A prediction model for advanced colorectal neoplasia in an asymptomatic screening population

Applying Natural Language Processing to Textual Data From Clinical Data Warehouses: Systematic Review

BACKGROUND

In recent years, health data collected during the clinical care process have been often repurposed for secondary use through clinical data warehouses (CDWs), which interconnect disparate data from different sources. A large amount of information of high clinical value is stored in unstructured text format. Natural language processing (NLP), which implements algorithms that can operate on massive unstructured textual data, has the potential to structure the data and make clinical information more accessible.

OBJECTIVE

The aim of this review was to provide an overview of studies applying NLP to textual data from CDWs. It focuses on identifying the (1) NLP tasks applied to data from CDWs and (2) NLP methods used to tackle these tasks.

METHODS

This review was performed according to the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines. We searched for relevant articles in 3 bibliographic databases: PubMed, Google Scholar, and ACL Anthology. We reviewed the titles and abstracts and included articles according to the following inclusion criteria: (1) focus on NLP applied to textual data from CDWs, (2) articles published between 1995 and 2021, and (3) written in English.

RESULTS

We identified 1353 articles, of which 194 (14.34%) met the inclusion criteria. Among all identified NLP tasks in the included papers, information extraction from clinical text (112/194, 57.7%) and the identification of patients (51/194, 26.3%) were the most frequent tasks. To address the various tasks, symbolic methods were the most common NLP methods (124/232, 53.4%), showing that some tasks can be partially achieved with classical NLP techniques, such as regular expressions or pattern matching that exploit specialized lexica, such as drug lists and terminologies. Machine learning (70/232, 30.2%) and deep learning (38/232, 16.4%) have been increasingly used in recent years, including the most recent approaches based on transformers. NLP methods were mostly applied to English language data (153/194, 78.9%).

CONCLUSIONS

CDWs are central to the secondary use of clinical texts for research purposes. Although the use of NLP on data from CDWs is growing, there remain challenges in this field, especially with regard to languages other than English. Clinical NLP is an effective strategy for accessing, extracting, and transforming data from CDWs. Information retrieved with NLP can assist in clinical research and have an impact on clinical practice.

Prediction algorithm for gastric cancer in a general population: A validation study

BACKGROUND

Worldwide, gastric cancer is a leading cause of cancer incidence and mortality. This study aims to devise and validate a scoring system based on readily available clinical data to predict the risk of gastric cancer in a large Chinese population.

METHODS

We included a total of 6,209,697 subjects aged between 18 and 70 years who have received upper digestive endoscopy in Hong Kong from 1997 to 2018. A binary logistic regression model was constructed to examine the predictors of gastric cancer in a derivation cohort (n = 4,347,224), followed by model evaluation in a validation cohort (n = 1,862,473). The algorithm's discriminatory ability was evaluated as the area under the curve (AUC) of the mathematically constructed receiver operating characteristic (ROC) curve.

RESULTS

Age, male gender, history of Helicobacter pylori infection, use of proton pump inhibitors, non-use of aspirin, non-steroidal anti-inflammatory drugs (NSAIDs), and statins were significantly associated with gastric cancer. A scoring of ≤8 was designated as "average risk (AR)". Scores at 9 or above were assigned as "high risk (HR)". The prevalence of gastric cancer was 1.81% and 0.096%, respectively, for the HR and LR groups. The AUC for the risk score in the validation cohort was 0.834, implying an excellent fit of the model.

CONCLUSIONS

This study has validated a simple, accurate, and easy-to-use scoring algorithm which has a high discriminatory capability to predict gastric cancer. The score could be adopted to risk stratify subjects suspected as having gastric cancer, thus allowing prioritized upper digestive tract investigation.

Colorectal Cancer Risk Assessment and Precision Approaches to Screening: Brave New World or Worlds Apart?

Current colorectal cancer (CRC) screening recommendations take a "one-size-fits-all" approach using age as the major criterion to initiate screening. Precision screening that incorporates factors beyond age to risk stratify individuals could improve on current approaches and optimally use available resources with benefits for patients, providers, and health care systems. Prediction models could identify high-risk groups who would benefit from more intensive screening, while low-risk groups could be recommended less intensive screening incorporating noninvasive screening modalities. In addition to age, prediction models incorporate well-established risk factors such as genetics (eg, family CRC history, germline, and polygenic risk scores), lifestyle (eg, smoking, alcohol, diet, and physical inactivity), sex, and race and ethnicity among others. Although several risk prediction models have been validated, few have been systematically studied for risk-adapted population CRC screening. In order to envisage clinical implementation of precision screening in the future, it will be critical to develop reliable and accurate prediction models that apply to all individuals in a population; prospectively study risk-adapted CRC screening on the population level; garner acceptance from patients and providers; and assess feasibility, resources, cost, and cost-effectiveness of these new paradigms. This review evaluates the current state of risk prediction modeling and provides a roadmap for future implementation of precision CRC screening.

