Development and validation of case-finding algorithms for recurrence of breast cancer using routinely collected administrative data

Xu, Yuan; Kong, Shiying; Cheung, Winson Y.; Bouchard‐Fortier, Antoine; Dort, Joseph C.; Quan, Hude; Buie, Elizabeth M.; McKinnon, Geoff; Quan, May Lynn

doi:10.1186/s12885-019-5432-8

Cited by 30 publications

(34 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…In some chronic conditions, multiple drugs are indicated, for example, there are half dozen statins. 43 In some patients, a certain drug may be discontinued and another drug may The features presented here, including F i (the length of use), describe complex drug use which could also be used as an alternative input for complex algorithms and other machine learning applications 45,46 that use a wide variety of input data.…”

Section: Discussionmentioning

confidence: 99%

Classification of drug use patterns

Righolt

Zhang

Mahmud

2020

Pharmacology Res & Perspec

View full text Add to dashboard Cite

Characterizing long‐term prescription data is challenging due to the time‐varying nature of drug use. Conventional approaches summarize time‐varying data into categorical variables based on simple measures, such as cumulative dose, while ignoring patterns of use. The loss of information can lead to misclassification and biased estimates of the exposure‐outcome association. We introduce a classification method to characterize longitudinal prescription data with an unsupervised machine learning algorithm. We used administrative databases covering virtually all 1.3 million residents of Manitoba and explicitly designed features to describe the average dose, proportion of days covered (PDC), dose change, and dose variability, and clustered the resulting feature space using K‐means clustering. We applied this method to metformin use in diabetes patients. We identified 27,786 metformin users and showed that the feature distributions of their metformin use are stable for varying the lengths of follow‐up and that these distributions have clear interpretations. We found six distinct metformin user groups: patients with intermittent use, decreasing dose, increasing dose, high dose, and two medium dose groups (one with stable dose and one with highly variable use). Patients in the varying and decreasing dose groups had a higher chance of progression of diabetes than other patients. The method presented in this paper allows for characterization of drug use into distinct and clinically relevant groups in a way that cannot be obtained from merely classifying use by quantiles of overall use.

show abstract

Section: Discussionmentioning

confidence: 99%

Classification of drug use patterns

Righolt

Zhang

Mahmud

2020

Pharmacology Res & Perspec

View full text Add to dashboard Cite

show abstract

“…[13][14][15]17,18 Studies from North America have validated algorithms to identify patients diagnosed with recurrence of breast, colorectal, and lung cancer. [6][7][8][9][10]12 In these studies, indicators of recurrence were based on procedure codes, diagnosis codes, and hospice and oncology visits; the sensitivity ranged from 70% to 94%, and the specificity ranged from 70% to 98%. Only two studies reached a sensitivity above 90%.…”

Section: Comparison With Literaturementioning

confidence: 99%

“…Only two studies reached a sensitivity above 90%. 6,9 The accurate performance of the algorithms described in Danish studies seems to result from the inclusion of pathology data as indicators of recurrence. [13][14][15]17 Restrictions to the pathology coding prevented false positives, and the pathology data reached high positive predictive values, ranging from 91% to 100% across the studies by Rasmussen et al 15,17,18 Studies from the United States based their algorithms on specific medical claims combined with cancer-related register data and other administrative data restricted to specific populations, geographic areas, or insurance groups.…”

Section: Comparison With Literaturementioning

confidence: 99%

“…3 Internationally, algorithms based on routinely collected data like medical claims and cancer-related register data have been developed to identify patients diagnosed with cancer recurrence, mainly from breast and colorectal cancer; the validation of these algorithms has shown mixed results. [6][7][8][9][10][11][12] Recent studies have shown that cancer-specific algorithms based on Danish national health registries can accurately identify patients diagnosed with recurrence of colorectal cancer, 13 breast cancer, [14][15][16] bladder cancer, 17 or endometrial cancer, 18 with little selection bias. However, no similar approach exists to identify patients diagnosed with recurrence of malignant melanoma.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

A Validated Register-Based Algorithm to Identify Patients Diagnosed with Recurrence of Malignant Melanoma in Denmark

