Graduate Education

My drive to help others and insatiable curiosity about the world have fueled a life-long learning journey.

Thesis Details

Project Summary:  Using Machine Learning to Improve Care for Homeless Patients

Code Highlights: 

Accomplishments:

• Established a new academic / institutional partnership between the University of New Hampshire Department of Health Management & Policy and Catholic Medical Center.

• Designed and conducted a retrospective real world evidence (RWE) study using electronic health record data (EMR), claims data and Collective Medical portal data. 

• Study protocol written and approved in Spring 2020, and Institutional Review Board (IRB) approval received in Fall 2021. 

• Collaborated with Health Care for the Homeless' (HCH) leadership to identify and outline the research's business case and goals. 

• Created analysis plan in collaboration with thesis committee.

• Coded, evaluated and presented a biostatistics and machine learning analysis capable of identifying and describing patient service groups within the larger clinic population. 

• Successfully presented the results to key leaders and stakeholders at Catholic Medical Center and Health Care for the Homeless. 

• Successfully defended my approach, methods and results to thesis committee members and UNH Health Data Science faculty.

This in-depth, collaborative analysis helped the Health Care for the Homeless team by providing actionable insights capable of improving tailored care for homeless patients. 

Related Positions

Evaluation Analyst, Elson S. Floyd College of Medicine

Washington State University, Spokane WA (Remote)

Internship under Dr. Erin Griffin, Director of Evaluation & Scholarly Associate Professor 

Jul 2021 - Mar 2022 

Mentored analysts in Evaluation and Assessment departments in the R programming language and basic statistical methods

Designed and created an 8 module course with videos, presentations and R notebook assignments to get analysts started with R programming, help them understand the programmer's context - including computing, the internet and the cloud, and teach them the basics of nonparametric statistical computing using the 14th most popular scripting language in the world

Summary and References:  Teaching R to Analysts

Course Materials:  GitHub repository, R-cs-course

Created, deployed and analyzed qualitative / Likert instruments in E-Flo for measurement of medical student's self-assessment and program assessment under the direction and guidance of the Director of Evaluation

Wrote SQL extraction jobs to create daily .csv extracts from E-Flo system

Wrote python code to:

• Scrape data from the web-based curriculum catalog

• Automatically generate graphics for inclusion in compliance reports

• Mine comments collected in qualitative instruments to perform network analysis, topic modeling and sentiment analysis with Latent Dirichlet allocation (LDA) and the Natural Language Toolkit (NLTK)

• Automate the cleaning, merging, manipulation and summary of results collected from surveys within the E-Flo and Qualtrics systems

Under the direction and guidance of the Director of Evaluation, authored detailed course review reports for both compliance / administration and course designers and instructors, including succinct summary sections synthesizing key analysis findings.

Teaching Assistant, Health Data Systems

Heath Data Science (HDS) 804

Professor: John McInally MBA

Installed and configured Apache cTAKES (clinical Text Analysis Knowledge Extraction System) on both Ubuntu and Windows platforms.

Set up and configured API access to the National Institute of Health's Unified Medical Language System (UMLS), a compendium of many controlled vocabularies in the biomedical sciences, to extract the ICD-10 vocabulary, and create, configure, and install a cTAKES dictionary able to derive meaning from extracted ICD-10 codes.

Contributed code improvements to an existing python library on GitHub for creating and processing  pandas Data Frames mined from cTAKES output.

Developed an instructional video and guide for students using the cTAKES and ICD-10 dictionary tools, along with the python ingestion library, to set up a data pipeline for natural language processing (NLP) of clinical text, including admission notes and discharge summaries.

Created installation procedure documentation for UNH's IT group to easily install Data Science Lab instance of cTAKES.

