Week 6 Flashcards
AHRQ
anrq.gov
Agency for Healthcare Research and Quality
- safer
*patient centered - timely
- effective
- accessible
- efficiently provided
- equitably distributable
NCQA
ncqa.org
National Committee for Quality Assurance
use measurement, transparency and accountability to highlight top performers and drive improvement.
HEDIS
The Healthcare Effectiveness Data and Information Set (HEDIS)
Tracking over 90 CMS measures this tool provides a way to compare health plans, providers and
CMS Quality Measures
CMS implements quality initiatives to assure quality health care for Medicare Beneficiaries through accountability and public disclosure. CMS uses quality measures in its various quality initiatives that include quality improvement, pay for reporting, and public reporting.
CMS
Centers for Medicare and Medicaid Services
Medicare vs Medicaid
Medicare is healthcare for individuals 65 and up
Medicaid is for individuals with limited income and resources.
NQF
qualityforum.org
National Quality Forum
The Joint Commission
A organization responsible for certification of health organizations CMS measure quality. It is required y most states for Medical and Medicaid reimbursement.
ANNs
FFNNs vs RNNs
Artificial Neural Networks
Feed / Fast Forward Neural Network vs. Recurrent Neural Network
Data modeling patterns influenced by Biological systems that are sometimes capable of finding patterns in a “input layer” of features and producing an “output layer” of desired categories.
FFNNs has the “signal” pass from the input layer to the output layer though n number of hidden layers in one direction.
RNNs also have an input layer of features and an output layer of categories, but the signal state is remembered at each hidden layer and may travel in both directions before arriving at the output layer. Most healthcare models are of this type and are also the most expensive to train and deploy.
AI Features vs Predications and Classification
AI Features are the input values to a statistical model that are considered to be relevant in calculating the Predictions or Classifications in the output. Classifications refer to assigning a discrete label to the computation of the features.
Deep Learning CNNs
Deep Learning Convolutional Neural Networks have many hidden layers and are capable of being better at finding patterns (ie. in X-rays) than clinicians. They are the de facto standard in image processing.
Overfitting
A phenomena in ANNs were too many features are used and the model becomes good at predicting in the training set but bad at predicting unseen data from the test set
ANN Regularization
Artificial Neural Network Regularization is a set of techniques that help prevent overfitting and improve the models ability to predict generalized data (ie. from the test set).
Ridge regression vs LASSO
Ridge regression is a feature optimization technique that does not remove features but decreases some features weight to decrease over-fitting.
LASSO is also a feature optimization technique that may reduce some features weights to zero excluding them from the model.
Which of the following is not a common characteristic of deep learning models?
A.Computationally intensive
B.Use of convolution functions
C.Fewer layers but with more nodes
D.Can identify more obscure patterns in the
The answer is C. Fewer layers but with more nodesand here is why. The way deep learning models can capture deeper, morecomplex patterns in the data is by using many layers where each layer can capture some abstract characteristic of the data. Deep learning models commonly use convolutions,and if they do, they are called convolutional neural networks. Because of their complex structure and the use of complex functions such as convolutions, deep learning models are computationally more intensive than traditional models such as general linear models or even basic ANNs