Deep learning approach to detection of colonoscopic information from unstructured reports

BACKGROUND

Colorectal cancer is a leading cause of cancer deaths. Several screening tests, such as colonoscopy, can be used to find polyps or colorectal cancer. Colonoscopy reports are often written in unstructured narrative text. The information embedded in the reports can be used for various purposes, including colorectal cancer risk prediction, follow-up recommendation, and quality measurement. However, the availability and accessibility of unstructured text data are still insufficient despite the large amounts of accumulated data. We aimed to develop and apply deep learning-based natural language processing (NLP) methods to detect colonoscopic information.

METHODS

This study applied several deep learning-based NLP models to colonoscopy reports. Approximately 280,668 colonoscopy reports were extracted from the clinical data warehouse of Samsung Medical Center. For 5,000 reports, procedural information and colonoscopic findings were manually annotated with 17 labels. We compared the long short-term memory (LSTM) and BioBERT model to select the one with the best performance for colonoscopy reports, which was the bidirectional LSTM with conditional random fields. Then, we applied pre-trained word embedding using large unlabeled data (280,668 reports) to the selected model.

RESULTS

The NLP model with pre-trained word embedding performed better for most labels than the model with one-hot encoding. The F1 scores for colonoscopic findings were: 0.9564 for lesions, 0.9722 for locations, 0.9809 for shapes, 0.9720 for colors, 0.9862 for sizes, and 0.9717 for numbers.

CONCLUSIONS

This study applied deep learning-based clinical NLP models to extract meaningful information from colonoscopy reports. The method in this study achieved promising results that demonstrate it can be applied to various practical purposes.

A deep learning and natural language processing-based system for automatic identification and surveillance of high-risk patients undergoing upper endoscopy: A multicenter study

BACKGROUND

Timely identification and regular surveillance of patients at high risk are crucial for early diagnosis of upper gastrointestinal cancer. However, traditional manual surveillance method is time-consuming, and current surveillance rate is below 50%. Here, we aimed to develop a surveillance system named ENDOANGEL-AS (automatic surveillance) for automatic identification and surveillance of high-risk patients.

METHODS

7874 patients from Renmin Hospital of Wuhan University between May 1 and July 31, 2021 were used as the training set, 6762 patients between August 1 and October 31, 2021 as the internal test set, and 7570 patients from two other hospitals between August 1 and October 31, 2021 as the external test sets. We first extracted descriptions of abnormalities from endoscopic and pathological reports based on natural language processing techniques to identify individuals. Then patients were classified at nine risk levels according to endoscopic and pathological findings, and a deep learning model was trained to identify demarcation line (DL) in gastric low-grade intraepithelial neoplasia (LGIN) using 1561 white-light still images for risk stratification of gastric LGIN. Finally, patients undergoing upper endoscopy were classified and assigned one of ten surveillance intervals according to guidelines. The performance of ENDOANGEL-AS was evaluated and compared with physicians.

FINDINGS

Patient identification module achieved an accuracy of 100% and 99.91% in internal and external test sets, respectively. Risk level classification module achieved an accuracy of 100% and 99.85% in the internal and external test sets, respectively. DL identification module achieved an accuracy of 87.88%. ENDOANGEL-AS on surveillance interval assignment achieved an accuracy of 99.23% and 99.67% in internal and external test sets, respectively. ENDOANGEL-AS had significantly higher accuracy compared with physicians (99.00% vs 38.87%, p < 0.001). The accuracy (63.67%, p < 0.001) of endoscopists with the assistance of ENDOANGEL-AS was significantly improved.

INTERPRETATION

We established a surveillance system that can automatically identify patients and assign surveillance intervals with high accuracy and good transferability.

FUNDING

This work was partly supported by a grant from the Hubei Province Major Science and Technology Innovation Project (2018-916-000-008) and the Fundamental Research Funds for the Central Universities (2042021kf0084).

External validation of models for predicting risk of colorectal cancer using the China Kadoorie Biobank

BACKGROUND

In China, colorectal cancer (CRC) incidence and mortality have been steadily increasing over the last decades. Risk models to predict incident CRC have been developed in various populations, but they have not been systematically externally validated in a Chinese population.  This study aimed to assess the performance of risk scores in predicting CRC using the China Kadoorie Biobank (CKB), one of the largest and geographically diverse prospective cohort studies in China.

