Introduction
Colonoscopy is essential for diagnosis and treatment of colonic diseases and plays
a key role in colorectal cancer (CRC) screening, resulting in effective reduction
of incidence and related mortality [1]
[2]. However, poor-quality colonoscopy is associated with missed lesions, contributing
to increased rates of interval cancers [3]. This has led to many recent publications on quality in colonoscopy and attempts
to implement quality monitoring in endoscopy units worldwide. Various national and
international professional societies have defined several QI to measure the quality
of practice and compare local performance and benchmarks. This offers the possibility,
by feedback to improve colonoscopy quality [4]
[5]
[6] and finally patient outcomes [7].
Initiation of continuous quality monitoring suffers from constraints related to both
technical and human resources. Recently, the European Society of Gastrointestinal
Endoscopy (ESGE) published a list of requirements and standards for endoscopy reporting
systems to ensure adequate quality assessment and improvement [8]. They recommend restricted use of free text and also structured and standardized
data entry. They also discourage separation of databases and support automatic linkage
between an endoscopy reporting system and other databases, with inclusion of information
on histopathology of detected lesions. The drawback of the adenoma detection rate
(ADR) monitoring, considered as one of the main QI in lower gastrointestinal endoscopy
[9], is the need to obtain this automatic linkage to avoid manual data entry of pathological
data, which is a clear human and economic limit in QI monitoring implementation. This
critical point has been considered as an urgent need in colonoscopy quality monitoring
[10]. However, until now, there have been no published data on implementation of a system
in endoscopy units that automatically links endoscopy report results and pathologic
data from resected polyps to enhance quality monitoring in colonoscopy.
In the present study, we report on the feasibility of implementing a colonoscopy QI
recording system by automatically extracting data from two separate databases to obtain
endoscopic and pathological information.
Material and methods
Electronic reporting system use and adaptation
During 2016 and 2017, the reporting system used in our endoscopy unit was the version
13.5 Endobase software system (Olympus, Winter & Ibe GmbH, Hamburg, Germany). It was
configured to include patientʼs general characteristics, technical information, and
results from the procedure, incorporated by the endoscopist. The pathology software
used was the Diamic software system, version 8 (Infologic-santé, Valence, France).
Before the procedure, the software automatically enters information on patient identification
and the type of scope (with serial number) used, whereas information on the physicians
and staff involved were recorded manually. During the endoscopy procedure, photographs
of anatomic landmarks and lesions are taken and saved in the software. After completion
of the colonoscopy, endoscopists choose photographs to include in the final report.
Then they add in appropriate tabs information on the indication of colonoscopy, type
of sedation, and other medication administered during colonoscopy, type and quality
of bowel preparation, normal and pathological findings, procedure indicators (PI),
biopsies, endoscopic treatments, conclusions, and clinical advice. Data are systematically
entered by selecting options from a drop-down menu, with predefined and structured
terminology, limiting free text to a minimum. In the case of polyps, different structured
data are added in dedicated tabs to record their size, morphology (using different
classifications as Paris, NICE, laterally spreading tumor type), and also information
about the polypectomy technique and results. Once all description information has
been entered, a written report is generated and displayed in a word processor to allow
for possible modifications and selected endoscopy photographs.
To ensure continuous measurement of colonoscopy performance, this commercially available
reporting system was locally adapted by adding a dedicated tab, named quality monitoring,
with a specific tab for each selected QI. Some are automatically added from previously
entered items (i. e., type of sedation, level of progression and quality of bowel
preparation); others are added at the end of the procedure by the endoscopist (indication
for colonoscopy and number of polyp(s) resected).
Quality indicators
Based on the current literature and international guidelines, we selected several
endoscopy QI that are recognized to be key performance measures, associated with clinically
relevant outcomes, and available in routine clinical practice. They include ADR (i. e.
the rate of colonoscopies with at least one adenoma identified), cecal intubation
rate (CIR) (i. e. the percentage of complete colonoscopies, defined as reaching and
visualizing the whole cecum and its landmarks) and the proportion of colonoscopies
with adequate bowel preparation. We used the Boston Bowel Preparation Scale (BBPS)
to evaluate the quality of preparation, with adequate preparation defined by a score
better than 5 [4]
[11]. We also included the polyp detection rate (PDR) (i. e. the rate of colonoscopies
in which at least one polyp was identified).
The ADR was broken down into several indications. As recommended [4], we measured ADR for all patients aged 50 years or older (i. e. ADR 50). We also
calculated the ADR for screening colonoscopies performed in asymptomatic patients
with average risk for colorectal cancer (i. e. ADR screening) and also in all colonoscopies
irrespective of the indication for the procedure (including screening colonoscopy,
follow-up colonoscopy, and diagnostic and therapeutic colonoscopy, i. e. ADR total).
