Background: Performance Status (PS) is prognostic for poorer survival in oncology. Poor PS (pPS) may lead to inferior outcomes both as an independent factor and by limiting the range of available treatment options and the ability of patients to tolerate those regimens. Despite high real-world prevalence, patients with pPS are consistently underrepresented in clinical trials. Limited information is available regarding current treatment patterns for patients with pPS and advanced or metastatic non-small cell lung cancer (mNSCLC). Methods: We conducted a retrospective analysis using the nationwide de-identified Flatiron Health electronic health record (EHR)-derived database, selecting patients with mNSCLC diagnosed between 2017-2022 without EGFR or ALK genomic tumor aberrations. We estimated the prevalence of pPS (ECOG 2-3) within this cohort, and described common systemic treatment patterns by year and by PD-L1 status. For the subset of patients treated with pemetrexed/pembrolizumab/carboplatin (PPC), we estimated the association between pPS versus good PS (gPS - 0/1) and early discontinuation ( < 4 cycles) or dose reduction (AUC < 4) of carboplatin using logistic regression controlling for observable confounders. Results: One quarter (24%) of the 22,575 patients in our sample had pPS. Relative to patients with gPS, patients with pPS were more likely to be older (mean age 72 vs. 69y, p < 0.001), to have stage IV cancer (62% vs 56%, p < 0.001) and less likely to have known PD-L1 status (60% vs. 63%, p = 0.002). The fraction of patients with no documented treatment was substantially higher among pPS versus gPS (37% vs. 21%). In the subset of patients with PS3, 53% of patients had no documented treatment. The most common treatment regimens among pPS patients were mono-immunotherapy (mono-IO - 19%) and PPC (15%). Use of mono-IO was positively correlated with higher PD-L1 expression (44% vs. 14% vs. 5.9% in PD-L1 high/low/negative, respectively). No documented treatment was negatively correlated with PD-L1 status (26% vs. 35% vs. 42%). Among patients with PS3, higher rates of no documented treatment were observed (39% vs. 50% vs 58% for PD-L1 high/low/negative, respectively). Anti-CTLA-4 combination use was low regardless of PS. pPS was associated with greater likelihood of early carboplatin discontinuation, (adjusted odds ratio, aOR 2.06), and dose reduction (aOR 1.75), both p < 0.001. Conclusions: One quarter of patients with mNSCLC have pPS. These patients are almost twice as likely to receive no treatment as those with gPS, and within the subset treated with the most common platinum containing regimen (PPC), pPS patients are also more likely to both initiate carboplatin at a reduced dose, and to discontinue it prior to completion. This suggests a substantial unmet need for this important subset of patients with mNSCLC.