Covid19 Survival

This document does the analysis of the patient level data set. The data was cloned from :https://github.com/beoutbreakprepared/nCoV2019 on April 2, 2020

Loading the libraries

library("FRESA.CAD")

Loading the data

I curated the original data: https://github.com/beoutbreakprepared/nCoV2019/tree/master/latest_data/latestdata.csv

The reported symtoms and chronic diseases were reclasified using the following keywords

Symptoms:

fever_chills: fever : chills
cough: cough
breath_dyspnea: severe: pneumonia: grasp: dyspnea: breath: acute: respiratory: gasp
fatigue: weakness: fatigue
pain: headache: myalgias: sore: mialgia: pain

Chronic Diseases:

hypertension: hypertension
renal: kidney renal
diabetes: diabetes
asthma: asthma

SymptomsOutcome <- read.delim("../Data/symptomsOutcome.txt", stringsAsFactors=FALSE)
SymptomsOutcome$Chronic_Disease <- NULL

nochronic <- SymptomsOutcome
nochronic$hypertension <- NULL;
nochronic$diabetes <- NULL;
nochronic$renal <- NULL;
nochronic$other_chronic <- NULL;
nochronic$asthma <- NULL;
nochronic$Country <- NULL;

nochronic.mat <- as.data.frame(model.matrix(Death ~ Age*.,nochronic))
nochronic.mat$`(Intercept)` <- NULL
nochronic.mat$Death <- as.numeric(nochronic$Death)
fnames <- colnames(nochronic.mat)
fnames <- str_replace_all(fnames," ","_")
fnames <- str_replace_all(fnames,"/","_")
fnames <- str_replace_all(fnames,":",".")
colnames(nochronic.mat) <- fnames

repetitions = 300
trainFraction = 0.90

##Basic Description


pander::pander(sum(table(SymptomsOutcome$Sex)))

174


plot(density(subset(SymptomsOutcome,Death==1)$Age),xlim=c(20,90),col="red",main="Age Distribution",xlab="Age",ylab="p(x)")
lines(density(subset(SymptomsOutcome,Death==0)$Age),xlim=c(20,90),col="blue")
legend("topleft",legend = c("Died","Discharged"),bty = "n",pch = 20,col = c("Red","Blue"))


op <- par(no.readonly = TRUE)


pander::pander(table(SymptomsOutcome$Country))

Table continues below
Brazil	China	France	Gambia	Germany	Guyana	Italy	Japan	Malaysia
4	59	2	1	1	2	1	10	2

Table continues below
Nepal	Philippines	Romania	Singapore	South Korea	Taiwan	Thailand
2	61	1	6	4	2	2

Vietnam
14

pander::pander(summary(SymptomsOutcome$Age))

Min.	1st Qu.	Median	Mean	3rd Qu.	Max.
0.25	44	61	57.18	72	89

pander::pander(table(SymptomsOutcome$Death))

0	1
73	101

pander::pander(table(SymptomsOutcome$Sex))

0	1
61	113

pander::pander(table(SymptomsOutcome$fever_chills))

0	1
84	90

pander::pander(table(SymptomsOutcome$cough))

0	1
109	65

pander::pander(table(SymptomsOutcome$breath_dyspnea))

0	1
126	48

pander::pander(table(SymptomsOutcome$fatigue))

0	1
160	14

pander::pander(table(SymptomsOutcome$pain))

0	1
153	21

pander::pander(table(SymptomsOutcome$other_symthom))

0	1
144	30

pander::pander(table(SymptomsOutcome$hypertension))

0	1
121	53

pander::pander(table(SymptomsOutcome$renal))

0	1
162	12

pander::pander(table(SymptomsOutcome$diabetes))

0	1
135	39

pander::pander(table(SymptomsOutcome$asthma))

0	1
170	4

pander::pander(table(SymptomsOutcome$other_chronic))

0	1
160	14



SymptomsOutcome$Country <- NULL

logistic Model

lmodel <- glm(Death~.,SymptomsOutcome,family=binomial(link=logit))

Logistic Results


cStats <- predictionStats_binary(cbind(SymptomsOutcome$Death,predict(lmodel,SymptomsOutcome)),plotname = "Logit");

Logit


sm <- summary(lmodel)
pander::pander(sm$coefficients)

