Health data science Flashcards
Why is health data science important?
Helps us understand causes, prevalence, characteristics and outcomes of disease
Identify new treatment methods
What are different types of health data?
Patient data Specific instruments ie questionnaires Data from blood or tissue samples Data from images Health and fitness devices
How can health data be collected?
Observational studies - case-control study, etc
Intervention study
Big data - Routinely collected data, data linkage, AI
What is a cross-sectional study?
Study that measures variable interests - classically exposures and outcomes - at the same time
What are strengths of cross-sectional studies?
Relatively easy and cheap
Provide important information
What are weaknesses of health data studies?
Oly measure prevalence, not incidence
Can be difficult to establish sequence of events ie did risk factor come before disease
What are strengths of case control studies?
Quick and relatively cheap
Good for studying rare disease
Good for disease with long latent periods between exposure and outomce
What are weaknesses of case-control studies?
Prone to selection bias
Prone to information bias
Can’t establish sequence of events
What is a cohort study?
Studies begin with a group of people who either all have or don’t have the disease, looking at exposures or characteristics respectively affecting disease
What are strengths of cohort studies?
Exposure or prognostic factors are measured at stat of study
Can provide data on course of development of outcomes
Multiple outcomes can be examined
What are weaknesses of cohort studies?
Slow and expensive
Inefficien for rare disease
Exposure status may change during study
Differential loss to follow-up may introduce bias
What is the gold standard interventional study?
Randomised control trial
What needs to be considered for risk of bias in RCTs?
Was generation of randomisation sequence unbiased?
Was allocation concealed until enrolment?
Were participants/assessors aware of treatment group?
Is there missing data which could introduce bias
Was measurement of outcome unbiased?
Was the pre-specified primary outcome reported?
What are challenges of big data?
Privacy and security
Quality of the data
What is data linkage?
Linking datasets harnessing the breadth of data that are available