Background Underreporting of road traffic injuries (RTIs) is common across the world and more so in low-and-middle-income countries (LMICs). The estimates of underreporting are often done using linkages between hospital and police records. However, access to population-level hospital data is challenging in LMICs, such as India.

Objective This study aims to estimate underreporting of RTIs by linking police reports and emergency call data.

Methods We used data from two sources: First Information Reports (FIRs) of road crashes registered by police and calls made to the Emergency Response Support System and forwarded to the Police Control Room for police response (PCR calls). In PCR data, we removed duplicate calls and non-transport related incidents, and classified severity of crash (fatal, injury, and property-damage only). We present descriptive statistics and initial results from data linkage for one month of data (April 2022).

Results Over one month, there were 9453 PCR calls for road crashes, with 7726 unique events after removing duplicates (n=1262), non-transport incidents (n=22), and cases with inadequate information to determine if an RTI occurred (n=443). Among those unique events, 0.3% were fatal, 26.5% with injury, and 73.2% property damage only (PDO). For the same period, there were 520 police FIRs of road crashes. Thus, the number of PCR calls was 15 times the number of police reports. The ratio of fatal to injury cases was 1:39 in PCR calls, while it was 1:4 in police reports. Of the total injury or fatal crashes reported in PCR, 12% were linked to police FIR. The ratio of fatal to PDO crashes was 1:109 in PCR and 9:1 in police reports.

Conclusions Crude comparisons show a high level of underreporting of traffic crashes in police reports, indicating that despite police responding to the incidents, seven in eight PCR calls regarding road crashes had no associated police report (FIR). Emergency call records are a promising data source for more complete capture of road traffic injury cases. Given the strong dependency between the two datasets, we will use log-linear model to estimate the level of underreporting by vehicle type.