Datas extraction and analysis
On a regular basis or on request, PI encoded during each colonoscopy are automatically
extracted from the Endobase software to a first database (the endoscopic database)
([Table 1]). In parallel, pathology results from resected polyps are extracted from Diamic
software in a second database (the pathology database) ([Table 2]) through the use of SNOMED CT codes to recognize each type of lesion and its site.
A list of “trigger” SNOMED CT codes corresponding to different adenomatous lesions
were chosen after discussion with our pathology team ([Table 3]). When one of these codes is selected by the pathologist, they automatically trigger
a positive result on the corresponding colonoscopy. Only the codes corresponding to
an adenomatous lesion can trigger this result. Structured query language (SQL) is
used to extract data from both databases. Each PI extracted from the databases is
used to calculate a corresponding QI. To automatically calculate the ADR, merging
of two databases is necessary and performed based on the patient identifier (unique
for each patient and used by all software in our hospital) and the date of the procedure.
Merger of the two databases creates a third database ([Table 4]), from which all QI (including ADR) are automatically calculated for each physician
and for the whole unit. All of these steps are summarized in [Fig. 1].
Table 1
Example of a procedure in the endoscopic database.
|
Patient ID
|
Age
|
Procedure Date
|
Procedure Type
|
Endoscopist ID
|
PI type
|
PI result
|
|
Patient 1
|
51
|
dd/mm/yy
|
Colonoscopy
|
17
|
Type of sedation
|
Propofol sedation
|
|
Patient 1
|
51
|
dd/mm/yy
|
Colonoscopy
|
17
|
Level of progression
|
Cecum
|
|
Patient 1
|
51
|
dd/mm/yy
|
Colonoscopy
|
17
|
Indication
|
Screening (average risk)
|
|
Patient 1
|
51
|
dd/mm/yy
|
Colonoscopy
|
17
|
Number of polyp resected
|
2
|
|
Patient 1
|
51
|
dd/mm/yy
|
Colonoscopy
|
17
|
Preparation
|
9
|
|
Patient 1
|
51
|
dd/mm/yy
|
Colonoscopy
|
17
|
Planned procedure
|
Total Colonoscopy
|
Table 2
Example of a pathology result (corresponding to [Table 1]) in the pathology database.
|
Patient ID
|
Reception date
|
Morphology code
|
Site code
|
|
Patient 1
|
dd/mm/yy
|
M-74002
|
T-59300
|
|
Patient 1
|
dd/mm/yy
|
M-74002
|
T-59300
|
Table 3
List of “trigger” SNOMED CT codes.
|
Morphology code SNOMED CT
|
Corresponding pathology
|
|
M-74000
|
Dysplasia
|
|
M-74001
|
Mild dysplasia
|
|
M-74002
|
Moderate dysplasia
|
|
M-74003
|
Severe dysplasia
|
|
M-81403
|
Adenocarcinoma
|
|
M-84803
|
Colloid adenocarcinoma
|
|
M-82633
|
Papillary adenocarcinoma
|
Table 4
Example of result from merging of databases.
|
Date of procedure
|
Patient ID
|
Polyp
|
Number of polyp(s)
|
Total Colonoscopy
|
Indication
|
Morphology Code
|
|
dd/mm/yy
|
Patient 1
|
Y
|
2
|
Y
|
Y
|
M-74002
|
Fig. 1 Data extraction and merging of databases.
To avoid mistakes during database merger, two internal quality controls were performed
by a staff member. For each control, the staff member reviewed hundred successive
pathological results and compared them with the results obtained by the automatic
extraction and calculated a rate of concordance, which was 100 %.
Training (second semester of 2016)
Before the initiation of continuous quality recording, the project was presented to
the endoscopic unit, with staff awareness about the quality in endoscopy implementation
program and the importance of its recording. During the final months of 2016, endoscopists
were encouraged to voluntarily insert information into the dedicated tab and the completeness
of QI recording was evaluated. An exam was considered fully filled if all of the PI
tabs were filled.
During 2017, the filling was turned to be mandatory so that the report cannot be finalized
before the filling of each tab. Performance measures for all endoscopists were compared
with global results for our department and with published targets.
