Question

how Data is the foremost character required for healthcare data analytics. Discuss the various data quality characteristics or dimensions that govern data quality.n

Answer 1

Predictive Modeling is about using historical data to look at the model to predict future outcomes. It is not a single algorithm, but rather a computational function sequence. There are many different options for every step in this pipeline. All these combined give us a lot of different pipelines to assess and compare. The key steps are to define the prediction target, build the right patient cohort, build the right features (observation window, index date, prediction window and date of diagnosis) and then make a prediction (with a predictive model). A predictive model maps the patient's input characteristics to the output target (diagnosis, treatment, laboratory outcome).

Computational Phenotyping is about turning chaotic eHRs into practical clinical concepts. Computational phenotyping input is raw patient data from numerous sources such as diagnosis of demographic information, drug, treatment, laboratory test and clinical notes. This raw patient data is transformed into scientific terms or phenotypes by the phenotyping algorithms. The principal use of this data is to support clinical operations such as billing, or genomic studies.

Phenotyping approaches can use Supervised Learning — data is labeled and algorithms need to perform a "Function Approximation" such as classification or Unsupervised Learning — data is unlabeled and algorithms need to provide a "Profile" or "Short Summary" like clustering. For the most part, phenotyping uses clustering algorithms to segment, let's say a case-by-disease matrix to help Case Stratification of various patient classes.