Biomarkers for the prediction of cardiac readmission

The Affordable Care Act of 2010 identified hospital 30-day readmissions and mortality to reduce reimbursement and create incentives for better hospital care [1]. Unplanned readmissions account for 17% of all Medicare reimbursement to hospitals, with annual estimated cost of $17 billion dollars [2]. 20% of Medicare fee-for-service discharges result in an unplanned readmission within 30 days [3]. On average, each readmission costs hospitals $9.923 [4]. The solutions will likely take the form of bundled payments for procedures and index hospitalizations with no reimbursement for readmissions. However, there has been poor adoption of readiness for discharge protocols, follow-up with primary care, continuity of care and patient safety.

Hospitals seek to maximize reimbursement by developing systems to reduce readmissions. Hospitals are developing protocols to improve readiness for discharge and establish routine follow-up protocols for continuity of care with primary care physicians. To do so prediction models must be developed to aid clinical care teams in discharge readiness and allocate resources for continuity of care for patients at highest risk of readmission. Patients readmitted are at significant increased risk of death, thus we propose the combined end point of readmission or death [5]. However, there is extremely limited understanding of how to predict readmission and death and whether or not objective measures such as single or multiple biomarkers could improve our prediction of 30-day readmission or mortality. Economically improving prediction of 30-day readmission and death with biomarkers, hospitals can reduce spending on transitional care. Improved risk prediction could target more resources to higher risk patients and prevent readmissions now costing hospitals on average $9.923 [3].

Current readmission models using clinical factors from registries and claims data predict readmission poorly [5–11]. However, we know the primary causes of readmission are pleural effusions, infection, heart failure and angina in adult cardiac surgery [12]. Pleural effusions, infection and cardiac-related problems are primary factors for readmission in congenital heart surgery [13]. We also have clinically available and novel biomarkers as early signals for identifying increased risk for causes of readmission and mortality. Externally validated risk tools for readmission and mortality could have large benefits to routine clinical practice, but to date have not been developed.

The performance of clinical and novel biomarkers must be evaluated for prediction of 30-day end points, including readmission and death. Readmission and death are two relevant end points in the current era of cardiac surgery. To date, there has been no evaluation of clinical or novel biomarkers in adult or pediatric surgery for readmission or death. ST2 [14], galectin-3′ [15], B-type natriuretic peptide (BNP) [16–18] and cardiac troponins [19,20] have been evaluated for mortality, but not readmission. For biomarkers to be leveraged as useful clinical tools for risk assessment, we must evaluate their effectiveness in predicting hard clinical end points such as readmission and death.

Novel biomarkers (ST2, galectin-3 and cystatin C) and established biomarkers (BNP, cardiac troponin T and IL-6) that are linked to the primary causes of readmission (pleural effusions, infection, heart failure and angina), may have prognostic utility related to 30-day readmission or death [12]. ST2 and galectin-3 are new cardiovascular biomarkers characteristics of ventricular remodeling and have the potential to be proxy markers for patients readmitted for heart failure, inflammation and infection [21–24]. Recent evidence reported ST2 independent of other novel markers predicted death, heart failure and cardiovascular outcomes in the Framingham Heart Study and significantly improved the C-statistic and net reclassification index [25]. Galectin-3, produced by activated cardiac macrophages, may be directly involved in impairing cardiac function and predictive of readmission and death [26]. Novel and clinically available biomarkers are likely significant predictors of readmission and mortality and have the potential to improve risk models [21–25].

While existing models to predict the occurrence of 30-day readmission with clinical characteristics has had limited success, few have examined the application of less conventional risk factors for readmission such as serum sodium and discharge hemoglobin. One frequent adverse outcome in open heart surgery patients is acute kidney injury (AKI), which has been directly associated with short- and long-term mortality after cardiac surgery. Additionally, AKI has been correlated with readmissions in patients hospitalized with acute myocardial infarction or heart failure.

We evaluated AKIN staging criteria as a marker for identifying early risk of 30-day readmission after cardiac surgery. Based on evidence associating AKI with end-organ damage and readmission, we explored the relationship of postcardiac surgery AKI stage (no AKI, AKI stage 1, stage 2 and stage 3) and the association with 5-year readmission. We evaluated patients undergoing coronary artery bypass grafting (CABG) surgery or valve surgery at any of hospitals providing readmission data to the Northern New England Cardiovascular Disease Study Group. AKI severity was defined by the last serum creatinine prior to cardiac surgery and highest postoperative serum creatinine prior to discharge [27].