Results
A total of 3762 colonoscopies were performed in Erasme Hospital, a tertiary university
hospital. The medical team is composed of 22 gastroenterologists who regularly perform
colonoscopies (more than 40 procedures a year [12]). There is no organized colonoscopy-based screening program in our country, and
in our institution, the majority (59 %) of colonoscopies are diagnostic procedures,
performed for clinical, biological or radiological abnormalities. The others are performed
for “opportunity” CCR screening and follow-up (38 %) or for scheduled therapeutic
procedures (3 %). The endoscopic profile of each endoscopist varies, depending on
his clinical activity.
During the quality monitoring training period of 6 months (July 1, 2016 to December
31, 2016), 1935 colonoscopies were performed with a QI tab fully completed in 63.1 %
of cases. Of the QI data, automatically extracted QI (bowel preparation
quality and type of sedation) were entered into the specific QI tab in 99.9 % and
97.5 % of cases, respectively, whereas manually filled QI data (progression, number
of polyps resected, and indication) were entered into the specific tab in 79.6 %,
76.6 %, and 76.3 % of cases, respectively.
After this training period, completion of the tab became mandatory and 1827 colonoscopies
were performed from July 1, 2017 to December 31, 2017. The QI tab was fully filled
for all colonoscopies. For the entire 6-month period, the following results were revealed
by the QI monitoring.
The majority of colonoscopies were performed under propofol sedation (93.8 %). The
global mean BBPS was 7.4, with 90.7 % of colonoscopies performed with adequate preparation.
That was more the case for outpatient (92.2 %) than for inpatient (87.2 %) colonoscopies
(P = 0.005). The procedure was complete, with reach to and visualization of the whole
cecum and its landmarks, in 93.8 % of all colonoscopies and 96.5 % of screening colonoscopies.
In 39.2 % of all colonoscopies, at least one polyp was resected, regardless of histology,
with a mean of 0.96 polyp resected per colonoscopy. Of all colonoscopies, 29.8 % had
at least one adenoma and for colonoscopies performed in people older than age 50 years,
36.6 % presented at least one resected adenoma. For screening colonoscopies performed
among patients with average risk, the global ADR was 28.8 %.
The data table with the list of individual quality scores and scores for the department
were shared with the clinicians responsible for the endoscopy unit and the head of
the department ([Table 5]). Individual feedback was provided to each endoscopist, with a comparison between
his scores and those for the department and recommended by guidelines ([Table 6]).
Table 5
Quality analysis of six endoscopists (July to December 2017).
|
Endoscopist (no. of colonoscopies)
|
Average BBPS
|
APR (%)
|
CIR (%)
|
PDR (%)
|
ADR 50 (%)
|
ADR screening (%)
|
ADR (all colonoscopies) (%)
|
|
Endoscopist 1 (63)
|
8.2
|
96.8
|
90.5
|
68.3
|
66
|
50
|
60.3
|
|
Endoscopist 2 (172)
|
7.1
|
90.1
|
94.2
|
28.5
|
22.3
|
28.6
|
18
|
|
Endoscopist 3 (77)
|
7.6
|
96.1
|
100
|
51.9
|
42.2
|
41.7
|
35.1
|
|
Endoscopist 4 (38)
|
7.6
|
94.7
|
100
|
39.5
|
25
|
33.3
|
23.7
|
|
Endoscopist 5 (56)
|
7.6
|
92.9
|
94.6
|
39.3
|
41.3
|
25
|
33.9
|
|
Endoscopist 6 (110)
|
6.7
|
89.1
|
90.9
|
17.3
|
21.9
|
12.5
|
12.7
|
BBPS, Boston Bowel Preparation Score; APR, adequate preparation rate; CIR, cecal intubation
rate; PDR, polyp detection rate; ADR, adenoma detection rate (ADR50, ADR for colonoscopies
performed in patients aged 50 years or older; ADR screening, ADR of colonoscopies
performed in asymptomatic patients with average risk for CCR).
Table 6
Quality analysis at the level of the department (July to December 2017).
|
Quality Indicators (July 2017–December 2017)
|
|
|
Colonoscopies (n)
|
1827
|
|
Average BBPS
|
7.4
|
|
APR (all patients)
|
90.7 %
|
|
APR (outpatients)
|
92.2 %
|
|
APR (inpatients)
|
87.2 %
|
|
CIR (all colonoscopies)
|
93.8 %
|
|
CIR (screening colonoscopies)
|
96.5 %
|
|
PDR
|
39.2 %
|
|
Average number of polyps resected per colonoscopy
|
0.96
|
|
ADR (all colonoscopies)
|
29.8 %
|
|
ADR 50
|
36.6 %
|
|
ADR screening
|
28.8 %
|
BBPS, Boston Bowel Preparation Score; APR, adequate preparation rate; CIR, cecal intubation
rate; PDR, polyp detection rate; ADR, adenoma detection rate (ADR50, ADR for colonoscopies
performed in patients aged 50 years or older; ADR screening, ADR of colonoscopies
performed in asymptomatic patients with average risk for CCR).