METHODS

Nine models were externally validated in 512,415 participants in CKB and included 2976 cases of CRC. Model discrimination was assessed, overall and by sex, age, site, and geographic location, using the area under the receiver operating characteristic curve (AUC). Model discrimination of these nine models was compared to a model using age alone. Calibration was assessed for five models, and they were re-calibrated in CKB.

RESULTS

The three models with the highest discrimination (Ma (Cox model) AUC 0.70 [95% CI 0.69-0.71]; Aleksandrova 0.70 [0.69-0.71]; Hong 0.69 [0.67-0.71]) included the variables age, smoking, and alcohol. These models performed significantly better than using a model based on age alone (AUC of 0.65 [95% CI 0.64-0.66]). Model discrimination was generally higher in younger participants, males, urban environments, and for colon cancer. The two models (Guo and Chen) developed in Chinese populations did not perform better than the others. Among the 10% of participants with the highest risk, the three best performing models identified 24-26% of participants that went on to develop CRC.

CONCLUSIONS

Several risk models based on easily obtainable demographic and modifiable lifestyle factor have good discrimination in a Chinese population. The three best performing models have a higher discrimination than using a model based on age alone.

Head-to-head comparison of a risk-adapted screening strategy using various risk prediction models in detecting colorectal neoplasm

BACKGROUND AND AIM

The risk-adapted screening strategy showed satisfying colorectal cancer (CRC) screening yield and efficiency. We therefore further explored the diagnostic performance variation of this strategy using different risk prediction models.

METHODS

A literature search was conducted to identify studies evaluating risk models for advanced colorectal neoplasm (ACN). The included models were retrospectively validated in a subset sample (N = 3219) from a population-based CRC screening trial in China. Diagnosis-related indictors were compared between the risk-adapted screening strategy using different models and the fecal immunochemical test (FIT)-only strategy. For simulated populations with ACN prevalence of 3% to 12%, the trade-off of additional false positives for each additional true positive were calculated.

RESULTS

We included 14 eligible risk models with the area under the curves ranging 0.570 to 0.652 in the validation set. The overall sensitivities of the risk-adapted screening strategy using different risk models for ACN varied from 46.0% to 69.8%, higher than FIT (21.9%), but at the expense of specificities (51.6% to 78.3% vs 97.1%). For population having ACN prevalence of 3%, risk-adapted screening strategies needed 20.5 to 31.1 additional false positives for each additional true positive compared with FIT, and respective number would substantially reduce (4.7 to 7.1) as the ACN prevalence increased to 12%.

CONCLUSIONS

Risk-adapted screening strategy using the current risk models showed improved sensitivity for ACN compared with FIT, at the cost of increased colonoscopy workload. The optimal strategy for screening practice should be tailored considering the disease burden and availability of healthcare resources.

Risk factors and prediction algorithm for advanced neoplasia on screening colonoscopy for average-risk individuals

BACKGROUND

Screening with colonoscopy for all average-risk population is probably not cost-effective due to the limited sources and over-generalization of the risk, and risk stratification can be used to optimize colorectal cancer screening.

OBJECTIVES

We aimed to assess risk factors for advanced neoplasia (AN) and a classification tree algorithm to predict the risk.

DESIGN

This is a retrospective cross-sectional study.

METHODS

This study was composed of consecutive asymptomatic average-risk individuals undergoing first screening colonoscopy between 2008 and 2019. Detailed characteristics including background diseases, habits, and medications were collected. We used multivariable logistic regression to investigate the associations between clinical variables and the presence of AN and built a classification algorithm to predict AN.

RESULTS

A total of 3856 patients were included (73.2% male, median age 55). Adenoma and AN detection rate were 15.8% and 3.4%, respectively. On multivariable analysis, predictors of AN [odds ratio (OR), 95% confidence interval (CI)] were age (1.04, 1.01-1.06, p = 0.003), male sex (2.69, 1.56-4.64, p < 0.001), and smoking (1.97, 1.38-2.81, p  < 0.001). A classification tree algorithm showed that smoking was the most important risk factor for prediction of AN (4.9% versus 2.4%, p < 0.001), followed by age with a cutoff value of 60 in the smokers (8.4% versus 3.8%, p = 0.001) and 50 in the non-smokers (2.9% versus 0.9%, p = 0.004).

CONCLUSION

Smoking habits, old age, and male gender are highly associated with an increased risk for AN and should be incorporated in the individualized risk-assessment to adapt a screening program.

