ABSTRACT
BACKGROUND AIMS: Offering LT to frail patients may reduce waitlist mortality but may increase post-LT mortality. LT survival benefit is the concept of balancing these risks. We sought to quantify net survival benefit with LT by liver frailty index (LFI). APPROACH RESULTS: We analyzed data in the multi-center Functional Assessment in LT (FrAILT) Study from 2012-2021. Pre-LT cohort included ambulatory patients with cirrhosis awaiting LT, without hepatocellular carcinoma; post-LT cohort included those who underwent LT. Primary outcomes were pre-LT and post-LT mortality. We computed 1-, 3-, and 5-year restricted mean survival times (RMST) from adjusted Cox models. Survival benefit was calculated as net gain in life-years with LT. Pre-LT cohort included 2628 patients: median MELDNa was 18 (IQR 14-22); 731 (28%) were frail; 440 (17%) died pre-LT. Post-LT cohort included 1335 patients: median MELDNa was 20 (IQR 14-24); 325 (24%) were frail; 103 (8%) died post-LT. Pre-LT RMST decreased substantially as LFI increased. Post-LT RMST also decreased as LFI increased but only modestly. There was no LFI threshold at which pre-LT and post-LT RMST intersected-patients had net survival benefit at all LFI values. CONCLUSION: Pre-LT and, to a lesser degree, post-LT mortality increased as LFI increased. Transplant offered a survival benefit at all LFI values, driven by a reduction in pre-LT mortality. No threshold of LFI was identified at which the risk of post-LT mortality exceeded pre-LT mortality. LT offers net survival benefit even in the presence of advanced frailty among those selected for LT.
ABSTRACT
BACKGROUND: Low neighborhood income is linked with increased hospitalizations for central line-associated bloodstream infections (CLABSIs) in pediatric short bowel syndrome (SBS). We assessed whether this relationship varies by hospital center. METHODS: We performed a retrospective cohort study using the Pediatric Health Information System (2018-2023) database for patients <18 years old with SBS (N = 1210) at 24 hospitals in the United States. Using 2015 US Census data, we determined the estimated median household income of each patient's zip code. Hospital-level neighborhood income was defined as the median of the estimated median household income among patients at each hospital. We applied an extension of Cox regression to assess risk for CLABSI hospitalization. RESULTS: Among 1210 children with 5255 hospitalizations, most were <1 year on initial admission (53%), male (58%), and publicly insured (69%). Hospitals serving low-income neighborhoods served more female (46% vs 39%), Black (29% vs 22%), and Hispanic (22% vs 16%) patients with public insurance (72% vs 65%) residing in the southern United States (47% vs 21%). In univariate analysis, low hospital-level neighborhood income was associated with increased risk of CLABSI hospitalization (rate ratio [RR], 1.48; 95% CI, 1.21-1.83; P < 0.001). These findings persisted in multivariate analysis (RR, 1.43; 95% CI, 1.10-1.84; P < 0.01) after adjusting for race, ethnicity, insurance, region, and patient-level neighborhood income. CONCLUSION: Hospitals serving predominantly low-income neighborhoods bear a heavier burden of CLABSI hospitalizations for all their patients across the socioeconomic spectrum. Hospital initiatives focused on CLABSI prevention may be pivotal in addressing this disparity.
Subject(s)
Catheter-Related Infections , Hospitals , Short Bowel Syndrome , Socioeconomic Factors , Humans , Short Bowel Syndrome/epidemiology , Short Bowel Syndrome/complications , Retrospective Studies , Male , Female , Catheter-Related Infections/epidemiology , Infant , United States/epidemiology , Child, Preschool , Hospitals/statistics & numerical data , Child , Adolescent , Hospitalization/statistics & numerical data , Catheterization, Central Venous/adverse effects , Risk Factors , Cohort Studies , Infant, NewbornABSTRACT
We discuss 12 misperceptions, misstatements, or mistakes concerning the use of covariates in observational or nonrandomized research. Additionally, we offer advice to help investigators, editors, reviewers, and readers make more informed decisions about conducting and interpreting research where the influence of covariates may be at issue. We primarily address misperceptions in the context of statistical management of the covariates through various forms of modeling, although we also emphasize design and model or variable selection. Other approaches to addressing the effects of covariates, including matching, have logical extensions from what we discuss here but are not dwelled upon heavily. The misperceptions, misstatements, or mistakes we discuss include accurate representation of covariates, effects of measurement error, overreliance on covariate categorization, underestimation of power loss when controlling for covariates, misinterpretation of significance in statistical models, and misconceptions about confounding variables, selecting on a collider, and p value interpretations in covariate-inclusive analyses. This condensed overview serves to correct common errors and improve research quality in general and in nutrition research specifically.
