Data Science and Healthcare Analytics
Introduction
I will graduate with a Master of Data Science in 2023 and am confidently looking to combine data analysis with my long history of experience in healthcare surrounding my Doctor of Pharmacy. I am most experienced with SQL, Python, Excel, and R with additional knowledge using SAS, Git, and PowerBI. I also enjoy consultant-based problem solving and how it can influence better use of data to guide business decisions.
ABOUT ME
I am lucky to have found a career that allows for constant growth and creativity. I love to learn! Whether I am working on projects or independently adding a new skill or hobby, I am constantly showing interest in new ideas. This not only includes working on a computer, but also cooking, playing musical instruments, or relaxing with an open book. Whether for leisure or at work, problem solving is one of my favorite things to do!
With great pleasure I can say that I have worked with some of the most intelligent professionals in the world of data science and healthcare around the Saint Louis area. I was on many teams as a student pharmacist in several hospitals and pharmacies before graduating with a Doctor of Pharmacy in 2017 from the University of Health Sciences and Pharmacy. Now, I am finishing a Master of Data Science from Maryville University of Saint Louis. There is no better way to gain experience and truly assess your own knowledge than working in the field you are studying. Fortunately, I have been part of an Advanced Analytics team with Midwest Employers Casualty Company where I have experienced tremendous growth. The enthusiasm that is present at work drives my independent study of data science and business analytics. My home learning to-do list is always getting longer, but I only continue to enjoy being part of this field.
January 2022 - Present
Data Scientist Intern, Midwest Employers Casualty Company
As a Data Science Intern on the Predictive Analytics team, I complete projects in an Agile workflow. A typical day includes using SQL, Python, Excel, and Word to problem solve and learn new tactics constantly. The team communicates using MS Teams and Webex, along with logging work in Confluence with Jira tickets. Additional use of PowerPoint, PowerBI, Git, and VSCode are utilized intermittently as needed. I work alongside data science graduates of Washington University in St. Louis and Maryville University of St Louis. There are currently 3 adjunct professors for Maryville University on my team who teach courses in data science and economics.
May 2010 - January 2021
Student Pharmacist, CVS/ Walgreens/ OptumRx - Adecco
I performed every pharmacy task including counseling patients, contacting patients' medical providers, inventory management, training technicians, and resolving escalations.
General Locations:
-
Saint Louis, MO
-
Bourbonnais, IL
-
Remote (Downer's Grove, IL)
Master of Data Science,
Maryville University of Saint Louis
May 2021 - August 2023
Maryville University of Saint Louis
Degree: Master of Science
Major: Data Science
Departments: Mathematics
GPA: 3.7
Data Science Courses Taken:
-
R Programming
-
Predictive Modeling
-
Python
-
Machine Learning
-
SQL
-
Text Mining
-
SAS Programming
-
Experimental Design
-
Data Visualization
-
Statistical Modeling
Post Bachelors Certificate in Machine Learning,
Maryville University of Saint Louis
May 2021 - January 2022
-
In addition to the Master's Degree program, I received this Certificate in Machine Learning while completing associated courses in machine learning and predictive modeling.
Doctor of Pharmacy (PharmD), Saint Louis College of Pharmacy
A 6 year program (0-6) covering all required education and training for a pharmacist to professionally practice.
August 2011 - June 2017
Some courses taken of the 208 credit hours completed during this program:
-
Therapeutics I-IV
-
Evidence-Based Medicine
-
Pharmacokinetics
-
Clinical Prevention and Population Health
-
Pharmacology I-II
-
Clinical Epidemiology
-
Health Systems Management and Economics
-
Pharmacy Law and Ethics
-
Biostatistics
-
Physics
-
Calculus
-
Top Pharm Care: Community Service
Certificates and Specializations (Paid Courses)
Management Consulting Specialization
Completed January 2023
Emory University
Course Instructor: John Kim
A non-credit Specialization that included the following 5 courses: Introduction to Management Consulting, Getting a Consulting Job, Consulting Approach to Problem Solving, Consulting Tools and Tips, Consulting Presentations and Storytelling
Pharmacy Internships
Advanced Pharmacy Practice Experiences
Barnes Jewish Hospital Pharmacy Medical Intensive Care Unit
Saint Louis, Missouri
Paul Juang, Pharm.D., BCPS
Elective - Pt. Care Elective
Family Care Health Center Pharmacy
Saint Louis, Missouri
Barry Wilson, R.Ph.
