Abstract 2341: Predictive analysis of gynecologic cancer risk factors using decision tree analysis
| Autor: | Christine Richardson, Larissa Brunner Huber, Wei Sha, Zahra Bahrani-Mostafavi |
|---|---|
| Rok vydání: | 2020 |
| Předmět: | |
| Zdroj: | Cancer Research. 80:2341-2341 |
| ISSN: | 1538-7445 0008-5472 |
| DOI: | 10.1158/1538-7445.am2020-2341 |
| Popis: | Introduction: Gynecologic cancer (GYNC) accounts for 6.3% of all cancers and is the fourth leading cause of cancer death for women in the U.S. Autoimmune diseases (AD) are among the top ten leading causes of death among U.S. women age 65 and under. This high mortality rate among women with AD has been linked to cancer and other causes. Considering cancer and AD are the accumulative effect of genetics and lifestyle, and 90-95% of all cancers are linked to lifestyle and environmental factors, a study to investigate the association between patient-level predictors and GYNC among AD patients in the U.S. was imperative. Methods: 2007-2013 data from Florida State Inpatient samples of the Healthcare Cost and Utilization Project were used. The study population (n=836,717) was restricted to women who had any AD. The outcome variable was diagnosis of GYNC. 36 categorical transformed variables including race/ethnicity, age, insurance type, income level, GYN procedures, and comorbidities were analyzed as predictive variables (PV). Bootstrap Forest (JMP pro13) were was used to create a 10,000-decision tree (DT) model to identify independent predictors of GYNC. Multiple outputs such as confusion matrix, column contribution, and ROC was created for predictive analytics. Results: The analytical model used was confirmed by evaluating criteria such as AUC (0.82), sensitivity (0.68), specificity (0.82), false positive rate (0.18), wrongly predicted value (0.24), and overall accurate prediction (0.76) of the model. Using column contribution and single DT, the PV hysterectomy was the highest predictor of GYNC, followed by PV age;, and comorbidities such as diabetes, and obesity; along with race/ethnicity; and income level. Also, subpopulations of AD patients at risk for GYNC based on unique combinations of risk factors were established, including subpopulations such as women with AD, involved glandular disorders who had hysterectomy, and comorbidities such as diabetes and fluid and electrolyte disorders. Discussion: Both GYNC and AD are major chronic diseases among women. Because of high prevalence of AD among women, more elucidation on the association between GYNC and AD is essential for GYNC management. The unique combinations of characteristics that describe subgroups of patients at risk for GYNC among AD patients can be used as a potential risk assessment, as well as an early detection and/ or prevention tools for GYNC. Citation Format: Zahra Bahrani-Mostafavi, Larissa B. Huber, Wei Sha, Christine Richardson. Predictive analysis of gynecologic cancer risk factors using decision tree analysis [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 2341. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|