Dynamic prediction of left ventricular assist device pump thrombosis based on lactate dehydrogenase trends.

AIMS: The risk of HeartMate II (HMII) left ventricular assist device (LVAD) thrombosis has been reported, and serum lactate dehydrogenase (LDH), a biomarker of haemolysis, increases secondary to LVAD thrombosis. This study evaluated longitudinal measurements of LDH post-LVAD implantation, hypothesizing that LDH trends could timely predict future LVAD thrombosis. METHODS AND RESULTS: From October 2004 to October 2014, 350 HMIIs were implanted in 323 patients at Cleveland Clinic. Of these, patients on 339 HMIIs had at least one post-implant LDH value (7996 total measurements). A two-step joint model combining longitudinal biomarker data and pump thrombosis events was generated to assess the effect of changing LDH on thrombosis risk. Device-specific LDH trends were first smoothed using multivariate boosted trees, and then used as a time-varying covariate function in a multiphase hazard model to analyse time to thrombosis. Pre-implant variables associated with time-varying LDH values post-implant using boostmtree were also investigated. Standardized variable importance for each variable was estimated as the difference between model-based prediction error of LDH when the variable was randomly permuted and prediction error without permuting the values. The larger this difference, the more important a variable is for predicting the trajectory of post-implant LDH. Thirty-five HMIIs (10%) had either confirmed (18) or suspected (17) thrombosis, with 15 (43%) occurring within 3 months of implant. LDH was associated with thrombosis occurring both early and late after implant (P < 0.0001 for both hazard phases). The model demonstrated increased probability of HMII thrombosis as LDH trended upward, with steep changes in LDH trajectory paralleling trajectories in probability of pump thrombosis. The most important baseline variables predictive of the longitudinal pattern of LDH were higher bilirubin, higher pre-implant LDH, and older age. The effect of some pre-implant variables such as sodium on the post-implant LDH longitudinal pattern differed across time. CONCLUSIONS: Longitudinal trends in surveillance LDH for patients on HMII support are useful for dynamic prediction of pump thrombosis, both early after implant and late. Incorporating upward and downward trends in LDH that dynamically update a model of LVAD thrombosis risk provides a useful tool for clinical management and decisions.

Biatrial maze procedure versus pulmonary vein isolation for atrial fibrillation during mitral valve surgery: New analytical approaches and end points.

OBJECTIVE: To use novel statistical methods for analyzing the effect of lesion set on (long-standing) persistent atrial fibrillation (AF) in the Cardiothoracic Surgical Trials Network trial of surgical ablation during mitral valve surgery (MVS). METHODS: Two hundred sixty such patients were randomized to MVS + surgical ablation or MVS alone. Ablation was randomized between pulmonary vein isolation and biatrial maze. During 12 months postsurgery, 228 patients (88%) submitted 7949 transtelephonic monitoring (TTM) recordings, analyzed for AF, atrial flutter (AFL), or atrial tachycardia (AT). As previously reported, more ablation than MVS-alone patients were free of AF or AF/AFL at 6 and 12 months (63% vs 29%; P < .001) by 72-hour Holter monitoring, without evident difference between lesion sets (for which the trial was underpowered). RESULTS: Estimated freedom from AF/AFL/AT on any transmission trended higher after biatrial maze than pulmonary vein isolation (odds ratio, 2.31; 95% confidence interval, 0.95-5.65; P = .07) 3 to 12 months postsurgery; estimated AF/AFL/AT load (ie, proportion of TTM strips recording AF/AFL/AT) was similar (odds ratio, 0.90; 95% confidence interval, 0.57-1.43; P = .6). Within 12 months, estimated prevalence of AF/AFL/AT by TTM was 58% after MVS alone, and 36% versus 23% after pulmonary vein isolation versus biatrial maze (P < .02). CONCLUSIONS: Statistical modeling using TTM recordings after MVS in patients with (long-standing) persistent AF suggests that a biatrial maze is associated with lower AF/AFL/AT prevalence, but not a lower load, compared with pulmonary vein isolation. The discrepancy between AF/AFL/AT prevalence assessed at 2 time points by Holter monitoring versus weekly TTM suggests the need for a confirmatory trial, reassessment of definitions for failure after ablation, and validation of statistical methods for assessing atrial rhythms longitudinally.