Designing a Predictive Model for Colorectal Neoplasia Diagnosis Based on Clinical and Laboratory Findings in Colonoscopy Candidate Patients

BACKGROUND

Health authorities have expanded two strategies to diminish CRC-related influence: CR screening and improve diagnostic process in symptomatic patients. The aim of the current study is to design a predictive model to identify the most important risk factors that can efficiently predict patients who have high risk of colorectal neoplasia.

METHOD

A cross-sectional study was constructed to include all patients who had positive test for FIT or had one or more risk factors for colorectal cancer based on the guidelines of detecting high-risk groups for colorectal cancer in Iran. Multivariable binary logistic regression model was constructed for prediction of colorectal neoplasia. We used sensitivity, specificity, positive and negative predictive value, and positive and negative likelihood ratio to check the accuracy. The Hosmer-Lemeshow test, chi-square test, and p value were used to determine the precision of model.

RESULT

Following an AIC stepwise selection model, only nine potential variables, namely gender, watery diarrhea, IBD, abdominal pain, melena, body mass index, depression drug, anti-inflammatory drug, and age, were found to be a predictor of colorectal neoplasia. The best cut-point probability in the final model was 0.27 and results of sensitivity and specificity, based on maximizing these two criteria, were 66% and 62%, respectively.

CONCLUSION

Overall, our model prediction was comparable with other risk prediction models for colorectal cancer. It had a modest discriminatory power to distinguish an individual's neoplasia colorectal risk.

Prediction model to discriminate leptospirosis from hantavirus

OBJECTIVE

The aim of this study was to build a prediction model to discriminate precociously hantavirus infection from leptospirosis, identifying the conditions and risk factors associated with these diseases.

METHODS

A logistic regression model in which the response variable was the presence of hantavirus or leptospirosis was adjusted.

RESULTS

As a result, the method selected the following variables that influenced the prediction formula: sociodemographic variables, clinical manifestations, and exposure to environmental risks. All variables considered in the model presented statistical significance with a p<0.05 value. The accuracy of the model to differentiate hantavirus from leptospirosis was 88.7%.

CONCLUSIONS

Concluding that the development of statistical tools with high potential to predict the disease, and thus differentiate them precociously, can reduce hospital costs, speed up the patient's care, reduce morbidity and mortality, and assist health professionals and public managers in decision-making.

Research and Application of Artificial Intelligence (AI) based on Electronic Health Records from Patients with Cancer: a Systematic Review (Preprint)

Background

With the accumulation of electronic health records and the development of artificial intelligence, patients with cancer urgently need new evidence of more personalized clinical and demographic characteristics and more sophisticated treatment and prevention strategies. However, no research has systematically analyzed the application and significance of artificial intelligence based on electronic health records in cancer care.

Objective

The aim of this study was to conduct a review to introduce the current state and limitations of artificial intelligence based on electronic health records of patients with cancer and to summarize the performance of artificial intelligence in mining electronic health records and its impact on cancer care.

Methods

Three databases were systematically searched to retrieve potentially relevant papers published from January 2009 to October 2020. Four principal reviewers assessed the quality of the papers and reviewed them for eligibility based on the inclusion criteria in the extracted data. The summary measures used in this analysis were the number and frequency of occurrence of the themes.

Results

Of the 1034 papers considered, 148 papers met the inclusion criteria. Cancer care, especially cancers of female organs and digestive organs, could benefit from artificial intelligence based on electronic health records through cancer emergencies and prognostic estimates, cancer diagnosis and prediction, tumor stage detection, cancer case detection, and treatment pattern recognition. The models can always achieve an area under the curve of 0.7. Ensemble methods and deep learning are on the rise. In addition, electronic medical records in the existing studies are mainly in English and from private institutional databases.

Conclusions

Artificial intelligence based on electronic health records performed well and could be useful for cancer care. Improving the performance of artificial intelligence can help patients receive more scientific-based and accurate treatments. There is a need for the development of new methods and electronic health record data sharing and for increased passion and support from cancer specialists.

Derivation and validation of a predictive model for advanced colorectal neoplasia in asymptomatic adults

OBJECTIVE

Knowing risk for advanced colorectal neoplasia (AN) could help patients and providers choose among screening tests, improving screening efficiency and uptake. We created a risk prediction model for AN to help decide which test might be preferred, a use not considered for existing models.

DESIGN

Average-risk 50-to-80-year olds undergoing first-time screening colonoscopy were recruited from endoscopy units in Indiana. We measured sociodemographic and physical features, medical and family history and lifestyle factors and linked these to the most advanced finding. We derived a risk equation on two-thirds of the sample and assigned points to each variable to create a risk score. Scores with comparable risks were collapsed into risk categories. The model and score were tested on the remaining sample.