Ambulatory Care
Veterans Affairs Medical Center (JB)
Jefferson Barracks
Saint Louis, Missouri
Sarah Kay, Pharm.D., BCPS
Community Care
March 27, 2017 - April 28, 2017
February 13, 2017 - March 17, 2017
January 09, 2017 - February 10, 2017
Patient Care Selective
Shop ’n Save #4533
Maplewood, Missouri
Kimberly Lorenz, Pharm.D.
November 14, 2016 - December 16, 2016
Elective - Pt. Care Elective
Schnuck’s Specialty #361
Transplant Patient Care
at Various Saint Louis Hospitals
Saint Louis, Missouri
Katie Akins, R.Ph.
Health System management
Mercy Hospital Jefferson
Crystal City, Missouri
Jamie Glore-McCarthy, Pharm.D., BCPS
October 10, 2016 - November 11, 2016
September 06, 2016 - October 07, 2016
Acute Care - General Medicine
Department of Veterans Affairs Medical Center
John J. Cochran Veterans Hospital
Saint Louis, Missouri
Besu Teshome, Pharm.D., MSPS, BCPS
July 25, 2016 - August 26, 2016
Introductory Pharmacy Practice Experiences
Interprofessional Patient Care
Saint Louis University
Saint Louis, Missouri
Pat Rafferty, Pharm.D., BCPS, CDE
Safe Medication Practices: Hospital
Kindred Hospital Saint Louis at Mercy
Saint Louis, Missouri
Natalie Vasant, Pharm.D.
Clinical Prevention and Population Health
Cardinal Ritter Adult Daycare
Saint Louis, Missouri
Amy Tiemeier, Pharm.D., BCPS
Safe Medication Practices: Community Care Pharmacy
Schnuck's #258
Saint Louis, Missouri
Kenneth Simpson, R.Ph
Foundations of Pharmacy Practice
Saint Louis College of Pharmacy
Saint Louis, Missouri
Nicole Gattas, Pharm.D., BCPS
August 2015 - May 2016
December 2015 - January 2016
August 2014 - May 2015
June 2014 - July 2014
January 2014 - May 2014
PROFESSIONAL HISTORY
ACADEMIC EXPERIENCE
University Graduate Programs
SKILLS
Technical Skills
Python
-
Professional: I frequently work with Jupyter notebooks and other Python files primarily in Visual Studio Code and occasionally in JupyterLab. This includes manipulating datasets, writing SQL queries to get data directly in Python, creating predictive models, analyzing existing or new models, QA/UA testing, data visualization, and and sharing work in collaborative projects.
-
Academic: I used Jupyter notebooks in the following courses: Python, Machine Learning, Text Mining, Data Visualization, Statistical Modeling.
-
I have been using and learning python for 2 years.
SQL
-
Professional: I use SQL every day at work. This includes gathering data, researching our data for specific requests or projects, creating new tables for repeat use or sharing, connecting to reporting tools in Excel and PowerBI, and code review.
-
Academic: I used SQL in the following courses: SQL, SAS
-
I have been using and learning SQL for 2 years.
R
-
Professional: I do not use R at work except when viewing previously built models that were written in R. I have done very small tasks in R Studio at work just for continued use of R while working with data.
-
Academic: I used R Studio in the following courses: R Programming, Predictive Modeling, Machine Learning, Experimental Design, Data Visualization, Statistical Modeling.
-
I have been using and learning R for 2 years.
Excel
-
Professional: Excel is used every day at work as well. Several documents are preferred to be created in Excel, especially when working interdepartmentally. Most business professionals are used to viewing data here.
-
Academic: In my data science program, Excel is only used for providing datasets, exporting data from python, R, or SAS to csv files when requested, and infrequently used for mathematical modeling of financial data. In pharmacy school, I would use Excel for data visualization, especially for regression analysis of a drug's clearance over time.
-
I have been using and learning excel for 12 years.
