Left ventricular assist device (LVAD) implantation has been shown to improve short-term (1 year) survival in stage D heart failure patients (1), and newer devices are providing improved durability for longer-term support (2). Thus, their use is increasing, specifically as destination therapy (DT) (3), and an increasing number of medical centers are involved in following patients supported with LVADs. The REMATCH (Randomized Evaluation of Mechanical Assistance for the Treatment of Congestive Heart Failure) experience (1) suggested that readmissions after LVAD implantation are frequent. The long-term burden of recurring admissions to the implanting hospital in patients with the contemporarily used axial flow pumps may be of specific interest to medical centers involved with this approach to the treatment of end-stage heart failure.

Patients supported in the long term with axial flow devices generally feel better after device implantation as a result of improved hemodynamics and end-organ perfusion. However, caring for these patients may challenge the clinician with a unique set of medical problems. Previously described morbidities include gastrointestinal and cerebrovascular bleeding episodes, infections including those of the device and its driveline, thromboembolism including thrombus formation on the pump, arrhythmias and right ventricular dysfunction. A comparative analysis of the major causes of readmissions after LVAD implantation may therefore be useful in providing perspective and categorizing the relative importance of morbidity associated with ongoing LVAD support.

The aim of this study was to determine the occurrence, causes, trends over time, and possible predictors of readmissions to the implanting hospital after LVAD implantation.

All patients surviving to discharge after implantation of a HeartMate II (Thoratec Corporation, Pleasanton California) LVAD at our institution between January 2008 and July 2011 were screened for readmissions.

We conducted a retrospective analysis of readmissions to our facility based on chart review. Baseline and follow-up characteristics of patients were retrieved from the electronic chart. The glomerular filtration rate was calculated based on the Modification of Diet in Renal Disease equation. Due to a general improvement during optimization before the operation, we differentiated baseline values (at the time of admission) from preoperative values (the morning before operation). Patients were censored for death, transplantation, or last follow-up. The study was approved by the Mayo Clinic institutional review board.

Evaluation and follow-up for LVAD patients at our facility includes comprehensive laboratory testing, cardiopulmonary exercise test and the 6-min walk test (for patients who can perform it), imaging (chest x-ray, abdominal ultrasound, carotid ultrasound, computed tomography when needed), electrocardiogram, echocardiogram as previously described (4), hemodynamic right heart catheterization and coronary angiography, renal function evaluation including iothalamate clearance, pulmonary function testing, fecal hemoglobin, and colonoscopy (unless performed within the previous 10 years). Specialty evaluation is also performed including assessment for the need for social services, rehabilitation, and palliative care. The data obtained on prospective candidates is then carefully reviewed and selection performed in a multidisciplinary conference. Preoperatively, patients are usually hospitalized for right heart catheterization, inotropic support, and intra-aortic balloon pump as needed.

Follow-up for LVAD recipients includes pre-scheduled outpatient visits and telephone follow-up. Visits are scheduled monthly for the first 3 months, every 3 months until 1 year, every 4 months in the second year, and every 6 months in the third and fourth years after implantation. Telephone assistance is available 24/7 by an LVAD coordinator for any patient-related questions or concerns with occasional coordinator-initiated follow-up calls. If a patient requires hospitalization, our usual preference is to recommend hospitalization at our facility. If the patient's medical situation does not allow that, we then recommend stabilization at a local medical facility with subsequent transfer. On arrival, patients are hospitalized, appropriate treatment initiated, and followed by the dedicated multidisciplinary LVAD service.

LVAD patients are treated with aspirin and warfarin with a goal international normalized ratio of 1.5 to 2.5, a proton pump inhibitor, and iron supplementation. If a major bleeding episode occurs, all anticoagulation is withheld for a month and then gradually resumed. An effort is made to identify the possible bleeding source (including upper and lower endoscopy, extended balloon endoscopy, and capsule enteroscopy in the case of gastrointestinal bleeding) and to treat it locally if possible. Lactate dehydrogenase is measured routinely during follow-up visits and an acute increase, when accompanied by symptoms, may prompt further evaluation for hemolysis due to pump thrombosis. Prophylactic antibiotic treatment is given before LVAD implantation and until 48 h after the operation. We do not use routine antibiotic prophylaxis thereafter unless otherwise indicated for endocarditis prophylaxis. Patients are instructed to meticulously clean the drive-line exit site daily.

