MCQ Quiz: Machine Learning in Pharmacy Practice

Machine Learning (ML), a powerful subset of Artificial Intelligence, enables computer systems to learn from vast amounts of data to make predictions and improve processes. Its role in pharmacy practice is rapidly expanding, with applications ranging from personalizing medication regimens to predicting adverse drug events and optimizing pharmacy operations. This quiz explores the fundamental concepts and applications of Machine Learning in the pharmacy profession.

1. Machine Learning is best defined as a field of study that:

• a. Gives computers the ability to learn without being explicitly programmed.
• b. Involves creating physical robots for dispensing.
• c. Is only used for playing chess.
• d. Focuses on designing computer hardware.

Answer: a. Gives computers the ability to learn without being explicitly programmed.

2. A pharmacist uses a system that was trained on thousands of patient records to predict the 30-day hospital readmission risk for a specific patient. This is an application of:

• a. Manual data entry
• b. Standard dispensing software
• c. Machine learning (predictive analytics)
• d. Sterile compounding

Answer: c. Machine learning (predictive analytics)

3. In “supervised learning,” a machine learning model is trained on data that is:

• a. Completely random and has no organization.
• b. Unlabeled, requiring the model to find its own patterns.
• c. Labeled with the correct outcomes.
• d. From a single patient only.

Answer: c. Labeled with the correct outcomes.

4. A hospital wants to use machine learning to group patients into distinct clinical phenotypes based on their EHR data, without any pre-defined categories. This would require which type of machine learning?

• a. Supervised learning
• b. Reinforcement learning
• c. Unsupervised learning (clustering)
• d. All of the above

Answer: c. Unsupervised learning (clustering)

5. The quality and performance of any machine learning model are most dependent on:

• a. The speed of the computer it runs on.
• b. The quality, size, and representativeness of the data it was trained on.
• c. The brand of the software used.
• d. The number of alerts it can generate.

Answer: b. The quality, size, and representativeness of the data it was trained on.

6. “Natural Language Processing” (NLP) is a type of machine learning that allows computers to:

• a. Process and understand human language from sources like clinical notes.
• b. Analyze medical images.
• c. Control robotic dispensing arms.
• d. Predict stock market trends.

Answer: a. Process and understand human language from sources like clinical notes.

7. A major challenge for using machine learning in pharmacy is the “black box” problem, which means:

• a. The data is kept in a secure, black-colored box.
• b. The models are too simple to be useful.
• c. It can be difficult to understand the reasoning behind a specific prediction made by a complex model.
• d. The models require a dark room to function.

Answer: c. It can be difficult to understand the reasoning behind a specific prediction made by a complex model.

8. Which of the following is an application of machine learning in personalized medicine?

• a. Using a patient’s genomic and clinical data to predict their response to a specific drug.
• b. Dispensing the same dose of a drug to every patient.
• c. Using a standard dosing nomogram.
• d. Recommending a multivitamin.

Answer: a. Using a patient’s genomic and clinical data to predict their response to a specific drug.

9. The pharmacist’s role in the era of machine learning will be to:

• a. Be replaced by the computer.
• b. Act as a clinical expert who validates, interprets, and applies the outputs of ML models to individual patient care.
• c. Write the code for all the machine learning models.
• d. Manually enter all data.

Answer: b. Act as a clinical expert who validates, interprets, and applies the outputs of ML models to individual patient care.

10. An ML model is trained on data from a single academic hospital. When applied to a rural community hospital population, its performance may decrease. This is an issue of:

• a. Too much data
• b. Model generalizability
• c. The model being too simple
• d. The model being perfect

Answer: b. Model generalizability

11. An ML model that performs perfectly on the data it was trained on but fails to make accurate predictions on new, unseen data is said to be:

• a. Underfitting
• b. Overfitting
• c. A generalized model
• d. A simple model

Answer: b. Overfitting

12. “Feature engineering” in the machine learning process involves:

• a. Designing the user interface of an application.
• b. Selecting and transforming the most relevant variables (features) from the raw data to be used by the model.
• c. Adding more memory to the computer.
• d. Choosing the color scheme for the output graphs.

Answer: b. Selecting and transforming the most relevant variables (features) from the raw data to be used by the model.

13. A pharmacy using an ML algorithm to analyze purchasing patterns and predict demand for flu vaccines is using it for:

• a. Clinical decision support
• b. Inventory and operations management
• c. Patient counseling
• d. Drug discovery

Answer: b. Inventory and operations management

14. If a machine learning model is trained on historically biased data (e.g., data where certain populations received poorer care), the model will likely:

• a. Correct the bias automatically.
• b. Learn and perpetuate the existing bias in its predictions.
• c. Refuse to make any predictions.
• d. Perform equally well for all populations.