-
While Excel is a much more common skill, I will note that some uncommon formulas/functions I use often for work include VLOOKUP(), CONCATENATE(), EDATE(), IF(), COUNTIFS(), Macros, as well as Pivot Tables for analysis. Essentially, these are aggregations I do in SQL that I just need to find how to accomplish in Excel.
SAS
-
Professional: SAS is not used in my current workplace and few members of the team have experience using it.
-
Academic: I took a semester of SAS programming because of my interest in healthcare analytics and a handful of hospitals in the area want candidates that are familiar with SAS. In SAS, I used common procedures to import data, create tables and writing queries like SQL, create models and visualize data.
Project Management
Version Control
-
We typically use Git (Bitbucket / Stash) at work to manage collaboration on data science projects, share work, merge code changes, and manage read and write access.
-
Generally, the data scientists communicate updates on their work in Microsoft Teams to provide awareness of any relevant changes or completion of work.
Quality Assurance and User Acceptance Testing
-
QA / UA testing is performed anytime there is a change in code or new code being implemented into production.
-
Examples include when a SQL query was adjusted to create a monthly report in Excel or a new Python script was written to retrieve data for the new models.
-
While these tests are usually not time consuming on their own, I often take this time as an intern to read more of these scripts to better understand what my teammates are accomplishing overall.
Agile Workflow
-
Confluence is used for several workflow processes. We use this for OKRs (team strategy) and KPIs (progress measurements) throughout the week and refer to it during weekly meetings. We also store work references, final project documents, and training material on confluence.
-
Jira is used currently to track work with Jira tickets and visualize workflow on a virtual Kanban board.
-
Webex and Microsoft Teams are used for weekly team meetings where we provide updates on our projects.
Sharing Insights
Data Visualization
-
PowerBI
-
PowerBI is a wonderful tool for creating interactive presentations of data. I have been part of 3 major projects that utilized PowerBI and they have all been collaborative projects. I am continuously learning more ways to use PowerBI as it is commonly mentioned in data science roles.
-
-
Tableau
-
I have only used Tableau academically. It is great for building out reports and data storytelling. Currently, I am working on two dashboards for major projects in my Data Visualization course. One is for displaying business sales data with visuals for region, year, and specific product categories. The other is regarding healthcare data and patient treatment outcomes.
-
-
PowerPoint
-
The majority of my academic career has involved formal presentations using PowerPoint. While everyone is familiar with this tool, I still actively learn new techniques for presenting with slideshows. Most recently, I have increased my focus on constructing a narrative, improving structure, and utilizing an appendix section at the end of the slides. This helps fight the urge to show all the work I did for the project and focus on the main message I am trying to convey.
-
-
R Shiny
-
Shiny is also something used academically and independently. It requires writing in R, but comes with personalized visuals, data manipulation, and can utilize R's machine learning capabilities.
-
-
Additionally, visuals are created using Python, R, and Excel. Examples of projects that were formally presented are included below in the "Projects" section.
Data Storytelling
-
People like narratives and I have learned a lot about how to craft a data-driven presentation to be more engaging. Through books on data storytelling, presentations by Hans Rosling, and especially the Management Consulting Specialization I took, I have found a lot of useful information that I have and will continue to try implementing.
-
One way I have created a narrative for a major project was related current events to assumptions made by the executive board. This example involved creating a reasonable story applying the "yes-loading" strategy used in sales and leadership roles. My goal was to get the audience to agree on some pretty undeniable crime-related events and then prove to them why the assumptions that they are affecting our business are wrong.
-
While narratives are engaging and fun, I make use of the Pyramid Principle to lead with the answer or argument I am making. This helps ensure the main point I want the audience to take away is clear and the time spent supporting insights can be adapted for time and respect the busy schedules of our business leaders.
Communication
-
Not only is communicating one of my greatest strengths, but it is has been my favorite responsibility in my professional history.
-
Leadership and effective communication were tremendously utilized in the following scenarios:
-
Student Pharmacist educating patients on their medications and answering questions.
-
Conflict resolution and de-escalating problems that occurred in the pharmacy, regardless of who caused them or if the workplace even had control over the matter. I was great at explaining the problem (taking responsibility) and providing the steps on how we plan to correct this issue now and prevent it from happening again.
-
Recognizing my target audience and catering the presentation of information to them.
-
Serving as president of a professional pharmacy organization.