Probability of atrial fibrillation after ablation: Using a parametric nonlinear temporal decomposition mixed effects model.

Atrial fibrillation is an arrhythmic disorder where the electrical signals of the heart become irregular. The probability of atrial fibrillation (binary response) is often time varying in a structured fashion, as is the influence of associated risk factors. A generalized nonlinear mixed effects model is presented to estimate the time-related probability of atrial fibrillation using a temporal decomposition approach to reveal the pattern of the probability of atrial fibrillation and their determinants. This methodology generalizes to patient-specific analysis of longitudinal binary data with possibly time-varying effects of covariates and with different patient-specific random effects influencing different temporal phases. The motivation and application of this model is illustrated using longitudinally measured atrial fibrillation data obtained through weekly trans-telephonic monitoring from an NIH sponsored clinical trial being conducted by the Cardiothoracic Surgery Clinical Trials Network.

Machine-learning phenotypic classification of bicuspid aortopathy.

BACKGROUND: Bicuspid aortic valves (BAV) are associated with incompletely characterized aortopathy. Our objectives were to identify distinct patterns of aortopathy using machine-learning methods and characterize their association with valve morphology and patient characteristics. METHODS: We analyzed preoperative 3-dimensional computed tomography reconstructions for 656 patients with BAV undergoing ascending aorta surgery between January 2002 and January 2014. Unsupervised partitioning around medoids was used to cluster aortic dimensions. Group differences were identified using polytomous random forest analysis. RESULTS: Three distinct aneurysm phenotypes were identified: root (n = 83; 13%), with predominant dilatation at sinuses of Valsalva; ascending (n = 364; 55%), with supracoronary enlargement rarely extending past the brachiocephalic artery; and arch (n = 209; 32%), with aortic arch dilatation. The arch phenotype had the greatest association with right-noncoronary cusp fusion: 29%, versus 13% for ascending and 15% for root phenotypes (P < .0001). Severe valve regurgitation was most prevalent in root phenotype (57%), followed by ascending (34%) and arch phenotypes (25%; P < .0001). Aortic stenosis was most prevalent in arch phenotype (62%), followed by ascending (50%) and root phenotypes (28%; P < .0001). Patient age increased as the extent of aneurysm became more distal (root, 49 years; ascending, 53 years; arch, 57 years; P < .0001), and root phenotype was associated with greater male predominance compared with ascending and arch phenotypes (94%, 76%, and 70%, respectively; P < .0001). Phenotypes were visually recognizable with 94% accuracy. CONCLUSIONS: Three distinct phenotypes of bicuspid valve-associated aortopathy were identified using machine-learning methodology. Patient characteristics and valvular dysfunction vary by phenotype, suggesting that the location of aortic pathology may be related to the underlying pathophysiology of this disease.

Boosted Multivariate Trees for Longitudinal Data.

Machine learning methods provide a powerful approach for analyzing longitudinal data in which repeated measurements are observed for a subject over time. We boost multivariate trees to fit a novel flexible semi-nonparametric marginal model for longitudinal data. In this model, features are assumed to be nonparametric, while feature-time interactions are modeled semi-nonparametrically utilizing P-splines with estimated smoothing parameter. In order to avoid overfitting, we describe a relatively simple in sample cross-validation method which can be used to estimate the optimal boosting iteration and which has the surprising added benefit of stabilizing certain parameter estimates. Our new multivariate tree boosting method is shown to be highly flexible, robust to covariance misspecification and unbalanced designs, and resistant to overfitting in high dimensions. Feature selection can be used to identify important features and feature-time interactions. An application to longitudinal data of forced 1-second lung expiratory volume (FEV1) for lung transplant patients identifies an important feature-time interaction and illustrates the ease with which our method can find complex relationships in longitudinal data.

Estimating the prevalence of atrial fibrillation from a three-class mixture model for repeated diagnoses.