Discussion
The effectiveness of colonoscopy in preventing CRC depends on the quality of the procedure.
Several QI have been identified to measure performance of colonoscopy, with the aim
of improving its quality. The main outcome of our report was to assess the feasibility
and practicality of implementing an automated computer-based quality reporting system
for colonoscopy in a tertiary center endoscopy unit, including automatic calculation
of ADR.
As previously noted, adequate endoscopic reporting is the cornerstone for quality
monitoring, but it is often incomplete and not standardized. The development of electronic
reporting system (ERS) helps to fill these gaps by offering predefined and standardized
text blocks for procedure details and endoscopic findings and by placing mandatory
fields. These ERS also offer a solution for quality monitoring. A first experience
was described by van Doorn et al., who developed a completely structured colonoscopy
reporting system that generated a standardized and complete report [12]. In the present study, almost all colonoscopies (94 %) were fully reported, including
all key QI. In our experience, including a mandatory “quality tab” in the ERS led
to an increase in completeness reaching 100 % of procedures in the second part of
our implementation. Information for each QI for every colonoscopy performed was prospectively
and completely entered.
Of the different QI in the literature, we included three of that have been recognized
as key QI: ADR, preparation adequacy, and rate of cecal intubation. ADR is well studied
and is the most associated with patient outcomes. It has been inversely associated
with risk of interval CRC and CRC-related death [13]
[14], and its improvement is associated with better outcomes [7]. Adequate preparation allows for good visualization of the mucosa to detect potential
lesions, and it promotes complete examination of whole colon, with intubation of the
cecum. An incomplete examination leads to inconvenience because the procedure has
to be repeated, and to increased costs, and low CIR is associated with increased risk
of CCR [3]. Use of these QI underscores some dysfunction. For example, a low CIR may be caused
by inadequate preparation can be corrected by modifying the bowel cleansing regimen.
Beside these key QI, others are recommended for monitoring, such as the type and rate
of adverse events, patient experience, and adherence to post-polypectomy surveillance
recommendations [4]. We plan to improve our monitoring system in the future by including these QI.
International societies recommend calculating the ADR among 50-year-old patients with
average risk for CCR. We reported all indications, allowing individual ADR interpretation
according to endoscopy activity profile. For example, endoscopists with oncological
activity (who perform more follow-up colonoscopies) or with therapeutic endoscopy
activity both have higher ADRs than a gastroenterologist who follows young patients
with inflammatory bowel disease. In this way, we can better interpret results in relationship
to the individual profile of each endoscopist.
For ADR calculation, the Dutch experience reports the need for a dedicated research
nurse who subsequently reviewed all histopathology results and completed data to the
linked endoscopic findings [12]. This double manual entry of data is cumbersome, time-consuming, costly, a source
of mistakes. and limits implementation of continued quality recording and assessment
[8]. The experience on lower gastrointestinal endoscopy quality monitoring reported
by Kim et al. [15] was also based on endoscopy reporting system adaptation. They also reported the
need of postponed human intervention to add data on pathological results. To overcome
this issue, because data from endoscopy are readily available, PDR was evaluated as
an ADR surrogate. The initial studies showed a relatively good correlation between
the PDR and the ADR [16]
[17]
[18], but successive studies demonstrated that PDR can be inaccurate in some situations
and seems to be a suboptimal QI [19]
[20]
[21]. Beside these reasons, the fact that the PDR calculation is based on manually entered
data adds risk from possible human error.
What makes our reported experience original is automatic extraction of both endoscopic
and pathological data, which are merged together automatically or on an on-demand
basis, through patient identifiers. This system makes possible continuous quality
monitoring. It allows automated statistical analyses and quality assessment regularly
or on request for our whole endoscopy unit as well as for each endoscopist individually.
It also overcomes the problem of double entry of data and related issues.
The system we have described seems reliable because we conducted two manual controls
of the automatic extraction of pathology results, and each control showed perfect
concordance. Using this analysis, it is possible to evaluate all different QI over
any period of time and to provide feedback to unit heads or to individual endoscopists.
Conclusion
In conclusion, this study illustrates that continuous quality assessment of colonoscopy
including automatic ADR extraction is feasible and easily implemented in a Belgian
tertiary hospital endoscopy unit with limited human resources by adapting a commercial
reporting system and linking it to the pathology department database. Further work
must be done to evaluate the potential benefits of automatic computer-based colonoscopy
quality recording in terms of unit and individual quality improvement.
