Preoperative magnetic resonance imaging use and oncologic outcomes in premenopausal breast cancer patients.

Breast magnetic resonance imaging (MRI) delineates disease extent sensitively in newly diagnosed breast cancer patients, but improved cancer outcomes are uncertain. Young women, for whom mammography is less sensitive, are expected to benefit from MRI-based resection. We identified 512 women aged ≤50 years, undergoing breast-conserving treatment (BCT: tumor-free resection margins and radiotherapy) during 2006-2013 through Northwestern Medicine database queries; 64.5% received preoperative MRI and 35.5% did not. Tumor and treatment parameters were similar between groups. We estimated the adjusted hazard ratios (aHR) for local and distant recurrences (LR and DR), using multivariable regression models, accounting for important therapeutic and prognostic parameters. LR rate with MRI use was 7.9 vs. 8.2% without MRI, aHR = 1.03 (95% CI 0.53-1.99). DR rate was 6.4 vs. 6.6%, aHR = 0.89 (95% CI 0.43-1.84). In 119 women aged ≤40, results were similar to LR aHR = 1.82 (95% CI 0.43-7.76) and DR aHR = 0.93 (95% CI 0.26-3.34). Sensitivity analyses showed similar results. The use of preoperative MRI in women aged ≤50 years should be reconsidered until there is proof of benefit.

Identifying Breast Cancer Distant Recurrences from Electronic Health Records Using Machine Learning.

Accurately identifying distant recurrences in breast cancer from the Electronic Health Records (EHR) is important for both clinical care and secondary analysis. Although multiple applications have been developed for computational phenotyping in breast cancer, distant recurrence identification still relies heavily on manual chart review. In this study, we aim to develop a model that identifies distant recurrences in breast cancer using clinical narratives and structured data from EHR. We applied MetaMap to extract features from clinical narratives and also retrieved structured clinical data from EHR. Using these features, we trained a support vector machine model to identify distant recurrences in breast cancer patients. We trained the model using 1,396 double-annotated subjects and validated the model using 599 double-annotated subjects. In addition, we validated the model on a set of 4,904 single-annotated subjects as a generalization test. In the held-out test and generalization test, we obtained F-measure scores of 0.78 and 0.74, area under curve (AUC) scores of 0.95 and 0.93, respectively. To explore the representation learning utility of deep neural networks, we designed multiple convolutional neural networks and multilayer neural networks to identify distant recurrences. Using the same test set and generalizability test set, we obtained F-measure scores of 0.79 ± 0.02 and 0.74 ± 0.004, AUC scores of 0.95 ± 0.002 and 0.95 ± 0.01, respectively. Our model can accurately and efficiently identify distant recurrences in breast cancer by combining features extracted from unstructured clinical narratives and structured clinical data.

Using natural language processing and machine learning to identify breast cancer local recurrence.

BACKGROUND: Identifying local recurrences in breast cancer from patient data sets is important for clinical research and practice. Developing a model using natural language processing and machine learning to identify local recurrences in breast cancer patients can reduce the time-consuming work of a manual chart review. METHODS: We design a novel concept-based filter and a prediction model to detect local recurrences using EHRs. In the training dataset, we manually review a development corpus of 50 progress notes and extract partial sentences that indicate breast cancer local recurrence. We process these partial sentences to obtain a set of Unified Medical Language System (UMLS) concepts using MetaMap, and we call it positive concept set. We apply MetaMap on patients' progress notes and retain only the concepts that fall within the positive concept set. These features combined with the number of pathology reports recorded for each patient are used to train a support vector machine to identify local recurrences. RESULTS: We compared our model with three baseline classifiers using either full MetaMap concepts, filtered MetaMap concepts, or bag of words. Our model achieved the best AUC (0.93 in cross-validation, 0.87 in held-out testing). CONCLUSIONS: Compared to a labor-intensive chart review, our model provides an automated way to identify breast cancer local recurrences. We expect that by minimally adapting the positive concept set, this study has the potential to be replicated at other institutions with a moderately sized training dataset.

Evaluating Mastectomy Skin Flap Necrosis in the Extended Breast Reconstruction Risk Assessment Score for 1-Year Prediction of Prosthetic Reconstruction Outcomes.