Atrial fibrillation (AF) is an abnormal heart rhythm characterized by rapid and irregular heartbeat, with or without perceivable symptoms. In clinical practice, the electrocardiogram (ECG) is often used for diagnosis of AF. Since the AF often arrives as recurrent episodes of varying frequency and duration and only the episodes that occur at the time of ECG can be detected, the AF is often underdiagnosed when a limited number of repeated ECGs are used. In studies evaluating the efficacy of AF ablation surgery, each patient undergoes multiple ECGs and the AF status at the time of ECG is recorded. The objective of this paper is to estimate the marginal proportions of patients with or without AF in a population, which are important measures of the efficacy of the treatment. The underdiagnosis problem is addressed by a three-class mixture regression model in which a patient's probability of having no AF, paroxysmal AF, and permanent AF is modeled by auxiliary baseline covariates in a nested logistic regression. A binomial regression model is specified conditional on a subject being in the paroxysmal AF group. The model parameters are estimated by the Expectation-Maximization (EM) algorithm. These parameters are themselves nuisance parameters for the purpose of this research, but the estimators of the marginal proportions of interest can be expressed as functions of the data and these nuisance parameters and their variances can be estimated by the sandwich method. We examine the performance of the proposed methodology in simulations and two real data applications.

Insights Into Timing, Risk Factors, and Outcomes of Stroke and Transient Ischemic Attack After Transcatheter Aortic Valve Replacement in the PARTNER Trial (Placement of Aortic Transcatheter Valves).

BACKGROUND: Prior studies of stroke and transient ischemic attack (TIA) after transcatheter aortic valve replacement (TAVR) are limited by reporting and follow-up variability. This is a comprehensive analysis of time-related incidence, risk factors, and outcomes of these events. METHODS AND RESULTS: From April 2007 to February 2012, 2621 patients, aged 84±7.2 years, underwent transfemoral (TF; 1521) or transapical (TA; 1100) TAVR in the PARTNER trial (Placement of Aortic Transcatheter Valves; as-treated), including the continued access registry. Stroke and TIA were identified by protocol and adjudicated by a Clinical Events Committee. Within 30 days of TAVR, 87 (3.3%) patients experienced a stroke (TF 58 [3.8%]; TA 29 [2.7%]; P=0.09), 85% within 1 week. Instantaneous stroke risk peaked on day 2, then fell to a low prolonged risk of 0.8% by 1 to 2 weeks. Within 30 days, 13 (0.50%) patients experienced a TIA (TF 10 [0.67%]; TA 3 [0.27%]; P>0.17). Stroke and TIA were associated with lower 1-year survival than expected (TF 47% after stroke versus 82%, and 64% after TIA versus 83%; TA 53% after stroke versus 80%, and 64% after TIA versus 83%). Risk factors for early stroke after TA-TAVR included more postdilatations, pure aortic stenosis without regurgitation, and possibly more pacing runs, earlier date of procedure, and no dual antiplatelet therapy; high pre-TAVR aortic peak gradient was a risk factor for stroke early after TF-TAVR. CONCLUSIONS: Risk of stroke or TIA is highest early after TAVR and is associated with increased 1-year mortality. Modifications of TAVR, emboli-prevention devices, and better intraprocedural pharmacological protection may mitigate this risk. CLINICAL TRIAL REGISTRATION: URL: http://www.clinicaltrials.gov. Unique identifier: NCT00530894.

Importance of residual primary cancer after induction therapy for esophageal adenocarcinoma.

OBJECTIVES: To (1) assess the continuous distribution of the percentage of residual primary cancer in resection specimens after induction therapy for locally advanced esophageal adenocarcinoma, (2) determine the effects of residual primary cancer on survival after esophagectomy, (3) ascertain interplay between residual primary cancer and classical classifications of response to induction therapy (ypTNM), and (4) identify predictors of residual primary cancer. METHODS: From January 2006 to November 2012, 188 patients (78%) underwent accelerated chemoradiotherapy, and 52 patients (22%) underwent chemotherapy alone followed by esophagectomy for adenocarcinoma. Mean age was 61 ± 9.2 years, and 89% were male. Residual primary cancer, assessed as the percentage of residual primary cancer cells in resection specimens, was quantified histologically by a gastrointestinal pathologist. Random Forest technology was used for data analysis. RESULTS: Twenty-five specimens (10%) had no residual primary cancer (ypT0), 79 (33%) had 1% to 25% residual cancer, 91 (38%) had 26% to 75%, and 45 (19%) had >75%. Survival was worse with increasing residual primary cancer, plateauing at 75%. Greater residual primary cancer was associated with worse survival across the spectrum of higher ypTN. Higher ypT, larger number of positive nodes, and use of induction chemotherapy rather than induction chemoradiotherapy were associated with greater residual primary cancer. CONCLUSIONS: Less residual primary cancer in response to preoperative therapy is associated with a linear increase in survival after esophagectomy for locally advanced esophageal adenocarcinoma; however, survival is poorer than for resected early-stage cancers. Therefore, for patients with poor prognostic indicators, including higher percentage of residual primary cancer, the role of adjuvant therapy needs to be further examined in an attempt to improve survival.

