Traffic safety decisions are based predominantly on information from police collision reports. However, a number of studies suggest that such reports tend to underrepresent bicycle and pedestrian collisions. Underreporting could lead to inaccurate evaluation of crash rates and may under- or overestimate the effects of road safety countermeasures. This review examined ten studies that used data linkage to explore potential underreporting of pedestrian and/or bicyclist injury in police collision reports. Due to variations in definitions of reporting level, periods of study, and study locations, it was difficult to directly compare the studies. Even among the six studies using the hospital link definition, estimates of reporting levels ranged from 44 to 75 percent for pedestrian crashes, and from 7 to 46 percent for bicycle crashes, suggesting a severe underreporting problem. However, few of the studies provided estimates of the error around their reporting level estimates, and as a result, it is difficult to determine the true level of underreporting. It may be that bicycle and pedestrian crashes appear in both police and hospital datasets but are less likely to be linked. Due to linkage error, link rate can only be used to estimate reporting level. Without the variance of that estimate, the effect of underreporting on traffic safety analyses cannot be accurately determined. Future studies should include estimates of the error present in their data linkage process for greater accuracy of the underreporting in police data. Datasets should be designed for easier linkage with hospital data and other datasets.