 and 2011. Our results highlight that the choice of loci for MLST is crucial, as the discriminatory power of the method was highly variable from locus to locus. In all, the eight-locus-based scheme we used displayed a high discriminatory power (Hunter [H] index, 0.996). Based on our findings, a simple and alternative MLST scheme relying on three loci only (mt26S, CYB, and SOD) provides enough discriminatory power (H-index, 0.987) to be used for preliminary investigations of nosocomial clusters of PCP.