Learning curves for transapical transcatheter aortic valve replacement in the PARTNER-I trial: Technical performance, success, and safety.

OBJECTIVES: Introduction of hybrid techniques, such as transapical transcatheter aortic valve replacement (TA-TAVR), requires skills that a heart team must master to achieve technical efficiency: the technical performance learning curve. To date, the learning curve for TA-TAVR remains unknown. We therefore evaluated the rate at which technical performance improved, assessed change in occurrence of adverse events in relation to technical performance, and determined whether adverse events after TA-TAVR were linked to acquiring technical performance efficiency (the learning curve). METHODS: From April 2007 to February 2012, 1100 patients, average age 85.0 ± 6.4 years, underwent TA-TAVR in the PARTNER-I trial. Learning curves were defined by institution-specific patient sequence number using nonlinear mixed modeling. RESULTS: Mean procedure time decreased from 131 to 116 minutes within 30 cases (P = .06) and device success increased to 90% by case 45 (P = .0007). Within 30 days, 354 patients experienced a major adverse event (stroke in 29, death in 96), with possibly decreased complications over time (P âˆ¼ .08). Although longer procedure time was associated with more adverse events (P < .0001), these events were associated with change in patient risk profile, not the technical performance learning curve (P = .8). CONCLUSIONS: The learning curve for TA-TAVR was 30 to 45 procedures performed, and technical efficiency was achieved without compromising patient safety. Although fewer patients are now undergoing TAVR via nontransfemoral access, understanding TA-TAVR learning curves and their relationship with outcomes is important as the field moves toward next-generation devices, such as those to replace the mitral valve, delivered via the left ventricular apex.

Learning curves for transfemoral transcatheter aortic valve replacement in the PARTNER-I trial: Technical performance.

OBJECTIVES: To assess technical performance learning curves of teams performing transfemoral transcatheter aortic valve replacement (TF-TAVR). BACKGROUND: TF-TAVR is a new procedure for treating severe aortic stenosis. The number of cases required for procedural efficiency is unknown. METHODS: In the PARTNER-I trial, 1,521 patients underwent TF-TAVR from 4/2007-2/2012. Learning curve analysis of technical performance metrics was performed using institution-specific patient sequence number, interval between procedures, and institutional trial entry date. Learning curve characteristics were assessed using semi-parametric and parametric mixed-effects models. RESULTS: As patient sequence number increased, average procedure time decreased from 154 to 85 minutes (P < 0.0001), and fluoroscopy time from 28 to 20 minutes (P < 0.0001). Procedure time plateaued at an average of 83 minutes (range 52-140). Procedure time plateau was dynamic during the course of the trial, averaging 25 cases (range 21-52) by its end. The later institutions enrolled in the trial, the shorter the initial procedure time. During the trial, percutaneous rather than surgical access increased from 7.9% to 69%. CONCLUSIONS: Technical performance learning curves exist for TF-TAVR; procedural efficiency increased with experience, with concomitant decreases in radiation and contrast media exposure. The number of cases needed to achieve efficiency decreased progressively, with optimal procedural performance reached after approximately 25 cases for late-entering institutions. Knowledge and experience accumulated by early TF-TAVR institutions were disseminated, shortening the learning curve of late-entering institutions. Technological advances resulting from learning during the trial moved the field from initial conservative surgical cut-down to percutaneous access for most patients. © 2015 Wiley Periodicals, Inc.

Learning curves for transfemoral transcatheter aortic valve replacement in the PARTNER-I trial: Success and safety.