Answer: b. Learn and perpetuate the existing bias in its predictions.

15. The “garbage in, garbage out” principle applies strongly to machine learning.

• a. True
• b. False

Answer: a. True

16. Which of the following is a potential benefit of using machine learning in medication safety?

• a. Identifying patients at high risk for adverse drug events before they happen.
• b. Analyzing EHR data to discover previously unknown drug interactions.
• c. Powering more intelligent and specific clinical decision support alerts.
• d. All of the above.

Answer: d. All of the above.

17. What is a “training dataset” in machine learning?

• a. The data used to test the final performance of a model.
• b. The data used to build and “teach” the machine learning model.
• c. A set of instructions for the programmer.
• d. The output of the model.

Answer: b. The data used to build and “teach” the machine learning model.

18. A “validation dataset” is used to:

• a. Build the initial model.
• b. Tune the model’s parameters and assess its performance during development.
• c. Provide new predictions for clinicians.
• d. Store the final model.

Answer: b. Tune the model’s parameters and assess its performance during development.

19. A key role for a pharmacist on a team developing a clinical ML model is to:

• a. Ensure the clinical variables and outcomes are defined correctly and are meaningful.
• b. Write the final programming code.
• c. Set up the computer servers.
• d. Market the final product.

Answer: a. Ensure the clinical variables and outcomes are defined correctly and are meaningful.

20. A pharmacist should always trust the output of a machine learning model without applying their own clinical judgment.

• a. True
• b. False

Answer: b. False

21. In drug discovery, machine learning can accelerate the process by:

• a. Screening millions of potential molecules virtually to predict their activity.
• b. Synthesizing the final drug product.
• c. Conducting human clinical trials.
• d. Writing the manuscript for publication.

Answer: a. Screening millions of potential molecules virtually to predict their activity.

22. An ML model designed to predict if a patient will be adherent to their statin medication is an example of a(n):

• a. Unsupervised learning problem.
• b. Reinforcement learning problem.
• c. Classification problem (a type of supervised learning).
• d. Clustering problem.

Answer: c. Classification problem (a type of supervised learning).

23. Real-World Data (RWD) is the primary fuel for training most healthcare-related machine learning models.

• a. True
• b. False

Answer: a. True

24. The FDA is developing a framework to regulate AI/ML-based software as a medical device.

• a. True
• b. False

Answer: a. True

25. A pharmacist using an NLP tool to scan through progress notes to find mentions of a specific side effect is leveraging machine learning to analyze:

• a. Structured data
• b. Unstructured data
• c. Numerical data only
• d. Image data

Answer: b. Unstructured data

26. The most significant barrier to implementing effective machine learning in healthcare is often:

• a. A lack of complex algorithms.
• b. The difficulty in obtaining large, clean, well-structured, and interoperable data.
• c. A lack of computing power.
• d. Resistance from patients.

Answer: b. The difficulty in obtaining large, clean, well-structured, and interoperable data.

27. An AI chatbot that provides patient counseling based on machine learning could be a future tool in pharmacy practice. A major concern would be:

• a. Ensuring the information provided is always accurate and safe.
• b. The lack of human empathy and ability to read non-verbal cues.
• c. Both a and b.
• d. Neither a nor b.

Answer: c. Both a and b.

28. “Deep learning” is a more advanced type of machine learning that uses:

• a. Simple linear regression.
• b. Decision trees.
• c. Complex, multi-layered neural networks.
• d. Abacuses.

Answer: c. Complex, multi-layered neural networks.

29. As machine learning becomes more integrated into practice, a pharmacist’s skills in ____ will become even more important.

• a. manual pill counting
• b. critical thinking and clinical judgment
• c. data entry
• d. inventory management

Answer: b. critical thinking and clinical judgment

30. Which of the following pharmacy tasks is LEAST likely to be impacted by machine learning?

• a. Screening for drug-drug interactions.
• b. Predicting patient risk for adverse events.
• c. Providing an empathetic response to a distressed patient.
• d. Optimizing pharmacy inventory.

Answer: c. Providing an empathetic response to a distressed patient.

31. The ethical principle of “fairness” in machine learning means:

• a. The model performs equally well across different demographic groups.
• b. The model is available for free.
• c. The model is easy to understand.
• d. The model makes the same prediction for every patient.

Answer: a. The model performs equally well across different demographic groups.

32. The pharmacist of the future may need to be skilled in “data literacy,” which is the ability to:

• a. Read, understand, create, and communicate with data.
• b. Write computer programs.
• c. Repair computer hardware.
• d. Only use paper charts.

