Eric Xu sx2402@cumc.columbia.edu
Alexandria Taylor at3583@cumc.columbia.edu
Kevin Guzman kg3006@cumc.columbia.edu
Edmund Wong ecw2178@cumc.columbia.edu
Jiseung Lee jl6458@cumc.columbia.edu
This project aims to investigate survival outcomes in a pressing women’s health issue, emphasizing the demographic and biological factors that shape disease prognosis and health equity.
Primary Topic: Breast Cancer Survival and Risk
Patterns
Breast cancer remains one of the most prevalent disease worldwide,
accounting for a substantial proportion of cancer incidence and health
burden. Despite advances in early detection and treatment, disparities
in outcomes is often observed across demographic and biological groups.
Clinical outcomes are influenced by multiple factors, including tumor
size, stage, lymph node involvement, and hormone receptor status.
Understanding how patient characteristics, tumor biology, and receptor profiles influence survival and risk is essential for improving risk stratification and guiding treatment decisions. This project aims to explore the relationship between the following factors predicting survival outcomes among patients with breast cancer.
This project starts with a public, de-identified dataset derived from the 2017 SEER (Surveillance, Epidemiology, and End Results) Program, containing information from 4,024 breast cancer cases with 16 demographic, clinical, and molecular variables. The data include survival time, tumor stage and grade, hormone receptor status, lymph node involvement, and other key prognostic factors.
In further investigation, usage of this dataset or a full SEER dataset with more detailed covariates would be considered. The main goal is to explore survival differences by demographic group, tumor characteristics, and receptor status, as well as broader examination of disease burden and disparities in breast cancer outcomes.
The project may also incorporate publicly available SEER incidence data (1975–2022) to contextualize survival findings within trends in breast cancer occurrence and population-level risk patterns over time.
If SEER data set is not applicable or not informative, an alternative plan would be applied.
Alternative Topic: Infant and Maternal Mortality in the
United States
As an alternative or complementary direction, this project may
investigate patterns and disparities in infant and maternal mortality
using national data from the CDC/NCHS Birth–Infant Death and Maternal
Mortality datasets. These datasets provide detailed demographic and
health information on mothers and infants across the United States,
offering a rich foundation for understanding the social, economic, and
medical determinants of survival in early life and maternal health.
This topic is of major public health importance, as both infant and maternal mortality reflect the overall effectiveness of healthcare systems, access to care, and equity in health outcomes. Analyses could focus on identifying demographic patterns, geographic disparities, and population subgroups at elevated risk, contributing to ongoing national efforts to reduce preventable deaths and improve maternal–child health outcomes.
Project Vision
Both directions align closely with the goals of applied epidemiologic and data science research. The overarching aim is to use population data to describe health disparities across female health burden and systematic determinants. With this project, we aim to identify key determinants of disease outcomes, and highlight areas for intervention. By integrating data management, exploratory insight, and public health relevance, this project will contribute to understanding critical patterns in population health and inform future work in epidemiology and health equity.
Our team will produce a fully reproducible, Github-hosted project that identify predictors of survival in breast cancer. Deliverables will include.
Final Written Report: A comprehensive report detailing workflow following the required structure (Motivation, Related Work, Data, Exploratory Analysis, Additional Analysis, Discussion).
Interactive Project Website: Built using R Markdown, hosted via GitHub Pages, and design to communicate results to a broad audience.
Two-minute Video Summary: A concise video presentation summarizing key findings, methodology, and public health implications of the project that is embedded into the project webpage.
GitHub Repository: A well-organized repository containing all code, data processing scripts, and documentation to ensure full reproducibility of the analysis.
Descriptive statistic and distribution of demographics and clincial variables
Correlation and association test between tumor characteristics/lymph node characteristics/hormone receptor status and survival outcomes
Kaplan-Meier survival analysis
Cox proportional hazards regression
Kaplan-Meier survival curves stratified by key variables (e.g., tumor stage, hormone receptor status)
Boxplots of tumor size by stage
Heatmap of variables correlation
Data cleaning and preprocessing of SEER dataset (if able to be)
Cleaning categorial variables in publicly available data set
Reformating data for survival analysis
Managing missing or inconsistent data enteries.
| Week / Date | Milestone | Key Tasks |
|---|---|---|
| Week 1 (Nov 11–17) | Data Cleaning & Setup | Import data, recode variables, handle missing values,
set up GitHub repo & R/Quarto structure. |
| Week 2 (Nov 18–24) | Exploratory Analysis & Visualization | Descriptives & plots; correlations; refine
questions; draft report sections. |
| Week 3 (Nov 25–Dec 1) | Modeling & Interpretation | Kaplan–Meier & Cox models; test associations;
interpret results; finalize figures & tables. |
| Week 4 (Dec 2–Dec 6) | Final Deliverables & Submission | Finish report & website; record 2-min screencast;
polish repo; submit by Dec 6 (11:59 PM). |