Where the data comes from
Public diversion data is a record of what was caught, acted on and published. It is not a record of what happened. Four channels supply most of it:
- Board orders. State boards of pharmacy, nursing and medicine publish discipline orders: reprimand, probation, suspension, revocation or surrender. In the DivertGuard registry, 45 of 931 cases name a board and 44 state a board action, because the registry leans on criminal and federal sources.
- OIG and DOJ releases. Press releases on charges, pleas, sentences and civil settlements. Of 931 registry cases, 227 cite justice.gov, 79 cite oig.hhs.gov and 43 cite dea.gov.
- Court records. Dockets, indictments, plea agreements and judgments. They carry the detail press releases drop, but they are slow to search and often paywalled.
- Licensing actions. Look-up tools and monthly action lists. Formats differ by state, so a count from one state is not built the same way as a count from another.
How long it takes to appear
Nothing is public at the moment of diversion. A theft is found, investigated, reported to the employer and then to a board or prosecutor, and a public document appears only at a later step: a charge, a plea, a sentence or a board order. These steps are separate events that can sit months or years apart, and the same person can generate several documents. The registry does not record when each source was published, so it cannot give a measured delay, and you should not quote one.
What the registry does show is the effect on recent years: it holds 127 cases dated 2025 and 93 dated 2026 (through 2026-10-30). The current year looks smaller because it is incomplete, not because diversion fell. Never compare a partial year with a full one.
What it systematically misses
- Unreported and handled internally. Cases resolved by termination with no referral never appear.
- No charge, no order. Cases where nobody was charged or disciplined, or where a prosecutor declined.
- Settled quietly. Confidential settlements and sealed records.
- Uneven coverage by state. Boards publish differently, and the registry covers 52 states and territories at different depths. A state with more cases is not necessarily a state with more diversion.
- Who gets prosecuted. Large or unusual cases are more likely to reach a press release, so the published set tilts toward the severe.
- Missing fields. Of 931 records, a role is stated in 255, an outcome in 210, and 351 have a month-only date. Counts by role or outcome describe only the records that state them.
Source quality in the registry
687 of 931 records link to a primary source (a government or court document); 244 rest on news reports. Only 25 carry an archived copy of the source, so links can die. A charge is an allegation: records marked "charged" are not convictions.
How to state the limits in your own analysis
Put this beside any figure you publish, and adapt the brackets:
Source: [named sources], covering [date range], pulled [date]. [N] records. Counts include only cases that became public through [channels]. They exclude cases handled internally or not published, and they do not measure how common diversion is. [Field] is stated in [n] of [N] records; figures for it use those [n]. Recent months are incomplete.
Then follow four habits. State the denominator with every rate. Report missing fields as missing, never as zero. Do not rank states or organisations by count alone. And do not claim a trend from a period in which source coverage itself changed.
For the method behind each of the registry's published figures, see how each was counted, and for the records themselves, download the data.