- Email: [email protected]
Data Mining Electronic Medical Records to Determine Colonoscopy Quality Indicators
presence of polyps and diverticulosis in minority populations like African Americans and Hispanics. There was no correlation between diverticulosis and the histology of the polyps. Further study is needed to explain the correlation.
minutes (10 - 23 min.). CR was 96% (93% - 97%), reaching of terminal ileum 90% (87% - 93%), ADR 30% (21% - 40%), the average number of polyps in colonoscopies where polyps were found was 2,4 (2,1 - 2,7), detection of colorectal cancer 4% (3% - 5%), detection of flat lesions 8% (5% - 11%). Correspondence of histological findings with evaluation of polyps using KUDO classification was 91% (86% - 96%). The percentage of extracted significant polyps was 80% (66% - 93%). The histologically complete polypectomy rate (HCPR) of significant polyps was 51% (32% - 69%). The percentage of colonoscopies without pre-medication was 44% (30% - 58%). The time between polypectomy of significant lesion and check up colonoscopy was 15 months (6 - 24). The percentage of complications after therapeutic colonoscopies was 1% (0% - 2%). Conclusion: ADR is above 20% which is a tolerable lower limit. ADR of the individual experts appears to be a factor which is not dependent on the length of the procedure and on the CR. The HCPR significantly correlates with the ADR, the length of procedure and CR. The HCPR appears to be an ideal indicator of the quality of a colonoscopy.
AGA Abstracts
Mo1145 Interendoscopist Performance Variation Does Not Explain the Difference in Colonoscopy Effectiveness for Colorectal Cancer Prevention According to a Proximal or Distal Location Jean-François Bretagne, Stéphanie Hamonic, Christine Piette, Sylvain Manfredi, Gaud Mallard, Gérard Durand, Françoise Riou Objectives: There is consistent evidence that a colonoscopy is less effective for the prevention of advanced colorectal neoplasia on the right side compared to the left side. The respective roles of a missed neoplasia or the particular biology of cancer are unclear. We tested the hypothesis that interendoscopist variability in the detection of adenomas was greater for right- compared to left-sided neoplasias. Methods: Colonoscopy findings from a mass colorectal cancer screening program based on guaiac-fecal occult blood tests were analyzed in a French population of about one million individuals. The 19 endoscopists who participated in the two first rounds of screening and who performed at least 30 colonoscopies during each round were selected. Two adenoma detection rates (ADR; proportion of colonoscopies with at least one adenoma) were calculated: one for the proximal colon (above the splenic flexure; PADR) and the other for the distal colon (DADR). Multivariate analysis was used to determine whether the endoscopist was a predictor of ADR variability at both sites. The magnitude of interendoscopist variability was assessed by the variance for each site and was compared between both sites. The correlation between the adjusted PADR (PADRa) and adjusted DADR (DADRa) was calculated with a Spearman test. Results: At least one adenoma was detected in 1,238 of 3,487 colonoscopies (35.5%). Among the 19 endoscopists, the PADRa ranged from 5.5% to 22.8%, and the DADRa ranged from 21.4% to 38.2%. Patient gender and age, as well as the endoscopist, were predictors for detecting at least one distal or one proximal adenoma. Variances did not differ significantly between PADRa (29.3) and DADRa (21.3; p=0.22). PADRa/DADRa ratios ranged from 0.19 to 1.02 (mean 0.42). There was a highly significant correlation between PADRa and DADRa after exclusion of the single endoscopist with a PADRa/DADRa ratio > 1 (ρ = 0.6471; p = 0.0002). Endoscopist performance was not related to the volume of colonoscopies performed. Conclusions: These data do not demonstrate significantly greater variability for detection of right- versus leftsided adenomas. The intraendoscopist concordance for detection of either proximal or distal adenomas was highly significant. These findings do not support an effect of the difference in effectiveness of colonoscopy between the right and left sides in differences in neoplasia prevention.
