HAC Healthcare Analytics Predictive Analytics & Trends 2 — Questions and Answers
Question 1: Which evaluation metric is most appropriate for a predictive model identifying rare hospital-acquired infections where false negatives are especially costly?
- Accuracy
- Specificity
- Recall (Sensitivity) (Correct answer)
- F1 Score
Correct answer: Recall (Sensitivity)
Recall maximizes detection of true positive cases, minimizing false negatives, which is critical when missing an infection has severe consequences.
Question 2: A healthcare organization wants to forecast ICU bed demand 30 days in advance. Which modeling approach is best suited for this time-series forecasting task?
- Logistic regression
- ARIMA (Correct answer)
- K-means clustering
- Decision tree
Correct answer: ARIMA
ARIMA (AutoRegressive Integrated Moving Average) is specifically designed for time-series forecasting, capturing temporal patterns in sequential data.
Question 3: What does the term 'data drift' mean in the context of deployed predictive models in healthcare?
- Data being transferred between EHR systems
- A shift in the statistical properties of input data over time (Correct answer)
- Errors introduced during data entry
- Unauthorized access to patient data
Correct answer: A shift in the statistical properties of input data over time
Data drift occurs when the statistical distribution of model input data changes over time, causing model performance to degrade.
Question 4: Which technique is used to identify which features most influence a predictive model's output, improving interpretability in clinical settings?
- Bootstrapping
- SHAP (SHapley Additive exPlanations) (Correct answer)
- Oversampling
- Cross-validation
Correct answer: SHAP (SHapley Additive exPlanations)
SHAP values quantify each feature's contribution to individual predictions, making complex models more interpretable to clinicians.
Question 5: A predictive readmission model trained on data from urban hospitals is deployed in a rural hospital and performs poorly. This is an example of:
- Overfitting
- Underfitting
- Generalizability failure (Correct answer)
- Selection bias
Correct answer: Generalizability failure
Generalizability failure occurs when a model trained on one population performs poorly on a different population with distinct characteristics.
Question 6: In predictive analytics, what is the purpose of a validation dataset distinct from the test dataset?
- To retrain the model with new data
- To tune hyperparameters without leaking test data information (Correct answer)
- To handle class imbalance
- To remove outliers before training
Correct answer: To tune hyperparameters without leaking test data information
The validation set is used during model development to tune hyperparameters, keeping the test set untouched to provide an unbiased final evaluation.
Question 7: Which emerging trend involves using natural language processing to extract structured clinical information from physician notes for predictive modeling?
- Computer vision for radiology
- Clinical NLP / text mining (Correct answer)
- Federated learning
- Reinforcement learning
Correct answer: Clinical NLP / text mining
Clinical NLP extracts structured data such as diagnoses, medications, and symptoms from unstructured physician notes to enrich predictive models.
HAC Healthcare Analytics Predictive Analytics & Trends 2
Which evaluation metric is most appropriate for a predictive model identifying rare hospital-acquired infections where false negatives are especially costly?