Rasmussen¹,

Jensen²,

Virgilsen³

et al. 2021

CLEP

View full text Add to dashboard Cite

Information on cancer recurrence is rarely available outside clinical trials. Wide exclusion criteria used in clinical trials tend to limit the generalizability of findings to the entire population of people living beyond a cancer disease. Therefore, population-level evidence is needed. The aim of this study was to develop and validate a register-based algorithm to identify patients diagnosed with recurrence after curative treatment of malignant melanoma. Patients and Methods: Indicators of recurrence were diagnosis and procedure codes recorded in the Danish National Patient Register and pathology results recorded in the Danish National Pathology Register. Medical records on recurrence status and recurrence date in the Danish Melanoma Database served as the gold standard to assess the accuracy of the algorithm. Results: The study included 1747 patients diagnosed with malignant melanoma; 95 (5.4%) were diagnosed with recurrence of malignant melanoma according to the gold standard. The algorithm reached a sensitivity of 93.7% (95% confidence interval (CI) 86.8-97.6), a specificity of 99.2% (95% CI: 98.6-99.5), a positive predictive value of 86.4% (95% CI: 78.2-92.4), and negative predictive value of 99.6% (95% CI: 99.2-99.9). Lin's concordance correlation coefficient was 0.992 (95% CI: 0.989-0.996) for the agreement between the recurrence dates generated by the algorithm and by the gold standard. Conclusion: The algorithm can be used to identify patients diagnosed with recurrence of malignant melanoma and to establish the timing of recurrence. This can generate populationlevel evidence on disease-free survival and diagnostic pathways for recurrence of malignant melanoma.

show abstract

“…By splitting the neurons into two directions of a text sequence, BiLSTM could learn forward and backward information of input words, Furthermore, BiLSTM with CRF (BiLSTM-CRF), proved its validity that outperformed the traditional models especially in Chinese clinical NER tasks [17][18][19][20][21]. After information extraction to obtain the structured features, NLP can be further implemented on clinical tasks, such as disease studies [22][23][24], drug-related studies [25,26], and clinical workflow optimization [27]. Computer-aided diagnosis is an important research field in disease study, which aims to use computer algorithms to provide physicians a reference for disease diagnosis.…”

Section: Introductionmentioning

confidence: 99%

A Natural Language Processing Pipeline of Chinese Free-Text Radiology Reports for Liver Cancer Diagnosis

Liu

Zhang

et al. 2020

IEEE Access

View full text Add to dashboard Cite

Despite the rapid development of natural language processing (NLP) implementation in electronic medical records (EMRs), Chinese EMRs processing remains challenging due to the limited corpus and specific grammatical characteristics, especially for radiology reports. In this study, we designed an NLP pipeline for the direct extraction of clinically relevant features from Chinese radiology reports, which is the first key step in computer-aided radiologic diagnosis. The pipeline was comprised of named entity recognition, synonyms normalization, and relationship extraction to finally derive the radiological features composed of one or more terms. In named entity recognition, we incorporated lexicon into deep learning model bidirectional long short-term memory-conditional random field (BiLSTM-CRF), and the model finally achieved an F1 score of 93.00%. With the extracted radiological features, least absolute shrinkage and selection operator and machine learning methods (support vector machine, random forest, decision tree, and logistic regression) were used to build the classifiers for liver cancer prediction. For liver cancer diagnosis, random forest had the highest predictive performance in liver cancer diagnosis (F1 score 86.97%, precision 87.71%, and recall 86.25%). This work was a comprehensive NLP study focusing on Chinese radiology reports and the application of NLP in cancer risk prediction. The proposed NLP pipeline for the radiological feature extraction could be easily implemented in other kinds of Chinese clinical texts and other disease predictive tasks.

show abstract

Development and validation of case-finding algorithms for recurrence of breast cancer using routinely collected administrative data

Cited by 30 publications

References 21 publications

Classification of drug use patterns

Classification of drug use patterns

A Validated Register-Based Algorithm to Identify Patients Diagnosed with Recurrence of Malignant Melanoma in Denmark

A Natural Language Processing Pipeline of Chinese Free-Text Radiology Reports for Liver Cancer Diagnosis

Contact Info

Product

Resources

About