Outcomes after readmission at the index or nonindex hospital following acute myocardial infarction complicated by cardiogenic shock

Little is known about the prevalence and outcomes of readmission to nonindex hospitals after an admission for acute myocardial infarction complicated by cardiogenic shock (AMI‐CS). We aimed to determine the rate of nonindex readmissions following AMI‐CS and to evaluate its association with clinical factors, hospitalization cost, length of stay (LOS), and in‐hospital mortality rates.


| INTRODUCTION
Acute myocardial infarction complicated by cardiogenic shock (AMI-CS) is associated with respiratory failure, hemodynamic failure, and electrical instability, which may lead to hypoperfusion and organ failure. 1 AMI-CS was once a very deadly disease, with a survival rate of 40-44%, 2,3 however, in the past decade inpatient mortality has decreased significantly due to the use of reperfusion strategies and mechanical innovations in the treatment of AMI. 4,5 As survival rates have increased, improving post-hospital outcomes is essential for patients with AMI-CS. Rashmee et al. 6 reported that patients with AMI-CS have a high likelihood of adverse outcomes in the early stages after being discharged from the hospital.
With the escalating costs of healthcare, readmission has become an important health policy topic due to its relevance to the quality of medical services. Readmissions place a huge burden on medical resources and increase healthcare costs. 7 Of the survivors of AMI-CS who are discharged from the hospital, almost 20% are readmitted within 30 days. 8 The Centers for Medicare & Medicaid Services (CMS) in the United States has set a quality measure to reduce the 30-day readmission of patients with AMI. 9 One unintended consequence of advances in treatment technology is that many patients are readmitted to hospitals other than the one where they initially received treatment, potentially fragmenting follow-up care. Previous studies show that readmissions to a nonindex hospital by patients with acute stroke 10 and by patients who underwent percutaneous coronary intervention (PCI) 11 are associated with poorer outcomes. Similar results for patients undergoing different types of surgery have also been reported. [12][13][14] Little is known about the prevalence and outcomes of readmission to a nonindex hospital following in patients diagnosed with AMI-CS. Using the National Readmission Database (NRD) a nationally representative assessment of nonindex readmissions can be made. In this study, we determine the rate of readmission to nonindex hospitals following AMI-CS and evaluate the association of clinical factors, hospitalization cost, length of stay (LOS), and in-hospital mortality rates with nonindex readmission.

| Data source
The NRD is part of a set of databases and software tools developed for the Healthcare Cost and Utilization Project (HCUP). It is a unique and powerful database that supports various types of analyses on national readmission rates for all payers and for the uninsured. It estimates roughly 36 million discharges every year from more than 20 states in the United States. 15 The NRD is publicly available, so this study was exempt from formal institutional review board approval, and informed consent is not required.

| Study population
We used the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) diagnosis codes or ICD-10-CM diagnosis codes to capture the data of patients with a primary admission for AMI (ICD-9:410; ICD-10:I21) who had an additional diagnosis of CS (ICD-9:758.14, ICD-10: R57.0). These codes have a specificity of 99.5% and a sensitivity of 72.4% for AMI, 16 and a specificity of 99.3%, and a sensitivity of 60% for CS. 17 We used the NRD variable "HOSP_NRD" to identify whether a patient was readmitted to the index hospital. The exclusion criteria included: (1) patients younger than 18 years, (2) patients who died during the index hospitalization, (3) patients who were discharged in December (as NRD does not provide cross-year follow-up data), and (4) elective readmissions.

| Covariate assessment
We used the NRD variables to identify demographic characteristics such as age, sex, payer information, and income by postal code; and the hospital characteristics including the ownership of the hospital, the number of beds, and the location or teaching status. We used the Elixhauser Comorbidity Index (ECI) to account for the burden of 29 common comorbidities. 18 Comorbidities, in-hospital procedures, and in-hospital complications were identified using ICD-9-CM and ICD-10-CM diagnostic or procedural codes (Table S1). 6-8

| Primary and secondary outcomes
Hospitalization cost, LOS, and in-hospital mortality rate during unplanned readmissions within 30 days of the index admission were the primary study outcomes. The hospitalization cost was recalculated by cost-to-charge ratios, which was provided by HCUP. The secondary outcomes included the temporal trend in the prevalence of nonindex readmissions, in-hospital outcomes during the 60 and 90 days readmission periods, predictors of readmission to a nonindex hospital, and the specific causes of readmission. The readmission causes were identified using Clinical Classification Software according to a previous study. 19

