Probability Revision. MED INF 406 Assignment 5. Golkonda, Jyothi 11/4/2012

Similar documents
Diagnostic Reasoning: Approach to Clinical Diagnosis Based on Bayes Theorem

MAESTRO TRIAL FINAL RESULTS. Gisela L.G. Menezes, MD, PhD

Understanding diagnostic tests. P.J. Devereaux, MD, PhD McMaster University

Meta-analysis of diagnostic research. Karen R Steingart, MD, MPH Chennai, 15 December Overview

Sensitivity, Specificity, and Relatives

Worksheet for Structured Review of Physical Exam or Diagnostic Test Study

INTRODUCTION TO MACHINE LEARNING. Decision tree learning

Sensitivity, Specificity and Predictive Value [adapted from Altman and Bland BMJ.com]

Evidence-Based Medicine: Diagnostic study

sickness, disease, [toxicity] Hard to quantify

The recommended method for diagnosing sleep

the standard deviation (SD) is a measure of how much dispersion exists from the mean SD = square root (variance)

METHODS FOR DETECTING CERVICAL CANCER

Example - Birdkeeping and Lung Cancer - Interpretation. Lecture 20 - Sensitivity, Specificity, and Decisions. What do the numbers not mean...

Statistics, Probability and Diagnostic Medicine

BMI 541/699 Lecture 16

Evalua&ng Methods. Tandy Warnow

Evaluation of a Portable Blood Lead Analyzer as an Alternative to Graphite Furnace Atomic Absorption Spectrophotometer

Data that can be classified as belonging to a distinct number of categories >>result in categorical responses. And this includes:

Diagnostic methods I: sensitivity, specificity, and other measures of accuracy

Supplemental Information

Baseline peripheral prolactin levels were normal in all occult and EAS patients. Four

The cross sectional study design. Population and pre-test. Probability (participants). Index test. Target condition. Reference Standard

I got it from Agnes- Tom Lehrer

Figure 1: Design and outcomes of an independent blind study with gold/reference standard comparison. Adapted from DCEB (1981b)

PERFORMANCE MEASURES

Hayden Smith, PhD, MPH /\ v._

Phone Number:

An Introduction to ROC curves. Mark Whitehorn. Mark Whitehorn

7/17/2013. Evaluation of Diagnostic Tests July 22, 2013 Introduction to Clinical Research: A Two week Intensive Course

Applying Data Mining for Epileptic Seizure Detection

Various performance measures in Binary classification An Overview of ROC study

Diagnostic Algorithms in VTE

OCW Epidemiology and Biostatistics, 2010 Michael D. Kneeland, MD November 18, 2010 SCREENING. Learning Objectives for this session:

Predictive Models for Healthcare Analytics

Chapter 10. Screening for Disease

Health Studies 315: Handouts. Health Studies 315: Handouts

Screening (Diagnostic Tests) Shaker Salarilak

CHAPTER 5 DECISION TREE APPROACH FOR BONE AGE ASSESSMENT

Evaluation of the Coris Legionella V-Test. In comparison with the Trinity Biotech EIA Legionella Urinary Antigen test

Vol. 51, No.2, February 1989 Copyright The American Fertility Society

Christina Martin Kazi Russell MED INF 406 INFERENCING Session 8 Group Project November 15, 2014

Solid Breast Nodules: Use of Sonography to Distinguish between Benign and Malignant Lesions

Brain Tumor segmentation and classification using Fcm and support vector machine

Assessment of diabetic retinopathy risk with random forests

When Overlapping Unexpectedly Alters the Class Imbalance Effects

An Improved Algorithm To Predict Recurrence Of Breast Cancer

Predicting Breast Cancer Survivability Rates

Diagnostic research designs: an introductory overview

MEDICAL SCORING FOR BREAST CANCER RECURRENCE. Nurul Husna bt Jamian UiTM (Perak), Tapah Campus

