Principles of Diagnostic Yields

Now that we have examined how to get examples of Endoscopic performance by analysis of documentation, we can perform the more interesting task of determining endoscopic performance by determining the pathology taken at endoscopy. This means we have to merge two datasets and then extract pathology based on the endoscopy results.

We will determine the dysplasia detection rate in our sample of Barrett’s endoscopies and we will do this per endoscopist so we can get an overview of who is the best at detecting dysplasia in Barrett’s oesopahgus. Note this is very similar to the concept of an adenoma detection rate in colonoscopy.

We start by defining our datasets. We have done this before as part of the Surveillance page so I’m not going to repeat it here. We can just start using it.

This is what the merged dataset looks like (truncated for ease of viewing):

EndoHistoMerge<-source('EndoPathMerged_ExternalCode.R')
EndoHistoMerge<-data.frame(EndoHistoMerge)
#Neaten up the names
names(EndoHistoMerge)<-gsub("value.","",names(EndoHistoMerge),fixed=T)
kable(head(EndoHistoMerge,5))

EndoHospNumId	EndoReports	Date.x	Endoscopist	Midazolam	Fentanyl	Indication	Diagnosis	BarrC	BarrM	Histop_dfHospNumId	HistoReport	Date.y	Macro	Diagnoses	Days	visible
U22415	Date of Procedure 2013-12-01 Endoscopist: Dr Bugs Bunny Midazolam: 4mg Fentanyl: 75mcg Indication: Dysphagia/Odynophagia Diagnosis: Esophageal candidiasis .Oesophagitis. .Hiatus Hernia. .Food bolus obstructing the oesophagus..Extensive neoplastic looking esophageal lesion. .Post chemo-radiotherapy stricture .Possible achalasia. Barrett’s oesophagus length: C0M6	2013-12-01	Bugs Bunny	4	75	Dysphagia/Odynophagia	Diagnosis: Esophageal candidiasis .Oesophagitis. .Hiatus Hernia. .Food bolus obstructing the oesophagus..Extensive neoplastic looking esophageal lesion. .Post chemo-radiotherapy stricture .Possible achalasia. Barrett’s oesophagus length: C0M6	0	6	U22415	Date received: 2013-11-29 Macrosopic description: 8 specimens collected the largest measuring 5 x 5 x 4 mm and the smallest 4 x 5 x 3 mm Diagnoses PAS staining shows occasional spores, consistent with candida..The biopsies of oesophageal squamous mucosa show surface erosion and active chronic inflammation..The appearances are those of Candida oesophagitis..Neither dysplasia nor malignancy is seen..No Helicobacter are seen..Numerous Candida spores and hyphae are present admixed with ulcer slough.	2013-11-29	8 specimens collected the largest measuring 5 x 5 x 4 mm and the smallest 4 x 5 x 3 mm	PAS staining shows occasional spores, consistent with candida..The biopsies of oesophageal squamous mucosa show surface erosion and active chronic inflammation..The appearances are those of Candida oesophagitis..Neither dysplasia nor malignancy is seen..No Helicobacter are seen..Numerous Candida spores and hyphae are present admixed with ulcer slough.	2 days	FALSE
U22415	Date of Procedure 2013-12-01 Endoscopist: Dr Bugs Bunny Midazolam: 4mg Fentanyl: 75mcg Indication: Dysphagia/Odynophagia Diagnosis: Esophageal candidiasis .Oesophagitis. .Hiatus Hernia. .Food bolus obstructing the oesophagus..Extensive neoplastic looking esophageal lesion. .Post chemo-radiotherapy stricture .Possible achalasia. Barrett’s oesophagus length: C0M6	2013-12-01	Bugs Bunny	4	75	Dysphagia/Odynophagia	Diagnosis: Esophageal candidiasis .Oesophagitis. .Hiatus Hernia. .Food bolus obstructing the oesophagus..Extensive neoplastic looking esophageal lesion. .Post chemo-radiotherapy stricture .Possible achalasia. Barrett’s oesophagus length: C0M6	0	6	U22415	Date received: 2013-12-02 Macrosopic description: 4 specimens collected the largest measuring 1 x 3 x 5 mm and the smallest 5 x 3 x 1 mm Diagnoses There is no intercellular oedema in the surface epithelium..Intestinal metaplasia is present..Neither dysplasia nor malignancy is seen.	2013-12-02	4 specimens collected the largest measuring 1 x 3 x 5 mm and the smallest 5 x 3 x 1 mm	There is no intercellular oedema in the surface epithelium..Intestinal metaplasia is present..Neither dysplasia nor malignancy is seen.	1 days	FALSE
