Menu

159
Strained and Constrained: How Capacity Constraints Affect Physician Decision-Making in the ICU

Tuesday, June 25, 2019
Exhibit Hall C (Marriott Wardman Park Hotel)

Presenter: Manasvini Singh

Co-Author: David Howard


Overview: We exploit exogenous expansions in intensive care unit (ICU) capacity alongside variation in ICU demand (i.e., ICU “strain”) to estimate the effects of capacity constraints on physicians’ use of ICUs. We find substantial heterogeneity in physician response to capacity expansions depending on both patient acuity and ICU strain. We also see evidence of forward-looking behavior by physicians in which current ICU decisions appear to be informed by anticipated future demand and available capacity, with physicians adjusting ICU admissions to accommodate the patients with the greatest medical need.
Background: ICUs lack systematic clinical criteria for admission and discharge, leading to much unexplained variability in ICU care patterns across physicians, hospitals and regions. This ambiguity has led to significant concern about the misuse of ICU care, especially in light of the high costs and high risk of complications associated with ICU stays. Physicians, tasked with making the “appropriate” admission and discharge decisions, play a central role in this process as gatekeepers of scarce ICU resources. In this paper, we explore how physicians may drive some of this variation in ICU care and examine the extent to which organizational factors - such as ICU capacity - influence a physician’s propensity to make optimal ICU care decisions.
Methods: We use 100% inpatient EHR data from two academic hospitals (2015-2018) to examine this question. We exploit 2 separate ICU expansions in 2016 and 2017 to examine the influence of capacity constraints on physician decision-making under varying levels of ICU strain. As a secondary analysis, we also use administrative data for FL (2014-2015) to examine the relationship between capacity constraints (as measured by ICU bed share) and ICU admission rates, where we exploit a patient’s differential distance as an instrument for hospital selection.
Results: We find that that physicians referred more patients to the ICU following the expansion, not only during periods of peak demand but also during periods of slack demand when all patients could have been accommodated at the pre-expansion capacity level. Results are consistent with two explanations: dynamically optimizing ICU gatekeepers were more willing to admit patients during periods of slack demand post-expansion, knowing that the ICU was less likely to reach capacity in the near future; OR physicians lowered their threshold for admitting ICU patients, suggesting that expansion may have led to inappropriate use. Physicians responses to increased capacity also depended significantly on patient acuity, indicating that physicians differentiate more between patient types when making decisions at times of high demand. Our secondary analysis using Florida administrative data supports our main findings, where we again estimate a positive effect of ICU capacity on a patient’s probability of being admitted to the ICU.