OBJECTIVE To assess whether temporal trends in very early (within 48 hours) case-fatality rates may differ from those occurring between 48 hours and 30 days in patients with spontaneous intracerebral hemorrhage (ICH). METHODS All cases of ICH that occurred in Dijon, France (151,000 inhabitants), were prospectively collected between 1985 and 2011, using a population-based registry. Time trends in 30-day case fatality were analyzed in 3 periods: 1985-1993, 1994-2002, and 2003-2011. Cox regression models were used to evaluate associations between time periods and case fatality within 48 hours and between 48 hours and 30 days, after adjustments for demographics, risk factors, severity, and ICH location. RESULTS A total of 531 ICH cases were recorded (mean age 72.9 ± 15.8, 52.7% women). Thirty-day case fatality gradually decreased with time from 40.9% in 1985-1993 to 33.5% 1994-2002 and to 29.6% in 2003-2011 (adjusted hazard ratio [HR] 0.71, 95% confidence interval [CI] 0.47-1.07, p = 0.106, for 1994-2002, and adjusted HR 0.49, 95% CI 0.32-0.73, p < 0.001, for 2003-2011). Over the whole study period, 43.6% of 1-month deaths occurred within the first 48 hours following ICH onset. There was no temporal change in case fatality occurring within the first 48 hours but a decrease in deaths occurring between 48 hours and 30 days was observed with time (HR 0.53, 95% CI 0.31-0.90, p = 0.02, for 1994-2002, and HR 0.32, 95% CI 0.32-0.55, p < 0.01, for 2003-2011, compared with 1985-1993). CONCLUSION Although 30-day case fatality significantly decreased over the last 27 years, additional improvements in acute management of ICH are needed since very early case-fatality rates (within 48 hours) did not improve.