BACKGROUND: Rates of mastectomy for breast cancer treatment and immediate reconstruction continue to rise. With increasing scrutiny on outcomes and patient satisfaction, there is an impetus for providers to be more deliberate in appropriate patient selection for breast reconstruction. The Breast Reconstruction Risk Assessment (BRA) Score was developed for prediction of complications after primary prosthetic breast reconstruction, focusing on calculating risk estimations for a variety of complications based on individual patient demographic and perioperative characteristics. In this study, we evaluated mastectomy skin flap necrosis (MSFN) as a function of patient characteristics to validate the BRA Score. STUDY DESIGN: We examined our prospective intra-institutional database of prosthetic breast reconstructions from 2004 to 2015. The end point of interest was 1-year occurrence of MSFN after stage I tissue expander placement. RESULTS: Nine hundred and three patients were included; 50% underwent bilateral reconstruction. Median follow-up was 23 months. Mean 1-year complication rates were as follows: MSFN 12.4%, seroma 3.0%, infection 6.9%, dehiscence/exposure 7.1%, and explantation 13.2%. Statistically significantly higher rates of MSFN were found in older patients, smokers, patients with postoperative infections, patients with hypertension, and patients who used aspirin. Neoadjuvant or adjuvant chemotherapy and radiation, diabetes, and seroma formation did not have a statistically significant impact on necrosis rates. CONCLUSIONS: The BRA Score was expanded to estimate complication risk after tissue expander placement up to 1 year postoperatively. The risk of MSFN as calculated by the BRA Score: Extended Length is consistent with published studies demonstrating increased risk with specific comorbidities, and further validates expansion of the BRA score risk calculator.

Contralateral Breast Cancer Event Detection Using Nature Language Processing.

To facilitate the identification of contralateral breast cancer events for large cohort study, we proposed and implemented a new method based on features extracted from narrative text in progress notes and features from numbers of pathology reports for each side of breast cancer. Our method collects medical concepts and their combinations to detect contralateral events in progress notes. In addition, the numbers of pathology reports generated for either left or right side of breast cancer were derived as additional features. We experimented with support vector machine using the derived features to detect contralateral events. In the cross-validation and held-out tests, the area under curve score is 0.93 and 0.89 respectively. This method can be replicated due to the simplicity of feature generation.

Percutaneous endoscopic gastrostomy site metastasis from head and neck squamous cell carcinoma: case series and literature review.

OBJECTIVES: To present our experience with head and neck squamous cell carcinoma (HNSCC) seeding of percutaneous endoscopic gastrostomy (PEG) sites and to review all reported cases to identify risk factors and develop strategies for complication avoidance. MATERIALS AND METHODS: The records of 4 patients with PEG site metastasis from HNSCC were identified from the authors' institution. Thirty-eight further cases were reviewed following a PubMed search and evaluation of references in pertinent articles. RESULTS: Review of 42 cases revealed the average time from PEG to diagnosis of metastatic disease to be 8 months. Average time to death from detection of PEG disease was 5.9 months. One-year survival following PEG metastasis was 35.5% with an overall mortality of 87.1%. CONCLUSION: PEG site metastatic disease portends a poor prognosis. Early detection and aggressive therapy may provide a chance of cure. Changes in PEG technique or in timing of adjunctive therapies are possible avenues in further research to prevent this complication.

Alphanumeric paging: a potential source of problems in patient care and communication.

PURPOSE: In recent years, the use of numeric paging in many medical centers has been largely replaced by 1-way alphanumeric paging. There is currently no research studying the potential for alphanumeric paging to lead to problems in communication. The purpose of this article is to determine whether the use of alphanumeric pagers may lead to potential problems in patient care and/or communication. METHODS: Alphanumeric pages sent to residents on 3 surgical services at the Medical College of Virginia Hospital were collected over a 3-month period. The pages were classified according to reason for the page, amount of information provided, and follow-up required. RESULTS: A total of 52,384 alphanumeric pages were sent to residents on the surgical services over a 3-month period. There were 1037 pages (2.0% of total) that contained patient laboratory results. 11,844 pages (22.6% of total) contained a callback number with no sender information and 6198 (11.8% of total) contained a callback number and sender information. Trauma pages totaled 10,312 (19.7% of total). There were 2636 pages (5.0% of total) that contained identifying information, potentially violating HIPAA regulations. CONCLUSIONS: The authors have observed a significant number of occurrences in which alphanumeric pages lack sufficient information, do not indicate the urgency of the page, and still require immediate callback by residents. This potentially interrupts patient care and educational activities.