Data Science Portfolio

Programming

Daniel Linnen Data Science Portfolio

github: data2vizdom

2021

  • Web scraping/parsing project: JSON to python; show feature correlations; build model of selected feature variables correlation with outcome y
  • Data science project: data wrangling; build lists/dictionaries of user data;  describe and filter data, plot temporal results and scatter plot with subcategories

Screen Shot 2021-09-15 at 10.28.20 AM

2020

  • Kaiser Permanente COVID-19 Patient and Employee exposure surveillance analytics
  • Magnet Readiness quality performance across quality domains, patient care experience, and readiness assessment (21 NCAL medical centers)
  • Epic Acuity and Staffing (4 analyses/21 medical centers using Tableau)
  • Epic Acuity Enterprise dashboard: 21 medical centers (Tableau)

2019

KP Career Progression Program Benefits/Business Case (Regional HR dataset with >20,000 unique IDs): Identified $xM in program benefits

  1. Reduced turnover of program participants
  2. Improved internal hiring and promotion
  3. Impact of program participation on hospital staff nurse turnover

AHRQ Predictive Analytics Challenge: County-level population characteristics on hospital utilization (US dataset; >2,000 hospitals); R and Python; team lead

AHRQ DS challenge PPL Project Summary

ahrq density plots


2018

Synch or Swim Study, Kaiser Permanente (2018)

Which hospital characteristics reduce unexplained variation in pressure injury risk?

Stata code: CNR code Linnen

Interregional research study at Kaiser Permanent Division of Research: Hospital-acquired pressure injury risk across 35 hospitals and 750,000 patient hospitalizations in CA; identified new hospital-level risk actors and invalidated previous held assumptions – WIN Best Paper Award

cnr_figure

PhD Dissertation at the University of California San Francisco (UCSF): Alert-Driven Rescue

UCSF Distinguished Dissertation Award

  1. Published world’s first systematic review of AI in prediction of clinical deterioration or ICU transfer
  2. Chart review study; SQL and STATA
  3. Matched-pair cohort study of alert-driven interventions on sepsis survival –  SQL and STATA

Do sepsis interventions, relative to an advanced early warning system alert, improve patient survival between decedents and comparable (matched) survivors?

Stata code: adr_code

adr matching

Rural Inpatient Mortality Study, UCSF Does hospital rural status predict hospital mortality?

Stata code: cmsrural_stata

RIMS CA outcomes map1


RIM Study – Machine Learning extension, UCSF (2017)

Does hospital rural status predict hospital mortality?

R code: cmsrural

ruralcms_auc


Data Visualization

Tableau-R-h2o interface, Kaiser Permanente – Proof of Concept (2017)

How does the technical interface between Tableau, R, and h2o work, and do benefits outweigh the negatives?

Project lead working with two data scientists to explore the interface between Tableau, R, and h2o. exemplar by @aust10_powers

tableau_r_h2o


Mobile weight loss program evaluation, Kaiser Permanente (2017)

Do class attendance and interacting with a mobile weight loss management app using gamification reduce weight?

Analyzed user app behavior and developed bivariate correlation matrices and summary statistics of weight loss and class attendance metrics (proprietary)


Mobile Health App Rater, Kaiser Permanente (2017)

Health apps prescription analysis in primary care; multiple analyses (proprietary)

app_rater


Enterprise-Level Innovation Dashboard, Kaiser Permanente (2016)

National and regional knowledgebase of innovation projects, technology categories, KPIs, and funding (4 interactive dashboard layers); developed, led production cycle, and published on Tableau server (proprietary)

rito


National Patient HCAHPS Satisfaction Scores – What is the baseline?, Kaiser Permanente (2016)

Compared publicly reported HCAHPS scores by state and analyzed California performance by rating category (Tableau)

hcahps


Enterprise-Level Physiological Alarm Dashboard, Kaiser Permanente (2015)

National, regional, hospital, and hospital department analytics of alarm dispatch frequencies by type, location, and severity (4 interactive dashboard layers); developed, led production cycle, and published on Tableau server (proprietary)

alarms