Temporal Association Rule Mining: Race-Based Patterns of Treatment-Adverse Events in Breast Cancer Patients Using SEER-Medicare Dataset.

PURPOSE: Disparities in the screening, treatment, and survival of African American (AA) patients with breast cancer extend to adverse events experienced with systemic therapy. However, data are limited and difficult to obtain. We addressed this challenge by applying temporal association rule (TAR) mining using the SEER-Medicare dataset for differences in the association of specific adverse events (AEs) and treatments (TRs) for breast cancer between AA and White women. We considered two categories of cancer care providers and settings: practitioners providing care in the outpatient units of hospitals and institutions and private practitioners providing care in their offices. PATIENTS AN METHODS: We considered women enrolled in the Medicare fee-for-service option at age 65 who qualified by age and not disability, who were diagnosed with breast cancer with attributed patient factors of age and race, marital status, comorbidities, prior malignancies, prior therapy, disease factors of stage, grade, and ER/PR and Her2 status and laterality. We included 141 HCPCS drug J codes for chemotherapy, biotherapy, and hormone therapy drugs, which we consolidated into 46 mechanistic categories and generated AE data. We consolidated AEs from ICD9 codes into 18 categories associated with breast cancer therapy. We applied TAR mining to determine associations between the 46 TR and 18 AE categories in the context of the patient categories outlined. We applied the spark.mllib implementation of the FPGrowth algorithm, a parallel version called PFP. We considered differences of at least one unit of lift as significant between groups. The model's results demonstrated a high overlap between the model's identified TR-AEs associated set and the actual set. RESULTS: Our results demonstrate that specific TR/AE associations are highly dependent on race, stage, and venue of care administration. CONCLUSIONS: Our data demonstrate the usefulness of this approach in identifying differences in the associations between TRs and AEs in different populations and serve as a reference for predicting the likelihood of AEs in different patient populations treated for breast cancer. Our novel approach using unsupervised learning enables the discovery of association rules while paying special attention to temporal information, resulting in greater predictive and descriptive power as a patient's health and life status change over time.

Racial Disparities in Breast Cancer Treatments and Adverse Events in the SEER-Medicare Data.

Despite lower incidence rates, African American (AA) patients have shorter survival from breast cancer (BC) than white (W) patients. Multiple factors contribute to decreased survival, including screening disparities, later presentation, and access to care. Disparities in adverse events (AEs) may contribute to delayed or incomplete treatment, earlier recurrence, and shortened survival. Here, we analyzed the SEER-Medicare dataset, which captures claims from a variety of venues, in order to determine whether the cancer care venues affect treatment and associated adverse events. We investigated a study population whose claims are included in the Outpatient files, consisting of hospital and healthcare facility venues, and a study population from the National Claims History (NCH) files, consisting of claims from physicians, office practices, and other non-institutional providers. We demonstrated statistically and substantively significant venue-specific differences in treatment rates, drugs administered, and AEs from treatments between AA and W patients. We showed that AA patients in the NCH dataset received lower rates of treatment, but patients in the Outpatient dataset received higher rates of treatment than W patients. The rates of recorded AEs per treatment were higher in the NCH setting than in the Outpatient setting in all patients. AEs were consistently higher in AA patients than in W patients. AA patients had higher comorbidity indices and were younger than W patients, but these variables did not appear to play roles in the AE differences. The frequency of specific anticancer drugs administered in cancer- and venue-specific circumstances and their associated AEs varied between AA and W patients. The higher AE rates were due to slightly higher frequencies in the administration of drugs with higher associated AE rates in AA patients than in W patients. Our investigations demonstrate significant differences in treatment rates and associated AEs between AA and W patients with BC, depending on the venues of care, likely contributing to differences in outcomes.

A Framework for Dynamic Composition and Management of Emergency Response Processes.

An emergency response process outlines the workflow of different activities that need to be performed in response to an emergency. Effective emergency response requires communication and coordination with the operational systems belonging to different collaborating organizations. Therefore, it is necessary to establish information sharing and system-level interoperability among the diverse operational systems. Unlike typical e-government processes that are well structured and have a well-defined outcome, emergency response processes are knowledge-centric and their workflow structure and execution may evolve as the incident unfolds. It is impractical to define static plans and response process workflows for every possible situation. Instead, a dynamic response should be adaptable to the changing situation. We present an integrated approach that facilitates the dynamic composition of an executable response process. The proposed approach employs ontology-based reasoning to determine the default actions and resource requirements for the given incident and to identify relevant response organizations based on their jurisdictional and mutual aid agreement rules. The Web service APIs of the identified response organizations are then used to generate an executable response process that evolves dynamically. The proposed approach is implemented and experimentally validated using an example scenario derived from the FEMA Hazardous Materials Tabletop Exercises Manual.

Drug repositioning for cancer in the era of AI, big omics, and real-world data.

