Assessing Variation in Implantable Cardioverter Defibrillator Therapy Guideline Adherence With Physician and Hospital Patient-sharing Networks

Abstract

Associations between the hospital network measures and guideline adherence suggests new approaches to better disseminate clinical guidelines across health systems.

Physician network position and patient outcomes following implantable cardioverter defibrillator therapy

“…Using national Medicare claims, physician networks can be inclusive of the entire nation or bounded using previously developed algorithms to assign physicians to “groups” (eg, hospitals or hospital referral regions) . Degree centrality is a measure of the number of ties a physician has to other physicians within the network boundary.…”

“…The ICD Registry was used to determine the following demographic and disease severity risk‐adjustment variables: patient age, sex, race, ventricular tachycardia, New York Heart Association symptom class, left ventricular ejection fraction during admission, ischemic heart disease, nonischemic dilated cardiomyopathy, heart failure, duration of symptoms since initial onset, prior myocardial infarction, previous ICD therapy, coronary artery bypass grafting during admission or previously, percutaneous coronary intervention during admission or previously, and family history of sudden death. Additional patient‐level covariates were whether the patient met ICD therapy clinical guidelines, socioeconomic indicators of patient ZIP code, and travel distance between patient ZIP code and ICD hospital. Travel distances from each ZIP code to each regional hospital (ie, American Hospital Association (AHA) hospital centroid) were estimated with ArcGIS software using U.S. street‐level road network geographic data between all 2010 ZIP code‐weighted centroids (origins) and all 2010 AHA hospitals (destinations) within an 8‐hour (480‐minute) cutoff time.…”

“…The hospital‐level HHI was calculated as previously described and represents local market competition normalized so that 1 indicates a monopoly and values approaching zero indicate a perfectly competitive market . To adjust for ICD hospital network prominence, we aggregated the physician ties by hospital affiliation to calculate a hospital‐level degree centrality measure equal to the number of ties a hospital has to other hospitals in the nation …”

“…Using national Medicare claims, physician networks can be inclusive of the entire nation 19 or bounded using previously developed algorithms to assign physicians to "groups" (eg, hospitals or hospital referral regions). 15,20 Degree centrality is a measure of the number of ties a physician has to other physicians within the network boundary. A mathematically appealing feature of degree centrality is that it perfectly partitions into the sum of the measure evaluated for a subnetwork and on the complement of the subnetwork.…”

“…The ICD Registry was used to determine the following demographic and disease severity risk-adjustment variables: patient age, sex, race, ventricular tachycardia, New York Heart Association symptom class, left ventricular ejection fraction during admission, ischemic heart disease, nonischemic dilated cardiomyopathy, heart failure, duration of symptoms since initial onset, prior myocardial infarction, previous ICD therapy, coronary artery bypass grafting during admission or previously, percutaneous coronary intervention during admission or previously, and family history of sudden death. Additional patientlevel covariates were whether the patient met ICD therapy clinical guidelines, 15 <0.001…”