Methods
Data source
All data were obtained from the National Inpatient Sample (NIS), a publicly available, nationally representative database administered by the United States Agency for Healthcare Research and Quality (AHRQ) Healthcare Cost and Utilization Project (HCUP) [
17]. Data was utilized from 2003 to 2009. The NIS is a 20% stratified sample of discharges from about 1000 hospitals. For the years utilized for this analysis, all discharges from sampled hospitals were included. Each hospitalization in the NIS includes all reported diagnosis and procedure codes based on the
International Classifications of Diseases, Ninth Revision, Clinical Modification (
ICD-9-CM). Additional patient-level data included age, sex, and race. Hospital-level data included bed size, location (urban/rural), ownership, and teaching status. Prior to 2004, hospitals in a Metropolitan Statistical Area (MSA) were considered urban, while those outside an MSA were considered rural. Beginning with 2004, hospitals with a Core Based Statistical Area (CBSA) of Metropolitan or Division were considered urban, while those with a CBSA of Micropolitan or Rural were considered rural. Hospitals were considering teaching if they met one of the following criteria: 1) had a residency training approval by the Accreditation Council for Graduate Medical Education, 2) had membership in the Council of Teaching Hospitals, or 3) had a ratio of full-time equivalent interns and residents to beds of 0.25 or higher [
17].
The Hospital Market Structure (HMS) file was linked to the NIS database for information on hospital market structure. The HMS data are updated every 3 years and last were updated in 2009 (state and hospital identifiers required for this supplemental file were removed from the NIS in 2012 to enhance confidentiality). Hospital market competition was assessed using the Herfindahl-Hirschman Index (HHI). The HHI is sum of squared market shares for all hospitals in the market. Market share is calculated by the number of discharges for a hospital divided by the total discharges from all hospitals in a market [
18]. The market was defined using the variable radius market definition. For each hospital, the distance between the hospital and each patient’s ZIP code was calculated. ZIP codes were then ranked in ascending order according to distance and then were aggregated until 75% of the hospital’s discharges were attained. The distance between the hospital and the last ZIP code was used to determine the variable radius used to define that hospital’s market. Sensitivity analysis was conducted using variable radius of 90% of patient discharges to determine any effect of varying definitions of a hospital’s market [
17]. Standard definitions of HHI determined by the Department of Justice and Federal Trade Commission to evaluate horizontal mergers and used in prior research were utilized: greater than 2500 was considered highly concentrated; 1500 to 2500 was considered moderately concentrated; and less than 1500 was considered unconcentrated or competitive [
7,
19]. (See Figure S
1)
Institutional review board approval and informed consent was not required since data was derived from a deidentified administrative database.
Study population
From 2003 to 2009 all hospitalizations for patients with the diagnosis of cardiogenic shock (
ICD-9-CM 785.31) and AMI (
ICD-9-CM 410.11–410.91) in any position were queried from the NIS database [
20]. Hospitals performing less than 10 percutaneous coronary interventions (PCI) per year were excluded from analysis in order to select for hospitals with cardiac catheterization laboratories. Hospitals unable to perform PCI were assumed not to have the capacity to place MCS devices. Patients were classified based on demographics such as age, sex, and race. Comorbidities were identified using 29 Elixhauser clinical and procedural variables defined by the admitting ICD-9 codes. Administrative codes were used to determine if patients underwent coronary angiography, right heart catheterization (RHC), PCI, or coronary artery bypass grafting (CABG) during index hospitalization (see Table S
1).
Study exposures and outcomes
The primary exposure of interest was HHI (as described above). For our primary analysis, variable radius of 75% of hospital discharges were used. Sensitivity analysis was performed with variable radius of 90%. Patient-level covariates included age, sex, race, comorbidities, and receipt of procedures (coronary angiography, PCI, or CABG) as noted above. Hospital-level covariates included bed size, teaching status, location, and hospital control.
The primary outcome variable was the use of MCS. MCS devices included percutaneous devices (
ICD-9-CM 37.68), non-percutaneous devices (
ICD-9-CM 37.60 and 37.65), intra-aortic balloon pumps (IABP) (
ICD-9-CM 37.61), and extracorporeal membrane oxygenation (ECMO) (
ICD-9-CM 39.65 and 39.66). Percutaneous devices include devices such as Impella (Abiomed Inc., Danvers, MA) and TandemHeart (CardiacAssist, Pittsburgh, PA), while nonpercutaneous devices include Thoratec paracorporeal ventricular assist devices (Thoratec Inc., Pleasanton, CA), AB5000 (Abiomed, Inc., Danvers, MA), BVS5000 (Abiomed, Inc., Danvers, MA), and Centrimag (Thoratec Inc., Pleasanton, CA) [
12]. The secondary outcome variable was in-hospital mortality which was captured through the NIS database.
Statistical analysis
Patient-level characteristics such as age, sex, race, and comorbidities were compared across levels of HHI with 1-way ANOVA for continuous variables and chi-square tests for categorical variables. Admission data, including rates of coronary angiography, RHC, PCI, and CABG as well hospital characteristics were also compared across levels of HHI using chi-square tests. Receipt of MCS and in-hospital mortality were compared across levels of HHI using chi-square tests. A multivariable logistic mixed effect model was used to model the receipt of MCS according to patient, admission, and hospital characteristics, including HHI. Hospital was included as a random effect to account for clustering of hospitalizations by hospital. Given change in treatment approaches for CS during the study time period, analysis was re-conducted by year (2003, 2006, and 2009 chosen to align with HMS updates) to evaluate temporal trends. All analyses were performed with R version 3.6.3.
