Now hospitals are doing the same. Three accident and emergency units in Scotland are recording detailed figures for how many victims of violent crime they treat.
They are all in Lanarkshire - Hairmyres, Wishaw and Monklands - and they are all telling the same story: numbers are down, and down by a lot.
Take March of 2013. The three A&Es handled 120 assault victims, only 40 of whom were known to the police. But that was 56% fewer than the 276 they treated the same month a year before.
NHS Lanarkshire, in the first statistics of the kind ever produced, has comparable figures for March through to July 2012 and the same five months this year. The average reduction in violent injuries was 37%.
Analyst Megan MacPherson, of the police's Violence Reduction Unit, has crunched the numbers. Her conclusion: they corroborate the official reported crime stats showing Scotland is safer.
In fact, the fall appears steeper for unreported crimes, suggesting the proportion of people who are hurt who go to the police, in Lanarkshire at least, is higher.
Ms MacPherson said: "In March 2012, there were 180 victims identified through injury surveillance [who] did not report their assault to the police.
"However in March 2013, only 80 victims of violence did not report the assault to the police.
"Therefore, between March 2012 and 2013 unreported violence was down by 44% and this percentage decrease is sustained for each monthly comparison.
"Police Scotland crime data has confirmed that reported violence has decreased and injury surveillance helps to identify a similar trend in unreported violence."
This is exactly what Ms MacPherson's bosses at the VRU and elsewhere in the police wanted to hear. They have long suspected their official recorded crime statistics only tell part of the real story of what happens in the streets and - arguably - even less of what happens behind closed doors. Their perhaps contradictory aim? To get more of the unreported crime reported, and to get the overall level of violent crime down.
It appears to be working. The early results from NHS Lanarkshire show unreported crime falling faster than reported crime, but both going down steadily.
Ms MacPherson has worked out that just under 65% of violent incidents identified in the three A&Es in 2012, the first full year the injury surveillance was carried out, were unknown the police. That is roughly in line with international experience.
The VRU and other police agencies have been lobbying Scottish NHS boards hard to encourage injury surveillance - something pioneered in Wales.
Hospitals aren't collecting the data as part of some mere statistical exercise. Injury surveillance is designed to help the police, by gathering intelligence on the times and places where violent incidents take place. Half of all those injured said they were hurt in the street. Eleven percent were in a pub or a bar. Some 13% were at home.
The information is anonymous but in such bulk that it is helping police decide where and when to mobilise their forces.
That is already helping to reduce crime across the old Strathclyde, where intelligence-led community policing and a sixfold rise in stop-searches has, police say, driven down crime stats.
The Herald's sister paper, the Evening Times, has completed its annual analysis of detailed beat data for the city of Glasgow and the rest of the old Strathclyde force. The Crime on Your Streets series revealed knife assaults in Glasgow dropped a third last year alone. Serious violent crime in Glasgow has nearly halved and the robbery rate in the old Strathclyde force area, including Lanarkshire, is now lower than in Lothian and Borders.
Police figures analysed by the Evening Times show recorded serious violent crime, so called Group 1 offending, in Lanarkshire down 27% in 2012-13 from 2011-12. This isn't directly comparable to the data from the three hospitals. But the ballpark trends are clear, experts say.
Injury surveillance is routinely carried out in Lanarkshire by clerical rather than medical staff. Like official statistics, the data is not perfect. Not all the information provided by patients, moreover, appears entirely robust.
When asked who hurt them, more than half of all victims either don't say or claim it was an unknown assailant. A fifth say they were injured by a friend or an acquaintance. Barely one in 20 says the injury was caused by a partner or ex-partner.
The Lanarkshire data also includes a single fact that surprises few who deal with the consequences of violent crime. The hospitals have recorded what they call "substance aggressors" - whether the perpetrator was abusing drugs or alcohol. In 40% of cases, it is unknown whether the attacker was on drugs, drink or both. But in 42% of cases he or she was on drink and in 5.69% on drink and drugs. Only 10% of perpetrators were definitely not under the influence.