Subject(s)
Observational Studies as Topic , Research Design , Humans , Research Design/standards , Models, Statistical , Data Interpretation, StatisticalABSTRACT
OBJECTIVES: Neighborhood contextual factors are associated with liver transplant outcomes. We analyzed associations between neighborhood-level socioeconomic status and healthcare utilization for pediatric liver transplant recipients. METHODS: We merged the Pediatric Health Information System and Scientific Registry of Transplant Recipients databases and included liver transplant recipients ≤21 years hospitalized between January 2004 and May 2022. Outcomes were annual inpatient bed-days, risk of hospitalizations, and risk of liver biopsies. The primary exposure was zip code-based neighborhood income at transplant. We applied causal inference for variable selection in multivariable analysis. We modeled annual inpatient bed-days with mixed-effect zero-inflated Poisson regression, and rates of hospitalization and liver biopsy with a Cox-type proportional rate model. RESULTS: We included 1006 participants from 29 institutions. Children from low-income neighborhoods were more likely to be publicly insured (67% vs. 46%), Black (20% vs. 12%), Hispanic (30% vs. 17%), and have higher model for end-stage liver disease/pediatric end-stage liver disease model scores at transplant (17 vs. 13) than the remaining cohort. We found no differences in inpatient bed-days or rates of hospitalization across neighborhood groups. In univariable analysis, low-income neighborhoods were associated with increased rates of liver biopsy (rate ratio [RR]: 1.57, 95% confidence interval [CI]: 1.04-2.34, p = 0.03). These findings persisted after adjusting for insurance, race, and ethnicity (RR: 1.86, 95% CI: 1.23-2.83, p < 0.01). CONCLUSIONS: Children from low-income neighborhoods undergo more liver biopsies than other children. These procedures are invasive and potentially preventable. In addition to improving outcomes, interventions to mitigate health inequities among liver transplant recipients may reduce resource utilization.
Subject(s)
Income , Liver Transplantation , Patient Acceptance of Health Care , Humans , Liver Transplantation/statistics & numerical data , Child , Male , Female , Adolescent , Income/statistics & numerical data , Child, Preschool , Patient Acceptance of Health Care/statistics & numerical data , Infant , United States , Neighborhood Characteristics/statistics & numerical data , Residence Characteristics/statistics & numerical data , Hospitalization/statistics & numerical data , Young Adult , Retrospective Studies , Healthcare Disparities/statistics & numerical dataABSTRACT
Mean arterial blood pressure (MAP), which decreases as portal hypertension progresses, may be a modifiable risk factor among patients with cirrhosis. We included adults enrolled in the Functional Assessment in Liver Transplantation study. We completed latent class trajectory analyses to define MAP trajectories. We completed time-dependent Cox-regression analyses to test the association between outpatient MAP and 3 cirrhosis-related outcomes: (1) stage 2 acute kidney injury (AKI), defined as a ≥200% increase in serum creatinine from baseline; (2) a 5-point increase in the MELD-Na score, defined as the incidence of increase from initial MELD-Na; (3) waitlist mortality, defined as death on the waitlist. For each outcome, we defined MAP cut points by determining the maximally selected Log-rank statistic after univariable Cox-regression analyses. Among the 1786 patients included in this analysis, our latent class trajectory analyses identified 3 specific outpatient MAP trajectories: "stable-low," "stable-high," and "increasing-to-decreasing." However, >80% of patients were in a "stable-low" trajectory. We found in adjusted analyses that outpatient MAP was associated with each of our outcomes: Stage 2 AKI (adjusted hazard ratio 0.88 per 10 mm Hg increase in MAP [95% CI: 0.79-0.99]); 5-point increase in MELD-Na (adjusted hazard ratio: 0.91 [95% CI: 0.86-0.96]; waitlist mortality (adjusted hazard ratio: 0.89 [95% CI: 0.81-0.96]). For each outcome, we found that an outpatient MAP of 82 mm Hg was most associated with outcomes ( p <0.05 for all). Our study informs the association between outpatient MAP and cirrhosis-related outcomes. These findings, coupled with the identification of specific thresholds, lay the foundation for the trial of targeted outpatient MAP modulation in patients with cirrhosis.