We found that AKIN stage 1 and stages 2 and 3 were significantly associated with a 29–81% increased hazard of 5-year readmission. In addition, AKIN stage 1 and stages 2 and 3 were associated with a significant threefold increased hazard of 5-year mortality [28]. Our findings demonstrate that current and novel interventions to reduce even small increases in postoperative creatinine levels may have a significant long-term impact in reducing readmission and mortality up to 5 years [28].

Hospital readmission after CABG surgery is common and frequently associated with increased mortality. Theoretical and empirical evidence indicates that biomarkers could provide additional utility in predicting readmission events, in addition to clinical and demographic variables. However, there are limited data evaluating the relationship between hospital readmission and cardiac biomarkers in CABG patients. Novel cardiac biomarkers have demonstrated to predict mortality in heart failure patients and could potentially have a predictive role in cardiac surgery [29].

We aimed to measure the association between novel biomarker levels and long-term adverse events following cardiac surgery. We analyzed the relationship between novel postoperative biomarkers ST2, galectin-3 and NT-proBNP levels and hospital readmission or death 1 year after isolated CABG surgery.

In CABG surgery patients, we found that postoperative Gal-3 and NT-proBNP levels are associated with readmission and/or mortality at 1 year, independent of established risk factors. Above-median levels of postoperative galectin-3 (hazard ratio: 1.40; 95% CI: 1.08–1.80; p = 0.010) and NT-proBNP (hazard ratio: 1.42; 95% CI: 1.07–1.87; p = 0.014) were each significantly associated with 1-year readmission or mortality [29].

Our findings suggest postoperative Gal-3 and NT-proBNP provide additional clinical utility in identifying CABG patients at greater risk of readmission or mortality at 1 year after discharge.

The addition of biomarkers to risk modeling is hypothesized to provide substantial value through personalized medicine and improved patient care management. Our recent research has evaluated risk model improvement with the addition of biomarker panels to identify patients at high risk of readmission or mortality after discharge following CABG surgery. We evaluated the relationships of novel biomarkers – ST2, galectin-3, N-terminal probrain natriuretic peptide, cystatin-C, IL-6 and IL-10 – with both 30-day readmission and 30-day combined readmission or death [30]. In this analysis, we applied a biomarker panel to an augmented Society of Thoracic Surgeons preoperative 30-day readmission model.

Our findings demonstrate that biomarker panels have the ability to substantially improve risk prediction for 30-day readmission or mortality following discharge after CABG surgery. The addition of novel and clinically available biomarkers resulted in significant improvement in the predictive model performance area under the receiver operating characteristic curve (AUROC): 0.74; 95% CI: 0.69–0.79; p < 0.001). External validation in an international cardiac surgery cohort showed minimal improvement with the addition of biomarker panels. Furthermore, we evaluated the application of preoperative biomarker panel including cardiac troponin T, N-terminal probrain natriuretic peptide, high-sensitivity c-reactive protein and blood glucose to improve the prediction of in-hospital mortality. This panel demonstrated a moderate improvement in discrimination of mortality from an AUROC of 0.83 (95% CI: 0.74–0.92) to 0.87 (95% CI: 0.80–0.94) with biomarkers (p = 0.09) [30]. We observed that the augmented Society of Thoracic Surgeons clinical model accounted for 32% of the predictability of the model, whereas the biomarkers contributed 68% [30].

Our innovative research supports the addition of biomarkers to risk modeling to improve prediction of readmission and mortality. Beyond those already evaluated, there are other promising biomarkers that may have prognostic utility with readmission and mortality. To accomplish this, we should bring together collaborations of biorepositories in cardiac surgery and other conditions to analyze new risk marker candidates for predicting readmission and mortality, with the addition of external validation. We recommend surgical teams include model clinical variables and biomarkers available to them to identify the risk of readmission or death for each patient prior to discharge. Improved readmission risk models can establish a transitional care plan that is tailored to each individual patient to mitigate major adverse events and readmission [30].

In summary, the novel method of integrating biomarkers into risk models can significantly improve risk discrimination. These findings should stimulate future research in the timing and combinations of biomarkers, as well as the models themselves for improving risk prediction in cardiac surgery.