We recorded all readmissions to our treatment facility until January 31, 2012. The date of admission, duration of stay, and primary reason for readmission was recorded. Primary diagnosis on discharge note was used to identify the cause for readmission. Causes were grouped based on mechanism into major groups including cardiac causes (arrhythmia, heart failure, chest pain), bleeding (any bleeding, anemia), infections (ventricular assist device–related and unrelated infections), thrombotic causes (thromboembolism, suspected LVAD thrombosis), LVAD related (abnormal readouts or alarms), biliary related (biliary colic, cholecystitis, biliary surgery, ductal cancer), elective readmissions (mostly for procedures), and other miscellaneous causes. Major neurological events were recorded as either thrombotic or bleeding events based on the presentation. Due to the morbidity related to such events, they are also reported separately in the results.

Rate calculations were corrected for number of patients available and follow-up time normalized to derive yearly rates. We used the term readmission rate using readmission per patient-year units. The same was performed for length of stay to yield the term readmission duration rate using days per patient-year units. Monthly grouping (first 6 months) and 3-month grouping for later follow-up were performed.

Descriptive analysis was performed by presenting the mean ± SD for numerical data unless markedly non-normal, in which case the median and interquartile range (25th and 75th percentiles) were used unless specified otherwise. Such was the case with the readmission rate where, despite an abnormal distribution, we thought that the mean value would make better clinical sense. Assessment of normality for numerical variables was performed using the Shapiro-Wilks method. Comparison of readmission rates during different follow-up intervals was performed using the paired Wilcoxon rank test.

To account for repeated events, analysis for possible predictors of readmissions was performed using Andersen-Gill models (5). For parameters evaluated after discharge from the implantation hospitalization (1-month follow-up), we evaluated the predictive value affiliated with readmissions occurring after measurement (later than 1 month). For the multivariate analysis, only parameters with fewer than 20% missing values were used. The multivariable model considered marginally significant univariate variables (p < 0.10) with model selection using stepwise combined selection. For each step in the model-fitting process, we included all records with complete data for the particular terms being fit.

All p values were 2 sided, and p values ≤0.05 were considered to indicate statistical significance. All data were analyzed with the JMP System software version 8.0 (SAS Institute, Inc., Cary, North Carolina) and SAS version 9.2 (SAS Institute Inc.).

A total of 137 adult patients underwent implantation with a permanent continuous flow LVAD. Of these, 13 patients did not survive to be discharged and 9 were implanted with pumps other than the HeartMate II (6 Venrassist, 1HM XVE, 1 HeartWare, 1 Duraheart) and were excluded, leaving 115 patients for analysis. Patient characteristics are shown in (Table 1). The median age was 62 years, 83% were male, and 49% had an ischemic etiology for heart failure. Only 36% were bridged to transplantation. Our patient group included 11 patients who underwent implantation with an LVAD for heart failure with predominantly restrictive physiology; however, most patients had dilated left ventricles with reduced ejection fraction.

Patients were followed for 1.4 ± 0.9 years. During this time, 224 readmissions to our facility were recorded in 83 of 115 patients. Thirty-two patients (28%) were not readmitted, 33 (29%) had 1 readmission, 17 (15%) had 2, and 33 (29%) had >2 readmissions during their follow-up. In the first month, 25 of 115 patients (22%) and at 1 year, 75 patients (65%) were readmitted. Characteristics of patients who were readmitted and those who were not are also shown in (Table 1). Patients who were not readmitted were followed for a shorter duration than those who were readmitted. They also tended to be younger, with a nonischemic etiology and more frequently underwent implantation as a bridge to transplantation.

The rate of readmissions to our facility for the whole cohort by time after implantation is shown in (Figure 41_gr1). Due to the small number of patients with longer follow-up, we show only the results up to 3 years. Overall, the mean ± SD readmission rate was 1.64 ± 1.97, and the median (IQR) was 1.1 (0 to 2.4) per patient-year of follow-up. The rate decreased during the first 6 months from 2.6 to 0.9 readmissions per patient-year and remained stable thereafter. The mean ± SD readmission rate for the first 6 months was 2.0 ± 2.3, and the median (IQR) rate was 2 (0 to 4), and for the subsequent duration of follow-up, it was 1.2 ± 2.1 (mean ± SD), and the median (IQR) rate was 0.4 (0 to 1.7) per patient-year. The difference between rates during the 2 follow-up periods was statistically significant (p = 0.0006). The cumulative readmission duration rate by time after implantation is shown in (Figure 41_gr2). Trends were similar to those of the readmission rate, again with a decrease during the first 6 months after surgery and stabilization thereafter. The mean ± SD readmission duration rate during the first 6 months was 19.5 ± 39, and the median (IQR) rate was 4 (0 to 10) (and for the subsequent follow-up, the mean ± SD rate was 7.4 ± 13 and the median (IQR) rate was 0 (0 to 10) days per patient-year (p = 0.0017 for the difference between intervals). (Table 2) shows the readmission rates in patients who underwent implantation as a bridge to transplantation versus DT. Overall DT patients had significantly longer follow-up leading to more readmissions; however readmission rates were not significantly different between the groups.