Subject(s)
Acute Kidney Injury , Arterial Pressure , Liver Cirrhosis , Liver Transplantation , Waiting Lists , Humans , Acute Kidney Injury/etiology , Acute Kidney Injury/mortality , Acute Kidney Injury/diagnosis , Acute Kidney Injury/epidemiology , Acute Kidney Injury/blood , Male , Female , Middle Aged , Liver Cirrhosis/mortality , Liver Cirrhosis/complications , Liver Cirrhosis/diagnosis , Risk Factors , Waiting Lists/mortality , Outpatients/statistics & numerical data , Aged , Hypertension, Portal/diagnosis , Hypertension, Portal/mortality , Hypertension, Portal/etiology , Hypertension, Portal/complications , Severity of Illness Index , Proportional Hazards Models , Creatinine/blood , Adult , Prospective Studies , Disease Progression , IncidenceABSTRACT
BACKGROUND AND AIMS: Existing tools for perioperative risk stratification in patients with cirrhosis do not incorporate measures of comorbidity. The Hospital Frailty Risk Score (HFRS) is a widely used measure of comorbidity burden in administrative dataset analyses. However, it is not specific to patients with cirrhosis, and application of this index is limited by its complexity. APPROACH AND RESULTS: Adult patients with cirrhosis who underwent nontransplant abdominal operations were identified from the National Inpatient Sample, 2016-2018. Adjusted associations between HFRS and in-hospital mortality and length of stay were computed with logistic and Poisson regression. Lasso regularization was used to identify the components of the HFRS most predictive of mortality and develop a simplified index, the cirrhosis-HFRS. Of 10,714 patients with cirrhosis, the majority were male, the median age was 62 years, and 32% of operations were performed electively. HFRS was associated with an increased risk of both in-hospital mortality (OR=6.42; 95% CI: 4.93, 8.36) and length of stay (incidence rate ratio [IRR]=1.79; 95% CI: 1.72, 1.88), with adjustment. Using lasso, we found that a subset of 12 of the 109 ICD-10 codes within the HFRS resulted in superior prediction of mortality in this patient population (AUC = 0.89 vs. 0.79, p < 0.001). CONCLUSIONS: While the 109-component HFRS was associated with adverse surgical outcomes, 12 components accounted for much of the association between the HFRS and mortality. We developed the cirrhosis-HFRS, a tool that demonstrates superior predictive accuracy for in-hospital mortality and more precisely reflects the specific comorbidity pattern of hospitalized patients with cirrhosis undergoing general surgery procedures.
ABSTRACT
OBJECTIVES: The phenomenon of vocal fatigue and the types of patients that are at greatest risk for vocal fatigue are not fully understood. The goal was to investigate patient profiles such as voice disorder type, demographics (age and gender), singing identity, interoceptive awareness, and psychosocial impacts on the severity of vocal fatigue. STUDY DESIGN: Prospective cohort study. METHODS: Ninety-five subjects with voice disorders were asked to complete Part 1 of the Vocal Fatigue Index (VFI-Part1), the Voice Handicap Index-10 (VHI-10), and the Multidimensional Assessment of Interoceptive Awareness, version 2 (MAIA-2). The effects of voice disorder type (structural, neurological, functional), psychosocial impact, age, gender, self-reported singing identity, and interoceptive awareness on self-perceived vocal fatigue (VFI-Part1) were determined using multivariate linear regression. RESULTS: Vocal fatigue had a significant psychosocial impact on patients with voice disorders, as measured by the VHI-10 (P < 0.001). However, there were no significant effects of vocal fatigue across any of the three voice disorder types (P's >0.05). Age (P = 0.220), gender (P = 0.430), and self-identified singing experience (P = 0.360) also did not have significant effects on vocal fatigue. Additionally, there were no significant relationships between interoceptive awareness MAIA-2 sum scores (P = 0.056) or any of the MAIA-2 sub-scores (P's > 0.05) and vocal fatigue severity (VFI-Part1). CONCLUSION: Vocal fatigue has a significant psychosocial impact on patients with voice disorders. However, patient profiles, including voice disorder type, patient age, gender, singing identity, and level of interoceptive awareness do not appear to play a major role in vocal fatigue symptom reporting. These findings suggest caution should be exercised when attributing patient profiles to vocal fatigue presentation and severity. Studying pathophysiological mechanisms underlying vocal fatigue may help better distinguish unconscious bias in patient profiling from the etiology and severity of vocal fatigue.