Overview. Goals of Interpretation. Methodology. Reasons to Read and Evaluate

Knowledge Discovery and Data Mining. Testing. Performance Measures. Notes. Lecture 15 - ROC, AUC & Lift. Tom Kelsey. Notes

RELIABILITY OF OPERATORS DURING THE VISUAL INSPECTION OF PRODUCED PARENTERAL DRUGS

Confirmation and Certainty in Toxicology Screening

Ricardo Bettencourt da Silva

Biochemical investigations in clinical medicine

MITOCW conditional_probability

Clinical decision making

A parable in four visits. A parable THE ARCHITECTURE OF MEDICAL TEST EVALUATIONS. παραβολή

THERAPEUTIC REASONING

Data Imbalance in Surveillance of Nosocomial Infections

Collaborating to Implement Evidence-Based Medicine Tools. The St. John Sepsis Agent and the Interdisciplinary Sepsis Advisor

Diagnosis and (early) prediction of acidosis and ketosis using milk fatty acids as biomarkers

Clinical Utility of Likelihood Ratios

Biosta's'cs Board Review. Parul Chaudhri, DO Family Medicine Faculty Development Fellow, UPMC St Margaret March 5, 2016

CHAPTER 4 PREDICTIVE VALUE OF EXFOLIATIVE CYTOL- OGY IN PIGMENTED CONJUNCTIVAL LESIONS

Molecular Testing for Indeterminate Thyroid Nodules. October 20, 2018

! Mainly going to ignore issues of correlation among tests

Automatic Detection of Heart Disease Using Discreet Wavelet Transform and Artificial Neural Network

EBM Diagnosis. Denise Campbell-Scherer Stefanie R. Brown. Departments of Medicine and Pediatrics University of Miami Miller School of Medicine

Clinical Decision Analysis

Performance Analysis of Different Classification Methods in Data Mining for Diabetes Dataset Using WEKA Tool

The Latest Technology from CareFusion

CHAPTER 6 HUMAN BEHAVIOR UNDERSTANDING MODEL

COMP90049 Knowledge Technologies

Week 2 Video 3. Diagnostic Metrics

Automatic Detection of Epileptic Seizures in EEG Using Machine Learning Methods

Bioengineering and World Health. Lecture Twelve

DECA Outcomes Report Program Year

An Introduction to Diagnostic Tests. Stephen D. Simon. The Children s Mercy Hospitals and Clinics

Systematic Reviews and meta-analyses of Diagnostic Test Accuracy. Mariska Leeflang

Lesson 87 Bayes Theorem

Dottorato di Ricerca in Statistica Biomedica. XXVIII Ciclo Settore scientifico disciplinare MED/01 A.A. 2014/2015

Evidence-based guidelines for diagnosis of common bile duct stones Vanja Giljaca University Hospital Center Rijeka Department of Gastroenterology

NAÏVE BAYES CLASSIFIER AND FUZZY LOGIC SYSTEM FOR COMPUTER AIDED DETECTION AND CLASSIFICATION OF MAMMAMOGRAPHIC ABNORMALITIES

SYSTEMATIC REVIEWS OF TEST ACCURACY STUDIES

ROI DETECTION AND VESSEL SEGMENTATION IN RETINAL IMAGE

Validity of Colposcopy in the Diagnosis of Early Cervical Neoplasia. Dr. Olaniyan Olayinka Babafemi

From 57 to 6 strategies: Use of economic evaluation methods to identify efficient diagnostic strategies BERNICE TSOI

PREDICTION OF BREAST CANCER USING STACKING ENSEMBLE APPROACH

S4. Summary of the GALNS assay validation. Intra-assay variation (within-run precision)

2011 ASCP Annual Meeting

Prediction of micrornas and their targets

PRECISION IMAGING: QUANTITATIVE, MOLECULAR AND IMAGE-GUIDED TECHNOLOGIES

Clinical Perspective. Interpreting Validity Indexes for Diagnostic Tests: An Illustration Using the Berg Balance Test

