Table 1.

Summary of relationships between congestion and traffic accident frequency and the spatio-temporal resolution of the traffic data used. “Concave” relationships encompass the U-shaped relationship mentioned throughout the review as well as any positive second order polynomial terms in linear models. Inversely, “convex” relationships include the inverse U-shape as well as any negative second order terms. The number of data points were calculated by the multiplication of the “number of units” and “number of measurement locations” columns, if not explicitly stated in the publications.

Relationship	Temporal Unit	Duration	Number of Units	Number of Measurement Locations	Number of Data Points	Publication
Linear	Year	1954–1955	2	987	1974	[10]
Linear	Year	1985	1	399	399	[39]
Linear	Year	1954–1955	2	426	852	[11]
Linear	Day	1955	365	1374	501,510	[9]
Linear	Day	2010–2013	1095	167	182,865	[54]
Concave	Hour	1997–1998	17,520	92 (×2 directions)	2,900,000 (after filtering)	[16]
Concave	Hour	1959–1963	1825	1	43,800	[12]
Concave	Hour	1993–1995	21,900	54 (×2 directions)	2,365,200	[50]
Concave	Hour	1993–1994	768 (for weekdays)	1	768	[15]
Convex	Year	2015	1	7	7	[38]
Convex	Year	2003–2007	-	-	1391	[6]