Abstract
Risk-averse behavior has been reported among physicians and facilities treating cardiogenic shock in states with public reporting. Our objective was to evaluate if public reporting leads to a lower use of mechanical circulatory support in cardiogenic shock. We conducted a retrospective study with the use of the National Inpatient Sample from 2005 to 2011. Hospitalizations of patients ≥18 years old with a diagnosis of cardiogenic shock were included. A regional comparison was performed to identify differences between reporting and nonreporting states. The main outcome of interest was the use of mechanical circulatory support. A total of 13043 hospitalizations for cardiogenic shock were identified of which 9664 occurred in reporting and 3379 in nonreporting states (age 69.9 ± 0.4 years, 56.8% men). Use of mechanical circulatory support was 32.8% in this high-risk population. Odds of receiving mechanical circulatory support were lower (OR 0.50; 95% CI 0.43–0.57; ) and in-hospital mortality higher (OR 1.19; 95% CI 1.06–1.34; ) in reporting states. Use of mechanical circulatory support was also lower in the subgroup of patients with acute myocardial infarction and cardiogenic shock in reporting states (OR 0.61; 95% CI 0.51–0.72; ). In conclusion, patients with cardiogenic shock in reporting states are less likely to receive mechanical circulatory support than patients in nonreporting states.
1. Introduction
Public reporting of outcomes was developed to improve transparency in healthcare, patient safety, and quality of care. States that have adopted some form of public reporting in cardiovascular care include Massachusetts, New Jersey, New York, Pennsylvania, and Washington [1, 2]. Unintended consequences including risk-averse behavior among physicians and facilities have emerged [3–5]. This has been particularly true for patients presenting with acute myocardial infarction (AMI) complicated by cardiogenic shock (CS) who are less likely to receive percutaneous coronary intervention (PCI) in reporting states [6, 7]. Despite advances in management, in-hospital mortality of CS remains high [8]. Efforts to reduce mortality have focused primarily on early reperfusion in patients presenting with AMI and use of mechanical circulatory support (MCS) [9, 10]. The objective of this study is to analyze whether the use of public reporting leads to a lower use of MCS in CS.
2. Materials and Methods
The study was conducted between January 2018 and May 2018 with the use of the National Inpatient Sample (NIS), Healthcare Cost and Utilization Project, Agency for Healthcare Research and Quality. The NIS is the largest publicly available database on all-payer administrative claims in the United States and has been extensively used to analyze trends in cardiovascular care [11, 12]. NIS data from January 2005 through December 2011 were included in the study. The database was redesigned in 2012 as a 20% national patient-level sample with nonrepresentative sampling across hospitals. As a result, data collected after 2011 are not representative of a particular region or state and were therefore excluded from the analysis. Hospitalizations of adult patients (age ≥18 years) with a diagnosis of CS were identified by International Classification of Diseases, Ninth Revision codes. The primary exposure of interest was care in a public reporting vs. a nonreporting state. Implementation of public reporting of outcomes for percutaneous interventions began in 1995 in New York and 2003 in Massachusetts. As such, these were considered public reporting states. A regional comparison within the Northeast region was performed with Connecticut, Maine, Maryland, Rhode Island, and Vermont serving as nonreporting states. Following previously accepted methodology, data from New Jersey and Pennsylvania were excluded from the analysis due to inconsistent reporting [13]. The main outcome of interest was the use of MCS. The secondary outcome was in-hospital mortality. The main outcome was identified by codes for intra-aortic balloon pump (IABP), percutaneous ventricular assist device including Impella (Abiomed Inc., Danvers, MA) and TandemHeart (CardiacAssist, Inc., Pittsburgh, PA), extracorporeal membrane oxygenation (ECMO), percutaneous cardiopulmonary support, and cardiopulmonary bypass (CPB).
Propensity score matching was performed given differences in baseline characteristics. Patients with the nearest propensity scores in the two groups (public reporting vs. nonreporting states) were matched using 1 : 1 scheme without replacement using the Greedy method. Matching variables included sex, race, insurance status, household income, use of PCI, ST-segment elevation myocardial infarction, and 23 NIS-reported comorbidities. Maximum propensity difference (caliper width) allowed was 0.05. Patients without matched observation were excluded. Weights were generated with NIS specified sample weights. “Proc Survey Logistic” function was used in SAS, and “SVY” function was used in STATA. Adjusted logistical regression models were subsequently created to generate odds ratios for propensity score-matched cohorts. Stata IC 12.0 (Stata-Corp) and SAS 9.3 (SAS Institute, Inc.) were used for statistical analysis.
3. Results
A total of 13043 hospitalizations for CS were included of which 9664 occurred in reporting and 3379 in nonreporting states. Baseline characteristics of patients in reporting and nonreporting states are presented in Table 1. Both groups shared a similar mean age, slight male predominance, and similar Charlson Comorbidity Index. AMI was present in 54% of cases in reporting and 56% of cases in nonreporting states. Patients in reporting states presented to large urban teaching hospitals more frequently than those in nonreporting states. Overall MCS use was 32.8% in this high-risk population. Patients in reporting states received MCS less frequently than patients in nonreporting states (31.9% vs. 35.4%; ). IABP and CPB were the most common used forms of MCS.