OBJECTIVES: To identify number of cases needed to maximize device success and minimize adverse events after transfemoral transcatheter aortic valve replacement (TF-TAVR), and determine if adverse events were linked to the technical performance learning curve. BACKGROUND: TF-TAVR is a complex procedure with an incompletely characterized learning curve for clinical outcomes. METHODS: From 4/2007-2/2012, 1521 patients underwent TF-TAVR in the PARTNER-I trial. Outcomes learning curves were defined as number of cases needed to reach a plateau for device success, adverse events, and post-procedure length of stay. Institutional variation was accounted for by mixed-model non-linear techniques, which were also used to identify contribution of the procedure time learning curve to 30-day major adverse events and length of stay. RESULTS: Eighty percent device success was achieved after 22 cases; major vascular complications fell below 5% after 70 cases and major bleeding below 10% after 25 cases. It took an average of 28 cases to achieve a consistent low risk of 30-day major adverse events, but institutions entering in the middle of the trial achieved it after about 26. The most significant correlate of 30-day major adverse events and post-procedure length of stay was procedure time (P < 0.0001). However, this association was related to patient and unmeasured variables, not the procedure time learning curve (P = 0.6). CONCLUSIONS: By end of trial, a consistent low risk of adverse events was achieved after ∼26 cases. However, these improved results were due to change in patient risk profile; outcomes were not linked to the technical performance learning curve. © 2015 Wiley Periodicals, Inc.

Current risks of HeartMate II pump thrombosis: Non-parametric analysis of Interagency Registry for Mechanically Assisted Circulatory Support data.

BACKGROUND: Data from 3 institutions revealed an abrupt increase in HeartMate II (Thoratec) pump thrombosis starting in 2011, associated with 48% mortality at 6 months without transplantation or pump exchange. We sought to discover if the increase occurred nationwide in Interagency Registry for Mechanically Assisted Circulatory Support (INTERMACS) data, and if so (1) determine if accelerated risk continued, (2) identify predictors, (3) investigate institutional variability, and (4) assess mortality after pump thrombosis. METHODS: From April 2008 to June 2014, 11,123 HeartMate II devices were implanted at 146 institutions. Machine learning, non-parametric Random Forests for Survival was used to explore risk-adjusted thrombosis based on 87 pre-implant and implant variables, including implant date. RESULTS: A total of 995 pumps thrombosed, with risk peaking within weeks of implant. The risk-adjusted increase in pump thrombosis began in 2010, reached a maximum in 2012, and then plateaued at a level that was 3.3-times higher than pre-2010. Pump exchange, younger age, and larger body mass index were important predictors, and institutional variability was largely explained by implant date, patient profile, and duration of support. The probability of death within 3 months after pump thrombosis was 24%. CONCLUSIONS: Accelerated risk of HeartMate II thrombosis was confirmed by Interagency Registry for Mechanically Assisted Circulatory Support data, with risk subsequently leveling at a risk-adjusted rate higher than observed pre-2010. This elevated thrombosis risk emphasizes the need for improved mechanical circulatory support systems and post-market surveillance of adverse events. Clinicians cognizant of these new data should incorporate them into their and their patients' expectations and understanding of risks relative to those of transplantation and continued medical therapy.

Outcomes in Nonagenarians Undergoing Transcatheter Aortic Valve Replacement in the PARTNER-I Trial.

BACKGROUND: This study describes short-term and mid-term outcomes of nonagenarian patients undergoing transfemoral or transapical transcatheter aortic valve replacement (TAVR) in the Placement of Aortic Transcatheter Valve (PARTNER)-I trial. METHODS: From April 2007 to February 2012, 531 nonagenarians, mean age 93 ± 2.1 years, underwent TAVR with a balloon-expandable prosthesis in the PARTNER-I trial: 329 through transfemoral (TF-TAVR) and 202 transapical (TA-TAVR) access. Clinical events were adjudicated and echocardiographic results analyzed in a core laboratory. Quality of life (QoL) data were obtained up to 1 year post-TAVR. Time-varying all-cause mortality was referenced to that of an age-sex-race-matched US population. RESULTS: For TF-TAVR, post-procedure 30-day stroke risk was 3.6%; major adverse events occurred in 35% of patients; 30-day paravalvular leak was greater than moderate in 1.4%; median post-procedure length of stay (LOS) was 5 days. Thirty-day mortality was 4.0% and 3-year mortality 48% (44% for the matched population). By 6 months, most QoL measures had stabilized at a level considerably better than baseline, with Kansas City Cardiomyopathy Questionnaire (KCCQ) 72 ± 21. For TA-TAVR, post-procedure 30-day stroke risk was 2.0%; major adverse events 32%; 30-day paravalvular leak was greater than moderate in 0.61%; and median post-procedure LOS was 8 days. Thirty-day mortality was 12% and 3-year mortality 54% (42% for the matched population); KCCQ was 73 ± 23. CONCLUSIONS: A TAVR can be performed in nonagenarians with acceptable short- and mid-term outcomes. Although TF- and TA-TAVR outcomes are not directly comparable, TA-TAVR appears to carry a higher risk of early death without a difference in intermediate-term mortality. Age alone should not preclude referral for TAVR in nonagenarians.

