telemedicine has also been applied to two major areas of accident and emergency practice
(Benger, 2000). These are the transmission of computed tomography scans for urgent
neurosurgery and the ongoing support of minor injury units.
Given that medical technology is likely to have improved over the course of our time
series, it is necessary to find relevant measures that adequately track these changes over both
space and time. The safety literature gives little guidance on this with few studies attempting
to account for this sort of change. Lave (1985) used hospitals per square mile to attempt to
account for access to medical services (in the event of an accident). This would serve to
control for rural areas being less accessible to fast medical care for emergencies. He found
this variable to be significant, though his analysis suffers from not controlling for either
cross-sectional or time-series effects. Noland (2001a) also used this variable with cross-
sectional time-series data but did not find this to be significantly associated with reduced
fatalities. In fact this variable may be endogenous implying that those areas with more
traffic-related fatalities may have comparatively more hospital facilities. The response of the
health care profession in developing emergency medicine procedures would certainly suggest
this possibility (Sakr & Wardrope, 2000).
Alberman (1985) has suggested that the downward trend in perinatal and infant
mortality rates is a result of improved medical technology. While infant mortality is not
explicitly linked to the type of injuries associated with traffic accidents, it is likely that
underlying improvements in medical technology can explain trends in both. Therefore, infant
mortality rates might be an adequate and readily available proxy variable for improvements
in medical technology. Noland (2001a) used white infant mortality rates at the state level and
found this to be significantly related to reductions in fatalities but not injuries. White infant
mortality rates were used to minimize the correlation that total infant mortality rates typically
have with per capita income. Noland (2001b) could not find as strong an association using