What was actually done

Florida is the best-documented case because CDC investigators evaluated it directly. The sequence there ran: a 2010 requirement that pain clinics prescribing controlled substances register with the state; a joint DEA and state law-enforcement operation beginning February 2010; expanded pain-clinic regulation later that year; statewide raids in February 2011; a July 2011 statutory prohibition on physicians dispensing schedule II or III drugs from their offices, alongside a public health emergency declaration and regional strike forces; mandatory dispenser reporting to a new prescription drug monitoring program from September 2011; and in 2012, expanded regulation of wholesale distributors.

Around the same time, a reformulated, harder-to-tamper-with version of extended-release oxycodone replaced the original on the market.

It worked on the metric it targeted

In Florida between 2010 and 2012, drug overdose deaths fell 16.7% (3,201 to 2,666) and the rate fell 17.7%. Prescription drug overdose death rates fell 23.2%. Oxycodone overdose deaths fell 52.1% and oxycodone prescribing fell 24.0%. CDC described these as possibly the first substantial decline in drug overdose mortality in any state in a decade.

Nothing on this site disputes that. Pill mills were a real harm, closing them was right, and it saved lives in Florida in those years.

And it moved the problem

In the same state over the same period, heroin overdose deaths rose from 48 to 108 — a 122.4% increase in rate, from 0.3 to 0.6 per 100,000. CDC’s own framing was cautious: the absolute substitution effect was limited in that window (668 fewer opioid analgesic deaths against 60 more heroin deaths), heroin deaths had fluctuated widely for years, and other states without prescription declines also saw heroin increases.

The larger displacement was not visible yet. Purified, cheap heroin and then illicitly manufactured fentanyl entered a market of people whose supply had been cut off and whose underlying condition — in many cases an opioid use disorder — had not been treated. Within a few years the fentanyl supply would make the pill-mill era look small.

The limitation the report named itself

The data sources available for this investigation did not permit any assessment of potential unintended consequences of these policy changes, such as reduction of access to pain medication for legitimate prescribing indications.

Johnson et al., MMWR, 2014 — stated limitations

That sentence is the hinge of this entire argument. The evaluation that established the crackdown worked explicitly could not measure the harm to legitimate patients, because the data to measure it did not exist. It still largely does not. Supplying it is part of what this program is for.