Effect of an Electronic Health Record “Nudge” on Opioid Prescribing and Electronic Health Record Keystrokes in Ambulatory Care

Abstract

Background Multiple policy initiatives encourage more cautious prescribing of opioids in light of their risks. Electronic health record (EHR) redesign can influence prescriber choices, but some redesigns add to workload. Objective To estimate the effect of an EHR prescribing redesign on both opioid prescribing choices and keystrokes. Design Quality improvement quasi-experiment, analyzed as interrupted time series. Participants Adult patients of an academic multispecialty practice and a federally qualified health center (FQHC) who received new prescriptions for short-acting opioids, and their providers. Intervention In the redesign, new prescriptions of short-acting opioids defaulted to the CDC-recommended minimum for opioid-naïve patients, with no alerts or hard stops, such that 9 keystrokes were required for a guideline-concordant prescription and 24 for a non-concordant prescription. Main Measures Proportion of guideline-concordant prescriptions, defined as new prescriptions with a 3-day supply or less, calculated per 2-week period. Number of mouse clicks and keystrokes needed to place prescriptions. Key Results Across the 2 sites, 22,113 patients received a new short-acting opioid prescription from 821 providers. Before the intervention, both settings showed secular trends toward smaller-quantity prescriptions. At the academic practice, the intervention was associated with an immediate increase in guideline-concordant prescriptions from an average of 12% to 31% of all prescriptions. At the FQHC, about 44% of prescriptions were concordant at the time of the intervention, which was not associated with an additional significant increase. However, total keystrokes needed to place the concordant prescriptions decreased 62.7% from 3552 in the 6 months before the intervention to 1323 in the 6 months afterwards. Conclusions Autocompleting prescription forms with guideline-recommended values was associated with a large increase in guideline concordance in an organization where baseline concordance was low, but not in an organization where it was already high. The redesign markedly reduced the number of keystrokes needed to place orders, with important implications for EHR-related stress. Trial Registration www.ClinicalTrials.gov protocol 1710018646

Applications of Behavioral Economics to Pharmaceutical Policymaking: A Scoping Review with Implications for Best-Value Biological Medicines

“…Of all identified studies, three main areas were found where they were conducted. The largest group was related to influencing prescribing behavior among physicians [ 42 – 61 ] ( n = 20), a second group aimed to evaluate ways to stimulate vaccine uptake [ 62 – 75 ] ( n = 14), and a last group examined methods to increase medication adherence [ 76 – 85 ] ( n = 10). Identified studies were conducted in the USA ( n = 32) [ 42 , 44 – 53 , 56 , 58 – 61 , 64 , 66 , 68 – 72 , 74 , 75 , 78 – 84 ], the UK ( n = 6) [ 43 , 54 , 55 , 57 , 63 , 85 ], Singapore ( n = 1) [ 76 ], Tanzania ( n = 1) [ 77 ], Germany ( n = 1) [ 62 ], France ( n = 1) [ 73 ], Australia ( n = 1) [ 67 ], and the Netherlands ( n = 1) [ 65 ].…”

“…Multiple behavioral concepts were tested in the identified studies, but the majority involved social norms [ 44 – 46 , 48 – 50 , 54 , 56 , 57 , 59 , 62 – 64 , 69 , 77 , 81 ] ( n = 16), defaults [ 42 , 43 , 47 , 50 , 51 , 53 , 55 – 58 , 65 , 68 , 82 ] ( n = 13), availability [ 44 , 45 , 50 , 52 , 60 , 71 , 72 , 75 ] ( n = 8), framing [ 44 , 47 , 56 , 61 ] ( n = 4), loss aversion [ 76 , 79 , 80 ] ( n = 3), and present bias [ 76 , 79 ] ( n = 1). A variety of study designs was observed among all studies, but most were randomized controlled trials [ 44 , 46 , 47 , 49 , 50 , 54 , 59 , 60 , 63 – 65 , 67 – 70 , 73 – 76 , 78 – 85 ] ( n = 27).…”

“…A variety of study designs was observed among all studies, but most were randomized controlled trials [ 44 , 46 , 47 , 49 , 50 , 54 , 59 , 60 , 63 – 65 , 67 – 70 , 73 – 76 , 78 – 85 ] ( n = 27). Other study designs included pre-post longitudinal studies [ 42 , 43 , 51 , 52 , 55 , 56 , 61 , 72 , 77 ] ( n = 9), randomized trials without a control group [ 45 , 48 , 62 ] ( n = 3), retrospective [ 58 , 71 ] ( n = 2), quasi-experimental [ 53 , 66 ] ( n = 2), or qualitative multi-method studies [ 57 ] ( n = 1).…”

Implementation and Evaluation of Two Nudges in a Hospital’s Electronic Prescribing System to Optimise Cost-Effective Prescribing

“…A recent systematic review of 15 studies, evaluating 20 different intervention studies found that 20% of the interventions did not work [ 25 ]. Nudges are designed to non-coercively influence behaviour without forbidding options [ 26 , 27 , 28 , 29 ], and it is the clinicians who must interpret and act. Four potential reasons why our nudges did not change prescribing are explored here.…”

Implementation and Evaluation of Two Nudges in an Electronic Prescribing System to Optimise Cost-Effective Prescribing in the Hospital

“…A recent systematic review of 15 studies, evaluating 20 different interventions studies found that 20% of the intervention did not work [11]. Nudges are designed to non-coercively influence behaviour without forbidding options [12][13][14][15], and it is the clinicians who must interpret and act. If clinicians do not trust the message, delivered in the electronic prescribing system (or its assumed messenger), they may purposely choose to go against it.…”