Table 2 presents baseline characteristics after 1 : 1 propensity score matching. MCS was used in 32.2% of cases in reporting and 36.4% of cases in nonreporting states (). After multivariate adjustment (Figure 1), MCS was used significantly less often in reporting states among all groups of patients. The effect was more pronounced among patients >65 years of age and those who did not present with AMI. Adjusted in-hospital mortality was higher in reporting states (OR 1.19; 95% CI 1.06–1.34; ). The effect was more pronounced for patients who received MCS, those older than 65 years of age, and those who did not present with AMI.
(a)
(b)
Figure 2 presents the adjusted odds ratios for MCS use and in-hospital mortality for the subgroup of patients with AMI and CS. Odds of receiving MCS were lower in patients with AMI and CS in reporting states (OR 0.61; 95% CI 0.51–0.72; ), regardless of type of myocardial infarction. There was no difference in adjusted odds ratio for in-hospital mortality in this subgroup (OR 1.01; 95% CI 0.87–1.18; ).
(a)
(b)
4. Discussion
This is the first study to analyze the impact of public reporting of outcomes on MCS utilization in patients with CS. Patients hospitalized in reporting states received MCS less frequently than patients in nonreporting states. Patients in reporting states had a higher in-hospital mortality. In the subgroup of patients with AMI and CS, MCS utilization was lower in reporting states regardless of the type of myocardial infarction, but in-hospital mortality was not significantly different.
A previous analysis of the NIS on the use of short-term MCS revealed that the use of nonpercutaneous forms of MCS increased by 101% from 2007 to 2011. Meanwhile, the use of percutaneous forms of MCS increased by 1,511% during the same period of time [14]. A more recent study of patients that received MCS for CS revealed similar increases in MCS use. Overall MCS utilization increased by 160% from 2004 to 2014. While IABP utilization continued to grow (140%), its relative use has declined compared to other forms of MCS (1421% for ECMO and 1229% for percutaneous devices) [15].
Public reporting of outcomes has been instrumental to improvements in healthcare, but concerns regarding unintended consequences remain. Risk-averse behaviors have been reported in physicians and hospitals caring for critically ill patients in states with public reporting [3, 6]. A regional analysis of the Medicare Provider Analysis and Review Files of patients with a diagnosis of AMI revealed that use of PCI was lower for patients in reporting states. The difference was more pronounced among patients with ST-segment elevation myocardial infarction, CS, and cardiac arrest [7]. An analysis of the NIS database from 2005 to 2011 revealed similar findings. In addition, in-hospital mortality was noted to be higher among patients in public reporting states [13].
Registry data have consistently suggested that public reporting may contribute to withholding of potentially life-saving interventions in critically ill patients. This has led some authors to suggest excluding the highest-risk patients from registry data. New York became the first state to exclude patients with CS and hypoxic encephalopathy from publicly reported PCI outcomes. Other states like Massachusetts offer risk-adjustment modifiers as an attempt to mitigate risk aversion [16]. However, risk-averse behavior has been reported even after changes in policy. A study of hospitalizations for AMI complicated by CS from 2002 to 2012 revealed that the exclusion of CS from public reporting in New York was followed by an increase in the use of PCI for patients with CS and a decrease in mortality. Nevertheless, PCI use remained low when compared to nonreporting states even after the change in policy [17]. One possible explanation is that patients with CS can be adjudicated post hoc as not “refractory” shock, which in turn places the provider at risk. Lower MCS utilization in public reporting states was noted in our study, presumably due to procedural risk aversion in this high-risk population. One possible explanation for the lower use of MCS in reporting states is that public reporting nudges physicians and institutions towards a more “conservative” approach in this very high-risk population. However, this hypothesis warrants further investigation.
Different strategies to reduce the impact of public reporting in the care for patients with CS have been proposed. A shift from procedure-based to disease-based outcomes reporting has been recommended to allow for accurate public reporting while reducing procedural risk aversion [16]. In addition, public reporting of institutional rather than individual operator outcomes may present a better option in patients presenting with CS or out-of-hospital cardiac arrest. Development of institutional protocols emphasizing early invasive hemodynamic monitoring and rapid initiation of MCS in patients presenting with AMI complicated by CS has resulted in promising initial results [9]. Institutions with systems in place to offer advanced therapies and with a proven record of improved outcomes could receive such patients preferentially which might ultimately enable the long-term goal of lowering mortality and improving outcomes in this very high-risk population.
4.1. Limitations
The NIS sampling design is statistically sound and has been used to estimate national healthcare trends. Strengths of the study include the heterogeneity of the population studied and large case volume. Nevertheless, some limitations are worth noting. First, given the code-based nature of the NIS database, the chance for miscoding or overcoding is present. Second, due to changes in the NIS design, data collected after 2011 are not representative of a particular state or region and as such were not included in the analysis. Risk-averse behavior has been noted to persist even after major policy changes, and thus we believe that the trends described here are likely to persist.
5. Conclusion
In conclusion, patients hospitalized for CS in public reporting states receive MCS less frequently than patients in nonreporting states. Further research is needed to optimize current reporting models and mitigate some of its unintended consequences.
Data Availability
Data used for this study were extracted from the publicly available National Inpatient Sample from the Healthcare Cost and Utilization Project.
Conflicts of Interest
There are no conflicts of interest to disclose.
Authors’ Contributions
Vikas Singh and Rodrigo Mendirichaga contributed equally to this manuscript.
Acknowledgments
The publication of this study was supported by a generous grant from the Mary K. Oxley Foundation.