RESULTS

Among 3025 subjects in the derivation set (mean age 57.3 (6.5) years; 52% women), AN prevalence was 9.4%. The 13-variable model (c-statistic=0.77) produced three risk groups with AN risks of 1.5% (95% CI 0.72% to 2.74%), 7.06% (CI 5.89% to 8.38%) and 27.26% (CI 23.47% to 31.30%) in low-risk, intermediate-risk and high-risk groups (p value <0.001), containing 23%, 59% and 18% of subjects, respectively. In the validation set of 1475 subjects (AN prevalence of 8.4%), model performance was comparable (c-statistic=0.78), with AN risks of 2.73% (CI 1.25% to 5.11%), 5.57% (CI 4.12% to 7.34%) and 25.79% (CI 20.51% to 31.66%) in low-risk, intermediate-risk and high-risk subgroups, respectively (p<0.001), containing proportions of 23%, 59% and 18%.

CONCLUSION

Among average-risk persons, this model estimates AN risk with high discrimination, identifying a lower risk subgroup that may be screened non-invasively and a higher risk subgroup for which colonoscopy may be preferred. The model could help guide patient-provider discussions of screening options, may increase screening adherence and conserve colonoscopy resources.

Comparison of multiple statistical models for the development of clinical prediction scores to detect advanced colorectal neoplasms in asymptomatic Thai patients

A good clinical prediction score can help in the risk stratification of patients with colorectal cancer (CRC) undergoing colonoscopy screening. The aim of our study was to compare model performance of binary logistic regression (BLR), polytomous logistic regression (PLR), and classification and regression tree (CART) between the clinical prediction scores of advanced colorectal neoplasia (ACN) in asymptomatic Thai patients.We conducted a cross-sectional study of 1311 asymptomatic Thai patients to develop a clinical prediction model. The possible predictive variables included sex, age, body mass index, family history of CRC in first-degree relatives, smoking, diabetes mellitus, and the fecal immunochemical test in the univariate analysis. Variables with a P value of .1 were included in the multivariable analysis, using the BLR, CART, and PLR models. Model performance, including the area under the receiver operator characteristic curve (AUROC), was compared between the model types.ACN was diagnosed in 53 patients (4.04%). The AUROCs were not significantly different between the BLR and CART models for ACN prediction with an AUROC of 0.774 (95% confidence interval [95% CI]: 0.706-0.842) and 0.765 (95% CI: 0.698-0.832), respectively (P = .712). A significant difference was observed between the PLR and CART models in predicting average to moderate ACN risk with an AUROC of 0.767 (95% CI: 0.695-0.839 vs AUROC 0.675 [95% CI: 0.599-0.751], respectively; P = .009).The BLR and CART models yielded similar accuracies for the prediction of ACN in Thai patients. The PLR model provided higher accuracy for ACN prediction than the CART model.

Evolving Role and Future Directions of Natural Language Processing in Gastroenterology

In line with the current trajectory of healthcare reform, significant emphasis has been placed on improving the utilization of data collected during a clinical encounter. Although the structured fields of electronic health records have provided a convenient foundation on which to begin such efforts, it was well understood that a substantial portion of relevant information is confined in the free-text narratives documenting care. Unfortunately, extracting meaningful information from such narratives is a non-trivial task, traditionally requiring significant manual effort. Today, computational approaches from a field known as Natural Language Processing (NLP) are poised to make a transformational impact in the analysis and utilization of these documents across healthcare practice and research, particularly in procedure-heavy sub-disciplines such as gastroenterology (GI). As such, this manuscript provides a clinically focused review of NLP systems in GI practice. It begins with a detailed synopsis around the state of NLP techniques, presenting state-of-the-art methods and typical use cases in both clinical settings and across other domains. Next, it will present a robust literature review around current applications of NLP within four prominent areas of gastroenterology including endoscopy, inflammatory bowel disease, pancreaticobiliary, and liver diseases. Finally, it concludes with a discussion of open problems and future opportunities of this technology in the field of gastroenterology and health care as a whole.

Risk prediction rule for advanced neoplasia on screening colonoscopy for average-risk individuals

BACKGROUND

In resource-limited countries, risk stratification can be used to optimize colorectal cancer screening. Few prospective risk prediction models exist for advanced neoplasia (AN) in true average-risk individuals.

AIM

To create and internally validate a risk prediction model for detection of AN in average-risk individuals.

