Characteristics of papers included: long-term timeframe
| Author (year) | Country | Years | Focus (sleep/light) | By season | By time of day (morning/ evening) | Population/ outcome | Timeframe | Finding (narrative) |
| Chu37 (1976) | USA | 1974 | Light | January to March only | ✓ | All fatalities | 3 months | Overall estimate of 47 fatalities saved (8%) in the first half of 1974 that can be attributed to DST. A sharply higher fatality rate during morning rush hour and a sharply lower rate in the afternoon peak hour |
| Coate18 (2004) | USA | 1998 and 1999 | Light | ✓ | ✓ | Fatalities: pedestrians and motor vehicle occupants | 1 month before and 1 month after | Full-year DST would reduce pedestrian fatalities by 171 per year (13%), and motor vehicle occupant fatalities by 195 per year (3%). An hour later sunset would reduce evening pedestrian fatalities by about one-quarter and an hour later sunrise would increase morning fatalities by about one-third. No increased risk to school children from full-year DST |
| Crawley29 (2012) | USA | 1976–2010 | Both sleep and light | ✓ | X | All collisions | 13 weeks before and 9 weeks after. Also comparison of 1987–2003 to 1976–1986 | Significant fatal crash-saving effects of DST in the long run, shown particularly in the autumn test. ‘The spring test gave little evidence either way’ |
| Ferguson et al 9 (1995) | USA | 1987–1991 | Light | ✓ | ✓ | Fatal collisions: pedestrians and motor vehicle occupants | 13 weeks before and 9 weeks after | The most notable effects of changing light levels on fatal crashes were seen when light levels changed from light to twilight (collisions increased) and when twilight changed to light (collisions decreased). Benefits are smallest during the darkest winter months because the evening reduction is increasingly offset by increases during the morning. An estimated 901 fewer fatal crashes (727 involving pedestrians and 174 involving vehicle occupants) might have occurred had DST been retained year-round from 1987 to 1991 |
| Huang16 (2010) | USA | 2001–2007 | Both sleep and light | ✓ | ✓ | All collisions and fatal collisions | 8 weeks before and after | DST, all else equal, is associated with fewer RTCs and fatal RTCs for most day parts (except 09:00–-15:00 hours) |
| Meyerhoff17 (1978) | USA | 1973–1974 | Light | ✓ | ✓ | All fatal collisions | January–February and March–April 1974 (DST) and January–February and March–April 1973 (no DST) (long-term) | A net reduction of about 0.7% during the DST period, March and April 1974, compared with the non-DST period. March and April 1973, but little net DST effect on fatal accidents in winter. A marked decrease in evening fatalities is observed, but the morning increase is not seen as anticipated |
| Sarma and Carey20 (2016) | Ireland | Light | ✓ | ✓ | Collisions, injuries, fatalities for different road users (pedestrians, cyclists and all road users) | 5 and 7 weeks pretransition and post-transition into and out of DST | Transition into DST: increase in collisions in evening period for 5-week (12.6%) and 7-week (13.1%) analyses. Also increase in casualties in evening period at 5 (17.6% increase) and 7 weeks (19.5% increase). Overall (combining morning and evening peak periods) increase in casualties at 5-week (10.5% increase) and 7-week (12.7%) analyses. Transition out of DST: increase in morning casualties at 7 weeks (12.4% increase) and overall increase in casualties when morning and evening combined for 7 weeks (5.5% increase) | |
| Sood and Ghosh22 (2007) | USA | 1976–2003 | Sleep and light | Spring only | X | Fatal collisions: pedestrians and motor vehicle occupants | 13 weeks before and 8 weeks after | Long-term reduction of 8%–11% in RTCs involving pedestrians, and 6%–10% in RTCs involving vehicle occupants |
| Sullivan21 (2001) | USA | 1987–1997 | Light | ✓ | Evening only | Fatal collisions: pedestrians and motor vehicles | 3 weeks before and after | Pedestrian fatalities 4.14 times more likely in darkness (DST) than in daylight (ST). Interaction between light and alcohol use |
| Sullivan38 (2002) | USA | 1987–1997 | Light | ✓ | ✓ | Fatal collisions: pedestrians and motor vehicle occupants | 9 weeks before and after | Overall, pedestrian fatalities 3–6.75 times more likely in darkness (ST) than in daylight (DST), while other crashes were only marginally more likely in darkness. Spring am: twilight shows a decline in crashes from week 8 (39 crashes) to week 1 (8 crashes); at the changeover, when the period is returned to darkness, the crash level rises again. Spring pm: the crash frequency is high during the dark period just before the DST changeover, and drops to 54, the week after the changeover and declines more the following week to 32. Autumn am: 79 crashes before the transition and 29 after. Autumn pm: in the week before the transition there were 65 crashes, in the following week there were 227, an increase of three and a half times |
| Sullivan39 (2003 and 2004) | USA | 1987–2001 | Light | Autumn only | Evening only | Fatal collisions: motor vehicle occupants only | 5 weeks before and after | Rear-end collisions change from an average count of about 13 crashes in the light (DST) to an average of 37 in the dark (ST). Impact of light on crash risk varies across rear-end collision types |
| Sullivan6 (2007) | USA | FARS=1987–2004; NCDOT=1991–1999 | Light | ✓ | Evening only | Fatal and non-fatal collisions: pedestrian (child, adult, elderly) and motor vehicle occupants | 5 weeks before and after | Fatal crashes involving pedestrians, animals and other motor vehicles showed the most reliable increases in risk in low light levels (ST). Children show a reliably greater risk in darkness, but this risk is much smaller than the risk observed for adult and elderly pedestrians, which is nearly seven times greater in darkness. Even when the data are not separated by age, the apparent increase in pedestrian risk in the dark is very strong |
DST, daylight saving time; RCT, randomised controlled trial; ST, standard time.