Application of multi-variate receiver operating characteristic (ROC) curve analysis for clinical data Sarma K L A P1, Lakshmi P.2,* Department of Statistics, Sri Krishnadevaraya University, Anantapur-515003, Andhra Pradesh (State), India, email: klapsarma@yahoo.co.in 1Prof of Statistics, Department of Statistics, Sri Krishnadevaraya University, Anantapur-515003, Andhra Pradesh (State) 2Research Scholar, Department of Statistics, Sri Krishnadevaraya University, Anantapur-515003, Andhra Pradesh (State) *Email: lakshmi_sas@yahoo.com
Online published on 26 February, 2013. Abstract Earlier the authors proposed ROC curve analysis approach for analyzing 1000 Cancer Patients based on Blood Hemoglobin levels [1]. Usually condition of a Patient, is to be determined based on various characteristics like Blood Hemoglobin (BH) levels, Serum Protein (SP) levels, Serum Creatinine (SC) levels, Fasting Blood Glucose (FBG) levels, Blood Pressure (BPS) (Systolic) levels, Blood Pressure (BPD) (Diastolic) levels and so on. So far in literature Receiver Operating Characteristic Analysis is available using single variable in Medical Research for Decision Making [2] [3] and [4]. A Multi-Variate approach is essential to take valid decisions about the condition of the patient in many clinical trials. This leads us to propose a new approach known as “Multi-Variate Receiver Operating Characteristic Curve Analysis (MVROCCA)”. In this paper we have proposed the simplest Multi-Variate ROC curve approach based on three characteristics namely: Blood Hemoglobin (BH) Serum protein (SP) and Fasting Blood Glucose (FBG)
Data is collected on the above characteristics at two time points namely t0 and t1 where t1 = t0+6 months. Now Multi-Variate ROC curve approach for above data is considered in this paper and conclusions were drawn based on the results obtained for General Patients. Here General Patients implies both male and female Cancer patients. Multi-Variate ROC Curves were obtained separately at two time points t0 and t1 and comparisons were made critically on the Prognosis of Treatment in these 1000 cancer patients. Top Keywords ROC Curves, Threshold Value, TPR, FPR, and MVROCCA. Top |