Capstone Details

Analysis Partner: Dhara Kapoor, MSPT

Code Highlights: GitHub repository, medicare-analysis (currently private, please request access)

Team Accomplishments:

• Analyzed Medicare data, including beneficiary demographics, chronic conditions and inpatient and outpatient claims data including costs and ICD-9 codes

• Produced Elixhauser scores for each visit, using CMS SAS macro

• Performed extensive feature development on the data set with python pandas, using all the data to produce detailed aggregate profiles of the health of 2,909 out of 3,143 U.S. counties (92.6%), and added in the RWJF County Health Rankings data as additional measures of county-level health factors

• Examined U.S. counties with no Medicare beneficiaries, discovering that in many of these counties average life expectancy is well below the age of eligibility

• Produced a risk adjustment Gamma regression model against the outcome of cost per beneficiary, and identified 1,109 counties with higher than expected costs using SAS GLM

• Performed a detailed factor analysis in python using factor_analyzer on 1,042 of the counties identified as having higher than expected costs, and discovered three factors accounting for 52% of this higher than expected cost: 

  • A hospitalization factor, accounting for the largest amount of cost variance (58%)

  • An illness / primary care factor, and

  • A high utilization factor

We produced a wholly innovative analysis of Medicare data, and were able to place our discoveries within the larger context of important previous U. S. health care cost analyses, including the Dartmouth Atlas Project and Steven Brill's The Bitter Pill.

Related Positions

Instructor, Using Statistics in Medicine

Fall 2023- Spring 2024 Course Development; Spring 2024 Course Debut

Program: UC Davis CPE Health Professions Post-Baccalaureate

Developing a new online course in Health Care Statistics for UC Davis CPE's Health Professions Post-Baccalaureate program, focusing on how health care professionals can understand, apply, and constructively critique methods commonly used in clinical and health outcomes research.

Developing curriculum and lessons using Canvas in partnership with UC Davis CPE program directors and instructional designers, to include a variety of learning modalities including:

Videos and lectures

Discussion assignments

Interpretation assignments

Self-grading quizzes

Google Colab Notebooks with hands-on python examples, including:

• Bayes Theorem, applied to a healthcare example

• Random Walks and the Central Limit Theorem

• The Age Distribution of Cancer Incidence (Distribution Testing)

• Descriptive Statistics & Measures of Central Tendency with the Cleveland Heart Attacks data set

• T-Testing & ANOVA with the Cleveland Heart Attacks data set

• Hypothesis Testing and Confidence Intervals

• Correlation and Linear Regression

• Multivariate Analysis - Logistic, Poisson Regression, MANOVA

• Survival Analysis using the Framingham Data

• Categorical Techniques: Chi-Squared, CMH, Correspondence Analysis

Instructor, Quantitative Methods & Decision Analysis

Program: UC Davis CPE Healthcare Analytics

Spring 2023 and Fall 2023 (Present)

Continued to perform all duties performed as TA in Spring 2022 (see below)

Created review videos teaching on topics of Regularization, using LASSO for variable selection, Logistic Regression, risk and odds ratio calculation and interpretation, and deriving model metrics from the Confusion Matrix

Created lesson on Algorithmic Bias to provide students with current information on this important topic, including insight into reasons for ongoing bias and possible ways to mitigate bias in both models and data presentations

Summary & References: About Algorithmic Bias

Teaching Assistant, Quantitative Methods & Decision Analysis

Program: UC Davis CPE Healthcare Analytics

Professor: Dr. Brian Paciotti

Spring 2022

Mentored students in SAS programming and statistical methods via weekly drop-in Zoom tutoring sessions

Coached students during Capstone analysis planning and communication of their statistical model interpretations

Assisted students with cohort creation planning and provided SAS code as needed

Provided example code in SAS for creating training, validation and testing sets and performing model evaluations with confusion matrices

Provided example code in SAS performing descriptive and predictive modeling using a variety of statistical methods including CMH and chi-squared testing, linear and logistic regression, regularization and survival analysis

Teaching Assistant, Applied Biostatistics

Program: UC Davis CPE Healthcare Analytics 

Professor: Dr. Erin Griffin

July 2020 - Dec 2020

Mentored students in SAS programming and statistical methods via weekly drop-in Zoom tutoring sessions. 

Improved assignments and added additional SAS code examples as needed. 

Provided feedback for students on how to communicate statistical and model interpretations and improve precision in analytics writing.