By Jake Harper

In a refrigerator in the coroner’s office in Marion County, Ind., rows of vials await testing. They contain blood, urine and vitreous, the fluid collected from inside a human eye.

In overdose cases, the fluids may contain clues for investigators.

“We send that off to a toxicology lab to be tested for what we call drugs of abuse,” said Alfie Ballew, deputy coroner. The results often include drugs such as cocaine, heroin, fentanyl or prescription pharmaceuticals.

After testing, coroners typically write the drugs involved in an overdose on the death certificate — but not always.

Standards for how to investigate and report on overdoses vary widely across states and counties. As a result, opioid overdose deaths aren’t always captured in the data reported to the federal government. The country is undercounting opioid-related overdoses by 20 to 35 percent, according to a study published in February in the journal Addiction.

“We have a real crisis, and one of the things we need to invest in, if we’re going to make progress, is getting better information,” said Christopher Ruhm, the author of the paper and a health economist at the University of Virginia.

Data from death certificates move from coroners and medical examiners to states and eventually the Centers for Disease Control and Prevention, which publishes reports on overdose counts across the U.S. According to the CDC, more than 42,000 people died from opioid-related overdoses in 2016, a 30 percent increase from the year before.

But that number is only as good as the data states submit to the CDC. Ruhm said the real number of opioid overdose deaths is closer to 50,000. He came to the higher estimate through an analysis of overdoses that weren’t linked to specific drugs.

On a death certificate, coroners and medical examiners often leave out exactly which drug or drugs contributed to a death. “In some cases, they’re classifying it as a drug death, but they don’t list the kind of drug that was involved,” said Ruhm. In the years he reviewed in his paper, 1999 to 2015, investigators didn’t specify a drug in one-sixth to one-quarter of overdose deaths.

Some states do worse than others. In 14 states, between 20 and 48 percent of all overdose deaths weren’t attributed to specific drugs in 2016, according to a breakdown from FiveThirtyEight.