Factors associated with adverse outcomes from cardiovascular events in the kidney transplant population: an analysis of national discharge data, hospital characteristics, and process measures

Cardiovascular events are the leading cause of death after kidney transplantation (KT). Significant amounts of research have been aimed at reducing event rates, primarily aimed at understanding prevalent risk factors, defining outcomes, and application of guideline-based care [1‐4]. Event rates continue to be high and endanger long-term patient and transplant outcomes.

Post-KT cardiovascular event are among the most important drivers of post-kidney transplant health care utilization and mortality [5]. KT recipients have high rates of hospitalization for myocardial infarction (MI), congestive heart failure (CHF), dysrhythmias, stroke (CVA), malignant hypertension, and cardiac arrest. Mortality is as high as to 20% in some hospitals. Few studies have focused on the rescue of KT patients once these events occur [6]. Patient and hospital factors may contribute to adverse outcomes from CVD events. Hospitals are known to vary in cardiac care practices [7‐11], and structural features including teaching status, technology, and staffing patterns are associated with better outcomes [12]. KT patients bring an even greater challenge in this setting – rescue from an acute cardiovascular event requires facility resources and well-developed care processes, which can be leveraged from transplant programs. The presence of these resources may improve outcomes and reduce costs of cardiovascular care, but this idea remains unexplored.

In this analysis, we aimed to understand how hospitals perform in the management of cardiovascular disease in kidney transplant patients. We modeled hospital characteristics including structural factors and cardiovascular process measures as well as clinical factors to identify predictors of inpatient mortality and costs [13]. We hypothesized that transplant hospitals (TH) would have lower mortality and costs compared to non-transplant hospitals (NTH), after adjustment for their inherent characteristics and patient differences.

The final analysis sample consisted of 7803 hospital admissions from 275 hospitals from 2009 to 2011. Among the 275 hospitals, 28% (n = 78) were kidney transplant facilities and 72% (n = 197) were non-transplant facilities. Cardiovascular hospitalizations in the KT population were evenly distributed between NTHs (n = 3893, 49.8%) and THs (n = 3910, 50.1%). CHF and dysrhythmia were the leading causes of admission. The descriptive statistics for patient and hospital characteristics are shown in Table 1.

There were significant population differences in THs versus NTHs (Table 1). NTHs had more white patients. Multiple CVDs were coded in 24.3% of cases, more often at NTHs. NTHs had a significantly higher proportion of MI, CHF, and dysrhythmia admissions, while THs had more stroke, cardiac arrest, and malignant hypertension admissions. NTHs admitted significantly more patients with a high co-morbidity burden by Charlson score. Diabetes mellitus was commonly coded in hospitalizations at both types of facilities, and was present in the majority of admissions (> 50%). NTHs had a significantly greater proportion of emergent/urgent admissions. NTHs demonstrated longer lengths of stay versus THs.

Facility characteristics differed between TH and NTHs. Cardiac intensive care was significantly more prevalent in THs vs. NTHs, but there was similar prevalence of specialty cardiac services in both hospital types. THs had significantly more surgical volume and daily occupancy. THs had significantly more technology (presence of 64-slice CT scanners and interventional radiology therapy). Case-mix was significantly different between the two hospital types, with THs had more Medicaid patients and NTHs had more Medicare. THs had higher total expenses, staffing ratios, and major teaching efforts compared to their counterparts.

Four cardiovascular process of care measures were available for analysis: time to receipt of fibrinolytic therapy, time to receipt of electrocardiogram, receipt of aspirin on hospital admission, and time to transfer to another facility for acute coronary intervention. THs and NTHs were similar in these, but these were not fully reported during the study period.

Table 2 demonstrates differences in mortality and proportion of high cost admissions stratified by primary cardiovascular diagnosis. THs and NTHs had similar rates of mortality and high cost admissions in MI, stroke, and dysrhythmia. Mortality from CHF was significantly higher in NTHs compared to THs, but had a similar proportion of high cost admissions. For patients admitted in cardiac arrest, mortality was similar, between 31 to 34%, but 73.7% of those admissions were high cost in THs compared to 57.3% in NTHs.

We have previously identified two important trends in health care utilization for cardiovascular disease in the transplant population, which fueled our interest in this study. First, utilization of hospital services for cardiovascular disease in the kidney transplant recipients is growing, particularly in non-transplant hospitals and secondly, there was a trend toward higher mortality in these hospitals [5]. By studying a large database of hospitalization episodes and linking it to granular data on hospital characteristics, we were able to design models to identify clinical risk factors and hospital characteristics predictive of adverse clinical and financial outcomes.