O349253	Date of Procedure 2013-09-19 Endoscopist: Dr Sal Addin Midazolam: 2mg Fentanyl: 125mcg Indication: chronic abdo pain and constipaton Diagnosis: Hiatus Hernia. .Esophageal candidiasis .Ulcer- Oesophageal.	2013-09-19	Sal Addin	2	125	chronic abdo pain and constipaton	Diagnosis: Hiatus Hernia. .Esophageal candidiasis .Ulcer- Oesophageal.	NA	NA	O349253	Date received: 2013-09-17 Macrosopic description: 2 specimens collected the largest measuring 1 x 1 x 5 mm and the smallest 5 x 4 x 4 mm Diagnoses Neither dysplasia nor malignancy is seen..This is a dysplastic sample. There is no dysplasia and no invasive carcinoma..The appearances are consistent with, but not specific for Barrett’s (columnar lined) oesophagus..The biopsies of oesophageal squamous mucosa show surface erosion and active chronic inflammation..The appearances are consistent with the endoscopic diagnosis of Barrett’s oesophagus with active chronic inflammation..Basal hyperplasia is prominent.There is no significant increase in intraepithelial eosinophils.	2013-09-17	2 specimens collected the largest measuring 1 x 1 x 5 mm and the smallest 5 x 4 x 4 mm	Neither dysplasia nor malignancy is seen..This is a dysplastic sample. There is no dysplasia and no invasive carcinoma..The appearances are consistent with, but not specific for Barrett’s (columnar lined) oesophagus..The biopsies of oesophageal squamous mucosa show surface erosion and active chronic inflammation..The appearances are consistent with the endoscopic diagnosis of Barrett’s oesophagus with active chronic inflammation..Basal hyperplasia is prominent.There is no significant increase in intraepithelial eosinophils.	2 days	FALSE
O349253	Date of Procedure 2013-09-19 Endoscopist: Dr Sal Addin Midazolam: 2mg Fentanyl: 125mcg Indication: chronic abdo pain and constipaton Diagnosis: Hiatus Hernia. .Esophageal candidiasis .Ulcer- Oesophageal.	2013-09-19	Sal Addin	2	125	chronic abdo pain and constipaton	Diagnosis: Hiatus Hernia. .Esophageal candidiasis .Ulcer- Oesophageal.	NA	NA	O349253	Date received: 2013-09-17 Macrosopic description: 5 specimens collected the largest measuring 1 x 2 x 3 mm and the smallest 4 x 1 x 2 mm Diagnoses The appearances are consistent with the endoscopic diagnosis of Barrett’s oesophagus with active chronic inflammation..Basal hyperplasia is prominent.No Helicobacter are seen..There is no dysplasia or malignancy..Neither dysplasia nor malignancy is seen..The biopsies of oesophageal squamous mucosa show surface erosion and active chronic inflammation..There is some ulceration..This is a dysplastic sample	2013-09-17	5 specimens collected the largest measuring 1 x 2 x 3 mm and the smallest 4 x 1 x 2 mm	The appearances are consistent with the endoscopic diagnosis of Barrett’s oesophagus with active chronic inflammation..Basal hyperplasia is prominent.No Helicobacter are seen..There is no dysplasia or malignancy..Neither dysplasia nor malignancy is seen..The biopsies of oesophageal squamous mucosa show surface erosion and active chronic inflammation..There is some ulceration..This is a dysplastic sample	2 days	FALSE
P433224	Date of Procedure 2014-08-12 Endoscopist: Dr Elmo Fudd Midazolam: 1mg Fentanyl: 150mcg Indication: Reflux-like Symptoms/Atypical Chest Pain Diagnosis: Gastritis.Post chemo-radiotherapy stricture Barrett’s oesophagus length: C1M5	2014-08-12	Elmo Fudd	1	150	Reflux-like Symptoms/Atypical Chest Pain	Diagnosis: Gastritis.Post chemo-radiotherapy stricture Barrett’s oesophagus length: C1M5	1	5	P433224	Date received: 2014-08-10 Macrosopic description: 3 specimens collected the largest measuring 3 x 1 x 2 mm and the smallest 4 x 4 x 4 mm Diagnoses There is no dysplasia or malignancy..There is no significant increase in intraepithelial eosinophils..The biopsies of oesophageal squamous mucosa show surface erosion and active chronic inflammation..Intestinal metaplasia is present..There is some ulceration.. PAS staining shows occasional spores, consistent with candida..The appearances are consistent with, but not specific for Barrett’s (columnar lined) oesophagus..The appearances are those of Candida oesophagitis..No granulomas or viral inclusions are seen.	2014-08-10	3 specimens collected the largest measuring 3 x 1 x 2 mm and the smallest 4 x 4 x 4 mm	There is no dysplasia or malignancy..There is no significant increase in intraepithelial eosinophils..The biopsies of oesophageal squamous mucosa show surface erosion and active chronic inflammation..Intestinal metaplasia is present..There is some ulceration.. PAS staining shows occasional spores, consistent with candida..The appearances are consistent with, but not specific for Barrett’s (columnar lined) oesophagus..The appearances are those of Candida oesophagitis..No granulomas or viral inclusions are seen.	2 days	FALSE

Breakdown of analytical method for yield

So the task now involves extracting the presence of dysplasia in those patients where Barrett’s was detected at endoscopy and then grouping them by endoscopist

So to spell out where each element of this comes from:

1. Detect rows that mention Barrett’s in the endoscopy report.

2. Also detect rows that mention dysplasia in the pathology report.

3. Group by endoscopist.

4. Then get the total number of reports that mention Barrett’s in the endoscopy report whether dyplasia is metioned or not, and group by endoscopist.