	Estimate	Std. Error	z value	Pr(>\|z\|)
(Intercept)	-5.89	1.752	-3.361	0.000777
Age	0.1025	0.02326	4.408	1.046e-05
Sex	0.1624	0.617	0.2631	0.7924
fever_chills	-0.94	0.8155	-1.153	0.249
cough	-0.1283	0.6313	-0.2032	0.839
breath_dyspnea	2.621	1.198	2.188	0.02864
fatigue	0.2342	1.134	0.2066	0.8363
pain	0.2427	0.8971	0.2706	0.7867
other_symthom	-0.7833	1.196	-0.6549	0.5125
hypertension	1.594	1.017	1.568	0.1169
renal	17.73	3831	0.004629	0.9963
diabetes	18.78	2218	0.008467	0.9932
asthma	17.24	7148	0.002412	0.9981
other_chronic	0.02315	0.8975	0.02579	0.9794

Modeling

bm <- BSWiMS.model(Death~.,SymptomsOutcome,NumberofRepeats = -100)

Results

par(op)
pander::pander(bm$univariate)

	Name	RName	ZUni
Age	Age	Age	11.51
hypertension	hypertension	hypertension	8.914
breath_dyspnea	breath_dyspnea	breath_dyspnea	8.018
diabetes	diabetes	diabetes	7.931
fever_chills	fever_chills	fever_chills	5.846
cough	cough	cough	3.778
renal	renal	renal	3.672
pain	pain	pain	2.315
asthma	asthma	asthma	2.031
other_symthom	other_symthom	other_symthom	1.508
other_chronic	other_chronic	other_chronic	1.101
fatigue	fatigue	fatigue	0.6182
Sex	Sex	Sex	0.1304


cStats <- predictionStats_binary(cbind(SymptomsOutcome$Death,predict(bm,SymptomsOutcome)),plotname = "BSWiMS");

BSWiMS


sm <- summary(bm)
pander::pander(sm$coefficients)

Table continues below
	Estimate	lower	OR	upper	u.Accuracy
Age	0.09417	1.079	1.099	1.119	0.8213
breath_dyspnea	3.627	7.644	37.61	185	0.6724
diabetes	13.34	929.8	618842	411892132	0.6437
hypertension	1.455	2.04	4.285	9.001	0.7011
other_symthom	0.08956	1.033	1.094	1.158	0.4918
fever_chills	-0.08304	0.8688	0.9203	0.9749	0.6954
renal	0.3072	1.086	1.36	1.702	0.4885

Table continues below
	r.Accuracy	full.Accuracy	u.AUC	r.AUC	full.AUC
Age	0.7648	0.8711	0.8195	0.7894	0.8716
breath_dyspnea	0.8276	0.863	0.714	0.8323	0.8663
diabetes	0.831	0.8617	0.6931	0.835	0.8657
hypertension	0.8174	0.851	0.7388	0.8268	0.8572
other_symthom	0.8375	0.8651	0.5373	0.8376	0.8641
fever_chills	0.8377	0.8647	0.7034	0.8377	0.8666
renal	0.8332	0.8534	0.5594	0.8322	0.853

	IDI	NRI	z.IDI	z.NRI	Frequency
Age	0.2395	1.223	7.968	11.39	1
breath_dyspnea	0.08419	0.8704	4.371	8.299	1
diabetes	0.07193	0.7225	3.931	7.33	0.8202
hypertension	0.06645	0.8479	3.707	7.925	0.4941
other_symthom	0.03484	0.4061	2.898	3.613	0.07059
fever_chills	0.033	0.8195	2.666	6.498	0.08739
renal	0.03399	0.297	2.62	3.622	0.02185


par(op)

gplots::heatmap.2(bm$bagging$formulaNetwork,trace="none",mar = c(10,10),main = "B:SWiMS Formula Network",cexRow = 0.75,cexCol = 0.75)

Symptoms only Analysis

noChronicsmodel <- BSWiMS.model(Death~.,nochronic.mat,NumberofRepeats = -100)

Symptoms only Fit Results

pander::pander(noChronicsmodel$univariate)

	Name	RName	ZUni
Age	Age	Age	11.51
Age.breath_dyspnea	Age.breath_dyspnea	Age.breath_dyspnea	8.34
breath_dyspnea	breath_dyspnea	breath_dyspnea	8.018
fever_chills	fever_chills	fever_chills	5.846
cough	cough	cough	3.778
Age.Sex	Age.Sex	Age.Sex	2.878
Age.other_symthom	Age.other_symthom	Age.other_symthom	2.566
pain	pain	pain	2.315
Age.fever_chills	Age.fever_chills	Age.fever_chills	1.941
other_symthom	other_symthom	other_symthom	1.508
Age.cough	Age.cough	Age.cough	1.317
Age.pain	Age.pain	Age.pain	1.258
fatigue	fatigue	fatigue	0.6182
Age.fatigue	Age.fatigue	Age.fatigue	0.3703
Sex	Sex	Sex	0.1304


cStats <- predictionStats_binary(cbind(nochronic.mat$Death,predict(noChronicsmodel,nochronic.mat)),plotname = "BSWiMS");

