HCI3 Exam 3 - Big Data

big data

extremely large and complex databases that aggregate information of different types and scales from multiple sources and require data processing applications

volume

the amount of data; quantity of generated and stored data

variety

the different forms of data; the type and nature of the data such as text, images, audio, video, etc.

velocity

the speed that new data is generated and moves around; often available in real-time

veracity

the trustworthiness of the data; extended definition for big data that refers to the data quality and value

EHR, insurance claims, smartphone applications, wearable devices, social media, and population genomics data

big data in healthcare may include ______ (6)

text messages, videos, audios, and images

list the four types of data typically sent from a cell phone

structured data

refer to information that is stored in a consistent, organized manner and is typically reported using standard units and ranges like lab results, vital signs, and ICD-based diagnosis

unstructured data

data that is devoid of a clear organization and precision

free-text boxes

the inclusion of _______ in EHR poses a significant challenge for structuring data

NLP and machine learning techniques

used to extract written text from EHRs

natural language processing

borrows concepts and techniques from linguistics, computer science, and engineering to process naturally occurring language

machine learning models

enable computers to extract patterns in datasets and draw conclusions on their own

pharmacovigilance

EhR has already provided valuable insights into ______ such as identification of drug-related adverse events, discrepancies in patient's meds and errors in prescriptions based on comorbidities and risk factors

financial burden of patients

EHRs contain drug cost data to help palliate the ______ and lead to large-scale analyses to allow prescribers to offer the most cost-effective meds to patients

real world data

TriNetX provides access to de-identified EHR data, offering _______

federated data network

TriNetX operates a _______, meaning it connects various healthcare organizations and their data while maintaining data privacy and security

true

T/F: TriNetX is HIPAA (USA) and GDPR (Europe) compliant

notes, raw diagnostic reports, DICOM image objects, providers, departments/clinics, patient level data, and publication guidelines

what information is unavailable in TriNetX?

250 million

approximately how may patients are included in TriNetX?

prevalence and incidence

epidemiology and disease burden studies estimate the ________ of diseases across diverse populations

drug safety and pharmacogivilance

conduct signal detection for adverse drug events, compare incidence of side effects between treatment groups, and identify potential drug-drug interactions

clinical trial feasibility and recruitment

responsible for identifying eligible patient populations for prospective clinical trials, forecast recruitment rates based on historical patient flows, and locate clinical sites with enough patients meeting trial inclusion/exclusion criteria

spurious correlation

describes the phenomenon of important variables being highly correlated with variables with which they have no actual relations when a large number of variabels are evaluated

type 1

spurious correlation is an example of _______ error (false positive)

type 1

error examples include a compound has no effect but FDA approves it, an innocent person goes to jail, and a normal email was recognized as spam by a spam filter

alpha

probability of incorrectly rejecting the Ho

p

general consensus in scientific literature is an alpha level at ______

type 2

error examples include a compound has effect but FDA rejects it, a guilty person goes free, and a spam email was recognized as normal by a spam filter

false negative

a type 2 error is also known as a ______

beta

probability of making a type 2 error

noise accumulation

accumulation of estimation errors when prediction is based on a large number of parameters, leading to poor classification or prediction

true

T/F: when the number of predictors is large, the addition of more predictors improves the predictive power and reduces noise of the model

predictive analytics

a new method of analysis developed to overcome the challenges

predictive modeling, machine learning, and data mining

three statistical techniques used in predictive analytics

construct patient-level data set, reduce dimensionality, and build model with validation and selection

process of predictive analytics

cleansing and normalization of data, aggregation of multiple data sources, de-identification, and validation

steps involved in the construction of a patient-level data set

input

reducing the number of _____ (input/output) dimensions can improve the performance of predictive analytic models

univariate selection

selection acting on one trait in order to reduce dimensionality of predictive model

multivariable regression

statistical modeling that only uses variables strongly associated with the outcome to reduce dimensionality

training set

used for model building

validation set

used for model selection

test set

used to assess the performance of the model selected

overfitting

occurs when a model is excessively complex and describes random error and noise rather than the underlying relationship; represents the data rather than predicting it

randomly splitting the data set

used to mitigate the overfitting problem of the original data set

logistic regression

categorical continuous data uses ________

linear regression

continuous continuous data uses ________

artificial neuron networks, support vector machines, discriminant analysis, and classification trees

models commonly used in big data analytics include ______

binary

specificity, sensitivity, and ROC curve area under the curve are used for _____ (binary/continuous) outcomes

continuous

mean absolute error, mean squared error, and medical absolute error are used for _____ (binary/continuous) outcomes

sensitivity

measures the proportion of actual positive which are correctly identified and is complementary to the false negative rate

specificity

measures the proportion of negatives which are correctly identified and is complementary to the false positive rate

AUROC (area under the curve receiver operating characteristics curve)

one of the most important evaluation metrics for checking any classification model's performance

method 1

which method performed better?

data analyst, pharmacoepidemiologist or outcomes researcher, computer scientist, and system administrator

required resources for big data predictive analytics

computing machines, large storage space, and security protection

infrastructure requirements for predictive analytics