8. Healthcare Data Analytics & Stats Flashcards
what is an example of rapid data growth in hc
better PET scans (RPA)
watches, ear buds that can measure temperature
what are the 5Vs of big data [EXAM Q] & the new element to add in
- volume
- velocity
- value
- veracity
- variety
- complexity is the new element
what is complexity in big data (+1 C to the 5Vs)
data contains deeply embedded associations & hidden relationships that need to be extract using complex computerised analytical methods
what is an example of a digital twin in hc
drug effects on patient physiology
how do hc providers benefit from data analysis
- it’s labour intensive to interpret histopathic images (counting, delineation, staining). analytics can make the diagnosis more complete and timely
- pre-emptive care
- tailored care
- auditing
- strategic resource allocation (limited funding & resources)
do you need a tool to extract medical records
yes because different IS, EMR EHR etc.
what makes up a data warehouse cube in hc
variable x individual x time
what is the difference between experimental and observational studies
experimental = establish causality, randomised trials, has flaws e.g. publication bias, selection bias, unethical protocols e.g. thalidomide
observational = passive researcher participation, prospective studies, following a sample into the future, retrospective studies, collect information about samples on specific outcomes
what is a post-hoc analysis
testing hypotheses formulated after collecting data / confirm findings using experiments
what are type 1 & type 2 errors
1 = FP
2 = FN
what is paired and what is independent
paired means two populations share a mutual characteristic
e.g. popn with age, gender, smoking history and another popn has age, gender - these samples are not paired
attributes & types need to be exactly the same
what is the wilcoxon test
paired test for non-parametric data
- data is at least ordinal e.g. patient feedback
- n >= 20
what is t test
statistical test that is used to compare means
what is anova & it’s assumptions
similar to t-test, except instead of 2 means, multiple means can be compared
assumptions = independent samples, populations are normally distributed, std. dev are equal
what is kruskal-wallis & it’s assmptions
- non-parametric version of one-way ANOVA
assumptions
Samples are random samples, or allocation to treatment group is random. The two samples are mutually independent. The measurement scale is at least ordinal, and the variable is continuous.