ABSTRACT
With the availability of massive amounts of data from electronic health records and registry databases, incorporating time-varying patient information to improve risk prediction has attracted great attention. To exploit the growing amount of predictor information over time, we develop a unified framework for landmark prediction using survival tree ensembles, where an updated prediction can be performed when new information becomes available. Compared to conventional landmark prediction with fixed landmark times, our methods allow the landmark times to be subject-specific and triggered by an intermediate clinical event. Moreover, the nonparametric approach circumvents the thorny issue of model incompatibility at different landmark times. In our framework, both the longitudinal predictors and the event time outcome are subject to right censoring, and thus existing tree-based approaches cannot be directly applied. To tackle the analytical challenges, we propose a risk-set-based ensemble procedure by averaging martingale estimating equations from individual trees. Extensive simulation studies are conducted to evaluate the performance of our methods. The methods are applied to the Cystic Fibrosis Foundation Patient Registry (CFFPR) data to perform dynamic prediction of lung disease in cystic fibrosis patients and to identify important prognosis factors.
ABSTRACT
BACKGROUND: Colorectal cancer is a leading cause of morbidity and mortality across U.S. racial/ethnic groups. Existing studies often focus on a particular race/ethnicity or single domain within the care continuum. Granular exploration of disparities among different racial/ethnic groups across the entire colon cancer care continuum is needed. We aimed to characterize differences in colon cancer outcomes by race/ethnicity across each stage of the care continuum. METHODS: We used the 2010-2017 National Cancer Database to examine differences in outcomes by race/ethnicity across six domains: clinical stage at presentation; timing of surgery; access to minimally invasive surgery; post-operative outcomes; utilization of chemotherapy; and cumulative incidence of death. Analysis was via multivariable logistic or median regression, with select demographics, hospital factors, and treatment details as covariates. RESULTS: 326,003 patients (49.6% female, 24.0% non-White, including 12.7% Black, 6.1% Hispanic/Spanish, 1.3% East Asian, 0.9% Southeast Asian, 0.4% South Asian, 0.3% AIAE, and 0.2% NHOPI) met inclusion criteria. Relative to non-Hispanic White patients: Southeast Asian (OR 1.39, p < 0.01), Hispanic/Spanish (OR 1.11 p < 0.01), and Black (OR 1.09, p < 0.01) patients had increased odds of presenting with advanced clinical stage. Southeast Asian (OR 1.37, p < 0.01), East Asian (OR 1.27, p = 0.05), Hispanic/Spanish (OR 1.05 p = 0.02), and Black (OR 1.05, p < 0.01) patients had increased odds of advanced pathologic stage. Black patients had increased odds of experiencing a surgical delay (OR 1.33, p < 0.01); receiving non-robotic surgery (OR 1.12, p < 0.01); having post-surgical complications (OR 1.29, p < 0.01); initiating chemotherapy more than 90 days post-surgery (OR 1.24, p < 0.01); and omitting chemotherapy altogether (OR 1.12, p = 0.05). Black patients had significantly higher cumulative incidence of death at every pathologic stage relative to non-Hispanic White patients when adjusting for non-modifiable patient factors (p < 0.05, all stages), but these differences were no longer statistically significant when also adjusting for modifiable factors such as insurance status and income. CONCLUSIONS: Non-White patients disproportionately experience advanced stage at presentation. Disparities for Black patients are seen across the entire colon cancer care continuum. Targeted interventions may be appropriate for some groups; however, major system-level transformation is needed to address disparities experienced by Black patients.
