Increasing Implementation of Post-Operative Opioid Prescribing Guidelines

Learning Health System Clinician Feedback Using Patient-reported Data to Increase Implementation of Post-operative Opioid Prescribing Guidelines: a Stepped-wedge Cluster Randomized Trial

The objective of this research study is to evaluate the effect of a quality improvement initiative carried out by a health system opioid stewardship task force aiming to increase clinician post-operative prescribing adherence with procedure specific guidelines that were developed using patient reported data. The feedback compares the clinician's average number of opioid pills prescribed after a given procedure to other clinicians in the health system and to the health system guideline recommended amount based on patient reported data on opioid pills taken for that procedure. The feedback also provides historical data on mean patient reported number opioid pills taken following a given procedure and on patients' ability to manage pain among those who received guideline adherent prescriptions compared with patients who received greater than the guideline recommended amount.

This research study will analyze results from a Penn Medicine Opioid Task Force initiative using a two-arm stepped-wedge cluster randomized trial design. The feedback intervention will be rolled out across all participating Penn Medicine surgical departments and divisions currently using Penn Medicine's post-operative text-messaging learning health system platform. Timing of intervention roll-out within each department and division will be randomly assigned to permit evaluation of the effect of the intervention. With the stepped wedge design, not only will all attendings and associated prescribers ultimately receive the intervention, but by the end of the trial, each will be actively using the intervention, meaning its sustenance would occur automatically. A stepped-wedge design will be utilized, in which all clinicians begin in usual care and the order in which they adopt interventions is randomly assigned, because it has both practical and scientific advantages over traditional parallel-cluster randomized designs, in which randomization determines which clusters adopt the intervention at all. Importantly, the stepped-wedge design provides time to prepare surgical divisions for implementation of the intervention(s), enhances stakeholders' enthusiasm for participating by ensuring that each will receive the intervention(s), and typically increases statistical power. The primary analysis of study outcomes will be conducted at the patient-level. Secondary analyses will be conducted at the prescriber and attending surgeon level.

The feedback intervention will be rolled out across all participating Penn Medicine surgical departments and divisions currently using Penn Medicine's post-operative text-messaging learning health system platform. Timing of intervention roll-out within each department and division will be randomly assigned to permit evaluation of the effect of the intervention.

Procedure and prescribing providers will be randomized to when they will begin receiving individual reports on their opioid prescribing.

Initial comparison: Clinicians will receive an initial peer comparison feedback report by email in which their post-operative opioid prescribing for patients undergoing their top 3 procedures is compared to the guideline recommended doses and the mean prescribing doses by other University of Pennsylvania Health System (UPHS) prescribers for each of those 3 procedures since the start of the baseline period. Furthermore, in an attached report, for each of the 3 procedures, the mean patient reported ability to manage pain one week at home after the procedure will be displayed for those who received guideline compliant dosages of opioids relative to those who received higher than guideline recommended amounts. Monthly comparison: On a monthly basis clinicians will get a report providing similar peer comparison feedback as above but with a monthly trend line of their performance since the start of the intervention.

Inclusion Criteria: All surgical prescribers (surgeons, advanced practice providers or resident physicians) who have written at least 5 opioid prescriptions per month for patients undergoing an eligible procedure during the baseline period or attending surgeons who performed an eligible procedure during the baseline period. Procedures are eligible to be included if there are a minimum of 10 unique cases during any given month and post-operative opioids are prescribed. Data on patient reported opioid use, pain scores, and ability to manage pain is currently being collected by Penn Medicine's post-operative text messaging platform. As of February 15, 2022, there were 30 procedures defined by Current Procedural Terminology (CPT) groupings that meet these criteria. In the research study analysis, patients will be included in the primary sample if they (1) undergo an eligible surgical procedure by an eligible attending surgeon, and (2) the proportion of guideline compliant prescriptions written during the baseline period for that procedure was less than 90%. Exclusion Criteria: None

Hussey MA, Hughes JP. Design and analysis of stepped wedge cluster randomized trials. Contemp Clin Trials. 2007 Feb;28(2):182-91. doi: 10.1016/j.cct.2006.05.007. Epub 2006 Jul 7.

Hemming K, Taljaard M, Forbes A. Analysis of cluster randomised stepped wedge trials with repeated cross-sectional samples. Trials. 2017 Mar 4;18(1):101. doi: 10.1186/s13063-017-1833-7.

White H. A Heteroskedasticity-Consistent Covariance-Matrix Estimator and a Direct Test for Heteroskedasticity. Econometrica 1980;48:817-38.

Zeger SL, Liang KY. Longitudinal data analysis for discrete and continuous outcomes. Biometrics. 1986 Mar;42(1):121-30.

Starks MA, Sanders GD, Coeytaux RR, Riley IL, Jackson LR 2nd, Brooks AM, Thomas KL, Choudhury KR, Califf RM, Hernandez AF. Assessing heterogeneity of treatment effect analyses in health-related cluster randomized trials: A systematic review. PLoS One. 2019 Aug 12;14(8):e0219894. doi: 10.1371/journal.pone.0219894. eCollection 2019.

Hogan JW, Lancaster T. Instrumental variables and inverse probability weighting for causal inference from longitudinal observational studies. Stat Methods Med Res. 2004 Feb;13(1):17-48. doi: 10.1191/0962280204sm351ra.