There is no reliable national database for firearm injuries. The Centers for Disease Control and Prevention collect data from multiple sources. But they rely significantly on a system designed to monitor injuries from consumer products, using data from surveys of a relatively small sample of hospitals.

Turnover among the hospitals sampled can shift the data considerably, because gun violence is so concentrated geographically. Armed gangs tend to operate in specific neighborhoods, for instance, while gun suicides vary greatly by state. Thus, the official count on firearm injuries may rise or fall in ways that do not reflect national trends, or provide an accurate portrait of firearm injuries nationwide, leading researchers on gun violence agree.

Much health care data is regionally compartmentalized, because it’s collected by individual states from hospital emergency rooms and other sources. For firearms, the problem is compounded by variations in reporting. As Daniel Webster, director of the Center for Gun Policy and Research at Johns Hopkins University, wrote:

Crimes committed with firearms that do not result in gunshot wounds and occur in areas in which gun violence is relatively common are less likely to be reported to and recorded by the police than those occurring in areas where gun violence is less common. Consistency in reporting is also likely to differ between periods when gun crime is trending high and more peaceful times, as political pressures stemming from high rates of gun crime create incentives for police to under-report crimes in which there are no serious injuries to victims.

The FBI currently collects crime data through two different reporting systems. All states still using the older of the two, the Uniform Crime Reporting program, are scheduled to transition to the newer National Incident-Based Reporting System by 2021. That standardization should help, but only if police departments are encouraged to provide more comprehensive data, including important context on non-fatal shootings. For example, Uniform Crime Reporting standards currently document the type of weapon used in a crime, but how the gun was used – to threaten? to wound? to fire at and miss? – is unspecified.

Another improvement would be for hospital emergency rooms nationwide to code firearm injuries and fatalities in greater detail, uniformly classifying not only the type of injury, such as a “gunshot wound to lower extremity,” but also whether the gunshot resulted from accident, criminal assault or self-harm. That data could then be anonymized and submitted to a centralized database accessible to gun-violence researchers.

The big-picture data we do know for the U.S. are stark: roughly 35,000 gun deaths annually, along with about 90,000 firearm injuries. Gun violence is the leading cause of death for young American men.

Lowering these numbers requires investigating the precise ways in which guns are used to injure or kill people. But high-quality research can’t happen without high-quality data.

 Editorials are written by the Bloomberg View editorial board.