Ilona Verburg Nicolette de Keizer Niels Peek

Slides:



Advertisements
Similar presentations
Chapter 3 Properties of Random Variables
Advertisements

Regression Analysis Once a linear relationship is defined, the independent variable can be used to forecast the dependent variable. Y ^ = bo + bX bo is.
Systemic inflammatory response syndrome score at admission independently predicts mortality and length of stay in trauma patients. by R2 黃信豪.
Part V The Generalized Linear Model Chapter 16 Introduction.
Centre for Health Economics Modelling the impact of being obese on hospital costs Katharina Hauck Bruce Hollingsworth.
1-1 Regression Models  Population Deterministic Regression Model Y i =  0 +  1 X i u Y i only depends on the value of X i and no other factor can affect.
Regression Analysis. Unscheduled Maintenance Issue: l 36 flight squadrons l Each experiences unscheduled maintenance actions (UMAs) l UMAs costs $1000.
Basic Statistical Concepts Psych 231: Research Methods in Psychology.
Basic Statistical Concepts
Log-linear and logistic models
Generalised linear models Generalised linear model Exponential family Example: logistic model - Binomial distribution Deviances R commands for generalised.
The Impact of Insurance Status on Hospital Treatment and Outcomes David Card, Carlos Dobkin and Nicole Maestas.
Basic Statistical Concepts Part II Psych 231: Research Methods in Psychology.
Validation of predictive regression models Ewout W. Steyerberg, PhD Clinical epidemiologist Frank E. Harrell, PhD Biostatistician.
Regression Model Building Setting: Possibly a large set of predictor variables (including interactions). Goal: Fit a parsimonious model that explains variation.
Severity Distributions for GLMs: Gamma or Lognormal? Presented by Luyang Fu, Grange Mutual Richard Moncher, Bristol West 2004 CAS Spring Meeting Colorado.
DOES MEDICARE SAVE LIVES?
Recommendations on Minimum Data Recording Requirements in Hospitals from the Directorate of Health in Iceland: Is it possible to use Hospital Patient Registry.
A Primer on the Exponential Family of Distributions David Clark & Charles Thayer American Re-Insurance GLM Call Paper
Marketing Research Aaker, Kumar, Day and Leone Tenth Edition
© 2005 The McGraw-Hill Companies, Inc., All Rights Reserved. Chapter 12 Describing Data.
Simple Linear Regression
1 Is Managed Care Superior to Traditional Fee-For-Service among HIV-Infected Beneficiaries of Medicaid? David Zingmond, MD, PhD UCLA Division of General.
Early surgery for proximal femoral fractures is associated with lower complication and mortality rates Parag Kumar Jaiswal Arthroplasty Fellow.
Biostatistics Case Studies 2015 Youngju Pak, PhD. Biostatistician Session 4: Regression Models and Multivariate Analyses.
Chapter 3: Generalized Linear Models 3.1 The Generalization 3.2 Logistic Regression Revisited 3.3 Poisson Regression 1.
Randomized controlled trial to evaluate a focused communication intervention to reduce length of stay for critically ill children in a pediatric intensive.
Microeconometric Modeling William Greene Stern School of Business New York University.
1 Experimental Statistics - week 10 Chapter 11: Linear Regression and Correlation.
ALISON BOWLING THE GENERAL LINEAR MODEL. ALTERNATIVE EXPRESSION OF THE MODEL.
Basics of Regression Analysis. Determination of three performance measures Estimation of the effect of each factor Explanation of the variability Forecasting.
Specific Aim 1: Determine the impact of psychiatric disorders on the hospital length of stay (LOS) in pediatric patients diagnosed with SCD admitted for.
Research Process Parts of the research study Parts of the research study Aim: purpose of the study Aim: purpose of the study Target population: group whose.
Weekend & Night Outcomes in a Mature State Trauma System Brendan G. Carr, MD MS Department of Emergency Medicine Department of Biostatistics and Epidemiology.
Time – Immortal Bias in the analysis of “Influenza and COPD Mortality Protection as Pleiotropic, Dose-dependent effects of statins” by Floyd J, Frost et.
AUTHOR: MORAR ANICUȚA IONELA COORDINATOR: COPOTOIU MONICA COAUTHOR: ROMAN NICOLETA GRANCEA IULIA.
Prognostic models in the ICU From development to clinical practice L. Minne, MSc. Dr. S. Eslami, PharmD Dr. D.A. Dongelmans, MD Prof. Dr. S.E.J.A. de Rooij,
© Department of Statistics 2012 STATS 330 Lecture 20: Slide 1 Stats 330: Lecture 20.
June 11, 2008Stat Lecture 10 - Review1 Midterm review Chapters 1-5 Statistics Lecture 10.
September 18-19, 2006 – Denver, Colorado Sponsored by the U.S. Department of Housing and Urban Development Conducting and interpreting multivariate analyses.
Evaluating Risk Adjustment Models Andy Bindman MD Department of Medicine, Epidemiology and Biostatistics.
AUTHOR: MORAR ANICUȚA IONELA COAUTHOR: ROMAN NICOLETA GRANCEA IULIA COORDINATOR: COPOTOIU MONICA.
Econometrics Course: Cost as the Dependent Variable (I) Paul G. Barnett, PhD November 20, 2013.
Raghavan Murugan, MD, MS, FRCP Associate Professor of Critical Care Medicine, and Clinical & Translational Science Core Faculty, Center for Critical Care.
Advanced Statistical Methods: Continuous Variables REVIEW Dr. Irina Tomescu-Dubrow.
Psychology 202a Advanced Psychological Statistics October 22, 2015.
Love does not come by demanding from others, but it is a self initiation. Survival Analysis.
Armando Teixeira-Pinto AcademyHealth, Orlando ‘07 Analysis of Non-commensurate Outcomes.
Variance Stabilizing Transformations. Variance is Related to Mean Usual Assumption in ANOVA and Regression is that the variance of each observation is.
Assumptions of Multiple Regression 1. Form of Relationship: –linear vs nonlinear –Main effects vs interaction effects 2. All relevant variables present.
The Impact of Insurance Status on Hospital Treatment and Outcomes David Card, Carlos Dobkin and Nicole Maestas.
1 WIES Estimation Raymond Robbins. 2 WIES Estimation Objective Estimate WIES for uncoded episodes — Separated episodes — Current inpatients (WIP WIES)
Proportional Hazards Model Checking the adequacy of the Cox model: The functional form of a covariate The link function The validity of the proportional.
1/61: Topic 1.2 – Extensions of the Linear Regression Model Microeconometric Modeling William Greene Stern School of Business New York University New York.
Approaches to quantitative data analysis Lara Traeger, PhD Methods in Supportive Oncology Research.
Multi-state piecewise exponential model of hospital outcomes after injury DE Clark, LM Ryan, FL Lucas APHA 2007.
PANDHARIPANDE PP ET AL. N ENGL J MED 2013; 369: Long-Term Cognitive Impairment after Critical Illness.
CMS SAS Users Group Conference Learn more about THE POWER TO KNOW ® October 17, 2011 Medicare Payment Standardization Modeling using SAS Enterprise Miner.
Date of download: 6/26/2016 From: Variations in Mortality and Length of Stay in Intensive Care Units Ann Intern Med. 1993;118(10): doi: /
Biostatistics Class 3 Probability Distributions 2/15/2000.
 Naïve Bayes  Data import – Delimited, Fixed, SAS, SPSS, OBDC  Variable creation & transformation  Recode variables  Factor variables  Missing.
Bootstrap and Model Validation
135th Annual Meeting of APHA, November 3-7, 2007 Washington DC Session
Eastern Michigan University
Statistics in MSmcDESPOT
Microeconometric Modeling
Microeconometric Modeling
What is Regression Analysis?
Risk adjustment using administrative and clinical data: model comparison
Alcoholic liver disease in intensive care
Presentation transcript:

