The porcine reproductive and respiratory syndrome virus (PRRSV) is an enveloped RNA virus, with high mutation rates and genetic variability; which is evident by the large number of discrete strains that co-circulate in swine populations. Veterinary practitioners frequently identify certain discrete PRRSV strains as having a higher clinical impact on production. However, with exception of a few strains, production impact is not well characterized for the majority of PRRSV variants. Predictive analytics, coupled with routine diagnostic sequencing of PRRSV, provide opportunities to study the clinical impact of discrete PRRSV strains on production. Thus, the primary objective of this research was to evaluate clinical impact of discrete PRRSV clades observed in Ontario sow farms. PRRS viruses were classified into discrete clades using Bayesian analysis of the nucleotide sequences of the ORF-5 region of the genome. Production data were gathered through veterinary clinics from herds participating in the ongoing PRRSV surveillance system. Data about pre-weaning mortality, sow mortality, and abortion rates were measured up to 8 weeks post initial PRRSV outbreak. Through conventional regression analysis, results support that clinical impact of the viruses varied among clades over time for abortion rate (p = 0.05) and pre-weaning mortality (p < 0.01). Using predictive modelling approaches based on grouped K-fold cross-validation, it was identified that PRRSV clade designations and other measured factors showed low predictive performance for abortion (R2 = 0.07), pre-weaning mortality (R2 = 0.09), and sow mortality (R2 = 0.04). Clade designation consistently showed moderate importance for abortion and pre-weaning mortality, with clade 2 viruses being identified, on average, as having higher impact. These results demonstrate that the prediction of clinical impact, through production parameters, based on phylogenetic classification of PRRS viruses is possible. However, very high impact outbreaks were difficult to predict across production parameters. More surveillance-derived data are required to continue to improve predictive performance of the models.