Biomedical Engineering Reference
In-Depth Information
Clinical analytics is fraught with complexities in this area, and indeed the
market is in its infancy, especially in open source. The mathematical and
programmatic foundations are strong and companies like Google and
Twitter [25] are bringing meaning to massive amounts of dirty data.
A smorgasbord of topics like population health, pay for performance,
accountable care organizations, clinical trials, risk adjustment, disease
management, and utilization statistics are about to undergo a golden age of
understanding, intelligence, and adoption. The rollout and standardization
of medical records through initiatives like HL7, IHE, and Health Story are
producing immense amounts of structured data exactly for this purpose.
It is time to start analyzing these data in aggregate!
Data in aggregate come with a large set of challenges. Mathematicians
and statisticians have been hard at work for centuries developing
techniques to understand how to interpret small and large sets of data.
The following is a short list of common capabilities and methodologies
needed when studying clinical analytics.
■
Linear/non-linear/curvilinear regressions - different levels of standard
statistical methods to study the interdependency between variables.
Examples: what is the relationship between smoking and lung
cancer? What is the relationship between birth weight and health
outcomes?
■
Statistical classifi cation - algorithms for specifi cation of new sets based
on patterns learned from previous sets of data. Examples: used in
natural language processing of medical text. A new disease outbreak
has been discovered, what known patterns does it exhibit?
■
Clustering - algorithms to identify subpopulations and to study the
identities and characteristics of those subpopulations. Example: what
are the different lifestyles, cultures, or age groups associated with
different types of diabetes?
■
Pattern-mining - a method of data-mining to fi nd repeatable situations.
Example: what behavior patterns exhibit a likelihood of alcoholism?
■
Dimension reduction - process of removing dimensions to simplify a
problem for study. Examples: when determining the relative cost
of a disease, remove variables associated with geography or wealth.
Uncovering the molecular variation of certain cancers.
■
Evolutionary algorithms - understanding impacts of repeated
conditions. Examples: what impact on health do certain policy changes
cause? Could a certain method of continual treatment have been a
contributing factor over time?
Search WWH ::
Custom Search