If you have been shocked and frightened by recent government reports on the growth in the number of Americans addicted to and dying from heroin, brace yourself for some more bad news: The problem is even worse than government data indicate.

Consider the question of how many Americans have a diagnosable heroin use disorder. The 591,000 figure quoted virtually every day in the media and in political debates comes from the federal government’s annual National Household Survey on Drug Use and Health (NSDUH).

This survey provides accurate information for many substance use and mental health-related indicators but has two serious flaws that lead it to dramatically underestimate the prevalence of heroin use disorder. First, it excludes people who are incarcerated and people who are living on the street, both of whom have very high rates of drug use. Second, it relies on self-reports: While most people are comfortable telling government surveyors about their use of tobacco, alcohol and marijuana, many understandably fear acknowledging use of heroin.

The degree to which NSDUH underestimates the prevalence of heroin use disorder is enormous. In 2010, a research team combined NSDUH data with that from other studies to determine that NSDUH could only identify 60,000 of the 1 million Americans who used heroin daily or near daily heroin users. As most daily or nondaily heroin users would meet diagnostic criteria for heroin use disorder, NSDUH’s most recent estimate of 591,000 probably didn’t even capture the depth of the problem back in 2010, which was before the heroin problem exploded.

The true level of heroin use disorder today could easily be double or even triple NSUDH’s estimate, but no one can truly know. As successive Congresses have clamped down on federal spending, many government agencies have been forced to curtail their research capacity. National programs that gathered substance use data from people entering jails and from emergency room patients fell under the budget ax in 2014 and 2011, respectively.

Government-reported heroin overdose figures also understate the severity of the epidemic. Death certificates completed by county coroners are the root data source for state and federal government estimates of the extent of the overdose epidemic. Professor Christopher J. Ruhm of the University of Virginia recently demonstrated that many death certificates in drug overdoses cases do not specify the drug or drugs involved. Correcting for this problem, he estimated the true heroin death rate in the U.S. is 22 percent above the government’s official figures.

Asking state and federal government agencies to invest more in better heroin-related data collection systems may seem pedantic and nerdy, perhaps even a distraction from the desperate need for more treatment and prevention efforts. But without such an investment, the government and the nation will continue to fly blind on the heroin epidemic, unable to determine whether current policies are making the problem better, worse, or making no difference at all.