Abstracts S23 need for preoperative (preop) dialysis raised IHM to 48%. Among patients with an isolated LVAD the strongest partitioning factor was the need for preop dialysis with an IHM of 22%. In those with an LVAD alone who required preop dialysis and were also over 66 yrs old the IHM rose to 42%. Conclusion: The need for a simultaneous RVAD, preop dialysis and older age identifies PTS with an IHM of near 50%. The high risk of a poor short-term outcome may call into question the utility of using an LVAD in such patient populations. CART mortality analyses may be informative for PT selection.
4( 0) An Updated Bayesian Model for Predicting Mortality in Continuous Flow-Left Ventricular Assist Device Patients L.E. Lohmueller ,1 M.K. Kanwar,2 N.L. Loghmanpour,1 S. Murali,2 J.F. Antaki.1 1Biomedical Engineering, Carnegie Mellon University, Pittsburgh, PA; 2Cardiovascular Institute, Allegheny General Hospital, Pittsburgh, PA. Purpose: We previously published a Bayesian predictive model of postLVAD mortality at 30 days, 90 days, 1 year and 2 years. The model was derived using INTERMACS data for patients who received a CF-LVAD during Jun 2006-Apr 2013. In this study, we validated our original model for predictive accuracy and then updated the model with the more recent INTERMACS data and new variables. Methods: The existing model was developed using Tree-Augmented Naïve Bayes (TAN) algorithm on 8050 patients in the INTERMACS database and validated by 5-fold cross validation. It was validated for predictive accuracy using the receiver operating characteristic (ROC) in 3,803 CF-LVAD INTERMACS patients from Apr 2014 - Mar 2015. The new dataset was then used to generate a TAN model, with improvements made by refining discretization of continuous variables and incorporating new variables. The resulting model was evaluated by 5-fold cross validation with the new data set. Results: The predictive accuracy of mortality of the original (2006 - 2013) model was 93%, 90%, 84%, and 83% for 30 days, 90day, 1 year, and 2 years, respectively, when tested against the new (2014 - 2015) dataset. However, the ROC for each time-point was decreased to 67-73%, reflecting a loss in sensitivity. The predictiveness of mortality by the new model was improved to 94%, 93%, 89%, and 90% for 30 days, 90 days, 1 year, and 2 year mortality, and ROC was improved to 83-84%. (Table 1) Conclusion: This study demonstrated reasonably good predictive accuracy by the original 2006-2013 Bayesian predictive model with the most recent INTERMACS data set. The addition of new data elements and improved discretization was found to improve predictive accuracy. The clinical pathway for evaluating and implanting VADs has changed over time; therefore updates to the predictive models are necessary to provide optimal decision support. Results indicate that future automatically updating models will be important to maintain accuracy and utility for medical risk assessment. Predictive Accuracy (and ROC) Ranges
4( 1) Muscular Strength Predicts Functional Capacity in Left Ventricular Assist Device Supported Patients R.Y. Loyaga-Rendon ,1 S.V. Pamboukian,1 D. Acharya,1 J. Tallaj,1 F. Siric,2 C. Hoopes,2 W. Holman,1 J. Kirklin,1 E.P. Plaisance.3 1Division of Cardiology, University of Alabama at Birmingham, Birmingham, AL; 2Division of Cardiovascular Surgery, University of Alabama at Birmingham, Birmingham, AL; 3Department of Human Studies and Nutrition Obesity Research Center, University of Alabama at Birmingham, Birmingham, AL. Purpose: Muscle wasting and deconditioning are frequently present in advanced heart failure patients at the time of continuous flow left ventricular assist device (CF-LVAD) placement and may influence post LVAD outcomes and quality of life (QoL). We analyzed the contribution of muscular strength to functional capacity (FC) and QoL in patients supported by CF-LVAD. Methods: FC and QoL were measured with 6 minute walk test (6MWT) and Kansas City Quality Questionnaire (KCQQ). Maximum isometric strength of the elbow flexors and knee extensors was evaluated while EMG of the biceps brachii and rectus femoris were measured. A sit-to-stand test was performed using EMG to evaluate percentage of maximal knee strength required to sit and stand. Clinical, echocardiographic and hemodynamic information were obtained from clinical records. Results: Eighteen patients with a mean age of 55±14 years were enrolled in this study, 79% were male and 50 % were African Americans. The median time of support was 15.5 months (6 - 36). Baseline characteristics are shown in Table 1. Of the variables examined (Table 2), only elbow flexor and knee extensor strength (adjusted to BMI) correlated significantly with both 6MWT and KCQQ. Conclusion: We conclude that isometric muscle strength measured in stable ambulatory LVAD patients is associated with FC and QoL.
Basic clinical and functional characteristics Characteristics Echocardiography -Left ventricular diastolic dimension (LVDD (cm) -Left ventricular Ejection Fraction (LVEF) (%) Hemodynamics (n= 9) -Right Atrial (RA) pressure (mmHg) -Pulmonary artery (PA) mean pressure (mmHg) -Pulmonary capillary wedge pressure (PCWP) (mmHg) -Cardiac index (CI) (L/min/m2) -Mean arterial pressure (MAP) (mmHg) -HR (bpm) Functional Capacity -6MWTD (m) -Gate Speed -NYHA-FC I II III Quality of Life -KCCQ Muscular Strength -Knee -Elbow -Stand-up
Model Dataset
Original INTERMACS (2006-2013) 8,050 patients New INTERMACS (20132015) 3,803 patients
5.7±1.2 23±11 11±6 24±6 12±6 2.6±0.4 87±7 75±10
387±77 4.5±0.8 9 (50%) 7 (38.9%) 2 (11.1%)
53.4±7 369.8±109.5 214±66.7 54.2±24.8
Univariable analysis with outcomes of FC and QoL Verification Dataset
Accuracy (ROC)
Value
Original INTERMACS (2006-2013) 8,050 patients 81-91% (77-81%)
New INTERMACS (2013-2015) 3,803 patients 83-93% (67-73%)
89-94% (83-84%)
Age BMI RA/PCWP/CI/MAP LVEF/LVDD Elbow Knee Stand-up
6MWT
KCCQ
−0.6 (0.01) ns ns ns 0.58 (0.01) 0.47 (0.055) ns
ns 0.73 (0.001) ns ns 0.56 (0.018) 0.64 (0.007) −0.74 (0.001)