Mo1148 Data Mining Electronic Medical Records to Determine Colonoscopy Quality Indicators Victor Ciofoaia, Mark J. Metwally, Stephen O'Mahony Introduction: The extensive amounts of data stored in medical databases require the development of specialized approaches for data analysis and effective use of information. The traditional manual data analysis of the colonoscopy and associated pathology reports generated as part of routine clinical practice has become insufficient and methods for computerbased analysis are now indispensable. Methods: We analyzed 60802 colonoscopies and the associated 34682 anatomopathology reports that were performed in our 300 beds community Hospital University affiliated during 2001-2010 by querying the Oracle database that powered our EMR system Cerner Powerchart. By using Microsoft Excel and Access, we used basic families of keywords (ontologies) to generate from the text colonoscopy and pathology reports data fields we included in a second database. Then we computed extra variables such as age at the time of colonoscopy, type of colonoscopy according to indication, whether a tubular adenoma/tubullo-villous polyp or carcinoma was detected, the day of the week when the colonoscopy was done etc. Results: By querying the database thus generated we were able to determine for our community hospital the basic epidemiologic data of the patients who had a colonoscopy (race, gender, age at the time of colonoscopy), the percentage of screening colonoscopies versus colonoscopies done for other indications, the adenoma detection rate (ADR) for each GI attending or fellow for each year, the type of sedation used, the relationship of ADR to the day of the week/weekend when the colonoscopy was done and the percentage of follow-up findings after the initial index colonoscopy. By filtering, sorting and reviewing the database generated through data mining interesting cases became apparent, such as cases that required multiple repeat colonoscopies or cases that presented with more than 10 tubular adenomas. By using statistics software such as Graphpad Prism or IBM SPSS more information can be extracted and statistical significance better detected for faster knowledge discovery and presentation. Conclusions: We present step by step a solution that is easy to implement with generally available software in any hospital or GI practice that maintains electronic medical records for colonoscopy and pathology reports that requires only 20-50 man hours in order to obtain essential quality improvement information such as the adenoma detection rates, epidemiologic data (age, gender, race) or the incidence of interval colon cancer after the index colonoscopy. We discuss the advantages, the limitations and ways to further improve the reliability of this approach. By using this method to process the information included in the colonoscopy and pathology reports, relevant data becomes readily available and further clinical questions can be answered quickly.
Mo1146 Is There a Role for an Early Colonoscopy Following an Acute Non Bleeding Diverticulitis? A Single Center Experience Hemda Schmilovitz-Weiss, Evgen Yalunin, Eli Atar, Vered Shehayek-Shabat, Izhak Levin, Alexandra Chervinski, Yaron Niv, Haim Shirin Aim: To determine the impact of colonoscopy on the clinical management of patients with acute diverticulitis and their outcome. Patients and methods: This is was a retrospective single center chart review study. Medical records of 220 patients hospitalized for acute diverticulitis between 01.06.2002 through 01.09.2009 were reviewed. The diagnosis of acute diverticulitis has been made by clinical criteria and characteristic CT findings. 15 patients with questionable CT and 5 patients presenting with hematochezia were excluded, leaving 200 patients for further evaluation. The mean age was 61.8 ± 14.3 years, (61% females). Clinical parameters, laboratory results, imaging, endoscopic reports and long-term patients' outcome were analyzed. Results: Fourteen patients (7%) had complicated diverticulitis diagnosed by abscess or micro-perforation per CT. One hundred patients (age 61.8 ± 13.3 years, 54.1% females), underwent early (4-6 weeks) colonoscopy following hospital discharge. There were no significant differences in patients' characteristics between the two groups of patients (with or without colonoscopy). There were no cases of malignancy. However, in 32 patients (32%) at least one polyp was found. Of these only one was determined as an advanced adenomatous polyp. The admission diagnosis did not changed after colonoscopy in none of the patients in the colonoscopy group and there were no differences in the survival rate between the groups (4 ± 1.9 vs. 4.2 ± 2.1 years, p=0.62). In conclusion: Our results suggest that colonoscopy did not affect the management of acute diverticulitis patients and did not alter their outcome. The current practice of a routine colonoscopy following acute diverticulitis diagnosed by typical clinical symptoms and computed tomography (CT) needs to be re-evaluated.
