They used private insurance as the exclusive source for their data and extrapolated everything else from that. That is, all of the people with government or simply no funding were excluded from their data, but they acknowledge that glaring failure only in a separately published appendix. Specifically, they eliminated huge, extremely high risk populations to draw attention to a tiny difference in a very low risk group (the highest rate that they found in a high gun ownership state was 2.5, states where guns are essentially banned were lower).
Indeed, buried in the appendix (which looks a lot like what should've been the methods or discussion section), they hid this gem:
"Although the commercially insured population is not reflective of the general U.S.
population, which includes uninsured and those receiving public insurance, we required
health care utilization data with exact dates of service to identify firearm-related health
care utilization on the dates of NRA annual meetings and surrounding control dates.
Other administrative databases, such as the Healthcare Cost and Utilization Project
(HCUP) inpatient and emergency department databases,16 include firearm injuries for a
broader population of individuals requiring hospital care but do not provide exact dates of
service. Moreover, in a previous analysis of HCUP data, those with private insurance
accounted for nearly a third of emergency department visits for unintentional firearm
injury, making this an important population to consider.15"
Link to Appendix:
http://www.nejm.org/doi/suppl/10.1056/N ... pendix.pdf
By their own (well hidden) admission, they artificially cut out 67% of firearm injuries from their data while collecting firearm injury data only from the group most likely to be able to fly across the country for a firearm convention, and then hid their acknowledgement in an appendix that they knew most people wouldn't read.
This is deceitful. Even if one is biased, the cardinal sin for a scientist is to allow that bias to influence their work. It is worse still to attempt to use one's position to mislead laypeople who are inclined to inherently trust one's conclusions. Publishing such trash in an effort to mislead the public is unconscionable.
But we already know all of that: biased authors published biased letters in biased journals, and the bias against guns in the Northeastern medical establishment is well supported with direct quotes. So what is the real finding? Look at the actual numbers. Within the limitations of the study and without the unsupported conclusions, the authors did perform an interesting investigation using a 76 million data points and advanced statistics. They were wildly off base in presenting that data in a deliberately misleading way in the service of an obvious bias, but I have no cause to doubt their numeracy or stats program. The rate of all cause firearm injury resulting in private medical insurance provider payment for hospitalization or emergency department treatment among insured people in the US DURING THE SUMMER (WHEN CRIME IS AT ITS PEAK), hovers around 1.4/100,000.
The authors' failure here is in finding one thing, realizing that it ran contrary to their biases, and then misrepresenting their data in conclusions that the public isn't sufficiently curious or educated to question. This study proves something that we've been saying forever about how focal and divorced from normal gun owners the American problem of gun violence truly is. Those fools just made the error of formally publishing it.
I might even retract my criticisms if they would just tone down the rhetoric, put their methods in the paper where it belongs, and change the title:
"Gun Ownership Poses Much Smaller Health Risks To Insured Populations Than Previously Reported."
Edited many times, often with major content changes.