Drug repositioning in cancer has been pursued for years because of slowing drug development, increasing costs, and the availability of drugs licensed for other indications with anticancer effects in the laboratory. Repositioning has encountered obstacles due to generally insufficient single-agent clinical anticancer effects of licensed drugs and a subsequent reluctance by pharmaceutical companies to invest in phase III combination studies with them. Here we review potential machine learning/artificial intelligence (ML/AI) approaches for using real-world data (RWD) that could overcome the limitations of clinical trials and retrospective analyses. We outline a two-tiered filtering approach of identifying top-ranked drugs based on their drug-target binding affinity scores while considering their challenges and matching the top-ranked drugs with their top-ranked specific scenarios from among the multitude of real-world scenarios for efficacy and safety. This approach will generate RWD scenario-specific hypotheses that can be tested in randomized clinical trials with high probabilities of success.

Greater Survival Improvement in African American vs. Caucasian Women with Hormone Negative Breast Cancer.

Background: African American women have not benefited equally from recently improved breast cancer survival. We investigated if this was true for all subsets. Methods: We identified 395,170 patients with breast adenocarcinoma from the SEER database from 1990 to 2011 with designated race, age, stage, grade, ER and PR status, marital status and laterality, as control. We grouped patients into two time periods, 1990-2000 and 2001-2011, three age categories, under 40, 40-69 and ≥ 70 years and two stage categories, I-III and IV. We used the Kaplan-Meier and logrank tests to compare survival curves. We stratified data by patient- and tumor-associated variables to determine co-variation among confounding factors using the Pearson Chi-square test and Cox proportional hazards regression to determine hazard ratios (HR) to compare survival. Results: Stage I-III patients of both races ≥ 70 years old, African American widowed patients and Caucasians with ER- and PR- tumors had worse improvements in survival in 2001-2011 than younger, married or hormone receptor positive patients, respectively. In contrast, African Americans with ER- (Cox HR 0.70 [95% CI 0.65-0.76]) and PR- (Cox HR 0.67 [95% CI 0.62-0.72]) had greater improvement in survival in 2001-2011 than Caucasians with ER- (Cox HR 0.81 [95% CI 0.78-0.84]) and PR- disease (Cox HR 0.75 [95% CI 0.73-0.78]). This was not associated with changes in distribution of tumor or patient attributes. Conclusions: African American women with stage I-III ER- and PR- breast cancer had greater improvement in survival than Caucasians in 2001-2011. This is the first report of an improvement in racial disparities in survival from breast cancer in a subset of patients.

DIFFERENTIALLY PRIVATE OUTLIER DETECTION IN A COLLABORATIVE ENVIRONMENT.

Outlier detection is one of the most important data analytics tasks and is used in numerous applications and domains. The goal of outlier detection is to find abnormal entities that are significantly different from the remaining data. Often the underlying data is distributed across different organizations. If outlier detection is done locally, the results obtained are not as accurate as when outlier detection is done collaboratively over the combined data. However, the data cannot be easily integrated into a single database due to privacy and legal concerns. In this paper, we address precisely this problem. We first define privacy in the context of collaborative outlier detection. We then develop a novel method to find outliers from both horizontally partitioned and vertically partitioned categorical data in a privacy-preserving manner. Our method is based on a scalable outlier detection technique that uses attribute value frequencies. We provide an end-to-end privacy guarantee by using the differential privacy model and secure multiparty computation techniques. Experiments on real data show that our proposed technique is both effective and efficient.

Secure and Efficient k-NN Queries.

Given the morass of available data, ranking and best match queries are often used to find records of interest. As such, k-NN queries, which give the k closest matches to a query point, are of particular interest, and have many applications. We study this problem in the context of the financial sector, wherein an investment portfolio database is queried for matching portfolios. Given the sensitivity of the information involved, our key contribution is to develop a secure k-NN computation protocol that can enable the computation k-NN queries in a distributed multi-party environment while taking domain semantics into account. The experimental results show that the proposed protocols are extremely efficient.

Improving cancer patient emergency room utilization: A New Jersey state assessment.

