Healthcare Analyst Data Analysis & Interpretation Flashcards

A healthcare analyst observes that a new medication appears to reduce mortality in an observational study, but patients who received it were younger and had fewer comorbidities. This is an example of:

Answer: Healthy user bias

Healthy user bias occurs when healthier patients are more likely to receive a treatment, making the treatment appear more beneficial than it truly is.

When applying the LACE index (Length of stay, Acuity, Comorbidities, ED visits) to score readmission risk, which component represents the highest individual weight?

Answer: Number of ED visits in prior 6 months

In the LACE index, prior ED visits carry the highest possible sub-score (up to 4 points per visit, max 4), reflecting that frequent ED use is the strongest predictor of readmission.

An analyst builds a logistic regression model with an AUC of 0.91 on the training set but only 0.62 on the validation set. This pattern indicates:

Answer: The model is overfitting the training data

A large drop in AUC from training to validation is a classic sign of overfitting — the model learned noise in the training data rather than generalizable patterns.

A hospital system wants to compare its case-mix adjusted cost per episode to a national benchmark. The most appropriate comparison uses:

Answer: Observed costs divided by expected costs (O/E ratio)

The observed-to-expected (O/E) ratio standardizes performance by accounting for case-mix, enabling fair comparison to a benchmark population.

In a Kaplan-Meier survival analysis of time to hospital readmission, what does a 'censored' observation represent?

Answer: A patient whose follow-up ended before experiencing readmission (e.g., study end or lost to follow-up)

Censoring occurs when a patient's follow-up ends without experiencing the event of interest, contributing partial information to the survival estimate.

An analyst needs to present monthly ER visit trends to hospital leadership to highlight a 20% increase over 12 months. Which visualization best conveys this trend with appropriate context?

Answer: A line chart with control limits and trend annotation

A line chart with statistical process control limits shows the trend over time and distinguishes signal (real change) from noise, ideal for leadership communication.

A healthcare analyst is asked to evaluate the inter-rater reliability of two coders assigning quality flags to medical records. The most appropriate statistic is:

Answer: Cohen's kappa

Cohen's kappa measures agreement between two raters on categorical outcomes, correcting for chance agreement — the standard for coding reliability studies.