METHODS

Prospective study of asymptomatic individuals undergoing first screening colonoscopy. Detailed characteristics including diet, exercise and medications were collected. Multivariate logistic regression was used to elucidate risk factors for AN (adenoma ≥1 cm, villous histology, high-grade dysplasia or carcinoma). The model was validated through bootstrapping, and discrimination and calibration of the model were assessed.

RESULTS

980 consecutive individuals (51% F; 49% M) were enrolled. Adenoma and AN detection rates were 36.6% (F 29%: M 45%; P < 0.001) and 5.1% (F 3.8%; M 6.5%) respectively. On multivariate analysis, predictors of AN [OR (95%CI)] were age [1.036 (1.00-1.07); P = 0.048], BMI [overweight 2.21 (0.98-5.00); obese 3.54 (1.48-8.50); P = 0.018], smoking [< 40 pack-years 2.01 (1.01-4.01); ≥ 40 pack-years 3.96 (1.86-8.42); P = 0.002], and daily red meat consumption [2.02 (0.92-4.42) P = 0.079]. Nomograms of AN risk were developed in terms of risk factors and age separately for normal, overweight and obese individuals. The model had good discrimination and calibration.

CONCLUSION

The prevalence of adenoma and AN in average-risk Lebanese individuals is similar to the West. Age, smoking, and BMI are important predictors of AN, with obesity being particularly powerful. Though external validation is needed, this model provides an important platform for improved risk-stratification for screening programs in regions where universal screening is not currently employed.

Global Prevalence of Colorectal Neoplasia: A Systematic Review and Meta-Analysis

BACKGROUND & AIMS

Most colorectal cancers (CRC) arise from colorectal adenomas, yet there is not enough information on global prevalence to inform health care policy. We examined the prevalence of any type of adenomas, advanced adenomas (AADs), and CRC according to age, sex, ethnicity, geographic regions, and anatomic location (proximal vs distal).

METHODS

MEDLINE and Embase were searched from their inception through May 1, 2018, to identify population-based, observational studies that reported the prevalence of colorectal neoplasia. Studies on participants 15 years or older, with a sample size of 500 persons or more, were included. Metaprop (College Station, TX) was used to model within-study variability by binomial distribution and Freeman-Tukey Double Arcsine Transformation to stabilize the variances. The prevalence figures were presented by proportions and their 95% CIs using random-effects models.

RESULTS

Our meta-analysis included 70 studies involving 637,414 individuals. The overall prevalence rates of adenoma (23.9%; 95% CI, 22.2%-25.8%), AAD (4.6%; 95% CI, 3.8%-5.5%), and CRC (0.4%, 95% CI, 0.3%-0.5%) were calculated. Subgroup analysis indicated that prevalence values (adenomas, AADs, and CRCs) were higher among men (29.7%, 6.5%, and 0.8%, respectively) than women (19.3%, 3.8% and 0.4%, respectively), among older adults (25.9%, 5.2%, and 0.6%, respectively) than younger adults (14.6%, 1.6%, and 0.1%, respectively), among Caucasians (23.7%, 6.6%, and 0.5%, respectively) than other ethnicities, in European countries (25.9%, 8.4%, and 0.8%, respectively) than other countries, and among patients with proximal (25.9%, 5.3%, and 0.1%, respectively) vs distal neoplasia.

CONCLUSIONS

In a systematic review and meta-analysis, we found a high prevalence of colorectal neoplasia among some populations. This indicates a need to expand CRC screening programs for these groups. The pooled prevalence estimates can be used as quality indicators for established CRC screening programs.

Head-to-Head Comparison of the Performance of 17 Risk Models for Predicting Presence of Advanced Neoplasms in Colorectal Cancer Screening

OBJECTIVES

Many risk scores have been proposed to predict presence of advanced colorectal neoplasms, but a comprehensive comparison conducted in the same population is sparse. The aim of this study was to evaluate and directly compare the diagnostic performance of published risk prediction models for advanced colorectal neoplasms.

METHODS

Data were drawn from 2 cohorts of subjects undergoing screening colonoscopy in Germany, i.e., KolosSal (n = 16,195) and BliTz (n = 7,444). Absolute risks and relative risks were generated for the presence of at least 1 advanced neoplasm, taking the lowest risk group as the reference group. Performance of risk models was assessed by the area under the receiver operating characteristic curve (AUC) and compared by the net reclassification improvement.