INTRODUCTION: Due to its increasing incidence and its major contribution to healthcare costs, cancer is a major public health problem in the United States. The impact across different services is not well documented and utilization of emergency departments (ED) by cancer patients is not well characterized. The aim of our study was to identify factors that can be addressed to improve the appropriate delivery of quality cancer care thereby reducing ED utilization, decreasing hospitalizations and reducing the related healthcare costs. METHODS: The New Jersey State Inpatient and Emergency Department Databases were used to identify the primary outcome variables; patient disposition and readmission rates. The independent variables were demographics, payer and clinical characteristics. Multivariable unconditional logistic regression models using clinical and demographic data were used to predict hospital admission or emergency department return. RESULTS: A total of 37,080 emergency department visits were cancer related with the most common diagnosis attributed to lung cancer (30.0%) and the most common presentation was pain. The disposition of patients who visit the ED due to cancer related issues is significantly affected by the factors of race (African American OR=0.6, p value=0.02 and Hispanic OR=0.5, p value=0.02, respectively), age aged 65 to 75years (SNF/ICF OR 2.35, p value=0.00 and Home Healthcare Service OR 5.15, p value=0.01, respectively), number of diagnoses (OR 1.26, p value=0.00), insurance payer (SNF/ICF OR 2.2, p value=0.02 and Home Healthcare Services OR 2.85, p value=0.07, respectively) and type of cancer (breast OR 0.54, p value=0.01, prostate OR 0.56, p value=0.01, uterine OR 0.37, p value=0.02, and other OR 0.62, p value=0.05, respectively). In addition, comorbidities increased the likelihood of death, being transferred to SNF/ICF, or utilization of home healthcare services (OR 1.6, p value=0.00, OR 1.18, p value=0.00, and OR 1.16, p value=0.04, respectively). Readmission is significantly affected by race (American Americans OR 0.41, standard error 0.08, p value=0.001 and Hispanics OR 0.29, standard error 0.11, p value=0.01, respectively), income (Quartile 2 OR 0.98, standard error 0.14, p value 0.01, Quartile 3 OR 1.07, standard error 0.13, p value 0.01, and Quartile 4 OR 0.88, standard error 0.12, p value 0.01, respectively), and type of cancer (prostate OR 0.25, standard error 0.09, p value=0.001). CONCLUSION: Web based symptom questionnaires, patient navigators, end of life nursing and clinical cancer pathways can identify, guide and prompt early initiation of treat before progression of symptoms in cancer patients most likely to visit the ED. Thus, improving cancer patient satisfaction, outcomes and reduce health care costs.

Data science, learning, and applications to biomedical and health sciences.

The last decade has seen an unprecedented increase in the volume and variety of electronic data related to research and development, health records, and patient self-tracking, collectively referred to as Big Data. Properly harnessed, Big Data can provide insights and drive discovery that will accelerate biomedical advances, improve patient outcomes, and reduce costs. However, the considerable potential of Big Data remains unrealized owing to obstacles including a limited ability to standardize and consolidate data and challenges in sharing data, among a variety of sources, providers, and facilities. Here, we discuss some of these challenges and potential solutions, as well as initiatives that are already underway to take advantage of Big Data.

A Scalable Privacy-preserving Data Generation Methodology for Exploratory Analysis.

Big data coupled with precision medicine has the potential to significantly improve our understanding and treatment of complex disorders, such as cancer, diabetes, depression, etc. However, the essential problem is that data are stuck in silos, and it is difficult to precisely identify which data would be relevant and useful for any particular type of analysis. While the process to acquire and access biomedical data requires significant effort, in many cases the data may not provide much insight to the problem at hand. Therefore, there is a need to be able to measure the utility/relevance of additional datasets for a particular biomedical research task without direct access to the data. Towards this, in this paper, we develop a privacy-preserving approach to create synthetic data that can provide a firstorder approximation of utility. We evaluate the proposed approach with several biomedical datasets in the context of regression and classification tasks and discuss how it can be incorporated into existing data management systems such as REDCap.

African American Race is an Independent Risk Factor in Survival from Initially Diagnosed Localized Breast Cancer.

BACKGROUND: African American race negatively impacts survival from localized breast cancer but co-variable factors confound the impact. METHODS: Data sets were analyzed from the Surveillance, Epidemiology and End Results (SEER) directories from 1973 to 2011 consisting of patients with designated diagnosis of breast adenocarcinoma, race as White or Caucasian, Black or African American, Asian, American Indian or Alaskan Native, Native Hawaiian or Pacific Islander, age, stage I, II or III, grade 1, 2 or 3, estrogen receptor or progesterone receptor positive or negative, marital status as single, married, separated, divorced or widowed and laterality as right or left. The Cox Proportional Hazards Regression model was used to determine hazard ratios for survival. Chi square test was applied to determine the interdependence of variables found significant in the multivariable Cox Proportional Hazards Regression analysis. Cells with stratified data of patients with identical characteristics except African American or Caucasian race were compared. RESULTS: Age, stage, grade, ER and PR status and marital status significantly co-varied with race and with each other. Stratifications by single co-variables demonstrated worse hazard ratios for survival for African Americans. Stratification by three and four co-variables demonstrated worse hazard ratios for survival for African Americans in most subgroupings with sufficient numbers of values. Differences in some subgroupings containing poor prognostic co-variables did not reach significance, suggesting that race effects may be partly overcome by additional poor prognostic indicators. CONCLUSIONS: African American race is a poor prognostic indicator for survival from breast cancer independent of 6 associated co-variables with prognostic significance.

Privacy preserving integration of health care data.

For health care related research studies the medical records of patients may need to be retrieved from multiple sites with different regulations on the disclosure of health information. Given the sensitive nature of health care information, privacy is a major concern when patients' health care data is used for research purposes. In this paper, we propose an approach for integration and querying of health care data from multiple sources in a secure and privacy preserving manner.