Here is a simplified overview of how national OD numbers happen. Much of this process is dictated by vital statistics laws.
At the time of death, a doctor writes a suspected cause of death on a registration form. The death is registered with county government and becomes a death certificate.
Depending on professional suspicions, coroners and medical examiners do autopsies, part of which is to run tests on bodily fluids. Based on these findings, the death record is updated with a definitive cause-of-death. This process can result in refinement: A death with doctor-assigned suspected cause-of-death (COD) of “overdose” can be updated to “overdose involving fentanyl and alcohol intoxication.” If we are looking specifically at one drug or a class of drugs, this is obviously an upgrade. The process takes the better part of a year because the talented workforce that does the work is under-resourced. 😳
Regardless of COD, all local death records originating in counties, parishes, and cities get passed to state health departments.
State health departments often make dashboards and reports. These are the state data we will access below, because they are literally closer to the source truth.
State health departments are also required by law to send an abbreviated snippet of the records to CDC. If the autopsy investigation isn’t complete, states can send “provisional” data containing the suspected cause of death. Some states also report the provisional data in public reports, with asterisks. States send records to CDC in batches.
CDC takes these snippets of all deaths, and starts by standardizing. This mostly automated step takes into account spelling and other naming variations, returning ICD codes, an internationally accepted alphanumeric language for classifying all possible causes of death.
CDC extracts the records that are either definitively classified as an overdose, or have a suspected COD in provisionals.
So far, the whole process has take 4-18 months (with 8 a reasonable approximation). The next step is for CDC to make a national prediction.
Along the way, there is opportunity for delay and inconsistency. Yet, the data are invaluable.