Mo1149 Factors Influencing Adherence to Surveillance Colonoscopy Guidelines Benjamin W. Raile, Simeng Sun, Thomas Coyne, Scott Naples, Richard Huynh, Donald N. Tsynman, James C. Reynolds BACKGROUND The timing of surveillance colonoscopy is integral to colorectal cancer prevention. This study evaluates adherence to colonoscopy surveillance guidelines related to physician and patient factors. METHODS Retrospective chart reviews of outpatient nonemergent colonoscopy reports with polyp biopsy results were performed. Patients with Inflammatory Bowel Disease, hereditary polyposis syndromes or cancer were excluded. Endoscopist recommended time periods for patient follow-up surveillance colonoscopy were compared to guidelines outlined by the U.S. Multi-Society Task Force on Colorectal Cancer. The roles of patient demographics and co-morbidities were also evaluated. RESULTS Twenty five hundred paper charts from 2002-2009 were reviewed, resulting in a 215 patient sample past due for surveillance colonoscopy. The average patient age was 63. The average patient Charlson Co-morbidity Index adjusted for age was 3.5. Of the74 patients with normal colonoscopies, 51 (68.9%) were recommended surveillance per guidelines and 23 (31.1%) were recommended earlier surveillance. There were 50 patients satisfying the 5-10 year surveillance guideline category having 1 or 2 polyps with no more than low-grade dysplasia progression and no villous features. Within this group, endoscopists matched the guideline for 15 (30.0%) patients but recommended surveillance at less than 5 years for 35 (70.0%) patients. There were 32 patients fulfilling the 3-year surveillance guidelines of colonoscopies with either 3-10 polyps, polyps of at least 1 cm, high-grade dysplasia or villous features. Endoscopists followed this guideline for 19 (59.4%) patients and recommended earlier surveillance for 13 (40.6%). The 59 patients with poor bowel preparation leading to incomplete colonoscopies were appropriately directed to return within 1 year for repeat colonoscopies. Overall, guidelines were matched for 144 (67.0%) of the patients in this study. CONCLUSION Approximately one third of the included patients were directed to receive surveillance colonoscopy earlier than guidelines recommend. Endoscopists in this study followed the surveillance guidelines more closely for higher risk patients. Additional education regarding surveillance colonoscopy may increase compliance to recommended guidelines. Patient age and co-morbidities may impact compliance for receiving surveillance colonoscopy. A prospective study may further illustrate the relationship between co-morbidities and patient adherence to surveillance colonoscopy.
Mo1147 Are There Other Indicators of Quality of Colonoscopy Than Adenoma Detection Rate (ADR)? Marek Benes, Pavel Drastich, Petr Stirand, Julius Spicak Introduction: Caecum intubation rate (CR) is historically the most commonly used indicator of quality of colonoscopy. However, nowadays this indicator is not enough and it has been frequently replaced by adenoma detection rate (ADR) in combination with extraction time (ET). The aim of our study was to evaluate various parameters of colonoscopy in order to develop a new more precise indicator of quality of colonoscopy. Method: We have used a structured database used to generate medical reports about colonoscopies at our endoscopy unit for retrospective evaluation. This database is interconnected with a register of histological findings which provided data about histological examinations. The individual data was statistically evaluated using correlation tests (Spearman's coefficient). Results: We have evaluated 1360 consecutive colonoscopies altogether, which were carried out by 6 experts. Each one of the endoscopists had performed more than 250 colonoscopies in a year and more than 1000 colonoscopies during his carrier. The average time of the procedure was 16
AGA Abstracts
S-568
minutes (10 - 23 min.). CR was 96% (93% - 97%), reaching of terminal ileum 90% (87% - 93%), ADR 30% (21% - 40%), the average number of polyps in colonoscopies where polyps were found was 2,4 (2,1 - 2,7), detection of colorectal cancer 4% (3% - 5%), detection of flat lesions 8% (5% - 11%). Correspondence of histological findings with evaluation of polyps using KUDO classification was 91% (86% - 96%). The percentage of extracted significant polyps was 80% (66% - 93%). The histologically complete polypectomy rate (HCPR) of significant polyps was 51% (32% - 69%). The percentage of colonoscopies without pre-medication was 44% (30% - 58%). The time between polypectomy of significant lesion and check up colonoscopy was 15 months (6 - 24). The percentage of complications after therapeutic colonoscopies was 1% (0% - 2%). Conclusion: ADR is above 20% which is a tolerable lower limit. ADR of the individual experts appears to be a factor which is not dependent on the length of the procedure and on the CR. The HCPR significantly correlates with the ADR, the length of procedure and CR. The HCPR appears to be an ideal indicator of the quality of a colonoscopy.
