A new report from the University of Pittsburgh reveals that U.S. states have undercounted the number of opioid deaths within their borders. After reviewing drug overdoses from 1999 to 2015, researchers discovered that a large swath of opioid deaths were being categorized as “unclassified” overdoses. While the official number of U.S. opioid deaths from 1999 to 2015 sits at 255,527 people (58 percent of all overdoses in that period), another 70,000 opioid overdoses had accidentally been listed as “unclassified,” meaning that the number of national opioid deaths may be as much as 27 percent higher than previously reported.

According to the report’s lead author, University of Pittsburgh biostatistics professor Jeanine M. Buchanich, those increases weren’t spread evenly across U.S. states. While researchers recategorized only 9 overdoses as opioid-caused in Vermont, they found over 11,152 miscategorized cases in Pennsylvania. In Pennsylvania, along with in Alabama, Indiana, Louisiana, and Mississippi, the authors found such a high number of miscategorized opioid deaths that the overall number in those states more than doubles the previous assumption.

The reason some states have more unclassified overdoses than others has to do with the extent to which their medical system is centralized. Those with more centralized systems tend to rely on reports from medical examiners instead of from coroners — who are less likely to be doctors and often classify drug deaths irregularly — and therefore have more accurate opioid death counts.