Eiran Z. Gorodeski

Are All Readmissions Bad Readmissions

The authors examined the association between readmission and death within 30 days after hospitalization for heart failure. They found that a higher occurrence of readmissions was associated with lower 30-day mortality.

High-Dimensional Variable Selection for Survival Data

The minimal depth of a maximal subtree is a dimensionless order statistic measuring the predictiveness of a variable in a survival tree. We derive the distribution of the minimal depth and use it for high-dimensional variable selection using random survival forests. In big p and small n problems (where p is the dimension and n is the sample size), the distribution of the minimal depth reveals a “ceiling effect” in which a tree simply cannot be grown deep enough to properly identify predictive variables. Motivated by this limitation, we develop a new regularized algorithm, termed RSF-Variable Hunting. This algorithm exploits maximal subtrees for effective variable selection under such scenarios. Several applications are presented demonstrating the methodology, including the problem of gene selection using microarray data. In this work we focus only on survival settings, although our methodology also applies to other random forests applications, including regression and classification settings. All examples presented here use the R-software package randomSurvivalForest.

P wave duration and risk of longitudinal atrial fibrillation in persons ≥ 60 years old (from the Framingham Heart Study).

Long-term risk prediction is a priority for the prevention of atrial fibrillation (AF). P wave indices are electrocardiographic measurements describing atrial conduction. The role of P wave indices in the prospective determination of AF and mortality risk has had limited assessment. We quantified by digital caliper the P wave indices of maximum duration and dispersion in 1,550 Framingham Heart Study participants ≥ 60 years old (58% women) from single-channel electrocardiograms recorded from 1968 through 1971. We examined the association of selected P wave indices and long-term outcomes using Cox proportional hazards regression incorporating age, gender, body mass index, systolic blood pressure, treatment for hypertension, significant murmur, heart failure, and PR interval. Over a median follow-up of 15.8 years (range 0 to 38.7), 359 participants developed AF and 1,525 died. Multivariable-adjusted hazard ratios (HRs) per SD increase in maximum P wave duration were 1.15 (95% confidence interval [CI] 0.90 to 1.47, p = 0.27) for AF and 1.02 (95% CI 0.96 to 1.08, p = 0.18) for mortality. The upper 5% of P wave maximum duration had a multivariable-adjusted HR of 2.51 (95% CI 1.13 to 5.57, p = 0.024) for AF and an HR of 1.11 (95% CI 0.87 to 1.40, p = 0.20) for mortality. We found no significant associations between P wave dispersion with incidence of AF or mortality. In conclusion, maximum P wave duration at the upper fifth percentile was associated with long-term AF risk in an elderly community-based cohort. P wave duration is an electrocardiographic endophenotype for AF.

Multidetector Computed Tomographic Angiography in Planning of Reoperative Cardiothoracic Surgery

BACKGROUND Redo cardiothoracic surgery is associated with increased morbidity and mortality compared with primary operations. Multidetector computed tomographic angiography (MDCTA) delineates the course of previous coronary artery bypass grafts (CABG) and proximity of mediastinal structures to the chest wall. We sought to determine if high-risk preoperative MDCTA findings were associated with greater use of preventive surgical strategies during redo cardiac surgery in patients with prior CABG. METHODS We studied 167 patients (mean age 69 +/- 9 years, 79% men) with prior CABG, referred for redo cardiac surgery, who underwent contrast-enhanced MDCTA to assess CABG location and mediastinal relationship to chest wall. Preoperative risk was determined. Prevalence of high-risk MDCTA findings, use of preventive surgical strategies, frequency of severe intraoperative bleeding, and postoperative mortality were recorded. RESULTS Mean risk score was high (7.5 +/- 3). High-risk MDCTA findings included proximity (<1 cm) of right ventricle/aorta to chest wall (24%) or CABG crossing midline in close proximity (<1 cm anteroposteriorly) to sternum (38%). Preventive surgical strategies included surgery cancelled (4%), nonmidline incision (8%), deep hypothermic circulatory arrest (5%), initiation of peripheral cardiopulmonary bypass (11%) and extrathoracic vascular exposure before incision (53%). These strategies were used at a higher frequency in patients with high-risk MDCTA findings versus those without (88% versus 28%, p < 0.0001). Frequency of severe bleeding, graft injuries, and 1-month mortality were 4.4%, 5%, and 2.5%, respectively. CONCLUSIONS Routine use of preoperative MDCTA to detect high-risk findings has a strong association with adoption of preventive surgical strategies in high-risk patients undergoing redo cardiac surgery.