-
Speaking about predictive analytics projects to other departments in my current office.
-
Communicating to people of various ages, ethnicities, and backgrounds in several healthcare environments.
-
Data Analysis
-
Analysis is a broad description of forming ways to understand data.
-
Pivoting information in excel to view counts and distributions is quick and helpful.
-
Visualizing data for analysis often requires reverse-engineering the problem to identify what we are trying to see. When that is decided, we can easily select which graph is most appropriate to use, such as a Line, Bar, Scatter, histogram, boxplot, etc.
-
-
Data architecture is managed by another team, but I am responsible for understanding what data I have access to and reporting issues that inhibit data science projects from being completed due to locations of data.
-
Data governance is an ongoing inter-departmental project at my current workplace and I have taken part in collecting documentation and reporting on the various levels of granularity within the company. This helped leadership decide of the best standards to implement across all departments and delegate responsibility.
-
Data preparation and cleaning often consumes a major portion of time involved in a project but is undeniably required before any reliable analysis is conducted.
Projects
Examples of projects below are intentionally vague for non-disclosure of company strategy*
Major Projects
Problem
The Named Entity Recognition Model that uses natural language processing to identify drugs from our data needed evaluated on performance to determine how effective the model is doing with new data. The current model's accuracy had been questioned and the team is wondering if a revision is necessary.
Action
A Model Evaluation
I used SQL to identify samples of data from every client whose data is seen by this NER model. I manually assessed the text and compared it to the accuracy of the model using Excel. When completed, I created a PowerPoint and shared the results.
Results
I presented my PowerPoint to the Director of Data Science, Director of Data Strategy, and Data Scientists. The results were accepted as useful for the team and the decision was made to continue using the current model based on its remarkable accuracy of 96.7% nearly two years after implementation. Additionally, small token errors identified by my analysis were addressed and resolved, improving the secondary function of the model from 80% up to 86% accuracy.
Model Revision and Feature Analysis
A Random Forest Model that performs binary classification on future cost predictions is a major priority for the company. Leading up to a large revision process, the VP of Predictive Analytics wanted me to examine the false negatives and identify patterns in the results with false predictions. Another goal of this project was to find which features are most associated with the errors that have occured.
I created SQL queries to identify the targeted instances of false negatives and false positives, searched extensively through many observations, and created tables of the observations with suspected causal features in a designated database. The project was then escalated, and a data scientist was paired with me to create a model to better identify patterns in these observations. We each created classification models in Python (Jupyter Notebooks) and used SHAP and LIME to explain feature importance.
This project was presented to the VP of Predictive Analytics, Director of Data Science, and Director of Data Strategy. In many projects, large amounts of work get stored in an appendix, but here it was requested to be shown extensively in our final Python notebook. All the areas we explored were discussed and shown visually with graphs and code supporting our logic. The end results ended up saving weeks of work for the team when they eventually started feature analysis for model revisions with a head start on which areas would be adjusted.
Creating Predictive Models
As an intern, I am not individually creating models getting deployed into production. However, I was able create models to provide insight on new methods. Three features related to injury details have been scored on their association with costs. The Director of Data Science needed to know how powerful these new scores are in predicting correct outcomes.
In a python notebook, I created two predictive models using XGBoost. Each model had specific dollar thresholds and the classification was based on those amounts being surpassed or not. I created the subset of data used in SQL and pulled it directly from the server into the Python so the model was trained on the most current data. The evaluation of the two models were done in the same notebook covering various metrics of model performance.
The model results were presented to the Director of Data Science and Director of Data Strategy. My explanation informed them of how effective these three features are on their own. The ROC curve indicated 83% accuracy which is not enough to use just these features alone. The final outcome was that these features with the new scoring system were beneficial and would be a new addition to the future models that are now in effect.
Market Research and Benchmarking
Members of the Executive Board have observed articles regarding analytics used in the Insurtech space. Some have even been from our competitors. The board wanted to know where our company ranks among our competition and what analytical services we might be able to offer to our clients to improve their business outcomes.