The readmission rate by cause of readmission for various time intervals after LVAD implantation is shown in (Figure 41_gr3). The major causes of readmission in the first 6 months include bleeding and cardiac-related causes. These decrease during the second 6 months of follow-up (together with the miscellaneous causes), accounting for an overall decrease in readmissions. During the second year, there was an increased rate of readmissions for bleeding and fewer cardiac readmissions compared with the previous interval. During the third year, there were more cardiac admissions and fewer related to bleeding with an increase in the rate of thrombosis and of hospitalizations for elective procedures.

Because the relative burden of readmission may be influenced by both the rate of readmission and the duration of stay during readmission, we compared the proportional impact of the different causes on these 2 variables (Figure 41_gr4). Bleeding and cardiac causes accounted for more readmissions but accounted for fewer days spent in the hospital compared with thrombotic and infectious causes. Taken together the 4 leading causes, accounting for the majority of readmissions (∼75%) were bleeding, cardiac, infection, and thrombosis.

The proportion of patients presenting with a defined cause for readmission is shown in (Figure 41_gr5). Although the analysis is limited because some events were treated in other facilities and hence missed, the figure gives a reasonable estimate based on the readmissions to the implanting hospital. Approximately one third of the patients were readmitted due to a cardiac event. Of these, 18 (51%) underwent implantation for ischemic etiology. Their preoperative estimates of right ventricular function were similar to those of the total group (right index of myocardial performance, 0.55 ± 0.21; right atrial pressure, 16.3 ± 6.6; and right ventricular stroke work index, 8.5 ± 4.5). One third of the patients were readmitted due to bleeding. Most bleeding events were gastrointestinal (n = 51), 5 were cerebral, 2 hematuria, 2 uterine, 2 epistaxis, 2 for anemia of unknown cause, and the rest due to hemoptysis, hemothorax, and a hematoma (1 each). Bleeding tended to recur despite our policy to stop anticoagulation. Within 3 months of the index bleeding event, 3 patients had 3 subsequent thrombotic episodes and 8 patients had 13 subsequent bleeding episodes. For the duration of follow-up, 5 patients had 5 subsequent thrombotic episodes and 12 patients had 24 subsequent bleeding episodes. Infection as the primary cause of readmission occurred in 22% and thrombosis in 14%. A detailed description of the major causes of specific readmissions is shown in (Table 3), specifying the numbers and proportions among DT and bridge to transplantation patients.

A univariate analysis of the association with readmissions after LVAD implantation using the Anderson-Gill model was performed for all variables presented in (Table 1); p values are included in the table. Significantly associated variables are presented in (Table 4).

Variables significantly associated with (fewer) readmissions in a multivariate model included residence within our hospital extended referral zone of Minnesota (MN) and the neighboring states (MN: 1 for yes, 0 for no) (beta = −0.45, hazard ratio [HR]: 0.64; 95% confidence interval [CI]: 0.48 to 0.85; p = 0.002), pre-operative hemoglobin (beta = −0.113; HR: 0.89; 95% CI: 0.83 to 0.96; p = 0.004), preoperative N-terminal pro–B-type natriuretic peptide (NT-proBNP) per 1,000 U (beta = −0.0198; HR: 0.98; 95% CI: 0.96 to 1.0; p = 0.044), and closed aortic valve observed on echocardiography in the first month [aortic valve: 1 for closed, 0 for no] (beta = −0.74; HR: 0.48;95% CI: 0.34 to 0.67; p < 0.001). The C-statistic for the model was 0.632. The proportional hazards term of the Anderson-Gill model is exp(3.17 − 0.45 × MN – 0.113 × hemoglobin − 0.0198 × NT-proBNP – 0.74 × aortic valve) and can be multiplied by the average readmission rate to yield patient-specific hazard. To account for possible selection of a sicker population with preserved ejection fraction (causing aortic valve opening), we performed a second analysis excluding patients with an ejection fraction at baseline >25%. The results were similar, and a closed aortic valve at 1 month was still associated with fewer readmissions (HR: 0.68; 95% CI: 0.48 to 0.96; p = 0.031). Preoperative variables significantly associated with (fewer) readmissions in a multivariate model included residence within our hospital extended referral zone of Minnesota and the neighboring states (HR: 0.66; 95% CI: 0.48 to 0.91; p = 0.011), preoperative hemoglobin (HR: 0.91; 95% CI: 0.84 to 0.99; p = 0.027) and preoperative NT-proBNP (HR: 0.98; 95% CI: 0.96 to 1.0 per 1,000-unit increase; p = 0.022); the C-statistic was 0.626.