Prognosis on Chronic Dobutamine or Milrinone Infusions for Stage D Heart Failure

Background—There are no published clinical trials comparing dobutamine with milrinone in outpatients with stage D heart failure on continuous inotropes. Methods and Results—In a retrospective analysis of 112 inotrope-dependent patients with stage D heart failure who were not transplant candidates at enrollment, we investigated the relationship between choice of dobutamine or milrinone and mortality. Half the patients were on dobutamine (mean dose, 5.4±2.5 &mgr;g/kg per minute) and half on milrinone (mean dose, 0.4±0.2 &mgr;g/kg per minute). Those on dobutamine tended to be older (63 years old versus 54 years old), male (86% versus 79%), and fewer had implantable cardioverter-defibrillators (57% versus 74%). During a median follow-up time of 130 days (range, 2 to 2345 days), there were 85 deaths (76% of cohort) and 55 rehospitalizations. Use of dobutamine compared with milrinone was associated with higher all-cause mortality in an unadjusted analysis (hazard ratio [HR], 1.63; 95% CI, 1.06 to 2.52; P<0.03). However, this association was not significant after adjustment for baseline characteristics in the full cohort (N=112; HR, 0.99; 95% CI 0.5 to 1.97; P=0.98) or propensity-matched cohort (N=70; HR, 0.94; 95% CI 0.48 to 1.85; P=0.86). Conclusions—In this single-center retrospective study, there were no mortality differences between chronic intravenous dobutamine or milrinone in patients with stage D heart failure being discharged from the hospital. The high mortality in this group selected for inotrope dependence warrants careful consideration of all options and priorities for further care.

Application of the Seattle Heart Failure Model in Ambulatory Patients Presented to an Advanced Heart Failure Therapeutics Committee

Background—We sought to assess the predictive value of the Seattle Heart Failure Model (SHFM) when applied to ambulatory patients with advanced heart failure (HF) presented to an advanced HF therapeutics committee at a tertiary care US institution. Methods and Results—We evaluated model discrimination and calibration in 215 consecutive ambulatory patients who were presented to the Cleveland Clinic advanced HF therapeutics committee between 2004 to 2007 for evaluation for advanced options including transplantation and ventricular assist device (VAD). Analyses were stratified by committee decision (not listed versus listed United Network of Organ Sharing [UNOS] Status 2). Eighty-five percent had 1 or no missing SHFM variables. The primary outcome was a composite of all-cause mortality, VAD, or urgent (UNOS Status 1) transplantation. During a median follow-up of 24 months, 68 died, 18 received VAD support, and 81 underwent heart transplantation. Discrimination was modest both for those not listed (c-index, 0.683 at 1 year and 0.648 at 2 years), and for those listed UNOS status 2 (c-index, 0.629 at 1 year and 0.628 at 2 years). Calibration was acceptable among those patients not listed for heart transplantation but with substantial underestimation of risk (ie, overestimation of survival free of VAD or urgent transplantation) among UNOS status 2 patients. Conclusions—In ambulatory patients presented to an advanced HF therapeutics committee for evaluation for heart transplantation, the SHFM offers modest discrimination of risk for the primary composite outcome of mortality, VAD, or urgent transplantation, with underestimation of risk in those patients listed for nonurgent transplantation. Interpretation of risk prediction by the SHFM in this patient population must be done with caution.

Electrocardiographic prediction of abnormal genotype in congenital long QT syndrome: Experience in 101 related family members

Prediction of Congenital Long QT Syndrome. Introduction: Previous studies showed that diagnosing congenital long QT syndrome (LQTS) is difficult due to variable penetrance and genetic heterogeneity, especially when subjects from multiple families with diverse mutations are combined. We hypothesized that a combination of clinical and ECG techniques could identify gene carriers within a single family with congenital LQTS.

