Flashcards in Health Data Deck (20)
Name the types of health data?
- appointment dates, test results, treatment details
- questionnaires, rating scales
Data from blood/tissue samples e.g. DNA, Histology
Data from images e.g. X-Rays, CT, MRI
Health and Fitness Data e.g. heart rate, activity monitoring
Name two methods of data collection?
Conventional study designs
- Observational Studies
- cross-sectional study
- case-controls study
- cohort study
- Intervention Study
What is a cross sectional study?
Measures variables of interest - risk factors and disease at the same time
What are the strengths and weaknesses of a cross sectional study?
- Relatively easy and cheap to conduct
- Provide important information on the distribution and burden of exposures and outcomes
- only measure prevalent, not incident cases. Therefore, only limited value for identifying risk factors
- It can be difficult to establish the time sequence of events
What is a case control study?
Starts with groups with and without a disease and looks back to see who had the exposure in the past. Used to identify risk factors for diseases with long latent periods
What are the strengths and weaknesses of a case-control study?
- Quick and relatively cheap
- Good for studying rare diseases
- Good for diseases with ling latent periods of time between exposure and outcome
- Prone to selection bias
- Prone to information bias
- Cannot establish the sequence of events - was risk factor present before disease?
What are cohort studies?
Studies begin with a group of people
- without the disease, measure exposures and then follow up over time to see who gets disease
- with a particular disease, measure characteristics and then see who gets particular outcomes
What are the strengths and weaknesses of a cohort studies?
- exposure/prognostic factors are measured at the start of study before outcome occurs, so measurement is not biased by the presence or absence of outcome
- can provide data on time course of the development of the outcomes
- multiple outcomes can be explained
- slow and potentially expensive
- inefficient for rare diseases
- exposure status may change during study
- differential loss to follow-up may introduce bias
What is Randomised Controlled Trial?
Gold standard interventional study
Comparison gorups should be similar with respect to cofounders
Prevents bias in allocation of participants to treatment/control
Difference between the intervention and control groups will be whether or not they received the intervention therefore any difference in outcome should be attributable to the intervention
What opportunities does Big Data present?
More comprehensive data
More detailed data
What challenges does Big Data present?
Privacy and security
- it is vital that identifiable personal data is held securely to protect
Quality of Data
- Researchers generally have less control of data
- Missing Data
- Risk of change findings due to multiple comparisons
What is data linkage?
Linking datasets that harness the breadth of data are available
What is Artificial Intelligence?
Science of mimicking human intelligence
What is machine learning?
An automated way to find patterns in data without being explicitly programmed where to look or what to conclude
What is deep learning?
A subset of machine learning using more computational techniques to learn computational techniques to learn complex patterns in large amounts of data
What is supervised machine learning ?
Train a machine by showing it examples instead of programming it
Name some applications of deep learning
Content understanding for filtering e.g. hate speech
Anomaly detection in financial security
Applications of deep learning in healthcare?
Automated fracture detection on wrist x-rays
Categorisation of lung nodules as benign
Classification of histology samples
Determination of nature of skin lesions
Data monitoring in ICU
Prediction of bowel cancer survival
Name the ethical principles relating to health data?