Ilona Verburg Nicolette de Keizer Niels Peek Comparison of different statistical methods to predict Intensive Care Length of Stay Ilona Verburg Nicolette de Keizer Niels Peek Dept. Of Medical Informatics Academic Medical Center University of Amsterdam The Netherlands ESCTAIC 2012,Timisoara

Background and objective Intensive Care Units (ICUs) assess their performance to improve quality and reduce costs Background Efficiency of care Effectiveness of care Case mix mortality length of stay 4-10-2012

Background and objective ICU Length of stay is influenced by case mix. Example: Length of stay (mean) 10 days 5 days Age (mean) 68 57 Medical vs surgical 80% medical 40% medical admission type (%) 20% surgical 60% surgical 4-10-2012

Background and objective ICU Observed outcome Compare Case mix Predictive model Case mix Expected outcome 4-10-2012

Background and objective Models exist to predict ICU mortality (example APACHE IV) Few models exist to predict ICU Length of Stay (LoS) No consensus about best modelling method Objective Compare the performance of different statistical regression methods to predict ICU LoS. 4-10-2012

Data NICE registry Dutch National Intensive Care Evaluation (NICE) Registry of ICU admissions in the Netherlands (since 1996) All admissions from (voluntary) participating ICUs (>90%) Evaluating (systematically) the effectiveness and efficiency of ICUs in the Netherlands Identifying quality of care problems Quality assurance Database 4-10-2012

Data Data Patients admitted to ICUs participating NICE 2009 - 2011 Included patients 94,251 (42.4%) admissions Exclusion criteria APACHE IV exclusion criteria elective surgery 81,190 (86.1%) survivors 13,061 (13.9%) non-survivors 4-10-2012

Length of stay Distribution of Length of Stay in fractional days ICU non-survivors (n= 13,061) ICU survivors (n= 81,190) Median: 2.4 (days) Mean: 5.9 Standard deviation: 10.2 Maximum: 139.0 Median: 1.7 (days) Mean: 4.2 Standard deviation: 8.2 Maximum: 326.6 Plaatjes nog vervangen en mediaan en mean etc nog invullen…. 4-10-2012

ICU Length of Stay Distribution of discharge time

Modeling ICU length of stay Different methods to model ICU length of stay (in fractional days) Ordinary least square (OLS) regression LoS and Log-transformed LoS Most frequently used method in literature 4-10-2012

