Opioid and Naloxone Prescribing Following Insertion of Prompts in the Electronic Health Record to Encourage Compliance With California State Opioid Law

This cohort study examines changes in naloxone and opioid prescribing rates in the outpatient setting at 1 integrated health care system after the launch of decision-support prompts in the electronic health record (HER) in compliance with the State of California Assembly Bill 2760.


A. Definition of high-risk opioid overdose
The law requires that the prescribers offer a prescription for naloxone hydrochloride among patients who receive a dosage is at least 90 milligram morphine equivalents, an opioid medication prescribed concurrently with a prescription for benzodiazepine, and who presents with an increased risk of overdose as determined by the patient's history. Throughout this paper we refer to patients that meet any or all of these criteria "high risk".

B. Definition of Endpoints
For the primary endpoints, we define naloxone prescription rate as orders placed per month; naloxone possession rate as percentage of patients who were prescribed with naloxone at discharge within the previous year. We define outpatient opioid prescription rate as the number of opioid orders for patients at discharge per prescriber-month or per prescriber-year in different model settings. We assessed the immediate and long-term effects of the prompts on clinicians' prescription patterns measured monthly. We also assessed average annual changes, comparing pre-and post-intervention at clinicians' level.
For the secondary endpoints, we classify a concomitant prescription as when an opioid and a specified medication were placed either 30 days before or after the order of the other. For other endpoints, we define initial opioid orders as when patients who were prescribed an opioid without having previously had such a prescription at least 90 days prior to the index date; we define renewal orders as when those patients who received an opioid prescription within 90 days of the index date, but had not been prescribed opioids more than 90 days prior to the index date; we define chronically high-dose orders when dosage levels reached MME ≥ 50: this group had also received two or more prescriptions with two different start dates within the 90 days preceding the index date, as well as at least one prescription between 91 and 180 days prior to the index date.

C. Statistical Model
We used an interrupted time series study design with segmented regression and generalized linear mixed models.
δ I δ T δ I * T , ℎ ϵ ~N 0, and μ ~N 0, ; i: Prescriber; t: time (month); The analysis takes the form of an interrupted time series mixed model with segmented regression using monthly repeated measures. The outcome is a function of the expected value of , where i is the prescriber, and t is time.
The right side of the equation is composed of the following elements: 1) the intercept term; 2) an indicator variable for when the prompt warning was put in place; 3) a linear term of a trend that occurs over the course of the data period; 4) the change of the slope term at the point of the intervention. The and β vectors are all the fixed covariates at prescriber level that might be additional explanatory factors for the outcome variable; 5) a random intercept term for each prescriber; and 6) an error term that gives us the error in every time period within each prescriber.
As a sensitivity analysis, we analyzed the model with a quadratic time term. The non-linear model yielded compatible estimates as the linear model.

D. Definition of Variables
We define an adult primary care provider as one who specializes in internal medicine, family medicine, or general The model was adjusted for within-provider clustering, nested within medical center areas. Data is collected at clinician's level.
Model estimates are reported as a scale of rate ratio (RR).
Column names represent outcomes measured; raw names represent covariates in the model. 0.819 (0.679 to 0.988) d a Data is collected annually at clinicians' level. Rate ratios (RR) represent the ratio of post-intervention rate / pre-intervention rate. Interaction term was used for the comparison between clinicians' characteristics. b The reference group in the subgroup analysis. c p<0.001 for rate ratio in annual pre-post analysis. d p<0.05 for rate ratio in annual pre-post analysis. e p<0.001 for difference between the group with specified trait and reference group. P value for difference is derived from multiple comparison method. a Data is collected annually at clinicians' level. Rate ratios (RR) represent the ratio of post-intervention rate / pre-intervention rate. Interaction term was used for the comparison between primary care physicians and non-primary care physicians. P value for difference is derived from multiple comparison method. b The reference group in the subgroup analysis. c p<0.001 for rate ratio in annual pre-post analysis. d p<0.001 for difference between the group with specified trait and reference group. P value for difference is derived from multiple comparison method.