RESULTS

The 2 cohorts included 1,917 (11.8%) and 848 (11.4%) participants with advanced neoplasm, respectively. Absolute risks were mostly between 5% and 10% among participants in the lowest risk group and between 15% and 20% among participants in the highest risk group, and relative risks mostly ranged from 2.0 to 4.0 across the risk models in both cohorts. The AUCs ranged from 0.58 to 0.65 in KolosSal and from 0.57 to 0.61 in BliTz for all risk scores. Compared to models with lower AUC, classification was significantly improved in most models with higher AUC.

DISCUSSION

Risk models for advanced colorectal neoplasms generally yielded modest discriminatory power, despite some variation in performance between models. Future studies should evaluate the performance of these risk models in racially diverse populations and investigate possible extensions, such as combination with polygenic risk scores.

Is it possible to automatically assess pretreatment digital rectal examination documentation using natural language processing? A single-centre retrospective study

OBJECTIVES

To develop and test a method for automatic assessment of a quality metric, provider-documented pretreatment digital rectal examination (DRE), using the outputs of a natural language processing (NLP) framework.

SETTING

An electronic health records (EHR)-based prostate cancer data warehouse was used to identify patients and associated clinical notes from 1 January 2005 to 31 December 2017. Using a previously developed natural language processing pipeline, we classified DRE assessment as documented (currently or historically performed), deferred (or suggested as a future examination) and refused.

PRIMARY AND SECONDARY OUTCOME MEASURES

We investigated the quality metric performance, documentation 6 months before treatment and identified patient and clinical factors associated with metric performance.

RESULTS

The cohort included 7215 patients with prostate cancer and 426 227 unique clinical notes associated with pretreatment encounters. DREs of 5958 (82.6%) patients were documented and 1257 (17.4%) of patients did not have a DRE documented in the EHR. A total of 3742 (51.9%) patient DREs were documented within 6 months prior to treatment, meeting the quality metric. Patients with private insurance had a higher rate of DRE 6 months prior to starting treatment as compared with Medicaid-based or Medicare-based payors (77.3%vs69.5%, p=0.001). Patients undergoing chemotherapy, radiation therapy or surgery as the first line of treatment were more likely to have a documented DRE 6 months prior to treatment.

CONCLUSION

EHRs contain valuable unstructured information and with NLP, it is feasible to accurately and efficiently identify quality metrics with current documentation clinician workflow.

Colorectal cancer screening from 45 years of age: Thesis, antithesis and synthesis

Colorectal cancer incidence and mortality in patients younger than 50 years are increasing, but screening before the age of 50 is not offered in Europe. Advanced-stage diagnosis and mortality from colorectal cancer before 50 years of age are increasing. This is not a detection-bias effect; it is a real issue affecting the entire population. Three independent computational models indicate that screening from 45 years of age would yield a better balance of benefits and risks than the current start at 50 years of age. Experimental data support these predictions in a sex- and race-independent manner. Earlier screening is seemingly affordable, with minimal impediments to providing younger adults with colonoscopy. Indeed, the American Cancer Society has already started to recommend screening from 45 years of age in the United States. Implementing early screening is a societal and public health problem. The three independent computational models that suggested earlier screening were criticized for assuming perfect compliance. Guidelines and recommendations should be derived from well-collected and reproducible data, and not from mathematical predictions. In the era of personalized medicine, screening decisions might not be based solely on age, and sophisticated prediction software may better guide screening. Moreover, early screening might divert resources away from older individuals with greater biological risks. Finally, it is still unknown whether early colorectal cancer is part of a continuum of disease or a biologically distinct disease and, as such, it might not benefit from screening at all. The increase in early-onset colorectal cancer incidence and mortality demonstrates an obligation to take actions. Earlier screening would save lives, and starting at the age of 45 years may be a robust screening option.

An Automated Feature Engineering for Digital Rectal Examination Documentation using Natural Language Processing

Digital rectal examination (DRE) is considered a quality metric for prostate cancer care. However, much of the DRE related rich information is documented as free-text in clinical narratives. Therefore, we aimed to develop a natural language processing (NLP) pipeline for automatic documentation of DRE in clinical notes using a domain-specific dictionary created by clinical experts and an extended version of the same dictionary learned by clinical notes using distributional semantics algorithms. The proposed pipeline was compared to a baseline NLP algorithm and the results of the proposed pipeline were found superior in terms of precision (0.95) and recall (0.90) for documentation of DRE. We believe the rule-based NLP pipeline enriched with terms learned from the whole corpus can provide accurate and efficient identification of this quality metric.

Risk Scores for Predicting Advanced Colorectal Neoplasia in the Average-risk Population: A Systematic Review and Meta-analysis

OBJECTIVES

A systematic review and meta-analysis was performed to summarize the available evidence on risk scores for predicting advanced colorectal neoplasia (advanced adenomas and cancer) in average-risk and asymptomatic populations undergoing screening colonoscopy.

