Using machine learning to parse breast pathology reports.

PURPOSE: Extracting information from electronic medical record is a time-consuming and expensive process when done manually. Rule-based and machine learning techniques are two approaches to solving this problem. In this study, we trained a machine learning model on pathology reports to extract pertinent tumor characteristics, which enabled us to create a large database of attribute searchable pathology reports. This database can be used to identify cohorts of patients with characteristics of interest. METHODS: We collected a total of 91,505 breast pathology reports from three Partners hospitals: Massachusetts General Hospital, Brigham and Women's Hospital, and Newton-Wellesley Hospital, covering the period from 1978 to 2016. We trained our system with annotations from two datasets, consisting of 6295 and 10,841 manually annotated reports. The system extracts 20 separate categories of information, including atypia types and various tumor characteristics such as receptors. We also report a learning curve analysis to show how much annotation our model needs to perform reasonably. RESULTS: The model accuracy was tested on 500 reports that did not overlap with the training set. The model achieved accuracy of 90% for correctly parsing all carcinoma and atypia categories for a given patient. The average accuracy for individual categories was 97%. Using this classifier, we created a database of 91,505 parsed pathology reports. CONCLUSIONS: Our learning curve analysis shows that the model can achieve reasonable results even when trained on a few annotations. We developed a user-friendly interface to the database that allows physicians to easily identify patients with target characteristics and export the matching cohort. This model has the potential to reduce the effort required for analyzing large amounts of data from medical records, and to minimize the cost and time required to glean scientific insight from these data.

Estrogen switches pure mucinous breast cancer to invasive lobular carcinoma with mucinous features.

Mucinous breast cancer (MBC) is mainly a disease of postmenopausal women. Pure MBC is rare and augurs a good prognosis. In contrast, MBC mixed with other histological subtypes of invasive disease loses the more favorable prognosis. Because of the relative rarity of pure MBC, little is known about its cell and tumor biology and relationship to invasive disease of other subtypes. We have now developed a human breast cancer cell line called BCK4, in which we can control the behavior of MBC. BCK4 cells were derived from a patient whose poorly differentiated primary tumor was treated with chemotherapy, radiation and tamoxifen. Malignant cells from a recurrent pleural effusion were xenografted in mammary glands of a nude mouse. Cells from the solid tumor xenograft were propagated in culture to generate the BCK4 cell line. Multiple marker and chromosome analyses demonstrate that BCK4 cells are human, near diploid and luminal, expressing functional estrogen, androgen, and progesterone receptors. When xenografted back into immunocompromised cycling mice, BCK4 cells grow into small pure MBC. However, if mice are supplemented with continuous estradiol, tumors switch to invasive lobular carcinoma (ILC) with mucinous features (mixed MBC), and growth is markedly accelerated. Tamoxifen prevents the expansion of this more invasive component. The unexpected ability of estrogens to convert pure MBC into mixed MBC with ILC may explain the rarity of the pure disease in premenopausal women. These studies show that MBC can be derived from lobular precursors and that BCK4 cells are new, unique models to study the phenotypic plasticity, hormonal regulation, optimal therapeutic interventions, and metastatic patterns of MBC.

Identifying Targets for Innovation in Amazon Reviews of Bedwetting Alarms: Thematic Analysis.

BACKGROUND: Nocturnal enuresis (NE) is a frequent diagnosis in pediatric and adolescent populations with an estimated prevalence of around 15% at the age of 6 years. NE can have a substantial impact on multiple health domains. Bedwetting alarms, which typically consist of a sensor and moisture-activated alarm, are a common treatment. OBJECTIVE: This study aimed to determine areas of satisfaction versus dissatisfaction reported by the parents and caregivers of children using current bedwetting alarms. METHODS: Using the search term "bedwetting alarms" on the Amazon marketplace, products with >300 reviews were included. For each product, the 5 reviews ranked the "most helpful" for each star category were selected for analysis. Meaning extraction method was applied to identify major themes and subthemes. A percent skew was calculated by summing the total number of mentions of each subtheme,+1 for a positive mention, 0 for a neutral mention, and -1 for a negative mention, and dividing this total by the number of reviews in which that particular subtheme was observed. Subanalyses were performed for age and gender. RESULTS: Of 136 products identified, 10 were evaluated based on the selection criteria. The main themes identified across products were long-term concerns, marketing, alarm systems, and device mechanics and features. The subthemes identified as future targets for innovation included alarm accuracy, volume variability, durability, user-friendliness, and adaptability to girls. In general, durability, alarm accuracy, and comfort were the most negatively skewed subthemes (with a negative skew of -23.6%, -20.0%, and -12.4% respectively), which are indicative of potential areas for improvement. Effectiveness was the only substantially positively skewed subtheme (16.8%). Alarm sound and device features were positively skewed for older children, whereas ease of use had a negative skew for younger children. Girls and their caretakers reported negative experiences with devices that featured cords, arm bands, and sensor pads. CONCLUSIONS: This analysis provides an innovation roadmap for future device design to improve patient and caregiver satisfaction and compliance with bedwetting alarms. Our results highlight the need for additional options in alarm sound features, as children of different ages have divergent preferences in this domain. Additionally, girls and their parents and caretakers provided more negative overall reviews regarding the range of current device features compared to boys, indicating a potential focus area for future development. The percent skew showed that subthemes were often more negatively skewed toward girls, with the ease of use being -10.7% skewed for boys versus -20.5% for girls, and comfort being -7.1% skewed for boys versus -29.4% for girls. Put together, this review highlights multiple device features that are targets for innovation to ensure translational efficacy regardless of age, gender, or specific family needs.

Fracture of the Costal Cartilage: Presentation, Diagnosis, and Management.

A 52-year-old woman sustained a fracture of the left 7th costal cartilage after a ski injury. She presented complaining of painful clicking over the area. Initial imaging studies were negative for fracture; however, 3-dimensional reconstruction of a chest computed tomography scan, formatted to costal cartilage, revealed the fracture. She was offered and underwent surgical fixation of the fracture with a plate and 4 screws using a biaxial, convergent construct. Postoperatively, her symptoms resolved. In this case report, we review the rationale for fixation of costal cartilage, including a summary of previous literature pertaining to this relatively rare thoracic injury.

A Qualitative Study of Family Experience With Hospitalization for Neonatal Abstinence Syndrome.

BACKGROUND AND OBJECTIVES: Although the incidence of neonatal abstinence syndrome (NAS) in the United States quintupled between 2000 and 2012, little is known about the family perspective of the hospital stay. We interviewed families to understand their experiences during the newborn hospitalization for NAS and to improve family-centered care. METHODS: A multidisciplinary team from 3 hospital units composed open-ended interview questions based on a literature review, clinical experience, and an internal iterative process. Trained investigators conducted semi-structured interviews with 20 families of newborns with NAS at hospital discharge. Interviews were recorded and transcribed verbatim. Two investigators independently analyzed each transcript, identified themes via an inductive qualitative approach, and reached a consensus on each code. The research team sorted the themes into broader domains through an iterative process that required consensus of 4 team members. RESULTS: Five domains of family experience were identified: parents' desire for education about the course and treatment of NAS; parents valuing their role in the care team; quality of interactions with staff (supportive versus judgmental) and communication regarding clinical course; transfers between units and inconsistencies among providers; and external factors such as addiction recovery and economic limitations. CONCLUSIONS: Families face many challenges during newborn hospitalization for NAS. Addressing parental needs through improved perinatal education, increased involvement in the care team, consistent care and communication, and minimized transitions in care could improve the NAS hospital experience. The results of this qualitative study may allow for improvements in family-centered care of infants with NAS.