An opioid use disorder simulation was calibrated separately to two large urban, two intermediate and two rural Pennsylvania counties, then used to project overdose mortality to 2029 under 10%, 20% and 30% increases in naloxone distribution and buprenorphine dispensing. A 30% naloxone increase was projected to cut 2029 deaths by 70% (55 to 81%) in Allegheny, 28% (2 to 64%) in Clearfield and 11% (4 to 17%) in Erie, against 10 to 23% for buprenorphine. Responsiveness tracked each county's historical variability in dispensing, so the spread is a property of the calibration rather than a measured difference.
Why it is interesting: It sets out how far the return on an identical harm-reduction increase may vary within one state, on model calibration alone.