Data Analytics and Data Management Flashcards

Which statistical test is appropriate for comparing the mean length of stay across three or more independent hospital units when data are normally distributed?

Answer: One-way ANOVA

One-way ANOVA tests whether means differ significantly across three or more independent groups when the normality and homogeneity of variance assumptions are met.

In master data management (MDM), which process resolves the problem of the same patient appearing as multiple records across different source systems?

Answer: Entity resolution (record linkage)

Entity resolution uses deterministic or probabilistic matching algorithms to identify and merge duplicate representations of the same real-world patient across systems.

A clinical data warehouse team wants to track slowly changing attributes such as a patient's insurance plan over time without losing historical values. Which SCD type should they use?

Answer: SCD Type 2 (add new row)

SCD Type 2 inserts a new row with effective and expiration dates each time an attribute changes, preserving full history of all prior values.

Which privacy-preserving technique adds calibrated statistical noise to query results to protect individual records while still allowing accurate aggregate analysis?

Answer: Differential privacy

Differential privacy injects carefully calibrated noise into outputs so that the presence or absence of any single individual cannot be inferred from the results.

In a Hadoop ecosystem, which component is responsible for distributed storage of large health datasets across commodity hardware nodes?

Answer: HDFS (Hadoop Distributed File System)

HDFS splits large files into blocks and replicates them across multiple nodes, providing fault-tolerant distributed storage for big data workloads.

A health system wants to share de-identified data for research under the HIPAA Safe Harbor method. Which of the following must be removed?

Answer: All 18 specified PHI identifiers including ZIP codes and dates

HIPAA Safe Harbor requires removal of all 18 specific identifiers, including geographic subdivisions smaller than a state, dates other than year, and contact information.

Which concept describes the phenomenon where a predictive model performs well on training data but poorly on new, unseen patient data due to learning noise rather than true patterns?

Answer: Overfitting

Overfitting occurs when a model memorizes training data noise, resulting in high training accuracy but poor generalization to new data.