Appropriate patient selection or health care rationing? Lessons from surgical aortic valve replacement in the Placement of Aortic Transcatheter Valves I trial.

OBJECTIVES: The study objectives were to (1) compare the safety of high-risk surgical aortic valve replacement in the Placement of Aortic Transcatheter Valves (PARTNER) I trial with Society of Thoracic Surgeons national benchmarks; (2) reference intermediate-term survival to that of the US population; and (3) identify subsets of patients for whom aortic valve replacement may be futile, with no survival benefit compared with therapy without aortic valve replacement. METHODS: From May 2007 to October 2009, 699 patients with high surgical risk, aged 84 ± 6.3 years, were randomized in PARTNER-IA; 313 patients underwent surgical aortic valve replacement. Median follow-up was 2.8 years. Survival for therapy without aortic valve replacement used 181 PARTNER-IB patients. RESULTS: Operative mortality was 10.5% (expected 9.3%), stroke 2.6% (expected 3.5%), renal failure 5.8% (expected 12%), sternal wound infection 0.64% (expected 0.33%), and prolonged length of stay 26% (expected 18%). However, calibration of observed events in this relatively small sample was poor. Survival at 1, 2, 3, and 4 years was 75%, 68%, 57%, and 44%, respectively, lower than 90%, 81%, 73%, and 65%, respectively, in the US population, but higher than 53%, 32%, 21%, and 14%, respectively, in patients without aortic valve replacement. Risk factors for death included smaller body mass index, lower albumin, history of cancer, and prosthesis-patient mismatch. Within this high-risk aortic valve replacement group, only the 8% of patients with the poorest risk profiles had estimated 1-year survival less than that of similar patients treated without aortic valve replacement. CONCLUSIONS: PARTNER selection criteria for surgical aortic valve replacement, with a few caveats, may be more appropriate, realistic indications for surgery than those of the past, reflecting contemporary surgical management of severe aortic stenosis in high-risk patients at experienced sites.

Aortic Dissection in Patients With Bicuspid Aortic Valve-Associated Aneurysms.

BACKGROUND: Data regarding the risk of aortic dissection in patients with bicuspid aortic valve and large ascending aortic diameter are limited, and appropriate timing of prophylactic ascending aortic replacement lacks consensus. Thus our objectives were to determine the risk of aortic dissection based on initial cross-sectional imaging data and clinical variables and to isolate predictors of aortic intervention in those initially prescribed serial surveillance imaging. METHODS: From January 1995 to January 2014, 1,181 patients with bicuspid aortic valve underwent cross-sectional computed tomography (CT) or magnetic resonance imaging (MRI) to ascertain sinus or tubular ascending aortic diameter greater than or equal to 4.7 cm. Random Forest classification was used to identify risk factors for aortic dissection, and among patients undergoing surveillance, time-related analysis was used to identify risk factors for aortic intervention. RESULTS: Prevalence of type A dissection that was detected by imaging or was found at operation or on follow-up was 5.3% (n = 63). Probability of type A dissection increased gradually at a sinus diameter of 5.0 cm--from 4.1% to 13% at 7.2 cm--and then increased steeply at an ascending aortic diameter of 5.3 cm--from 3.8% to 35% at 8.4 cm--corresponding to a cross-sectional area to height ratio of 10 cm(2)/m for sinuses of Valsalva and 13 cm(2)/m for the tubular ascending aorta. Cross-sectional area to height ratio was the best predictor of type A dissection (area under the curve [AUC] = 0.73). CONCLUSIONS: Early prophylactic ascending aortic replacement in patients with bicuspid aortic valve should be considered at high-volume aortic centers to reduce the high risk of preventable type A dissection in those with aortas larger than approximately 5.0 cm or with a cross-sectional area to height ratio greater than approximately 10 cm(2)/m.