Readmissions to the hospital are of interest to the medical community for multiple reasons. Primarily they reflect patient morbidity and hence quality of life. They are also important because of the cost incurred on the health care system. In an era of increasing awareness of cost for health care delivery, an appreciation of the readmission rate for a particular medical condition is important for planning of resource allocation. Among Medicare beneficiaries with an implanted LVAD, 55.6% were readmitted within 6 months. On average, these patients spent 29.8 ± 45.0 days in the hospital during the subsequent 2 years. These were considered significant contributors to the cost associated with LVAD treatment (6).

We report an institutional readmission rate of 1.64 ± 1.97 with a cumulative incidence of 22% at 1 month and 65% at 1 year after axial flow LVAD implantation. In the HeartMate II clinical trial, a higher readmission rate of 2.64 events per patient-year is reported during a median follow-up of 1.7 years; this was significantly reduced compared with the HeartMate XVE pulsatile pump (2), which is generally no longer used. The rate of readmissions that we report after axial flow LVAD implantation is comparable to reported rates after cardiac transplantation (7) and cardiac surgery (8), to those after hospitalizations for heart failure ((9),10), or to those of general hospitalizations in a sick population (11). Rates may be at first higher than reported after a diagnosis of heart failure (0.87 readmissions per patient-year) (12) but become comparable to these after the initial 6-month postoperative period.

An important part of our analysis is related to the causes of readmissions in patients supported by axial flow LVAD. Understanding the causes may be important both from the institutional standpoint (knowing why, when, and how many patients get readmitted) as well as the patient perspective (what is the likelihood of being readmitted for a specific reason). Using readmission cause as the case definition may be helpful in avoiding inconsistencies in reporting between centers. As an example, the incidence of bleeding in axial flow LVAD patients was reported to range between 5% and 81% (13). In this regard, a comparative analysis of the different causes of readmissions may be useful in directing future research because it may highlight areas of greater importance to patient care.

In our analysis, the major causes for readmission were bleeding, cardiac, infection, and thrombosis. Thus, our findings were similar to the major late adverse events in the HeartMate II clinical trial, which were bleeding, infection, and arrhythmias. However, higher bleeding and higher infection rates were previously reported. Cerebral thrombotic events were noted mostly within the first 30 days (14). These differences may reflect our longer follow-up. In our analysis, the thrombotic event rate seems to increase during the third year of follow-up. We also acknowledge that the analysis reflects in part our clinical practice and patient population, which may be different from those in the clinical trial.

The causes of readmission in patients supported with an LVAD is unique to this population. For comparison, after hospitalization for heart failure in an elderly population, the most common reasons for readmission were heart failure (28%) and respiratory infections (6%). Gastrointestinal bleeding and arrhythmias accounted for 2.5% each (10). After cardiac transplantation, the primary reason for hospitalization was infection (51.5%) (7). In patients supported by axial flow LVAD, bleeding complications are long known to be a significant cause of morbidity (15), the reason for which is unclear. Although the left ventricular output is maintained by the LVAD, cardiac complications may be caused by right ventricular dysfunction, arrhythmias, pump malfunction, or aortic incompetence. Interestingly, aortic regurgitation was not a frequent cause of readmissions, perhaps due to our practice of treating any aortic regurgitation during LVAD implantation. Pump-related infections were infrequent in our cohort. Other infections were more common; however, these are not specific to this patient population and are a leading cause of readmissions after surgery (8) or in heart failure (12). Thrombotic events, including cerebrovascular events and pump thrombosis, were among the leading causes of prolonged readmissions. We observed a trend toward more readmissions for late pump thrombosis with a decreased rate of bleeding complications during the third year of follow-up. It is possible that this may be related to less stringent anticoagulation with longer period of time post-implantation.

