Micha Mandel

Survival of critically ill patients hospitalized in and out of intensive care units under paucity of intensive care unit beds

Objective:The demand for intensive care beds far exceeds their availability in many European countries. Consequently, many critically ill patients occupy hospital beds outside intensive care units, throughout the hospital. The outcome of patients who fit intensive care unit admission criteria but are hospitalized in regular wards needs to be assessed for policy implications. The object was to screen entire hospital patient populations for critically ill patients and compare their 30-day survival in and out of the intensive care unit. Design:Screening teams visited every hospital ward on four selected days in five acute care Israeli hospitals. The teams listed all patients fitting a priori developed study criteria. One-month data for each patient were abstracted from the medical records. Setting:Five acute care Israeli hospitals. Patients:All patients fitting a priori developed study criteria. Interventions:None. Measurements and Main Results:Survival in and out of the intensive care unit was compared for screened patients from the day a patient first met study criteria. Cox multivariate models were constructed to adjust survival comparisons for various confounding factors. The effect of intensive care unit vs. other departments was estimated separately for the first 3 days after deterioration and for the remaining follow-up time. Results showed that 5.5% of adult hospitalized patients were critically ill (736 of 13,415). Of these, 27% were admitted to intensive care units, 24% to specialized care units, and 49% to regular departments. Admission to an intensive care unit was associated with better survival during the first 3 days of deterioration, after we adjusted for age and severity of illness (p = .018). There was no additional survival advantage for intensive care unit patients (p = .9) during the remaining follow-up time. Conclusions:The early survival advantage in the intensive care unit suggests a window of critical opportunity for these patients. Under economic constraints and dearth of intensive care unit beds, increasing the turnover of patients in the intensive care unit, thus exposing more needy patients to the early benefit of treatment in the intensive care unit, may be advantageous.

Survival of critically ill patients hospitalized in and out of intensive care

Objective:A lack of intensive care units beds in Israel results in critically ill patients being treated outside of the intensive care unit. The survival of such patients is largely unknown. The present studys objective was to screen entire hospitals for newly deteriorated patients and compare their survival in and out of the intensive care unit. Design:A priori developed intensive care unit admission criteria were used to screen, during 2 wks, the patient population for eligible incident patients. A screening team visited every hospital ward of five acute care hospitals daily. Eligible patients were identified among new admissions in the emergency department and among hospitalized patients who acutely deteriorated. Patients were followed for 30 days for mortality regardless of discharge. Setting:Five acute care hospitals. Patients:A total of 749 newly deteriorated patients. Interventions:None. Measurements and Main Results:Crude survival of patients in and out of the intensive care unit was compared by Kaplan-Meier curves, and Cox models were constructed to adjust the survival comparisons for residual case-mix differences. A total of 749 newly deteriorated patients were identified among 44,000 patients screened (1.7%). Of these, 13% were admitted to intensive care unit, 32% to special care units, and 55% to regular departments. Intensive care unit patients had better early survival (0–3 days) relative to regular departments (p = .0001) in a Cox multivariate model. Early advantage of intensive care was most pronounced among patients who acutely deteriorated while on hospital wards rather than among newly admitted patients. Conclusions:Only a small proportion of eligible patients reach the intensive care unit, and early admission is imperative for their survival advantage. As intensive care unit benefit was most pronounced among those deteriorating on hospital wards, intensive care unit triage decisions should be targeted at maximizing intensive care unit benefit by early admitting patients deteriorating on hospital wards.

Predicting short-term disability in multiple sclerosis

Objective: To develop covariate specific short-term disability curves to demonstrate the probability of progressing by Expanded Disability Status Scale (EDSS) at semiannual visits. Methods: Semiannual EDSS scores were prospectively collected in 218 relapsing-remitting (RR) and clinically isolated syndrome (CIS) patients as part of the Comprehensive Longitudinal Investigation of Multiple Sclerosis at the Brigham and Womens Hospital (CLIMB) study. Baseline brain parenchymal fraction (BPF) and T2 lesion volume were available on 205 patients. A partial proportional odds model determined the influence of covariates on the change in EDSS score at subsequent visits. A discrete second order Markov transitional model was fit and generated a probability matrix for each subject; the 6-month probabilities of EDSS change were graphically represented. Results: The univariate analysis demonstrated the lowest baseline BPF quartile (OR 1.99; p = 0.0203) and the highest T2 lesion volume quartile (OR 2.19; p = 0.0130) were associated with progression in EDSS. Covariate specific disability curves demonstrated the effect of BPF and T2 lesion volume on short-term progression. In subjects with a 6-month EDSS of 2, the probability of a sustained progression of an EDSS of 3 within 3 years was 0.277 for a subject with low BPF and a high T2 lesion volume vs 0.055 for a subject with high BPF and a low T2 lesion volume. Conclusions: Markov transitional models allow for the comparison of covariate specific short-term disability changes among groups of patients with multiple sclerosis.