Assessment of significance of features acquired from thyroid ultrasonograms in Hashimoto s disease

Studies reporting ROC curves of diagnostic and prediction data can be incorporated into meta-analyses using corresponding odds ratios

Abstract. Introduction

RANDOLPH E. PATTERSON, MD, FACC, STEVEN F. HOROWITZ, MD, FACC*

Machine learning II. Juhan Ernits ITI8600

Transcription:

Probability Revision MED INF 406 Assignment 5 Golkonda, Jyothi 11/4/2012

Problem Statement Assume that the incidence for Lyme disease in the state of Connecticut is 78 cases per 100,000. A diagnostic test for the disease has a sensitivity of 81% and a specificity of 96%. Select two different probability revision techniques and use both to calculate: a) The probability that a person being tested in Connecticut has Lyme disease given a positive result for the test. b) The probability that the person has Lyme disease given a negative test result. 2 x 2 table Analysis True Positive A patient has a disease and the test result was positive False Positive A patient does not have a disease but the test result is positive False Negative A patient has a disease but the test result was negative True Negative A patient does not have a disease and the test result was negative 78 cases per 100,000 had Lyme disease. Hence percent of patients with Lyme disease = (78/100000) * 100 = 0.078% =( TP + FN) Percent of patients who do not have Lyme disease = 100 0.078 = 99.922% = (TN + FP) Sensitivity = 81% = 0.81 = TP/(TP+FN) TP/(TP+FN) = 0.81 TP/0.078 = 0.81 TP = 0.81 * 0.078 = 0.063% TP + FN = 0.078% 0.063 + FN = 0.078 FN = 0.078 0.063 = 0.015% Specificity = 96% = 0.96 = TN /(TN+FP) TN /(TN+FP) = 0.96 TN / 99.922 = 0.96 TN = 0.96 * 99.922 = 95.925 % FP = 99.922 95.925 = 3.997%

(Positive test) 4.06% Percent with Lyme disease (Disease) 0.078% 0.063% True positive (TP) Percent without Lyme disease (No Disease) 99.922% 3.997% False positive (FP) (Negative test) 95.94% 0.015% False Negative (FN) 95.925% True Negative (TN) Positive predictive value (Probability that the person actually has Lyme disease when the test result is positive) = TP/(TP + FP) = 0.063/(0.063 + 3.997) = 0.0155 or 1.55 % Negative predictive value (Probability that the person is disease free when the test result is negative) = TN/(TN + FN) = 95.925/(95.925 + 0.015) = 0.9998 or 99.98% Probability that the person has Lyme disease given a negative test result = FN/(TN+FN) = 0.015/(95.925 + 0.015) = 0.015/95.94 = 0.000156 = 0.0156% 2x2 Alternate calculations 78 cases out of 100,000 had Lyme disease. Number of cases who did not have Lyme disease = 100,000 78 = 99,922 Sensitivity of diagnostic test = 81% i.e. 81% of 78 cases will have a positive test result = 78 * (81/100) = 63 = True positive The rest of the 78 cases will show negative. False negative = 78-63 = 15 Specificity of the diagnostic test = 96% i.e. 96 % of 99,922 cases will have negative test result = 99922 * (96/100) = 95925 = True negative The rest of the cases will show positive when there is no disease.

False positive = 99922 95925 = 3997 Lyme disease (D+) No Lyme disease (D-) Total Positive test (T+) 63 True positive (TP) 3997 False positive (FP) 4060 Negative test (T-) 15 95925 95940 False Negative (FN) True Negative (TN) Total 78 99,922 100,000 Positive predictive value (Probability that the person actually has Lyme disease when the test result is positive) = TP/(TP + FP) = 63/(63 + 3997) = 63/4060 = 0.0155 or 1.55 % Probability that the person has Lyme disease given a negative test result = FN/(TN+FN) = 15/(95925 + 15) = 15/95940 = 0.000156 = 0.0156% Bayes' formula Postpositive test probability = Sensitivity * pretest probability (Sensitivity * pretest probability) + ((1-specificity) * (1-pretest probability)) Sensitivity = 81% = 0.81 Pretest probability = 78/100000 = 0.00078 Specificity = 96% = 0.96