Subject(s)
Colonic Neoplasms , Ethnicity , Health Services Accessibility , Healthcare Disparities , Racial Groups , Female , Humans , Male , Black or African American/statistics & numerical data , Colonic Neoplasms/epidemiology , Colonic Neoplasms/ethnology , Colonic Neoplasms/mortality , Colonic Neoplasms/therapy , Ethnicity/statistics & numerical data , Healthcare Disparities/ethnology , Healthcare Disparities/standards , Healthcare Disparities/statistics & numerical data , Hispanic or Latino/statistics & numerical data , United States/epidemiology , Race Factors/statistics & numerical data , Treatment Outcome , Health Services Accessibility/statistics & numerical data , East Asian People/statistics & numerical data , Southeast Asian People/statistics & numerical data , South Asian People/statistics & numerical data , Native Hawaiian or Other Pacific Islander/statistics & numerical data , Asian/statistics & numerical data , Databases, Factual/statistics & numerical data , American Indian or Alaska Native/statistics & numerical data , Racial Groups/statistics & numerical dataABSTRACT
OBJECTIVES: Interoception may play a role in how individuals perceive their voice disorder. The first objective of this study was to investigate relationships between interoception and voice disorder class (functional, structural, neurological). The second objective was to determine relationships between interoception and voice-related outcome measures between patients with functional voice and upper airway disorders and typical voice users. The third objective was to determine whether patients with primary muscle tension dysphonia (a type of functional voice disorder) had different levels of interoceptive awareness than typical voice users. STUDY DESIGN: Prospective cohort study. METHODS: One hundred subjects with voice disorders completed the multidimensional assessment of interoceptive awareness-2 (MAIA-2). Voice diagnosis and singing experience were also acquired from each patient's medical chart. Voice handicap (VHI-10) and Part 1 of the vocal fatigue index (VFI-Part1) scores were obtained from patients diagnosed with functional voice and upper airway disorders. MAIA-2, VHI-10, VFI-Part1, and singing experience were also obtained from 25 typical voice users. Multivariable linear regression models were used to assess the association between response variables and voice disorder class, adjusting for singing experience, gender, and age. RESULTS: There were no significant group differences between voice disorder class (functional, structural, neurological) after adjusting for multiple comparisons. Participants with functional voice and upper airway disorders who scored significantly higher on the VHI-10 and VFI-Part1 had lower Attention Regulation sub-scores on the MAIA-2 (P's<0.05). Patients with primary muscle tension dysphonia scored significantly lower on the Emotional Awareness MAIA-2 subscale than typical voice users (P=0.005). CONCLUSION: Patients with functional voice disorders with lower capabilities to attend to body sensations may score higher on voice-related patient-reported outcome measures, like the VHI-10 and VFI-Part1. Patients with primary muscle tension dysphonia may also have lower capabilities in processing their body sensations than typical voice users.
ABSTRACT
Recurrent event analyses have found a wide range of applications in biomedicine, public health, and engineering, among others, where study subjects may experience a sequence of event of interest during follow-up. The R package reReg offers a comprehensive collection of practical and easy-to-use tools for regression analysis of recurrent events, possibly with the presence of an informative terminal event. The regression framework is a general scale-change model which encompasses the popular Cox-type model, the accelerated rate model, and the accelerated mean model as special cases. Informative censoring is accommodated through a subject-specific frailty without any need for parametric specification. Different regression models are allowed for the recurrent event process and the terminal event. Also included are visualization and simulation tools.
ABSTRACT
BACKGROUND: Recent developments to segment and characterize the regions of interest (ROI) within medical images have led to promising shape analysis studies. However, the procedures to analyze the ROI are arbitrary and vary by study. A tool to translate the ROI to analyzable shape representations and features is greatly needed. RESULTS: We developed SAFARI (shape analysis for AI-segmented images), an open-source R package with a user-friendly online tool kit for ROI labelling and shape feature extraction of segmented maps, provided by AI-algorithms or manual segmentation. We demonstrated that half of the shape features extracted by SAFARI were significantly associated with survival outcomes in a case study on 143 consecutive patients with stage I-IV lung cancer and another case study on 61 glioblastoma patients. CONCLUSIONS: SAFARI is an efficient and easy-to-use toolkit for segmenting and analyzing ROI in medical images. It can be downloaded from the comprehensive R archive network (CRAN) and accessed at https://lce.biohpc.swmed.edu/safari/ .