METHODS

PubMed, EMBASE, and Web of Science databases were searched up to 28 March 2018. Studies that developed or validated a risk score to predict the risk of advanced colorectal neoplasia were included. Two reviewers independently extracted study characteristics including diagnostic performance indicators and assessed risk of bias and applicability in the included studies. Meta-analyses were conducted to determine the overall discrimination of risk scores evaluated by more than 1 study.

RESULTS

A total of 22 studies including 17 original risk scores were identified. Risk scores included a median number of 5 risk factors. Factors most commonly included were age, sex, family history in first-degree relatives, body mass index and smoking. The area under the receiver operating characteristic curve of risk scores ranged from 0.62 to 0.77 in the individual studies and from 0.61 to 0.70 in the meta-analyses.

CONCLUSIONS

Although the majority of available risk scores had relatively weak discriminatory power, they may be of some use for risk stratification in CRC screening. Rather than developing more risk scores based on environmental risk factors, future research should focus on exploring possibilities of enhancing predictive power by combining risk factor data with novel laboratory matters, such as polygenetic risk scores.

POPCORN: A web service for individual PrognOsis prediction based on multi-center clinical data CollabORatioN without patient-level data sharing

BACKGROUND AND OBJECTIVE

Clinical prognosis prediction plays an important role in clinical research and practice. The construction of prediction models based on electronic health record data has recently become a research focus. Due to the lack of external validation, prediction models based on single-center, hospital-specific datasets may not perform well with datasets from other medical institutions. Therefore, research investigating prognosis prediction model construction based on a collaborative analysis of multi-center electronic health record data could increase the number and coverage of patients used for model training, enrich patient prognostic features and ultimately improve the accuracy and generalization of prognosis prediction.

MATERIALS AND METHODS

A web service for individual prognosis prediction based on multi-center clinical data collaboration without patient-level data sharing (POPCORN) was proposed. POPCORN focuses on solving key issues in multi-center collaborative research based on electronic health record systems; these issues include the standardization of clinical data expression, the preservation of patient privacy during model training and the effect of case mix variance on the prediction model construction and application. POPCORN is based on a multivariable meta-analysis and a Bayesian framework and can construct suitable prediction models for multiple clinical scenarios that can effectively adapt to complex clinical application environments.

RESULTS

POPCORN was validated using a joint, multi-center collaborative research network between China and the United States with patients diagnosed with colorectal cancer. The performance of the models based on POPCORN was comparable to that of the standard prognosis prediction model; however, POPCORN did not expose raw patient data. The prediction models had similar AUC, but the BMA model had the lowest ECI across all prediction models, indicating that this model had better calibration performance than the other models, especially for patients in Chinese hospitals.

CONCLUSIONS

The POPCORN system can build prediction models that perform well in complex clinical application scenarios and can provide effective decision support for individual patient prognostic predictions.

The association between Helicobacter pylori infection and the risk of advanced colorectal neoplasia may differ according to age and cigarette smoking

BACKGROUND

The association between Helicobacter pylori infection and advanced colorectal neoplasia (ACN) remains controversial. This study aimed to clarify the association between H. pylori infection and ACN according to age groups.

METHODS

We retrospectively analyzed the association between H. pylori infection and ACN in patients aged <50 and ≥50 years receiving a health checkup that included colonoscopy. Helicobacter pylori positivity was determined by the results of serum anti-H. pylori immunoglobulin G or rapid urease test, if the anti-H. pylori immunoglobulin G was in the borderline range.

RESULTS

Among the 19 337 patients who were included, 56.2% and 3.4% were positive for H. pylori and ACN, respectively. Helicobacter pylori infection independently increased the risk of ACN in patients aged <50 years (odds ratio [OR], 1.602; 95% confidence intervals [CI], 1.194-2.150) but not in patients aged ≥50 years (OR, 1.046; 95% CI, 0.863-1.268). The positive association between H. pylori infection and ACN was affected by smoking history. When stratified by age and smoking history, H. pylori infection conferred an increased risk of ACN in patients aged <50 years with a history of smoking (OR, 1.926; 95% CI, 1.336-2.775) but not in the other 3 groups (3-way interaction test P = .023). Among patients aged <50 years with ACN, ACN in the left colon was found more frequently in patients with H. pylori infection and a history of smoking than in those without (69.3% vs 54.4%, respectively; P = .031).

CONCLUSIONS

Helicobacter pylori infection confers an increased risk of ACN, but the association may differ according to age and smoking history.