AGA Abstracts
Mo1145 Interendoscopist Performance Variation Does Not Explain the Difference in Colonoscopy Effectiveness for Colorectal Cancer Prevention According to a Proximal or Distal Location Jean-François Bretagne, Stéphanie Hamonic, Christine Piette, Sylvain Manfredi, Gaud Mallard, Gérard Durand, Françoise Riou Objectives: There is consistent evidence that a colonoscopy is less effective for the prevention of advanced colorectal neoplasia on the right side compared to the left side. The respective roles of a missed neoplasia or the particular biology of cancer are unclear. We tested the hypothesis that interendoscopist variability in the detection of adenomas was greater for right- compared to left-sided neoplasias. Methods: Colonoscopy findings from a mass colorectal cancer screening program based on guaiac-fecal occult blood tests were analyzed in a French population of about one million individuals. The 19 endoscopists who participated in the two first rounds of screening and who performed at least 30 colonoscopies during each round were selected. Two adenoma detection rates (ADR; proportion of colonoscopies with at least one adenoma) were calculated: one for the proximal colon (above the splenic flexure; PADR) and the other for the distal colon (DADR). Multivariate analysis was used to determine whether the endoscopist was a predictor of ADR variability at both sites. The magnitude of interendoscopist variability was assessed by the variance for each site and was compared between both sites. The correlation between the adjusted PADR (PADRa) and adjusted DADR (DADRa) was calculated with a Spearman test. Results: At least one adenoma was detected in 1,238 of 3,487 colonoscopies (35.5%). Among the 19 endoscopists, the PADRa ranged from 5.5% to 22.8%, and the DADRa ranged from 21.4% to 38.2%. Patient gender and age, as well as the endoscopist, were predictors for detecting at least one distal or one proximal adenoma. Variances did not differ significantly between PADRa (29.3) and DADRa (21.3; p=0.22). PADRa/DADRa ratios ranged from 0.19 to 1.02 (mean 0.42). There was a highly significant correlation between PADRa and DADRa after exclusion of the single endoscopist with a PADRa/DADRa ratio > 1 (ρ = 0.6471; p = 0.0002). Endoscopist performance was not related to the volume of colonoscopies performed. Conclusions: These data do not demonstrate significantly greater variability for detection of right- versus leftsided adenomas. The intraendoscopist concordance for detection of either proximal or distal adenomas was highly significant. These findings do not support an effect of the difference in effectiveness of colonoscopy between the right and left sides in differences in neoplasia prevention.
Mo1148 Data Mining Electronic Medical Records to Determine Colonoscopy Quality Indicators Victor Ciofoaia, Mark J. Metwally, Stephen O'Mahony Introduction: The extensive amounts of data stored in medical databases require the development of specialized approaches for data analysis and effective use of information. The traditional manual data analysis of the colonoscopy and associated pathology reports generated as part of routine clinical practice has become insufficient and methods for computerbased analysis are now indispensable. Methods: We analyzed 60802 colonoscopies and the associated 34682 anatomopathology reports that were performed in our 300 beds community Hospital University affiliated during 2001-2010 by querying the Oracle database that powered our EMR system Cerner Powerchart. By using Microsoft Excel and Access, we used basic families of keywords (ontologies) to generate from the text colonoscopy and pathology reports data fields we included in a second database. Then we computed extra variables such as age at the time of colonoscopy, type of colonoscopy according to indication, whether a tubular adenoma/tubullo-villous polyp or carcinoma was detected, the day of the week when the colonoscopy was done etc. Results: By querying the database thus generated we were able to determine for our community hospital the basic epidemiologic data of the patients who had a colonoscopy (race, gender, age at the time of colonoscopy), the percentage of screening colonoscopies versus colonoscopies done for other indications, the adenoma detection rate (ADR) for each GI attending or fellow for each year, the type of sedation used, the relationship of ADR to the day of the week/weekend when the colonoscopy was done and the percentage of follow-up findings after the initial index colonoscopy. By filtering, sorting and reviewing the database generated through data mining interesting cases became apparent, such as cases that required multiple repeat colonoscopies or cases that presented with more than 10 tubular adenomas. By using statistics software such as Graphpad Prism or IBM SPSS more information can be extracted and statistical significance better detected for faster knowledge discovery and presentation. Conclusions: We present step by step a solution that is easy to implement with generally available software in any hospital or GI practice that maintains electronic medical records for colonoscopy and pathology reports that requires only 20-50 man hours in order to obtain essential quality improvement information such as the adenoma detection rates, epidemiologic data (age, gender, race) or the incidence of interval colon cancer after the index colonoscopy. We discuss the advantages, the limitations and ways to further improve the reliability of this approach. By using this method to process the information included in the colonoscopy and pathology reports, relevant data becomes readily available and further clinical questions can be answered quickly.
