ENAR
2023-03-20

The Setting

History of Biostatistics at Vanderbilt

  • Chuck Federspiel was the first PhD biostatistician in 1959
  • Bill Dupont arrived in 1977
  • Yu Shyr arrived in 1994
  • Division of Biostatistics in Department of Preventive Medicine
  • Significant talent already present

VUMC

  • Academic medical centers: majority of biomedical research funding
  • Tremendous growth in NIH grant funding had started
  • Major cancer center
  • Difficult to recruit enough faculty biostatisticians into a Division
  • Operating fund, recruitment funds, P&T were not controlled by biostatisticians
  • VU School of Medicine had not decided to go “all in” on biostatistics

VUMC, continued

  • Vanderbilt School of Medicine has > 2200 faculty covering every area of biomedical research
  • Success in recruitment of biomedical research faculty (our collaborators) was stunning
  • School of Medicine rewarded team science in P&T
  • No funding for pre-grant development work outside the cancer center
  • Big push for new department from 3 research leaders
  • Big push from existing Biostatistics faculty

The Opportunity

Evolution

  • Originally position was for Vice Chair of Preventive Medicine
  • Hired as consultant to make suggestions for strengthening the division
  • Main suggestion: grow it into a department
  • Dean and Vice Chancellor agreed and asked me to be first head
  • Sought counsel from career mentor David DeMets
  • Opportunity to implement a number of ideas
              

The Plan

Hiring

  • Hire the best faculty and staff, using generous school support
  • Emphasize versatility when judging methods expertise
  • Positive correlation: methods research capabilities, collaboration, teaching skills
  • Viewed non-tenure-track positions as first-class citizens
  • Importance of administrative and IT staff
  • Able to attract top new graduates
    • Vanderbilt and Nashville attractive
    • Recruits excited to fully participate in formation of new department, setting priorities

Features

  • Wiki: manage department policies, procedures, collaborations, events, seminars, other Information
  • Great emphasis on reproducible research and open source computing tools, weaning from SAS, move to R
  • Major future goal: PhD and MS biostatistics graduate program with coverage of all 3 schools of thought
  • Most methods research is funded by meeting specific aims of large NIH grants … but
  • Tenure-track faculty exceeded expectations in becoming principal investigators on biostatistics research grants

Features, continued

  • Stable and ample funding for travel and computing costs through federal grant surcharge
  • All biostatisticians in the medical center are appointed and located in the Dept. of Biostatistics

Collaboration

  • Emphasize collaboration, de-emphasize consulting
  • Created Collaboration Protected Time Cost-Sharing Plan
  • Supported widely by clinical departments
  • Attached faculty and staff statisticians to them
  • Research development (pre-grant), K award mentoring

Collaboration, continued

  • Needed to rope off consulting to make clear to biomedical researchers the difference with collaboration
    • and need for 2 types of funding for collaboration
  • Created Biostatistics Clinics, offered each day since Oct 2005
  • No cost to investigators
  • Organized by daily area
    • HSR
    • High-D (omics, imaging)
    • Surgery/critical care
    • Clinical/Health research
    • Basic/Animal

Clinics, continued

  • We do anything to help the investigator that can be done in 1.25h
  • Led to multiple long-term collaborations
  • Route hallway conversations to clinic
  • Excellent collaboration training ground for new graduates
  • Made clear that grant development is a much longer-term process requiring funding

Graduate Program

Graduate Program

  • Took several years to find a highly innovative program designer/director trained in more than the classical tradition
  • Fortunate to recruit Jeffrey Blume to lead
  • Designed a modern non-denominational biostatistics curriculum
  • All courses new (VU has never had a Stat department)
  • Pushed through 7 levels of approval
  • Grad students significantly raised energy level of the department
  • Also made it easier to recruit new faculty

Challenges

Challenges

  • Defining “data science”
  • Ebb and flow of hard $ funding for research development
  • Constancy of federal funds (“soft” money = “hard” money)
  • Vanderbilt so successful in federal funding that we can never recruit enough faculty & analytical staff

Methods Research and Skills

  • Stay relevant through lifelong learning
    • New methods constantly invented
    • New biomedical data acquisition technologies; must be willing to change methodologic directions
    • Constantly learn new computing, interactive graphics, and reproducible reporting tools
  • Exposure to alternate schools of thought
  • Learning Bayesian modeling completely changes way of thinking

Methods, continued

  • Know limitations of data and methods
    • Don’t be afraid to call BS (some ML and omics research)
  • Concentrate research on widely applicable methods
    • semiparametric models
    • longitudinal models for a variety of response types
    • methods for high-dimensional data
    • …

Most Important Things

Most Important Things

  • The people
  • The culture
  • The institution, support from its leaders, their respect for our field

Thanks

Thanks

  • Our faculty and staff
  • Our predecessors
  • VUMC leaders and institutional culture
  • David DeMets
  • Liana, Charlotte, and Frank Harrell Sr.

David C. Hurst, 1928-2013

Founding Chair
Department of Biostatistics
University of Alabama at Birmingham

More Information

  • vumc.org/biostatistics
  • biostat.app.vumc.org (wiki)
  • hbiostat.org