5. Calculate proportion by endoscopist then visualise it.

1 & 2. Detect rows that mention term of interest in the endoscopy report.

This is done using grepl. We will do a combined grepl so we can get the subset we are interested in and that mention dysplasia in the pathology report.

DysplasticBarretts<-EndoHistoMerge[grepl("[Bb]arrett",EndoHistoMerge$EndoReports)&grepl("[Dd]ysplasi",EndoHistoMerge$HistoReport),]

So you will note that there are several reports where it is mentioned that there is no dysplasia so we have to get rid of these. This is tricky but for the purposes of this site we will use the brute force technique

DysplasticBarretts<-DysplasticBarretts[!grepl("[Nn]either dysplasia",DysplasticBarretts$Diagnoses)&!grepl("[Nn]o [Dd]ysplasia",DysplasticBarretts$Diagnoses),]

3. Group by endoscopist

There are two ways of doing this. The first way is to use dplyr as follows:

EndoscopistDDRBarretts<-DysplasticBarretts%>%group_by(Endoscopist)%>%summarise(n=n())

kable(EndoscopistDDRBarretts)

Endoscopist	n
Bilbo Baggins	95
Bugs Bunny	111
Charles Dickens	82
Chubby Checker	97
Davy Jones	98
Elmo Fudd	112
Elvis Presley	94
Florence Nightingale	109
Frank Sinatra	90
Jimminey Cricket	114
Jonny Begood	75
Joseph Conrad	95
King Richard III	87
Rara Rasputin	95
Sal Addin	119

Alternatively we can just use the table function which has a neater input but gives a messier output:

DDRtable<-table(EndoscopistDDRBarretts)
kable(DDRtable)

	75	82	87	90	94	95	97	98	109	111	112	114	119
Bilbo Baggins	0	0	0	0	0	1	0	0	0	0	0	0	0
Bugs Bunny	0	0	0	0	0	0	0	0	0	1	0	0	0
Charles Dickens	0	1	0	0	0	0	0	0	0	0	0	0	0
Chubby Checker	0	0	0	0	0	0	1	0	0	0	0	0	0
Davy Jones	0	0	0	0	0	0	0	1	0	0	0	0	0
Elmo Fudd	0	0	0	0	0	0	0	0	0	0	1	0	0
Elvis Presley	0	0	0	0	1	0	0	0	0	0	0	0	0
Florence Nightingale	0	0	0	0	0	0	0	0	1	0	0	0	0
Frank Sinatra	0	0	0	1	0	0	0	0	0	0	0	0	0
Jimminey Cricket	0	0	0	0	0	0	0	0	0	0	0	1	0
Jonny Begood	1	0	0	0	0	0	0	0	0	0	0	0	0
Joseph Conrad	0	0	0	0	0	1	0	0	0	0	0	0	0
King Richard III	0	0	1	0	0	0	0	0	0	0	0	0	0
Rara Rasputin	0	0	0	0	0	1	0	0	0	0	0	0	0
Sal Addin	0	0	0	0	0	0	0	0	0	0	0	0	1

4. Get the total number of specific endoscopies by endoscopist:

So now we know who is picking up dysplasia we can express this as a proportion of all the Barrett’s endoscopy they have done as follows:

AllBarretts<-EndoHistoMerge[grepl("[Bb]arrett",EndoHistoMerge$EndoReports),]
Endoscopist_All_Barretts<-AllBarretts%>%group_by(Endoscopist)%>%summarise(n=n())

5. DDR by endoscopist:

Now we just calulate the proportions to get the DDR. We have to bind Endoscopist_All_Barretts and EndoscopistDDRBarretts to calculate this. We merge by endoscopist

#For a bit of variety we are going to do the merge using dplyr join functions instead of base R merge functions:
DDRTable<-full_join(Endoscopist_All_Barretts, EndoscopistDDRBarretts, by = "Endoscopist")

And finally the proportions:

DDRTable$Prop<-(DDRTable$n.y/DDRTable$n.x)*100
DDRTable<-data.frame(DDRTable)
#Lets get rid of NA values by replacing with "0"
DDRTable$Prop[is.na(DDRTable$Prop)] <- 0

#Lets plot it out
barplot(DDRTable$Prop,names.arg=DDRTable$Endoscopist, ylab = "% dysplasia",
        cex.lab = 1.5,cex.axis=1.0,cex.main = 1.0,cex.names=1.0,main = "Barrett's Endoscopic Dysplasia detection proportion",las=2)