Mo1146 Is There a Role for an Early Colonoscopy Following an Acute Non Bleeding Diverticulitis? A Single Center Experience Hemda Schmilovitz-Weiss, Evgen Yalunin, Eli Atar, Vered Shehayek-Shabat, Izhak Levin, Alexandra Chervinski, Yaron Niv, Haim Shirin Aim: To determine the impact of colonoscopy on the clinical management of patients with acute diverticulitis and their outcome. Patients and methods: This is was a retrospective single center chart review study. Medical records of 220 patients hospitalized for acute diverticulitis between 01.06.2002 through 01.09.2009 were reviewed. The diagnosis of acute diverticulitis has been made by clinical criteria and characteristic CT findings. 15 patients with questionable CT and 5 patients presenting with hematochezia were excluded, leaving 200 patients for further evaluation. The mean age was 61.8 ± 14.3 years, (61% females). Clinical parameters, laboratory results, imaging, endoscopic reports and long-term patients' outcome were analyzed. Results: Fourteen patients (7%) had complicated diverticulitis diagnosed by abscess or micro-perforation per CT. One hundred patients (age 61.8 ± 13.3 years, 54.1% females), underwent early (4-6 weeks) colonoscopy following hospital discharge. There were no significant differences in patients' characteristics between the two groups of patients (with or without colonoscopy). There were no cases of malignancy. However, in 32 patients (32%) at least one polyp was found. Of these only one was determined as an advanced adenomatous polyp. The admission diagnosis did not changed after colonoscopy in none of the patients in the colonoscopy group and there were no differences in the survival rate between the groups (4 ± 1.9 vs. 4.2 ± 2.1 years, p=0.62). In conclusion: Our results suggest that colonoscopy did not affect the management of acute diverticulitis patients and did not alter their outcome. The current practice of a routine colonoscopy following acute diverticulitis diagnosed by typical clinical symptoms and computed tomography (CT) needs to be re-evaluated.
Mo1149 Factors Influencing Adherence to Surveillance Colonoscopy Guidelines Benjamin W. Raile, Simeng Sun, Thomas Coyne, Scott Naples, Richard Huynh, Donald N. Tsynman, James C. Reynolds BACKGROUND The timing of surveillance colonoscopy is integral to colorectal cancer prevention. This study evaluates adherence to colonoscopy surveillance guidelines related to physician and patient factors. METHODS Retrospective chart reviews of outpatient nonemergent colonoscopy reports with polyp biopsy results were performed. Patients with Inflammatory Bowel Disease, hereditary polyposis syndromes or cancer were excluded. Endoscopist recommended time periods for patient follow-up surveillance colonoscopy were compared to guidelines outlined by the U.S. Multi-Society Task Force on Colorectal Cancer. The roles of patient demographics and co-morbidities were also evaluated. RESULTS Twenty five hundred paper charts from 2002-2009 were reviewed, resulting in a 215 patient sample past due for surveillance colonoscopy. The average patient age was 63. The average patient Charlson Co-morbidity Index adjusted for age was 3.5. Of the74 patients with normal colonoscopies, 51 (68.9%) were recommended surveillance per guidelines and 23 (31.1%) were recommended earlier surveillance. There were 50 patients satisfying the 5-10 year surveillance guideline category having 1 or 2 polyps with no more than low-grade dysplasia progression and no villous features. Within this group, endoscopists matched the guideline for 15 (30.0%) patients but recommended surveillance at less than 5 years for 35 (70.0%) patients. There were 32 patients fulfilling the 3-year surveillance guidelines of colonoscopies with either 3-10 polyps, polyps of at least 1 cm, high-grade dysplasia or villous features. Endoscopists followed this guideline for 19 (59.4%) patients and recommended earlier surveillance for 13 (40.6%). The 59 patients with poor bowel preparation leading to incomplete colonoscopies were appropriately directed to return within 1 year for repeat colonoscopies. Overall, guidelines were matched for 144 (67.0%) of the patients in this study. CONCLUSION Approximately one third of the included patients were directed to receive surveillance colonoscopy earlier than guidelines recommend. Endoscopists in this study followed the surveillance guidelines more closely for higher risk patients. Additional education regarding surveillance colonoscopy may increase compliance to recommended guidelines. Patient age and co-morbidities may impact compliance for receiving surveillance colonoscopy. A prospective study may further illustrate the relationship between co-morbidities and patient adherence to surveillance colonoscopy.
Mo1147 Are There Other Indicators of Quality of Colonoscopy Than Adenoma Detection Rate (ADR)? Marek Benes, Pavel Drastich, Petr Stirand, Julius Spicak Introduction: Caecum intubation rate (CR) is historically the most commonly used indicator of quality of colonoscopy. However, nowadays this indicator is not enough and it has been frequently replaced by adenoma detection rate (ADR) in combination with extraction time (ET). The aim of our study was to evaluate various parameters of colonoscopy in order to develop a new more precise indicator of quality of colonoscopy. Method: We have used a structured database used to generate medical reports about colonoscopies at our endoscopy unit for retrospective evaluation. This database is interconnected with a register of histological findings which provided data about histological examinations. The individual data was statistically evaluated using correlation tests (Spearman's coefficient). Results: We have evaluated 1360 consecutive colonoscopies altogether, which were carried out by 6 experts. Each one of the endoscopists had performed more than 250 colonoscopies in a year and more than 1000 colonoscopies during his carrier. The average time of the procedure was 16
AGA Abstracts
S-568