HIV care cascades can evaluate programmatic success over time. However, methodologies for estimating cascade stages vary, and few have evaluated differences by demographic subgroups. We examined cascade performance over time and by age, sex, and race/ethnicity in Kaiser Permanente, providing HIV care in eight US states and Washington, DC. We created cascades for HIV+ members' age ≥13 for 2010-2012. We measured "linkage" (a visit/CD4 within 90 days of being diagnosed for new patients; ≥1 medical visit/year if established); "retention" (≥2 medical visits ≥60 days apart); filled ART (filled ≥3 months of combination ART); and viral suppression (HIV RNA <200 copies/mL last measured in year). The cascades were stratified by calendar year, sex, age, and race/ethnicity. We found men had statistically (p < 0.05) higher percent linkage, filled ART, and viral suppression for 2010 and 2011 but not for 2012. Women had significantly greater retention for all years. Annually, older age was associated (p < 0.05) with retention, filled ART, and viral suppression but not linkage. Latinos had greater (p < 0.05) retention than whites or blacks in all years, with similar retention comparing blacks and whites. Filled ART and viral suppression was increased (p < 0.05) for whites compared with all racial/ethnic groups in all years. Cascade methodology requiring success at upstream stages before measuring success at later stages (i.e., "dependent" methodology) underreported performance by up to 20% compared with evaluating each stage separately ("independent"). Thus, care results improved over time, but significant differences exist by patient demographics. Specifically, retention efforts should be targeted toward younger patients and blacks; women, blacks, and Latinos require greater ART prescribing.