Subject(s)
Artificial Intelligence , Image Processing, Computer-Assisted , Algorithms , Glioblastoma , HumansABSTRACT
In biomedical studies, survival data with a cure fraction (the proportion of subjects cured of disease) are commonly encountered. The mixture cure and bounded cumulative hazard models are two main types of cure fraction models when analyzing survival data with long-term survivors. In this article, in the framework of the Cox proportional hazards mixture cure model and bounded cumulative hazard model, we propose several estimators utilizing pseudo-observations to assess the effects of covariates on the cure rate and the risk of having the event of interest for survival data with a cure fraction. A variable selection procedure is also presented based on the pseudo-observations using penalized generalized estimating equations for proportional hazards mixture cure and bounded cumulative hazard models. Extensive simulation studies are conducted to examine the proposed methods. The proposed technique is demonstrated through applications to a melanoma study and a dental data set with high-dimensional covariates.
Subject(s)
Models, Statistical , Neoplasms , Humans , Proportional Hazards Models , Computer Simulation , Survival AnalysisABSTRACT
Truncated survival data arise when the event time is observed only if it falls within a subject specific region. The conventional risk-set adjusted Kaplan-Meier estimator or Cox model can be used for estimation of the event time distribution or regression coefficient. However, the validity of these approaches relies on the assumption of quasi-independence between truncation and event times. One model that can be used for the estimation of the survival function under dependent truncation is a structural transformation model that relates a latent, quasi-independent truncation time to the observed dependent truncation time and the event time. The transformation model approach is appealing for its simple interpretation, computational simplicity and flexibility. In this paper, we extend the transformation model approach to the regression setting. We propose three methods based on this model, in addition to a piecewise transformation model that adds greater flexibility. We investigate the performance of the proposed models through simulation studies and apply them to a study on cognitive decline in Alzheimer's disease from the National Alzheimer's Coordinating Center. We have developed an R package, tranSurv, for implementation of our method.
ABSTRACT
Single-index models have gained increased popularity in time-to-event analysis owing to their model flexibility and advantage in dimension reduction. We propose a semiparametric framework for the rate function of a recurrent event counting process by modelling its size and shape components with single-index models. With additional monotone constraints on the two link functions for the size and shape components, the proposed model possesses the desired directional interpretability of covariate effects and encompasses many commonly used models as special cases. To tackle the analytical challenges arising from leaving the two link functions unspecified, we develop a two-step rank-based estimation procedure to estimate the regression parameters with or without informative censoring. The proposed estimators are asymptotically normal, with a root-n convergence rate. To guide model selection, we develop hypothesis testing procedures for checking shape and size independence. Simulation studies and a data example on a hematopoietic stem cell transplantation study are presented to illustrate the proposed methodology.
ABSTRACT
There is often delayed entry into observational studies, which results in left truncation. In the estimation of the distribution of time-to-event from left-truncated data, standard survival analysis methods require quasi-independence between the truncation time and event time. Incorrectly assuming quasi-independence may lead to biased estimation. We address the problem of estimation of the survival distribution when dependence between the event time and its left truncation time is induced by shared covariates. We introduce propensity scores for truncated data and propose two inverse probability weighting methods that adjust for both truncation and dependence, if all of the shared covariates are measured. The proposed methods additionally allow for right censoring. We evaluate the proposed methods in simulations, conduct sensitivity analyses, and provide guidelines for use in practice. We illustrate our approach in application to data from a central nervous system lymphoma study. The proposed methods are implemented in the R package, depLT.
Subject(s)
Models, Statistical , Survival Analysis , Propensity Score , Computer SimulationABSTRACT
Two major challenges arise in regression analyses of recurrent event data: first, popular existing models, such as the Cox proportional rates model, may not fully capture the covariate effects on the underlying recurrent event process; second, the censoring time remains informative about the risk of experiencing recurrent events after accounting for covariates. We tackle both challenges by a general class of semiparametric scale-change models that allow a scale-change covariate effect as well as a multiplicative covariate effect. The proposed model is flexible and includes several existing models as special cases, such as the popular proportional rates model, the accelerated mean model, and the accelerated rate model. Moreover, it accommodates informative censoring through a subject-level latent frailty whose distribution is left unspecified. A robust estimation procedure which requires neither a parametric assumption on the distribution of the frailty nor a Poisson assumption on the recurrent event process is proposed to estimate the model parameters. The asymptotic properties of the resulting estimator are established, with the asymptotic variance estimated from a novel resampling approach. As a byproduct, the structure of the model provides a model selection approach among the submodels via hypothesis testing of model parameters. Numerical studies show that the proposed estimator and the model selection procedure perform well under both noninformative and informative censoring scenarios. The methods are applied to data from two transplant cohorts to study the risk of infections after transplantation.