Long-Term Outcomes in High-Risk Symptomatic Patients With Hypertrophic Cardiomyopathy Undergoing Alcohol Septal Ablation

OBJECTIVES We sought to assess outcomes of alcohol septal ablation (ASA) in high-risk patients. BACKGROUND Because surgical myectomy is the preferred treatment in patients with symptomatic hypertrophic obstructive cardiomyopathy (HOCM) at our institution, we perform ASA in patients who are at high risk for surgery. METHODS We studied 55 symptomatic HOCM patients (mean age 63 +/- 13 years, 67% women, mean follow-up 8 +/- 1 years), at high risk for surgery (as the result of age/comorbidities) who had ASA between 1997 and 2000. The following were recorded at baseline, 3 months, and 1 year: septal thickness, maximal (resting or provocable) left ventricular outflow tract gradient, Minnesota living with heart failure questionnaire score, and the presence of a permanent pacemaker. All-cause mortality was recorded. RESULTS No patients died at 48 h, 2 died at 1 year, 7 died at 5 years, and 13 died at 10 years. Only age >65 years at time of ASA predicted long-term mortality (log-rank p = 0.03). Mean maximal left ventricular outflow tract gradient (104 +/- 35 mm Hg vs. 49 +/- 28 mm Hg), septal thickness (2.4 +/- 0.4 cm vs. 1.8 +/- 0.6 cm), and Minnesota living with heart failure score (63 vs. 25) improved at 3 months, compared with baseline (all p < 0.001), with no significant changes at 1 year. New permanent pacemaker was present in 26% of patients. CONCLUSIONS In symptomatic HOCM patients who are at high risk for surgery, ASA is associated with symptomatic improvement and low short-term mortality; with long-term mortality only associated with older age at time of procedure. In symptomatic HOCM patients at high-risk for surgery, ASA is a viable option.

P wave duration is associated with cardiovascular and all-cause mortality outcomes: the National Health and Nutrition Examination Survey

BACKGROUND P wave indices are an intermediate phenotype modulated by atrial conduction and electrophysiology. Their clinical correlates and association with all-cause mortality have received limited scrutiny. OBJECTIVE To determine the relationship between P wave indices and cardiovascular and all-cause mortality in the National Health and Nutrition Examination Survey (NHANES), a highly representative United States sample. METHODS NHANES III (1988-1994) quantified PR interval and P wave duration and amplitude. Mortality data through 2006 were obtained from National Death Index (NDI) records. RESULTS Of 8,561 subjects with electrocardiograms (ECGs), 7,486 (mean age 60.0 ± 13.3 years., 51.9% women, 50.1% ethnic minorities) had ECGs in sinus rhythm, linked mortality data, and complete assessments. Over a median 8.6-year follow-up (range 0.1-12.2 years), there were 679 cardiovascular deaths and 1,559 all-cause mortality deaths. Older age, male sex, and higher body mass index were significantly associated with greater PR interval and P wave duration and with lower P wave amplitude. African Americans had higher mean values of all three P wave indices. In a multivariable model adjusting for cardiovascular risk factors, P wave duration was the only P wave index significantly associated with cardiovascular mortality (hazard ratio [HR] 1.13, per 1 standard deviation [SD], 95% confidence interval [CI] 1.04-1.23; P = .004) and all-cause mortality (HR 1.06 per 1 SD; 95% CI 1.00-1.13; P = .050). CONCLUSIONS In a highly representative U.S. sample, P wave duration was significantly associated with increased cardiovascular and all-cause mortality. P wave duration may reflect subclinical disease and merits elucidation as a marker of risk for adverse outcomes.

Identifying Important Risk Factors for Survival in Patient With Systolic Heart Failure Using Random Survival Forests

Background— Heart failure survival models typically are constructed using Cox proportional hazards regression. Regression modeling suffers from a number of limitations, including bias introduced by commonly used variable selection methods. We illustrate the value of an intuitive, robust approach to variable selection, random survival forests (RSF), in a large clinical cohort. RSF are a potentially powerful extensions of classification and regression trees, with lower variance and bias. Methods and Results— We studied 2231 adult patients with systolic heart failure who underwent cardiopulmonary stress testing. During a mean follow-up of 5 years, 742 patients died. Thirty-nine demographic, cardiac and noncardiac comorbidity, and stress testing variables were analyzed as potential predictors of all-cause mortality. An RSF of 2000 trees was constructed, with each tree constructed on a bootstrap sample from the original cohort. The most predictive variables were defined as those near the tree trunks (averaged over the forest). The RSF identified peak oxygen consumption, serum urea nitrogen, and treadmill exercise time as the 3 most important predictors of survival. The RSF predicted survival similarly to a conventional Cox proportional hazards model (out-of-bag C-index of 0.705 for RSF versus 0.698 for Cox proportional hazards model). Conclusions— An RSF model in a cohort of patients with heart failure performed as well as a traditional Cox proportional hazard model and may serve as a more intuitive approach for clinicians to identify important risk factors for all-cause mortality.