Postpositive test probability = probability that the person has Lyme disease when the test result is positive = (0.81 * 0.00078)/ ((0.81 * 0.00078) + ((1-0.96) *(1-0.00078))) = 0.00063/(0.00063 + 0.03997) = 0.00063/0.0405988 = 0.0155 or 1.55% Postnegative test probability = Probability that the person has Lyme disease when the test result is negative = (1-Sensitivity) * pretest probability ((1-Sensitivity) * pretest probability) + (specificity * (1-pretest probability)) = (1-0.81) *0.00078/(((1-0.81) * 0.00078) + (0.96 * (1-0.00078)) = 0.19 * 0.00078/((0.19 * 0.00078) + (0.96 * 0.99922)) = 0.0001482/(0.0001482 + 0.9592512) = 0.0001482/0.9593994 = 0.00015 = 0.015% Decision Tree 78 out of 100,000 cases have LymeDisease In the Decision tree the initial node has the total population which is 100,000 The two chance nodes are cases with Lyme disease which is 78 Cases without Lyme disease will be 100,000-78 = 99,922 Hence the probability for each node is 0.00078 and 0.99922 respectively The sensitivity of the test is 81% The probability of the branch with positive test results for the chance node with Lyme disease is 0.81 The probability of the branch with negative test results = 1-0.81 = 0.19 The specificity of the test is 96% The probability of the branch with a negative test result on the chance node with no Lyme disease is 0.96 The probability of the branch with positive test result = 1 0.96 = 0.04 True Positives (TP) = 78 * 0.81 = 63 False Negative (FN) = 78 * 0.19 = 15 False Positive (FP) = 99922 * 0.04 = 3997 True Negative (TN) = 99922 * 0.96 = 95925

Lyme disease 0.00078 100,000 0.99922 No Lyme disease Positive Test 0.81 78 Negative Test 0.19 Positive Test 0.04 99922 Negative Test 0.96 True positive = 78 *0.81 = 63 False Negative = 15 False Positive = 3997 True Negative = 95925 Positive predictive value (Probability that the person actually has Lyme disease when the test result is positive) = TP/(TP + FP) = 63/(63 + 3997) = 63/4060 = 0.0155 or 1.55 % Probability that the person has Lyme disease given a negative test result = FN/(TN+FN) = 15/(95925 + 15) = 15/95940 = 0.000156 = 0.0156% The odds-likelihood form of Bayes' Likelihood Ratio = Sensitivity/ (1 specificity) = 0.81/(1-0.96) = 20.25 Pretest odds = P/(1 P) Pretest probability = 78/100,000 = 0.00078 Pretest odds = 0.00078/(1-0.00078) = 0.00078/0.99922 = 0.00078 Posttest odds = 20.25 * 0.00078 = 0.0156 Converting odds to probability = O/ (1 + O) = 0.0156/(1+0.0156) = 0.0156/1.0156 = 0.015 = 1.55% Probability that the person actually has Lyme disease when the test result is positive = 1.55% Probability that the person has Lyme disease given a negative test result Likelihood Ratio = (1- Sensitivity)/ (specificity) = 0.19/(0.96) = 0.198 Pretest odds = P/(1 P)

Pretest probability = 78/100,000 = 0.00078 Pretest odds = 0.00078/(1-0.00078) = 0.00078/0.99922 = 0.00078 Posttest odds = 0.198 * 0.00078 = 0.00015 Converting odds to probability = O/ (1 + O) = 0.00015/(1+0.00015) = 0.00015/1.00015 = 0.00015 = 0.015% Probability that the person has Lyme disease given a negative test result = 0.015%