ABSTRACT
Tree-based methods are popular nonparametric tools in studying time-to-event outcomes. In this article, we introduce a novel framework for survival trees and ensembles, where the trees partition the dynamic survivor population and can handle time-dependent covariates. Using the idea of randomized tests, we develop generalized time-dependent receiver operating characteristic (ROC) curves for evaluating the performance of survival trees. The tree-building algorithm is guided by decision-theoretic criteria based on ROC, targeting specifically for prediction accuracy. To address the instability issue of a single tree, we propose a novel ensemble procedure based on averaging martingale estimating equations, which is different from existing methods that average the predicted survival or cumulative hazard functions from individual trees. Extensive simulation studies are conducted to examine the performance of the proposed methods. We apply the methods to a study on AIDS for illustration.
Subject(s)
Algorithms , Computer Simulation , ROC CurveABSTRACT
BACKGROUND: We conducted an observational study evaluating the association between uric acid, mean platelet volume (MPV), and high-density lipoprotein (HDL) with complications and outcomes of patients with sepsis in a critical care setting. METHODS: We followed patients with a diagnosis of severe sepsis and septic shock for a maximum of 28 days. Main outcomes assessed included length of stay (LOS), the need for renal replacement therapy (RRT), assisted mechanical ventilation (AMV), and vasopressor support as well as in-unit mortality. RESULTS: The overall average age of the 37 patients enrolled was 48.1 (19.8) years; among them, 37.8% were male. Abdominal related (43.2%) and pulmonary (29.7%) were the main sites of infection. The overall Acute Physiology and Chronic Health Evaluation 2 (APACHE-2) median score was 19 (9-24). Acute kidney injury (AKI) was observed in 46.9% of the sample. In all, 54.1% required vasopressor support, 54.1% AMV, and 35.1% RRT. Patients with bacteremia were significantly more likely to require vasopressor support and those with urinary tract infections were significantly younger. We found increasing ΔMPV levels, higher APACHE-2 scores, lower HDL values, and a reduced age to be associated with a longer LOS. Higher scores on the APACHE-2 scale and lower levels of HDL significantly associated with higher odds for developing AKI. The need for vasopressor support was significantly associated with higher values of 72-hour MPV and with higher levels of baseline uric acid and lower values of initial HCO3. Initial and 72-hour levels of MPV and higher scores in the APACHE-2 were all significantly correlated with the need for AMV. An increased probability of dying during follow-up was significantly correlated with increasing age. CONCLUSION: We were able to establish significant associations between our candidate biomarkers and relevant outcomes for patients with sepsis. Our results support the use of these low-cost biomarkers in the assessment of prognosis of patients with sepsis.
Subject(s)
Lipoproteins, HDL/blood , Mean Platelet Volume/mortality , Sepsis/blood , Sepsis/mortality , Uric Acid/blood , APACHE , Acute Kidney Injury/microbiology , Acute Kidney Injury/mortality , Adult , Aged , Biomarkers/blood , Critical Care/methods , Critical Care/statistics & numerical data , Critical Care Outcomes , Female , Humans , Length of Stay/statistics & numerical data , Male , Middle Aged , Renal Replacement Therapy/statistics & numerical data , Respiration, Artificial/statistics & numerical data , Sepsis/complicationsABSTRACT
PURPOSE: In several biomedical studies, one or more exposures of interest may be subject to nonrandom missingness because of the failure of the measurement assay at levels below its limit of detection. This issue is commonly encountered in studies of the metabolome using tandem mass spectrometry-based technologies. Owing to a large number of metabolites measured in these studies, preserving statistical power is of utmost interest. In this article, we evaluate the small sample properties of the missing indicator approach in logistic and conditional logistic regression models. METHODS: For nested case-control or matched case control study designs, we evaluate the bias, power, and type I error associated with the missing indicator method using simulation. We compare the missing indicator approach to complete case analysis and several imputation approaches. RESULTS: We show that under a variety of settings, the missing indicator approach outperforms complete case analysis and other imputation approaches with regard to bias, mean squared error, and power. CONCLUSIONS: For nested case-control and matched study designs of modest sample sizes, the missing indicator model minimizes loss of information and thus provides an attractive alternative to the oft-used complete case analysis and other imputation approaches.