Answer: a. Read, understand, create, and communicate with data.

33. An ML model could be used to optimize a patient’s warfarin dose by analyzing:

• a. Their past INR values.
• b. Their genetic information (e.g., CYP2C9, VKORC1).
• c. Their diet and concomitant medications.
• d. All of the above.

Answer: d. All of the above.

34. “Model drift” is a phenomenon where:

• a. The model’s performance improves over time.
• b. The model’s performance degrades over time as real-world data and practices change.
• c. The model’s predictions drift towards a single outcome.
• d. The computer hardware physically moves.

Answer: b. The model’s performance degrades over time as real-world data and practices change.

35. A key role for pharmacists is to be involved in the ____ of clinical machine learning models.

• a. design
• b. validation
• c. implementation and monitoring
• d. all of the above

Answer: d. all of the above

36. A machine learning algorithm could improve antimicrobial stewardship by:

• a. Predicting which patients are most at risk for a resistant infection.
• b. Recommending the most appropriate empiric antibiotic based on local resistance patterns and patient data.
• c. Identifying patients who could be switched from IV to PO therapy.
• d. All of the above.

Answer: d. All of the above.

37. The interpretation of machine learning outputs requires:

• a. No clinical knowledge.
• b. A deep understanding of the clinical context.
• c. Only statistical knowledge.
• d. Only computer programming knowledge.

Answer: b. A deep understanding of the clinical context.

38. The use of automated systems in the hospital, as mentioned in the HIPPE syllabus, is an early form of applying algorithms to improve safety.

• a. True
• b. False

Answer: a. True

39. A potential use of unsupervised learning in pharmacy would be:

• a. To predict a patient’s A1c next year.
• b. To identify previously unknown patient subgroups that respond differently to a medication.
• c. To verify the dose of a medication.
• d. To send a refill reminder.

Answer: b. To identify previously unknown patient subgroups that respond differently to a medication.

40. A pharmacist who questions an unusual recommendation from an AI tool is:

• a. Resisting technology.
• b. Performing a critical professional function.
• c. Lacking trust in the system.
• d. Wasting time.

Answer: b. Performing a critical professional function.

41. What type of data is essential for training an ML model to recognize adverse drug reactions from patient notes?

• a. A large volume of clinical notes with accurately labeled examples of ADRs.
• b. Only structured lab data.
• c. Only billing data.
• d. Only demographic data.

Answer: a. A large volume of clinical notes with accurately labeled examples of ADRs.

42. The “black box” nature of some complex ML models is a challenge for:

• a. Clinical explainability and trust.
• b. Regulatory approval.
• c. Identifying sources of error or bias.
• d. All of the above.

Answer: d. All of the above.

43. A pharmacist’s ability to critically appraise evidence will apply to ML-based studies in the future.

• a. True
• b. False

Answer: a. True

44. What is reinforcement learning?

• a. Training a model on labeled data.
• b. Finding patterns in unlabeled data.
• c. Training a model through a process of trial and error, using rewards and punishments.
• d. A type of human learning.

Answer: c. Training a model through a process of trial and error, using rewards and punishments.

45. An ML model to personalize cancer therapy would be an application of:

• a. Personalized medicine.
• b. Pharmacogenomics.
• c. Both a and b.
• d. Neither a nor b.

Answer: c. Both a and b.

46. The data used to train a machine learning model must be protected under what regulation?

• a. The Controlled Substances Act
• b. The Food, Drug, and Cosmetic Act
• c. HIPAA
• d. OBRA ’90

Answer: c. HIPAA

47. A key challenge for NLP models in reading clinical notes is understanding:

• a. Medical abbreviations.
• b. Sarcasm and nuanced language.
• c. Negated statements (e.g., “patient denies chest pain”).
• d. All of the above.

Answer: d. All of the above.

48. Machine learning is considered a subfield of Artificial Intelligence.

• a. True
• b. False

Answer: a. True

49. The overall goal of using machine learning in pharmacy is to:

• a. Augment the pharmacist’s abilities to improve the safety, efficiency, and effectiveness of patient care.
• b. Create a fully automated pharmacy with no human involvement.
• c. Make pharmacy practice more complicated.
• d. Focus only on the most profitable activities.

Answer: a. Augment the pharmacist’s abilities to improve the safety, efficiency, and effectiveness of patient care.

50. The ultimate reason for a student pharmacist to learn about machine learning is to:

• a. Be prepared for the future of healthcare and to be able to work effectively with new technologies.
• b. Become a computer programmer.
• c. Design their own AI.
• d. Pass an elective course.

Answer: a. Be prepared for the future of healthcare and to be able to work effectively with new technologies.