Here's one 2001 study that shows a small, but notable, increase:
Fatal accidents following changes in daylight savings time: the American experience
Conclusions: The sleep deprivation on the Monday following shift to DST in the spring results in a small increase in fatal accidents. The behavioral adaptation anticipating the longer day on Sunday of the shift from DST in the fall leads to an increased number of accidents suggesting an increase in late night (early Sunday morning) driving when traffic related fatalities are high possibly related to alcohol consumption and driving while sleepy. Public health educators should probably consider issuing warnings both about the effects of sleep loss in the spring shift and possible behaviors such as staying out later, particularly when consuming alcohol in the fall shift. Sleep clinicians should be aware that health consequences from forced changes in the circadian patterns resulting from DST come not only from physiological adjustments but also from behavioral responses to forced circadian changes.
This 2014 study from the University of Colorado, Boulder, shows that 10 years of data accrual shows a definite effect as well.
Spring Forward At Your Own Risk: Daylight Savings Time and Fatal Vehicle Crashes
Abstract
Despite mounting evidence that Daylight Saving Time (DST) fails in its primary goal of saving energy, some form of DST is still practiced by over 1.5 billion people in over 60 countries. I demonstrate that DST imposes high social costs on Americans, specifically, an increase in fatal automobile crashes. DST alters fatal crash risk in two ways: disrupting sleep schedules and reallocating ambient light from the morning to the evening. First, I take advantage of the discrete nature of the transitions between Standard Time and DST to measure the impact of DST on fatal crashes in a regression discontinuity design. Then, to measure the duration of the effect, I exploit variation in the coverage of DST created primarily by a 2007 policy change, in a day-of-year fixed effects model. Both models reveal a short-run increase in fatal crashes following the spring transition and no aggregate impact in the fall. Employing three tests, I decompose the aggregate effect into ambient light and sleep mechanisms. I find that shifting ambient light reallocates fatalities within a day, while sleep deprivation caused by the spring transition increases risk. The increased risk persists for the first six days of DST, causing a total of 302 deaths at a social cost of $2.75 billion over the 10-year sample period, underscoring the huge costs of even minor disruptions to sleep schedules.