Embed Size (px)
Citation preview
Creating preliminary data: Large dataset research
1
EAST Research Short Course
Wednesday, January 15th, 2020
Orlando, FL
Heena P Santry, MD MS FACS
Associate Professor of Surgery
Director, Center for Surgical Health Assessment, Research & Policy
Ohio State Wexner Medical Center
Nothing to disclose
2
Myths about large dataset research
• It’s quick
• It’s easy
• I have a med student who knows stats
3
Realities of dataset research
• Critical thinking
• Time
• Statistical expertise
4
Large dataset research is SCIENCE
• Robust research design
• Hypothesis driven
• Novel
5
Education
• MSHS
• MPH
• Certificate
• Epidemiology
• Biostatistics
6
Resources - Reading
7
Resources - Training
8
Types of data
• Administrative
• Registries
• Quality programs
• Survey
9
Administrative data
10
Registry data
11
Quality program data
12
Survey data
13
Analytic potential
• Epidemiology
• Outcomes research
• Social determinants of health
• Social network analysis
• Health behaviors
• Cost effectiveness
14
Step 1.
What is your research question?
Why do you need a large dataset to ask it?
What will you do with the findings?
15
Step 2.
Write your introduction – End with the why/so what
Template your tables
16
Step 3.
Understand the data – Review the data dictionary
What variables are collected?
How are they made available?
How can you identify your population of interest?
17
Step 4.
Which dataset can answer your research question?
What are the limitations of this choice dataset?
How can you acquire the data?
Does someone else on campus already own it?
18
Step 5.
Acquire the data
DUAs
IRBs
Data Security
19
Step 6.
Design your experiment
20
Step 7.
Conduct analyses
Fill your tables
21
Step 8.
Ask yourself did I find anything novel?
Did my findings support or refute the hypothesis?
Do not massage the data
22
Step 9.
Create compelling visuals
Write the results
Interpret the results
23
Step 10.
Craft conclusions, limitations, implications
So what?
24