BSWiMS


sm <- summary(noChronicsmodel)
pander::pander(sm$coefficients)

Table continues below
	Estimate	lower	OR	upper	u.Accuracy
Age	0.06828	1.059	1.071	1.083	0.8214
Age.fever_chills	0.03319	1.025	1.034	1.042	0.6525
Age.breath_dyspnea	0.07848	1.054	1.082	1.11	0.6724
breath_dyspnea	0.8563	1.542	2.354	3.596	0.6724
fever_chills	-2.838	0.02002	0.05855	0.1713	0.6954
Age.other_symthom	0.006055	1.003	1.006	1.009	0.5

Table continues below
	r.Accuracy	full.Accuracy	u.AUC	r.AUC
Age	0.7108	0.8621	0.8199	0.7234
Age.fever_chills	0.7221	0.813	0.6529	0.7336
Age.breath_dyspnea	0.7962	0.8259	0.714	0.7926
breath_dyspnea	0.8419	0.8616	0.714	0.8382
fever_chills	0.7915	0.8326	0.7034	0.8037
Age.other_symthom	0.7451	0.7884	0.556	0.7547

	full.AUC	IDI	NRI	z.IDI	z.NRI	Frequency
Age	0.864	0.2839	1.28	8.951	12.2	0.6327
Age.fever_chills	0.8088	0.182	1.008	6.711	8.777	0.2076
Age.breath_dyspnea	0.8304	0.1211	0.7915	5.087	7.383	0.7685
breath_dyspnea	0.8635	0.06985	0.7994	3.804	7.552	0.2036
fever_chills	0.8364	0.06905	0.7299	3.334	6.14	0.9142
Age.other_symthom	0.8009	0.0529	0.4554	3.139	4.128	0.2735


par(op)

gplots::heatmap.2(noChronicsmodel$bagging$formulaNetwork,trace="none",mar = c(10,10),main = "B:SWiMS Formula Network",cexRow = 0.75,cexCol = 0.75)

Cross Validation: Model with Chronic Disease


BSWiMScv <- randomCV(SymptomsOutcome,
                      "Death",
                      fittingFunction=BSWiMS.model,
                      trainFraction = trainFraction,
                      repetitions = repetitions,
                      NumberofRepeats = -1)

Chronic Disease Cross Validation Results

par(op)
cStats <- predictionStats_binary(BSWiMScv$testPredictions,plotname = "BSWiMS");

BSWiMS

BSWiMScv$jaccard

$Jaccard.SM [1] 0.8385182

$averageLength [1] 4.36

par(mar=c(4,10,4,4),pty="m")
barplot(BSWiMScv$featureFrequency/repetitions,xlim=c(0,1),las=2,cex.names =0.70,horiz = TRUE,main="Top Selected Features",xlab="Selection Frequency")

par(op)


fn <- baggedModel(BSWiMScv$selectedFeaturesSet,SymptomsOutcome,Outcome="Death")

…………………………


gplots::heatmap.2(fn$formulaNetwork,trace="none",mar = c(10,10),main = "B:SWiMS Formula Network",cexRow = 0.75,cexCol = 0.75)

BSWiMS Cross-Validation of only symptoms

SymBSWIMScv <- randomCV(nochronic.mat,
                        "Death",
                        fittingFunction=BSWiMS.model,
                        trainSampleSets=BSWiMScv$trainSamplesSets,
                        NumberofRepeats = -1)

Only Symptoms Results

par(op)
cStats <- predictionStats_binary(SymBSWIMScv$testPredictions,plotname = "BSWiMS");

BSWiMS

SymBSWIMScv$jaccard

$Jaccard.SM [1] 0.7740898

$averageLength [1] 5.346667

par(mar=c(4,10,4,4),pty="m")
barplot(SymBSWIMScv$featureFrequency/repetitions,xlim=c(0,1),las=2,cex.names =0.70,horiz = TRUE,main="Top Selected Features",xlab="Selection Frequency")

par(op)

fn <- baggedModel(SymBSWIMScv$selectedFeaturesSet,nochronic.mat,Outcome="Death")

…………………………


gplots::heatmap.2(fn$formulaNetwork,trace="none",mar = c(10,10),main = "B:SWiMS Formula Network",cexRow = 0.75,cexCol = 0.75)

SurvivalPredictor

Jose Tamez

4/3/2020

Covid19 Survival

Loading the data

logistic Model

Logistic Results

Modeling

Results

Symptoms only Analysis

Symptoms only Fit Results

Cross Validation: Model with Chronic Disease

Chronic Disease Cross Validation Results

BSWiMS Cross-Validation of only symptoms

Only Symptoms Results