Complex polymicrobial communities inhabit the lungs of individuals with cystic fibrosis (CF) and contribute to the decline in lung function. However, the severity of lung disease and its progression in CF patients are highly variable and imperfectly predicted by host clinical factors at baseline, CFTR mutations in the host genome, or sputum polymicrobial community variation. The opportunistic pathogen Pseudomonas aeruginosa (Pa) dominates airway infections in the majority of CF adults. Here we hypothesized that genetic variation within Pa populations would be predictive of lung disease severity. To quantify Pa genetic variation within whole CF sputum samples, we used deep amplicon sequencing on a newly developed custom Ion AmpliSeq panel of 209 Pa genes previously associated with the host pathoadaptation and pathogenesis of CF infection. We trained machine learning models using Pa single nucleotide variants (SNVs), clinical and microbiome diversity data to classify lung disease severity at the time of sputum sampling, and to predict future lung function decline over five years in a cohort of 54 adult CF patients with chronic Pa infection. The models using Pa SNVs alone classified baseline lung disease with good sensitivity and specificity, with an area under the receiver operating characteristic curve (AUROC) of 0.87. While the models were less predictive of future lung function decline, they still achieved an AUROC of 0.74. The addition of clinical data to the models, but not microbiome community data, yielded modest improvements (baseline lung function: AUROC=0.92; lung function decline: AUROC=0.79), highlighting the predictive value of the AmpliSeq data. Together, our work provides a proof-of-principle that Pa genetic variation in sputum is strongly associated with baseline lung disease, moderately predicts future lung function decline, and provides insight into the pathobiology of Pa's effect on CF.