Modeling ICU length of stay Different methods to model ICU length of stay (in fractional days) Ordinary least square (OLS) regression LoS and Log-transformed LoS General linear models (GLM) Gaussian - difference with OLS is the log link function Gamma - LoS time until discharge - depending on chosen parameters positively skewed Poisson - LoS count data `-depending on chosen parameters positively skewed - property: expectation = variance → overdispersion Negative binomial - count data -depending on chosen parameters positively skewed - generalisation of poisson 4-10-2012

Modeling ICU length of stay Different methods to model ICU length of stay (in fractional days) Ordinary least square (OLS) regression LoS and Log-transformed LoS General linear models (GLM) 4 different families Gaussian Gamma Poisson negative binomial Cox proportional Hazard (Cox PH) regression No assumptions on the shape of the distribution Omits the need of transform the outcome 4-10-2012

Modeling ICU length of stay Selection of covariates Starting with large set of variables Known relationship with LoS (literature) Stepwise backwards elimination of variables Included case mix Demographics Age Gender Admission type Diagnoses (APACHE IV) Severity of illness (APACHE IV severity-of-illness score) Different comorbidities (21) 4-10-2012

Validation Performance measures Good prediction Squared Pearson correlation = R2 = High ↑ Root Mean squared prediction error (RMSPE) = Low ↓ Low ↓ - or + Relative BIAS = Relative mean absolute prediction error (MAPE) = Low ↓ 4-10-2012

Validation Validation Performance measures calculated on original data Correcting for optimistic bias 100 bootstrap samples 4-10-2012

Results coefficients Covariates survivors OLS reg los OLS reg log los GLM: gaussian GLM: poisson GLM: negative binomial GLM: Gamma Cox PH chronic dialysis -1.04 -0.16 -0.25 -0.26 -0.28 0.31 cva 0.74 0.1 0.13 0.18 0.26 -0.3 diabetes -0.34 -0.01 -0.07 -0.06 -0.04 0.03 resperatory insufficient 0.38 0.06 0.09 0.15 -0.11 spline Aps (1) 5.55 0.64 1.74 1.65 1.61 -1.52 spline Aps (2) 11.07 1.09 3.16 2.78 2.64 -2.57 spline Aps (3) 15.98 0.99 2.07 2 2.08 -1.79 Covariates non-survivors OLS reg los OLS reg log los GLM: gaussian GLM: poisson GLM: negative binomial GLM: Gamma Cox PH chronic dialysis 0.15 0.08 cva -0.68 -0.18 -0.15 -0.12 0.09 diabetes 0.35 0.03 0.05 0.06 -0.05 resperatory insufficient -0.51 -0.03 -0.11 -0.1 -0.09 0.07 spline Aps (1) -5.59 -0.43 -0.94 -0.84 -0.8 0.7 spline Aps (2) -6.08 -0.73 -1.09 -1.26 -1.53 -1.55 1.54 spline Aps (3) -6.47 -1.64 -1.76 -1.87 -1.88 1.83 4-10-2012

Results validation ICU survivors Mean observed > mean expected   R2 RMSPE Relative BIAS Relative MAPE OLS regression (LoS) 0.174 7.448 0.008 0.812 OLS regression (log(LoS)) 0.183 7.714 -0.400 0.674 GLM Gaussian 0.197 7.335 0.001 0.771 GLM Poisson 0.194 7.349 0.000 0.769 GLM Negative Binomial 0.186 7.388 0.005 0.773 GLM Gamma 0.184 7.407 Cox PH regression 0.097 9.002 -0.693 0.938 Mean observed > mean expected Underestimation of mean LoS 4-10-2012

Results validation ICU non-survivors R2 RMSPE Relative BIAS   R2 RMSPE Relative BIAS Relative MAPE OLS regression (LoS ) 0.107 9.618 0.005 0.891 OLS regression (log(LoS)) 10.213 -0.510 0.762 GLM Gaussian 0.134 9.462 -0.009 0.868 GLM Poisson 0.128 9.504 0.000 0.872 GLM Negative Binomial 0.12 9.545 -0.001 GLM Gamma 0.112 9.602 0.877 Cox PH regression 0.075 11.388 -0.808 0.906 4-10-2012

Conclusion and discussion Difficult to predict ICU LoS Influenced by admission and discharge policy Seasonal pattern for admission and discharge time Skewed to the right GLM models shows best performance Poorest performance found for Cox PH regression Large relative bias was found for OLS regression of log-transformed LoS Differences in performance between models not statistically tested 4-10-2012

Conclusion and discussion Similar study for CABG patients (Austin et al.), with comparable results Different patient type Different distribution of length of stay Future research Different models for survivors and non-survivors combining with mortality in one prediction Statistical methods to predict ICU LoS developing a model for benchmarking purposes 4-10-2012

Thank you for your attention! Questions? 4-10-2012

APACHE IV Exclusiecriteria Age < 16 ICU admission < 4 hours Hospital admission >365 days Died during admission Readmissions Admissions from CCU/IC other hospital No diagnose Burns Transplantations Missing hospital discharge 4-10-2012