Spillovers Within and Between Physicians: Evidence from a Randomized Overprescribing Letter in Medicare

How to cite

Sacarny, Adam; Barnett, Michael; Olenski, Andrew. 2018. "Spillovers Within and Between Physicians: Evidence from a Randomized Overprescribing Letter in Medicare." AEA RCT Registry. AEARCTR-0003209. https://doi.org/10.1257/rct.3209-2.1

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Adam Sacarny* , Michael Barnett , Andrew Olenski

* Primary investigator

Registry ID
AEARCTR-0003209
DOI
https://doi.org/10.1257/rct.3209-2.1
First registered
Last updated
Status
Completed
Trial dates
2015-04-20 – 2021-01-01
Countries
United States of America ()
JEL codes
I10 I18
Registry record
http://www.socialscienceregistry.org/trials/3209

Abstract

In a fragmented healthcare system, efforts to change a physician’s practice patterns for the publicly insured may have implications for healthcare delivery that go beyond just the targeted physicians and insurers. Specifically, these efforts could change how physicians treat privately insured patients, and they could also influence the practice patterns of other physicians through peer effects. In this work, we aim to generate credible estimates of these spillovers by exploiting a randomized intervention that sought to reduce antipsychotic prescribing to publicly insured older and disabled adults. The original trial sent letters to high volume prescribers of the antipsychotic quetiapine in Medicare. The letters contained a peer comparison message and warned prescribers that they were under review by Medicare, and they reduced antipsychotic prescribing to Medicare patients by 16% over two years. We will estimate two spillovers. First, using commercial insurance data, we will estimate the effect of the letters on antipsychotic prescribing in private insurance, a within-prescriber spillover. Second, using Medicare data linked to practice and referral data, we will estimate the effect on antipsychotic prescribing by peers of the original study physicians, a between-prescriber spillover.

Intervention

Placebo Arm: The placebo arm received a letter describing a coming rule in Medicare that required prescribers to enroll in Medicare. The placebo arm then received a correction letter clarifying the rule. None of the letters mentioned prescribing of quetiapine.

Treatment Arm: The treatment arm prescribers received three letters, spaced roughly 3 months apart, about their quetiapine prescribing. Each letter included a peer comparison message and a notification that the prescriber was under review by Medicare.

Experimental design

In the original randomized trial, an analysis was conducted to identify outlier general-care prescribers of quetiapine in Medicare Part D. This analysis identified 5,055 prescribers, and they were then randomly allocated to the treatment or the control group. These prescribers are the "original study prescribers". The treatment prescribers were sent peer comparison review letters about their quetiapine prescribing; the control prescribers were sent placebo letters about an unrelated regulation.
This follow-up analysis considers two forms of spillover effects of the original letters. First, we examine whether the intervention altered prescribing by original study prescribers to patients covered by commercial insurance. For this analysis, we compare commercial prescribing in the original treatment and control groups.
Second, we examine whether the intervention influenced Medicare prescribing by peers of the original study participants. Peers are defined as either members of the same group practice as the original study prescribers or as those who share patients with original study prescribers. For this analysis, we study the Medicare prescribing of peers of physicians in the original treatment and control groups, exploiting the random assignment of the original study participants as a source of exogenous exposure to peer effects.

Design details

Randomization method
In the original study: randomization done by computer (in Stata)
Randomization unit
Prescriber
Clusters
5,055
Observations
Spillovers onto commercial insurance: 5,055 Spillovers onto peers: TBD
Arms
Spillovers onto commercial insurance: 2,528 control arm, 2,527 treatment arm Spillovers onto peers: TBD
Primary outcomes
Days of quetiapine supplied during the intervention period of April 20, 2015 to December 31, 2016

Cited by

No citation data available for this registration yet.

Keywords