Discussion
Our study demonstrated, that in addition to patient and hospital characteristics, regional market forces are associated with treatment decisions for patients admitted with AMI-CS. We showed that during our study period (2003–2009), in unadjusted analysis, patients in unconcentrated (or competitive) markets were more likely to receive MCS when admitted with AMI-CS than those in concentrated markets. In mixed-effect multivariable regression analysis, adjusting for patient and hospital factors, our study found that in more contemporary years (2009), market concentration was a significant factor in receipt of MCS.
Prior research has suggested that hospital market forces are associated with treatment decisions, especially with regard to diffusion of new technologies [
7,
21,
22]. Studies have shown that in unconcentrated markets, patients are more likely to undergo laparoscopic colectomy (compared to open) for colon cancer [
21], more likely to undergo endovascular repair (compared to open) for abdominal aortic aneurysm [
22], and more likely to undergo robotic-assisted surgery for genitourinary and gynecologic surgery [
7]. Notably, these studies controlled for both patient and hospital characteristics, suggesting a true effect from market concentration. Wright et al. showed that patients in unconcentrated markets were more likely to undergo robotic-assisted surgery. However, once analysis was restricted to hospitals that already performed robotic-assisted surgery, there was no association. This suggests that while market forces may influence decision to acquire technology—such as a surgical robot—it may not affect patient care decisions once the technology is acquired.
AMI-CS is an emergent condition requiring timely management. We hypothesized that hospitals in unconcentrated markets will have increased investment in MCS and be associated with higher utilization of MCS. Our study was conducted using data during a time of evolving management of AMI-CS with increasing use of percutaneous MCS and showed that overall, hospital market concentration did not have any effect on the receipt of MCS in the earlier years of the study period. However, in the later years of the study, while we saw an increase in utilization of MCS among all hospitals, patients in unconcentrated markets became more likely to receive MCS compared to those in concentrated ones. This effect was seen regardless of other hospital characteristics such as location, bed-size, and teaching status. Abiomed obtained 510(k) clearance from FDA for the Impella percutaneous device (Abiomed Inc., Danvers, MA) in May 2008. NIS data shows a doubling of use of percutaneous devices in patients with AMI-CS between 2008 and 2009 which may be driving early adoption in more competitive market (See Fig. S
2).
Health care policy over the past two decades encourages consolidation of markets—leading to increased market concentration—with the formation of accountable care organizations and use of bundled payments [
4]. While the goal of these policies has been improvement in coordination of care, there are likely unintended effects on costs of care and patient outcomes [
23]. One of these effects may be a decreased diffusion of new technologies, both proven (but underutilized) and unproven (and yet to demonstrate benefit). Expanded uses of proven technologies may improve patient outcomes, while increased use of unproven technologies may increase healthcare costs, without an effect on patient outcomes [
24]. Our study is the first, that we are aware, to evaluate these effects of hospital market concentration on use of cardiovascular technology. These findings may be useful in informing health care policy and federal antitrust efforts to mitigate negative effects of hospital consolidation.
There are a number of strengths to this study. First, the NIS is a nationwide database that is representative of the United States population. We were able to capture patients with cardiogenic shock in a crucial time period during national trends in increased market concentration and increased use of mechanical circulatory support. However, our study also has a number of limitations. First, while we were able to capture a unique period in the evolution of mechanical circulatory support, analysis of more contemporary data outside the NIS (which does not provide hospital market level data after 2009) is needed to analyze more recent effects. These recent effects will be better suited to evaluate the effect of market concentration once technologies (such as Impella) are more broadly used. Second, we used an administrative database and were unable to account for potential unmeasured confounders, including severity of presenting illness, medications, and socioeconomic factors. It is possible that there may be bias toward sicker patients in competitive markets that may not be realized in administrative data and might confound the effects of market concentration. Third, the NIS is limited to a sampling of only 20% of hospitals in the United States. Random sampling may have affected our results if hospitals selected in a particular market were biased toward more or less frequent MCS use compared to other representative hospitals in that market. Our study, however, captured over 800 hospitals and controlled for hospital factor such as bed size, teaching status, and location while likely minimized any effect of random sampling. Fourth, there is variation in the literature in identifying patients with AMI-CS from administrative data. We used a method from prior literature [
16,
20], that uses encounters that have
ICD-9-CM codes for cardiogenic shock or myocardial infarction in any position in the NIS databases. This method increases sensitivity for capturing cases of AMI-CS, but may capture cases of cardiogenic shock that are not due to myocardial infarction. Other literature, limits encounters to those that have myocardial infarction as the primary diagnosis [
8,
25]. Finally, HHI was used as an indicator for market concentration. While HHI is a validated metric to evaluate market concentration, it may not fully capture other important market factors, including hospital affiliations.
Open AccessThis article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit
http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (
http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.