Simultaneous confidence intervals based on the percentile bootstrap approach

This note concerns the construction of bootstrap simultaneous confidence intervals (SCI) for m parameters. Given B bootstrap samples, we suggest an algorithm with complexity of O(mB log(B)). We apply our algorithm to construct a confidence region for time dependent probabilities of progression in multiple sclerosis and for coefficients in a logistic regression analysis. Alternative normal based simultaneous confidence intervals are presented and compared to the bootstrap intervals.

Competitive exclusion, beta diversity, and deterministic vs. stochastic drivers of community assembly

Species diversity has two components - number of species and spatial turnover in species composition (beta-diversity). Using a field experiment focusing on a system of Mediterranean grasslands, we show that interspecific competition may influence the two components in the same direction or in opposite directions, depending on whether competitive exclusions are deterministic or stochastic. Deterministic exclusions reduce both patch-scale richness and beta-diversity, thereby homogenising the community. Stochastic extinctions reduce richness at the patch scale, but increase the differences in species composition among patches. These results indicate that studies of competitive effects on beta diversity may help to distinguish between deterministic and stochastic components of competitive exclusion. Such distinction is crucial for understanding the causal relationship between competition and species diversity, one of the oldest and most fundamental questions in ecology.

Simulation-Based Confidence Intervals for Functions With Complicated Derivatives

In many scientific problems, the quantity of interest is a function of parameters that index the model, and confidence intervals are constructed by applying the delta method. However, when the function of interest has complicated derivatives, this standard approach is unattractive and alternative algorithms are required. This article discusses a simple simulation-based algorithm for estimating the variance of a transformation, and demonstrates its simplicity and accuracy by applying it to several statistical problems.

The accelerated failure time model under biased sampling.

Chen (2009, Biometrics) studies the semi-parametric accelerated failure time model for data that are size biased. Chen considers only the uncensored case and uses hazard-based estimation methods originally developed for censored observations. However, for uncensored data, a simple linear regression on the log scale is more natural and provides better estimators.

Estimating time-to-event from longitudinal ordinal data using random-effects Markov models: application to multiple sclerosis progression

Longitudinal ordinal data are common in many scientific studies, including those of multiple sclerosis (MS), and are frequently modeled using Markov dependency. Several authors have proposed random-effects Markov models to account for heterogeneity in the population. In this paper, we go one step further and study prediction based on random-effects Markov models. In particular, we show how to calculate the probabilities of future events and confidence intervals for those probabilities, given observed data on the ordinal outcome and a set of covariates, and how to update them over time. We discuss the usefulness of depicting these probabilities for visualization and interpretation of model results and illustrate our method using data from a phase III clinical trial that evaluated the utility of interferon beta-1a (trademark Avonex) to MS patients of type relapsing-remitting.

Incidence and factors associated with treatment failure in the CLIMB multiple sclerosis cohort study

OBJECTIVE To determine the rate of treatment failure in patients outside of a controlled treatment trial and to ascertain the factors physicians used to make this decision. METHODS One hundred and thirty four patients with the diagnosis of relapsing-remitting (RR) multiple sclerosis (MS) or clinically isolated symptom (CIS) enrolled in the CLIMB study (Comprehensive Longitudinal Investigation of Multiple Sclerosis at the Brigham and Womens Hospital) were treated with either interferon beta or glatiramer acetate as their initial treatment for MS. RESULTS The probability of failing initial treatment within 3 years was 30%. Clinical activity, defined as relapses and/or progression in disability, determined treatment failure in 35.7% (n=10) of nonresponders. New T2 hyperintense or gadolinium-enhancing lesions on MRI was used to define treatment failure in 28.6% (n=8) and new MRI lesions were used in combination with clinical activity in 35.7% (n=10). Treatment failures had a higher T2 hyperintense lesion volume (p=0.015) and number of gadolinium-enhancing lesions (p=0.0001) on the enrollment MRI than responders. CONCLUSIONS These observations demonstrate that treating physicians use both clinical and MRI parameters to define a response to treatment and initiation of a treatment change and that baseline MRI identified those with increased risk of treatment failure.

Estimating Time to Event From Longitudinal Categorical Data: An Analysis of Multiple Sclerosis Progression.

The expanded disability status scale (EDSS) is an ordinal score that measures progression in multiple sclerosis (MS). Progression is defined as reaching EDSS of a certain level (absolute progression) or increasing EDSS by one point (relative progression). Survival methods for time to progression are not adequate for such data because they do not exploit the EDSS level at the end of follow-up. Instead, we suggest a Markov transitional model applicable for repeated categorical or ordinal data. This approach enables derivation of covariate-specific survival curves, obtained after estimation of the regression coefficients and manipulations of the resulting transition matrix. Large-sample theory and resampling methods are employed to derive pointwise confidence intervals, which perform well in simulation. Methods for generating survival curves for time to EDSS of a certain level, time to increase EDSS by at least one point, and time to two consecutive visits with EDSS greater than 3 are described explicitly. The regression models described are easily implemented using standard software packages. Survival curves are obtained from the regression results using packages that support simple matrix calculation. We present and demonstrate our method on data collected at the Partners Multiple Sclerosis Center in Boston. We apply our approach to progression defined by time to two consecutive visits with EDSS greater than 3 and calculate crude (without covariates) and covariate-specific curves.