To identify individual and systems-focused risk factors for pars plana vitrectomy among patients with proliferative diabetic retinopathy (PDR) in a diverse, urban, safety-net hospital setting. Single-center, retrospective, observational, case-control study at Zuckerberg San Francisco General Hospital and Trauma Center between 2017 and2022. Two hundred twenty-two patients with PDR over a 5-year span (2017-2022), consisting of 111 cases who underwent vitrectomy for vision-threatening complications (tractional retinal detachment, nonclearing vitreous hemorrhage, and neovascular glaucoma) and 111 controls with PDR with no history of vitrectomy or vision-threatening complications. Controls were matched 1:1 through incidence density sampling. Medical records were reviewed from time of entry into hospital system to vitrectomy date (or date-matched clinic visit for controls). Individual-focused exposures included age, gender, ethnicity, language, homelessness, incarceration, smoking status, area deprivation index, insurance status, baseline retinopathy stage, baseline visual acuity, baseline hemoglobin A1c, panretinal photocoagulation status, and cumulative anti-VEGF treatments. System-focused exposures included external department involvement, referral route, time within hospital and ophthalmology systems, interval between screening and ophthalmology appointment, interval between conversion to proliferative disease and panretinal photocoagulation or first treatment, and loss-to-follow-up in intervals of active proliferative disease. Odds ratios (ORs) for each exposure on vision-threatening diabetic complications requiring vitrectomy. The absence of panretinal photocoagulation was the primary significant individual-focused risk factor for vitrectomy in the multivariable analysis (OR, 4.78; P= 0.011). Systems-focused risk factors included longer interval between PDR diagnosis and initial treatment (weeks; OR, 1.06; P= 0.024) and greater cumulative duration of loss-to-follow-up during intervals of active PDR (months; OR, 1.10; P= 0.002). Greater duration in the ophthalmology system was the primary systems-focused protective factor against vitrectomy (years; OR, 0.75; P= 0.035). Largely modifiable variables modulate risk of complications requiring diabetic vitrectomy. Each additional month of loss-to-follow-up for patients with active proliferative disease increased odds of vitrectomy by 10%. Optimizing modifiable factors to promote earlier treatment and maintain critical follow-up in proliferative disease may reduce vision-threatening complications requiring vitrectomy in a safety-net hospital setting. Proprietary or commercial disclosure may be found after the references.