An important early finding in the analysis was a concerning trend toward higher inpatient cardiovascular mortality in NTHs. After adjusting for clinical and facility characteristics, THs and NTHs had similar mortality. Clinical factors largely mediated this difference. From the group of facility factors, only teaching status was associated with lower mortality, which was recently also observed by Silber et al. in a Medicare study on MI patients [12]. Predictors of mortality and HCH included age and cardiovascular disease burden (based on the number of coded cardiovascular diagnoses) and utilization of diagnostic cardiac catheterization. Machinicki et al. has previously shown that pre-existing cardiovascular disease burden can reliably predict Medicare mortality and costs in the 3 years following transplantation [20]. Cardiovascular procedures were associated with a lower risk of mortality and higher costs compared to non-procedural admissions, likely related to resources utilized and patient selection in these admissions versus others.

An interesting finding was related to costs of care. While NTHs had longer lengths of stay for the same diagnoses, other significant facility factors were associated with lower costs: higher Medicare payer-mix, lower physician staffing, and TH status. This may imply THs provide better value in managing CVD complications, considering THs and NTHs had similar odds of population-based mortality. Why would this be the case? THs are resource-intense facilities and typically carry significant resources and expertise. This may translate into better value by reducing unnecessary testing or care intensity [21], and could be related to patients being in their “transplant home” where they are a known entity. Practice patterns, in this context, likely differ between THs and NTHs and drive observed differences in HCHs. This finding is novel, and generates a hypothesis that warrants further analysis within specific diagnoses, and potentially with richer clinical data.

This study has direct implications for clinical practice and care models aimed at rescuing post-transplant patients in high-risk cardiovascular scenarios. Prevention of cardiovascular events is key. These events are increasingly recognized and inpatient mortality exceeds 3% [8‐11, 14, 20, 22‐24]. Risk factor modification should be a central tenet of post-transplant care. Increasing access to preventive cardiology, alterations in immunosuppression, adherence to cardio-protective medication regimens, and application of guideline-based cardiovascular medical therapy may improve outcomes in the post-transplant population [25, 26]. Secondly, our analysis suggests that certain clinical phenotypes are at high risk for mortality – older kidney transplant patients with multiple cardiovascular diagnoses who require invasive interventions. As observed here, transplant patients pursue complex care in all types of facilities – our study indicates that NTHs are associated with reasonable outcomes. This represents a shift from earlier years of clinical transplantation, likely related to greater prevalence of transplant patients in the community and the proliferation of well-resourced hospitals around the country. This shift also warrants the development of formal and informal care networks within communities to manage these patients. Non-transplant providers/facilities take on significant risk with these patients, and transplant providers/hospitals should support them. Formal and informal partnerships underscored by clear inter-facility communication are vital. The development of these networks requires earnest collaboration, and both transplant and non-transplant hospitals have the incentives to do so. Further research evaluating the effectiveness of these networks would be an interesting innovation in studying transplant health services.

This analysis has limitations. Since each record in the data represented a single unlinked hospitalization, the timing of the transplant relative to the cardiovascular event is unknown. Linkage of admissions would have enriched the observations in this analysis, and may have helped elucidate potential interventions for future studies. Administrative data inherently lack clinical granularity which limits our ability to see the true biological effects of documented co-morbidities on in-hospital mortality and costs, such as diabetes and dyslipidemia. The Donabedian model of health care quality prioritizes processes of care, but cardiovascular process measures were not associated with outcomes in this analysis. While these vary between hospitals, they may not be applicable to transplant patient outcomes, or have any effect on outcomes at all [27‐29]. All secondary data analyses provide the net effect of specific clinical and hospital-level covariates on outcomes across the entire population, and are subject to the ecological fallacy when evaluating individual outcomes. The effects observed here may also not reflect more recent practice patterns or hospital structural improvements that may affect mortality and costs today.

Adverse outcomes from cardiovascular events impact post-KT survival. Further research is needed to reduce the risk of mortality once an event occurs. Costs related to prevention and cost-effectiveness of event-based care warrant further analysis. These efforts will improve care for transplant patients, optimize rescue in acute settings, reduce post-transplant costs, and extend long-term post-transplant survival.

Using administrative data, this analysis indicates that transplant and non-transplant hospitals had similar risk-adjusted mortality when managing cardiovascular events in previous kidney transplant recipients. Transplant hospitals were less likely to have high cost episodes of care for these events, which may imply better value in post-transplant cardiovascular care delivery. These data have significant implications for patients, transplant and non-transplant providers, and payers.