| Statistical analysis
We used the survey analytical methods recommended by the HCUP for national estimates. 15 Figure S1). Table 1 compares the patient characteristics recorded during index admissions between index and nonindex readmissions. Patients of nonindex readmissions were more likely to be older and residents of same state as the admitting hospital, and they had higher rates of a history of PCI, prior coronary artery bypass graft (CABG), and a history of stroke, as well as higher ECI scores. These patients were also less likely to receive PCI and CABG during the index hospitalization. When we compared the characteristics of 30-day readmissions (Table S2), we found that patients of nonindex readmissions had higher ECI scores, and were more likely to go to private investor-owned (proprietary) hospitals, smaller hospitals, urban nonteaching hospitals, or hospitals located at large metropolitan

| Reasons for 30 day unplanned readmissions
Overall, the reasons of 30 day unplanned readmissions were similar between patients who rehospitalized to the index hospital and those who went to a different hospital. Table S3 shows the most common reasons for readmission to either index or nonindex hospitals.  Note: All models were adjusting for demographic characteristics, hospital characteristics, comorbidities, in-hospital procedures, and in-hospital complications at the time of index admission. Abbreviations: AMI, acute myocardial infarction; CAD, coronary artery disease; CI, confidence interval; PCI, percutaneous coronary intervention.

| DISCUSSION
When CMS focus on 30 days readmission and reduce Medicare payments for these patients, hospitals reduce excess readmissions. 20 However, hospital administrators can only identify readmissions to their own hospital, which may underestimate true readmission rates. 21 Using the NRD, we can identify the index and nonindex readmission rates. Readmission within 30 days is common for patients who survive AMI-CS. 6,7 In this study, approximately a fourth of those survivors were readmitted to a nonindex hospital.
Our study demonstrates an association between rehospitalization to a nonindex hospital and the less favorable outcomes for patients with AMI-CS. We found evidence of increased hospitalization costs, longer LOS, and increased in-hospital mortality rates associated with readmission to nonindex hospitals.
As many patients with AMI-CS survive their initial hospitalization, post-discharge outcomes are important. More than half of survivors are readmitted or die within 1 year of their index admission, and patients with CS have a higher mortality rate than those without CS. 6 In the early post-discharge period, patients with CS have a greater risk of poor outcomes than patients without CS. 6  When patients are readmitted to a local hospital without a cardiovascular specialist, the available provider may not have experience managing the patients' complications. 10 Fragmented care and repeated testing can lead to higher medical costs, but receiving treatment from a different doctor, especially one who specializes in cardiovascular disease, can lead to better outcomes for patients. 29 Patients who receive treatment from more than physician benefit from fewer missed diagnoses and medical errors. Followup care is also an important aspect of patient well-being. Dickinson et al. reported that more than 70% of unplanned readmissions were related to complications and were preventable, which suggests that improvement of follow-up care can reduce the risk of readmission and post-discharge mortality. 30 Furthermore, sharing electronic health records and having patient navigators may mitigate or offset some of the negative consequences of medical interruptions. 25 A better understanding of readmission patterns is needed to help reduce unnecessary readmissions, and could be accomplished by the development of a national database of demographic, clinical, and administrative data from different hospitals. 31 This study has several limitations. First, due to the use of administrative data, misclassifications or residual confounding may bias our results. Second, the NRD does not contain some important clinical information, such as medications or physiological data. Although we adjusted for several factors (patient characteristics and hospital characteristics) and performed some sensitivity analyses, some unknown bias is still possible. Third, only readmissions within the same state are calculated by the NRD. Thus, patients readmitted to hospitals in other states were not included in the dataset. In addition, we were unable to determine the impact or rate of inter-hospital transfers of AMI-CS patients readmitted to a nonindex hospital.

| CONCLUSIONS
This study highlights the readmission burden after AMI-CS. We found that readmission to a nonindex hospital occurs for more than a quarter of survivors of AMI-CS, and is associated with higher hospitalization costs, longer LOS, and higher in-hospital mortality rates.