I researched 8 competitors, 12 brokers, 20 vendors and insurtech companies, and 6 pharmacy benefit managers. I collected data in an Excel table by putting various services provided into "buckets" and noting which companies did or did not use that application of analytics. I put together screenshots, statistics, graphs, and wrote summaries on each company researched. The final product created was a detailed report in a Word document that grouped companies by their services and what unique offerings they have. Additionally, I compared our business to our competitors to show management where we stand in the market. I summarized what our analytics team was already accomplishing and other ideas that we could add for clients.
This report was given to the VP of Predictive Analytics who shared it with other members of the board. Our results found that we are highly competitive in using analytics with our clients' data to monitor and predict high-cost outcomes. This research also influenced leadership to start a new project to make customer acquisition a faster and easier process, which I was given the opportunity to begin. This is in collaboration with a company found from this research who specializes in natural language processing to automate applicant intake forms.
Formal Research Paper and Data Visualization
The executive board noted a higher prevalence in gun violence and questioned if this is affecting our business. If this can be proven true, we should consider adjusting policy premiums to reflect the growing risk of dangerous encounters involving a large portion of our clients.
The Data Operations &Reporting Analyst and I rolled out a major project with the primary hypothesis "gun violence has had a stastistically significant increase in our business". This involved internal research of our own data and national data (LEOKA) collected from 2010-2021. SQL was used to query data, fine-tune WHERE statements to reduce false positives, and create a database for the imported national data. PowerBI was used to connect to a SQL server and create visual comparisons. We created 27 graphs in PowerBI, comparing different featurs of the data such as aggregating observations by state, normalizing the data per capita, and comparing total deaths in our business to national data in a time series. I wrote the final version of the 8-page research paper and we decided to include 9 of the 27 PowerBI visuals that best told the trends we discovered, although including the full report in a stored appendix.
The finalized product was given to the VP of Predictive Analytics to share with the executive board. Our insights showed that overall gun violence has increased in the national database, but it has slightly decreased within our own business. The board decided to not pursue this issue further after we discovered it was not causing concern. Our work was saved for future analysis so that this could be monitored again in the future. We addressed this "control" phase of the project so internal queries linked to PowerBI will only need refreshed. National data will require data to be imported and decoded again, but those steps were well documented for replication.
Medical Research for Impact on Business
Research was required to investigate musculoskeletal surgeries and analyze outcomes related to treatment plans. I also was looking to identify trends in the course of injuries that lead to surgery. The purpose of this research was to look for similar trends in our business and find ways of predicting costly medical events.
Using research experience from pharmacy, I conducted searches on PubMed (U.S. National Institutes of Health's National Library of Medicine) to find literature on relevant surgeries. I found massive retrospective studies that included over 500,000 surgeries our Data Science team was interested in. I gathered statistics, summarized findings, and created a PowerPoint to share insights. This presentation included the natural progression of an injury that eventually requires surgical treatment.
I gave this presentation to the Predictive Analytics team. The medial topics were high-level to respect the audience and only dive into subjects that were really associated with our business. The results included substantial statistics on how long we should expect surgeries to last before requiring follow-up care. I also discussed significantly proven situations where workers' compensation negatively impacted recovery for reasons like avoiding returning to work and delayed scheduling of medical appointments.
Creating a Standard Operating Procedure
One of my ongoing responsibilities is to create 4 monthly reports for external departments and clients with SSMS, SSRS, and Excel. While the intended format and server connections have been created, the steps to make these reports, monitor for data issues, and adjust formatting issues were not clear. There was no document on steps to take and queries required to retrieve data were stored in displaced folders of the company's file explorer. This consumed a lot of time and the entire process was confusing.
I decided that a guide should be created and utilized some time to write detailed instructions in a word document. I made the steps to be so easy to follow that any future new hire could use this document to create the reports without having difficulty finding which query or server should be used in each scenario. I included screenshots of virtually every mouse click so users would see exactly the same steps as they complete them. I have written notes on how to verify if the results are indicating errors in the data we received. The end includes links to locations where the files should be saved and example emails of my team informing specific people that the reports were complete.
I am proud to say that this document now exists and is updated any time there is a change in the reporting process. I open this document for every report because everything is saved in one place, and I can copy and paste links and queries quickly. It also serves as a guide for some of the most complicated reports. I am confident that whoever takes over this responsibility in the future will get a lot of benefit from this document, even if they are in an early stage of their analytics career.
Albert Einstein
"If you cannot explain it simply, you don't understand it well enough"