It seems good old Dade County FLA has very good records concerning this.
Can Citizens Use Guns Competently?
Ordinary people, even if they have passed a firearms safety class, cannot be trusted to use guns competently, it is sometimes claimed. The guns will be taken away by criminals, or the gun-owners will shoot an innocent bystander by mistake, it is sometimes predicted. Wherever the concealed carry issue is raised in the future, it can be predicted with confidence that these objections will be raised by reform opponents, including many law enforcement professionals who claim expertise on the issue.The existing body of research provides no support for these fears. The best evidence we have about what happens when people have carry permits is the experience of the 1/3 of American states that issue such permits routinely. From these states, the most detailed data are those compiled by the Dade County (Miami) police. As discussed above, the police kept track of every known incident involving the county's more than 21,000 handgun carry permitees over a six-year period. In that six-year period, there was one known incident of a crime victim having his gun taken away by the criminal. There were no known incidents of a crime victim injuring an innocent person by mistake. In some cases the handgun permit holder was successful in preventing a crime, and in some cases not, but in no case was any innocent person injured as a result of mistake by a permit-holder.
Another study examined newspaper reports of gun incidents in Missouri, involving police or civilians. In this study, civilians were successful in wounding, driving off, capturing criminals 83% of the time, compared with a 68% success rate for the police. Civilians intervening in crime were slightly less likely to be wounded than were police. Only 2% of shootings by civilians, but 11% of shootings by police, involved an innocent person mistakenly thought to be a criminal. 
The Missouri research does not prove that civilians are more competent than police in armed confrontations. Civilians can often choose whether or not to intervene in a crime in progress, whereas police officers are required to intervene. Being forced to intervene in all cases, police officers would naturally be expected to have a lower success rate, and to make more mistakes. Attorney Jeffrey Snyder elaborates:
Rape, robbery, and attempted murder are not typically actions rife with ambiguity or subtlety, requiring special powers of observation and great book-learning to discern. When a man pulls a knife on a woman and says, "You're coming with me," her judgment that a crime is being committed is not likely to be in error. There is little chance that she is going to shoot the wrong person. It is the police, because they are rarely at the scene of the crime when it occurs, who are more likely to find themselves in circumstances where guilt and innocence are not so clear-cut, and in which the probability for mistakes is higher. 
In addition, the Missouri study was not restricted to "carry" situations, but also included self-defense in the home. Persons using a gun to defend their own home, who know its layout much better than does an intruder, might be expected to have a higher success rate than would persons using a gun in a less familiar public setting.
The most detailed information about civilian defensive gun use has been compiled by Professor Gary Kleck (a liberal Democrat, and member of the ACLU and Common Cause) in his book Point Blank: Guns and Violence in America. In 1992 the American Society of Criminology awarded the book the Hindelang Prize, as the most significant contribution to criminology in the previous three years. In Point Blank, Kleck studied computer tapes from the U.S. Department of Justice's National Crime Survey, for the years 1979-85. Analyzing the data from over 180,000 crime incidents in the National Crime Survey, as well from other studies, Kleck found the following:
- In no more than 1% of defensive gun uses was the gun taken away by a criminal.
- The odds of a defensive gun user accidentally killing an innocent person are less than 1 in 26,000.
- For robbery and assault victims, the lowest injury rates (17.4% for robberies, and 12.1% for assaults) were among victims who resisted with a gun.
- The next lowest injury rates were among persons who did not resist. Other forms of resistance (such as shouting for help, or using a knife), had higher injury rates than either passive compliance or resistance with a gun. 
It would seem that you need to worry more about the criminals and LEOs...