Given the amount of travel done across the state, it should be no surprise that injuries from traffic accidents represent a substantial amount of medical morbidity. Just how much is certainly a data question worth investigating.
One quick way to get an idea of the scale of their impact is to visualize the number of emergency department visits and hospital admissions with injury causes attributed to transportation-related incidents via their ICD-10 External Cause of Morbidity coding. This information typically provides context to the specific injury needing treatment for research and insurance purposes.
Below, I use public data made available by California’s Health Care Access and Information for emergency visits and hospital admissions in 2022 to generate approximate counts for transport-related injuries, and all accidents to provide additional context. I focus on codes labeled “initial encounter” to reduce the amount of double counting for injuries from a single accident. To provide some finer detail, I use the codes and accompanying descriptions to identify mode of transportation and specific vehicle types involved and whether the incident occurred on a road (“in traffic”).
*note: this work was performed as a coding exercise and not to be taken as an official writeup
Quick Summary
Between emergency admissions and hospitalizations, California recorded more than 2.5 million external causes of accident injuries in its EDV/hospitalization encounters, with more than 400,000 of those related specifically to transport accidents.
Codes for injuries from motor vehicle occupants comprised the majority of transportation codes reported (62%), and injuries from car occupants comprised the majority of those for vehicular occupant injuries (79%). But while only 7% of those involved admission to the hospital, codes for motorcycle riders involved hospitalization at more than 3 times the rate, which unsurprisingly provides evidence that the more unprotected nature of motorcyclists leads to a greater severity of injuries when something goes wrong. For most vehicle injuries, the percentage occurring on road was predictably high (e.g. cars at 93%, heavy vehicles at 72%). Only the miscellaneous transport category showing a majority not on the road, which makes sense given the types of vehicles subsumed by this category (e.g. trains, agricultural vehicles).
Additional Remarks
There are some considerations regarding the use of this methodology to measure impact. First, I’ve taken great care to label these counts as for the codes, specifically. This is because an encounter can take multiple ICD-10 external cause codes which could lead to some degree of double counting when creating totals. For example, vehicular accidents may also have airbag-related injuries have codes in the W-range, or smoke and fire injuries coded with one in the X-range. A complicated traffic crash involving multiple vehicle types may also take on multiple V-codes specifically. Because of this overlap, these counts approximate but don’t explicitly represent the number of persons with these injuries that had emergency medical care.
More broadly, one might be interested in how accurate the coding for transportation injuries is. The 16% of injuries among vehicle riders coded as unknown/unspecified with regard to mode of transport implies that there’s at least some difficulty with the data collection or interpretation. Fortunately, there’s been some recent research looking at that specifically for traffic and pedestrian accidents. A recent meta-analysis by Paleczny et al. examined two quality studies looking at transport injuries, concluding that coding for these pointed to high positive predictive value for motor vehicle accidents in one study and high specificity for pedestrian accidents in another. Interrater reliability was also fairly high for both studies (Cohen’s Kappa range: 0.88-0.98). Sensitivity was a bit more problematic, though the coding difficulties seem to be for pedestrian injuries (sensitivity range: 25-45%) and not motor vehicle or bicyclist injuries (sensitivity range: 87-98%). Both the studies are more than a decade and two decades old, respectively, so it should be that additional time with ICD-10 coding in place has improved the coding a bit.