It’s understandable that the Veterans Health Administration would be deeply concerned about opioid overdoses. From 1999 to 2019, almost 500,000 Americans died of such overdoses — and the opioid overdose rate among veterans is substantially higher than among other U.S. citizens.

In 2012, at the height of the prescription opioid trend in the United States, roughly 1 in 3 U.S. veterans was prescribed opioids. That statistic helped inspire the Department of Veterans Affairs to begin in 2013 an ambitious program to curb opioid prescriptions called the Opioid Safety Initiative. Treatment centers began to track veterans’ medication histories very closely, and doctors were strongly urged to try alternative pain-relief regimens. The program focused on high-risk groups: veterans taking high doses, those with long-term prescriptions and those taking other potent medications (such as benzodiazepines). It had a dramatic effect on treatment: By 2020, the number of VA patients prescribed an opioid had fallen by 64 percent from the peak.

Recent research we conducted, however, suggests that the program may have had unintended negative effects. There have long been concerns within and outside VA that the Opioid Safety Initiative, by curtailing access to painkillers, may have increased the vulnerability of some patients who suffer from severe chronic pain (or the effects of opioid withdrawal) even as it reduced opioid-related risks for others. Looking at trends in violent deaths from 2013 to 2018, we found evidence in favor of this hypothesis: namely, a significant increase in veteran suicides after the implementation of the safety initiative.

Our approach was to compare the trend in veteran suicides before and after 2013 (the year of the OSI’s implementation) to the same trend in non-veteran (civilian) suicides. Thus, civilians acted as the control group in our analysis — any broader changes affecting suicides in the U.S. population should be picked up by the trend for civilians (suicides in general have been steadily rising for years).

We used multiple data sets in the study: At the Centers for Disease Control and Prevention, the National Violent Death Reporting System’s Restricted Access Database links data from vital records, medical examiners and law enforcement agencies. It includes veteran status and county of residence for every violent death. (Seventeen states began contributing to the database before 2013, when the OSI began, so we focused on those states, which have a total population of more than 100 million people.) We supplemented that information with demographic and socioeconomic variables for those counties from the American Community Survey. Controlling for demographics helped show that it was indeed veteran status, not other characteristics, that drove the results. After all, veterans tend to be older and more often men than the general population, among other differences.

In our analysis, we did not just look at veteran vs. civilian status. We also differentiated between urban and rural veterans (and compared them with urban and rural civilians). This is because we expected any effects of the OSI on suicide to be particularly strong among rural veterans, who are more likely than urban veterans to receive health care through VA (and therefore be directly affected by the OSI). Rural veterans are also especially inclined to receive prescription opioids and to experience severe physical and mental health problems — yet they are less likely to receive mental health or substance-use-disorder treatment.

From 2010 to 2012, we did not see significant differences in the suicide trends of veterans relative to civilians (though veterans, as mentioned, had a higher overall suicide rate). Starting in 2013, however, the first year of reduced opioid prescriptions caused by the OSI, the rural veteran suicide rate rose markedly relative to other groups. We estimate that the OSI led to an increase, from 2013 to 2018, in the rural veteran suicide rate of about 12 per 100,000 relative to rural civilians. This represents a roughly 75 percent increase over the pre-OSI difference in suicide rates between rural veterans and civilians. (An increase of 12 suicides per 100,000 rural veterans also represents a jump of more than one-third in the average rural veteran suicide rate before the program’s introduction.)

The increase in suicides among urban veterans was also disturbing, although less dramatic. The increase took longer to begin, for one thing, becoming identifiable only in 2015. By 2018, the difference in the suicide rate in that population, relative to urban civilians, had increased by almost five suicides per 100,000 — an increase of one-third over the pre-OSI suicide rate difference between veterans and civilians.

Another way to describe the results we identified — based on other researchers’ estimates, in addition to ours — is that for every 100 unique veterans who were not prescribed opioids as a result of the OSI, just fewer than three died by suicide.

It is true that there have been efforts outside VA to curb the prescription of opioids: States have also begun to monitor such prescriptions and identify dangerous patterns, and opioid prescriptions have dropped generally. But the high base rate of prescriptions among veterans and the particular aggressiveness of the OSI led us to anticipate that the OSI would have a unique effect. The data bore that out.

The National Violent Death Reporting System does not contain data on accidental drug overdose deaths. Therefore, it’s possible that the OSI discouraged such deaths even as it led to an increase in suicides. But we view this as unlikely: First, other research has found that restrictions on prescription opioids appear to encourage use of more dangerous, illicit opioids; the use of illicit opioids, in turn, increases the risk of overdose. Second, other estimates in the literature suggest that while the introduction of opioids to new patients does increase their long-term risk of overdose, it is by an order of magnitude smaller than the suicide effects we identified after the introduction of the OSI. In short, the increased suicides we found were unlikely to have been offset by lives saved by other means as a result of the program.

It is important to recognize that the long-term consequences of VA’s Opioid Safety Initiative may be very different from the short-term ones: Fewer veterans are now exposed to dangerous opioid regimens to begin with. These long-run effects will become apparent only as additional data becomes available. Nevertheless, our estimates suggest that a sizable group of veterans, particularly those living in rural areas, have suffered as a result of a stricter opioid environment (after many years of relatively lax opioid policies). Our findings suggest that policies that swiftly and significantly curtail opioid prescriptions should be approached with great caution.