Biomedical Informatics Flashcards
Q:Define electronic health record (EHR)
A:Digital longitudinal repository of patient clinical demographic and administrative data intended for secure real time access and decision support
Q:List four structural models of EHRs
A:Integrated Source oriented Problem oriented Protocol oriented
Q:Name two major clinical coding systems
A:ICD ten and SNOMED
Q:Give three clinical advantages of moving from paper to EHRs
A:Simultaneous access Automated decision support Easier data audit
Q:State one major security risk introduced by EHRs
A:Large scale data breach potential
Q:What physiological waveform captures cardiac electrical activity
A:Electrocardiogram
Q:Why are window functions applied before an FFT
A:To taper edges of a time block and reduce spectral leakage caused by discontinuities
Q:Write the logistic regression logit equation
A:Log odds equals beta naught plus sum beta i times x i
Q:Interpret exp beta in logistic regression
A:It is the odds ratio for a one unit increase in the predictor
Q:Define survival function S t
A:Probability that event time exceeds t
Q:Express hazard function h t in terms of f t and S t
A:h t equals f t divided by S t
Q:What assumption underlies the Cox model
A:Proportional hazards constant hazard ratio over time
Q:Give the Kaplan Meier estimator formula
A:S t equals product over event times of one minus d i divided by n i
Q:Define sensitivity
A:True positives divided by true positives plus false negatives
Q:Define specificity
A:True negatives divided by true negatives plus false positives
Q:Explain AUROC
A:Probability that a random positive gets higher score than a random negative
Q:What is the curse of dimensionality
A:Exponential data requirement growth as feature count increases causing sparse coverage
Q:Describe mean imputation
A:Replacing missing entries with the variable mean so values become centered but variance is reduced
Q:Give one disadvantage of mean imputation
A:Bias toward normal values and understatement of uncertainty
Q:Explain k fold cross validation
A:Partition data into k equal folds iterate training on k minus one folds validate on the remaining fold average performance
Q:Define autoregressive AR p model
A:Current value equals weighted sum of previous p values plus noise
Q:Why is AIC used with AR models
A:To select order p by balancing goodness of fit and model complexity
Q:How can AR coefficients give spectral information
A:Transfer function poles derived from coefficients determine power spectral density
Q:State one limitation of AR models in ICU data
A:Assume stationarity within the modeled window