An interesting group of causes of readmissions is the biliary events. These may be related to ongoing low-grade hemolysis with increased bilirubin excretion. We now perform routine abdominal ultrasound before LVAD implantation and consider elective cholecystectomy. Last, hospitalizations due to purely pump-related issues such as technical difficulties, alarms, and abnormal readouts were not common and occurred mostly during the second half of the first year after implantation.

With the increasing growth of our LVAD program, we have had an evolution in our practice in the care of the LVAD recipient. We have taken a number of steps in the clinical care and follow-up of these patients, acknowledging their special needs. These now include enhancement of our follow-up with regular protocol-driven follow-up visits to the clinic and dedicated coordinator surveillance in between visits. We implemented follow-up procedures such as right heart catheterization, Vo₂ exercise, and 6-min walk tests. Anticoagulation monitoring was enhanced, and early involvement of subspecialists such as the infectious disease and the gastrointestinal services is now a routine in our practice.

Being able to predict which LVAD recipients will be at higher risk of readmission after implantation may be useful in discharge and after-care planning. Although different causes of readmission may have different and opposing predictors, a comprehensive analysis may provide a general assessment of patient well-being. However, any interpretation of the possible predictors in this paper must take into account that the population selected consists of patients who survived the operation to discharge. Interestingly, living far from the implanting facility was associated with more readmissions. This may be explained by a referral bias because patients who are referred from a different location are many times sicker than the local referrals. Decreased hemoglobin at admission also emerged as being strongly associated with readmissions and perhaps correlates with the high incidence of readmissions for bleeding after implantation. Anemia was also a strong predictor of readmissions in the heart failure population (12). Another observation was that a higher admission pro–B-type natriuretic peptide level was associated with fewer readmissions. Although this appears counterintuitive, similar findings of lower rather than higher levels of B-type natriuretic peptides were associated with mortality in patients with end-stage heart failure (16). This can be explained as a relative “exhaustion” of the compensatory natriuretic peptide system at the extremes of heart failure morbidity. Interestingly, we observed that a closed aortic valve observed during echocardiography at 1 month was associated with fewer readmissions. This appears to be in contrast to the general belief that maintaining at least partial opening of the valve by dialing down the pump speed may benefit future morbidity. Although longer follow-up may be needed, this observation argues against the utility of the intermittent aortic valve opening as a major determinant to guide the pump speed setting.

In summary, this analysis shows our single-center experience of the occurrence and causes of hospital readmissions over time after LVAD implantation. To the best of our knowledge, this is the first analysis of this kind. We demonstrate here for the first time that the rate decreases over the first 6 months after implantation from 2.5 readmissions per patient-year and then stabilizes at about 1 year. Although the contribution of various causes of readmission differed as a product of time after implantation, the 4 leading defined causes remained bleeding, cardiac, infection, and thrombosis. Bleeding and cardiac causes are the leading causes for the number of readmissions, and infection and thrombosis were the leading causes for cumulative hospital stay. Possible predictors of overall recurring admissions were evaluated and may be helpful in trying to reduce readmission rates.

Our study has the obvious limitations of a descriptive retrospective analysis and is not hypothesis driven. Because many of the patients were followed elsewhere before LVAD implantation, it is beyond the scope of this paper to compare hospital admissions before and after LVAD implantation. Being a single-centre study, it may not be representative of what occurs at other centers. Therefore, in interpreting the results, one should be conscious of the patient characteristics and management routines that we describe. Readmissions to other medical facilities were not recorded. Although it is our policy to refer patients with significant morbidity to be readmitted in our facility, the overall number of readmissions may be underestimated. Another possible shortcoming of this referral bias may be that patient readmissions included in the analysis may be of a more severe nature and hence longer and more complicated because shorter readmissions may have been dealt with at a local facility.

Readmission rates to the implanting facility for recipients of an axial flow LVAD are comparable to those previously observed in stable community heart failure patients.

Rates are higher in the first 6 months after surgery. Leading causes of readmission are bleeding, cardiac (heart failure and arrhythmia), infection, and thrombosis. Associated variables such as admission hemoglobin and NT-proBNP may help predict the general risk of readmissions.