Unexpected abrupt increase in left ventricular assist device thrombosis.

BACKGROUND: We observed an apparent increase in the rate of device thrombosis among patients who received the HeartMate II left ventricular assist device, as compared with preapproval clinical-trial results and initial experience. We investigated the occurrence of pump thrombosis and elevated lactate dehydrogenase (LDH) levels, LDH levels presaging thrombosis (and associated hemolysis), and outcomes of different management strategies in a multi-institutional study. METHODS: We obtained data from 837 patients at three institutions, where 895 devices were implanted from 2004 through mid-2013; the mean (±SD) age of the patients was 55±14 years. The primary end point was confirmed pump thrombosis. Secondary end points were confirmed and suspected thrombosis, longitudinal LDH levels, and outcomes after pump thrombosis. RESULTS: A total of 72 pump thromboses were confirmed in 66 patients; an additional 36 thromboses in unique devices were suspected. Starting in approximately March 2011, the occurrence of confirmed pump thrombosis at 3 months after implantation increased from 2.2% (95% confidence interval [CI], 1.5 to 3.4) to 8.4% (95% CI, 5.0 to 13.9) by January 1, 2013. Before March 1, 2011, the median time from implantation to thrombosis was 18.6 months (95% CI, 0.5 to 52.7), and from March 2011 onward, it was 2.7 months (95% CI, 0.0 to 18.6). The occurrence of elevated LDH levels within 3 months after implantation mirrored that of thrombosis. Thrombosis was presaged by LDH levels that more than doubled, from 540 IU per liter to 1490 IU per liter, within the weeks before diagnosis. Thrombosis was managed by heart transplantation in 11 patients (1 patient died 31 days after transplantation) and by pump replacement in 21, with mortality equivalent to that among patients without thrombosis; among 40 thromboses in 40 patients who did not undergo transplantation or pump replacement, actuarial mortality was 48.2% (95% CI, 31.6 to 65.2) in the ensuing 6 months after pump thrombosis. CONCLUSIONS: The rate of pump thrombosis related to the use of the HeartMate II has been increasing at our centers and is associated with substantial morbidity and mortality.

Impact of present-on-admission indicators on risk-adjusted hospital mortality measurement.

BACKGROUND: Benchmarking performance across hospitals requires proper adjustment for differences in baseline patient and procedural risk. Recently, a Risk Stratification Index was developed from Medicare data, which used all diagnosis and procedure codes associated with each stay, but did not distinguish present-on-admission (POA) diagnoses from hospital-acquired diagnoses. We sought to (1) develop and validate a risk index for in-hospital mortality using only POA diagnoses, principal procedures, and secondary procedures occurring before the date of the principal procedure (POARisk) and (2) compare hospital performance metrics obtained using the POARisk model with those obtained using a similarly derived model which ignored the timing of diagnoses and procedures (AllCodeRisk). METHODS: We used the 2004-2009 California State Inpatient Database to develop, calibrate, and prospectively test our models (n = 24 million). Elastic net logistic regression was used to estimate the two risk indices. Agreement in hospital performance under the two respective risk models was assessed by comparing observed-to-expected mortality ratios; acceptable agreement was predefined as the AllCodeRisk-based observed-to-expected ratio within ± 20% of the POARisk-based observed-to-expected ratio for more than 95% of hospitals. RESULTS: After recalibration, goodness of fit (i.e., model calibration) within the 2009 data was excellent for both models. C-statistics were 0.958 and 0.981, respectively, for the POARisk and AllCodeRisk models. The AllCodeRisk-based observed-to-expected ratio was within ± 20% of the POARisk-based observed-to-expected ratio for 89% of hospitals, which was slightly lower than the predefined limit of agreement. CONCLUSION: Consideration of POA coding meaningfully improved hospital performance measurement. The POARisk model should be used for risk adjustment when POA data are available.