Biostatistics Management Flashcards

Question 1

What is biostatistics in the context of health research and administration?

Accepted Answer

Biostatistics is the application of statistical methods to questions in health and public health. In management, it guides planning, data quality, analysis oversight, and clear communication of findings. It helps teams decide what to collect, how to summarize it, and how confident they can be in the conclusions.

Question 2

What is the difference between a population and a sample?

Accepted Answer

A population is the complete set of units a study aims to describe, such as all visits to a program in a defined period. A sample is the subset actually observed or measured. Administrators should define the population before collecting data, since conclusions apply to it, and an unrepresentative sample can mislead decision makers.

Question 3

What is the difference between descriptive and inferential statistics?

Accepted Answer

Descriptive statistics summarize the data you have, using counts, percentages, means, medians, and ranges. Inferential statistics use a sample to draw conclusions about a larger population, usually with confidence intervals or hypothesis tests. Managers need both: descriptive summaries show what happened, while inferential methods help judge whether differences are likely to be real.

Question 4

What are variables, and how do categorical and continuous variables differ?

Accepted Answer

A variable is any characteristic recorded for each unit in a dataset. Categorical variables place units into groups, such as sex, site, or appointment status. Continuous variables take numeric values across a range, such as length of stay or wait time in minutes. The type determines which summaries and tests are appropriate.

Question 5

What is a measure of central tendency, and which ones are commonly reported?

Accepted Answer

A measure of central tendency describes a typical value. The three most common are the mean (the arithmetic average), the median (the middle value when data are ordered), and the mode (the most frequent value). Skewed data often call for the median, since a few extreme values can distort the mean.

Question 6

Why is it important to define a study's outcome before collecting data?

Accepted Answer

A primary outcome is the main measure that a study is designed to evaluate, such as a rate, count, or score. Defining it in advance prevents selective reporting, where analysts choose whichever result looks favorable after the fact. Pre-specification also guides sample size, data collection forms, and the analysis plan.

Question 7

What does a confidence interval communicate about an estimate?

Accepted Answer

A confidence interval gives a range of plausible values for an unknown population quantity. A 95 percent interval is constructed so that, across many repeated samples, about 95 percent of such intervals would contain the true value. A wide interval signals imprecision, usually from a small sample or high variability.

Question 8

What is a p-value, and what does it not tell you?

Accepted Answer

A p-value is the probability of results at least as extreme as those observed, assuming the null hypothesis is true. It does not measure effect size or the chance that the null hypothesis is true. Read it alongside confidence intervals and study design, not as a simple pass or fail rule.

Question 9

What is the purpose of a study protocol in biostatistics management?

Accepted Answer

A study protocol is the written plan that defines the research question, population, design, outcomes, procedures, and analysis approach before a study begins. It serves as the reference point for staff, sponsors, and oversight bodies, so that changes can be documented and justified. Clear protocols reduce later disputes about intent.

Question 10

How does a randomized design differ from an observational design?

Accepted Answer

In a randomized design, investigators assign units to comparison groups by chance, which tends to balance known and unknown factors across groups. In an observational design, investigators record exposures and outcomes without assigning them, so groups may differ in ways that complicate comparisons. Design choice limits what conclusions the data can support.

Question 11

What is a sample size calculation, and what inputs does it require?

Accepted Answer

A sample size calculation estimates how many units a study needs to detect a specified effect. Typical inputs include the expected effect size, outcome variability, significance level, desired power, and anticipated dropout. Estimates depend on assumptions, so teams should document them and revisit them if conditions change.

Question 12

Why does a study need a clearly defined unit of analysis?

Accepted Answer

The unit of analysis is the entity to which the outcome and exposure data apply, such as an individual patient, a clinic, or a household. Choosing the wrong unit can inflate the apparent number of independent observations and produce overly narrow confidence intervals. Records and analysis plans should state the unit explicitly.

Question 13

What is a multisite study, and what management challenge does it create?

Accepted Answer

A multisite study collects data from several locations that may use different staff, equipment, and workflows. The main management challenge is standardization, since small differences in procedures or definitions can create variation that looks like a real effect. Teams address this with shared manuals, training, periodic site monitoring, and data quality checks across locations.

Question 14

What is a pilot study, and what can it establish?

Accepted Answer

A pilot study is a test of study procedures, recruitment, forms, or data systems before a full study begins. It can establish feasibility, reveal workflow problems, and provide rough estimates of variability that inform sample size planning. A pilot is not designed to confirm an effect, so its results are not conclusive.

Question 15

Why are data quality checks built into the study plan rather than added at the end?

Accepted Answer

Built-in checks catch errors while they are still inexpensive to fix, such as missing entries, impossible values, or mismatched identifiers. Waiting until analysis often means that problems cannot be corrected, because participants or sites may no longer be available. Early planning also assigns responsibility for each review and issue log.

Question 16

What is the difference between a protocol deviation and a protocol amendment?

Accepted Answer

A protocol deviation is a departure from the approved plan that happened without prior approval, such as a visit held outside its window. A protocol amendment is a formal, approved change to the plan, documented and approved before implementation. Both should be logged, categorized, and reported, since deviations can affect analysis populations.

Question 17

What is a data dictionary, and why is it useful?

Accepted Answer

A data dictionary is a reference document that defines each variable, its format, allowed values, units, and coding rules. It helps staff collect and interpret data consistently, and it allows analysts to understand a dataset without relying on memory or informal conversations. Keeping it current is a basic records management practice.

Question 18

What are case report forms, and how are they used?

Accepted Answer

A case report form is a structured document, paper or electronic, used to record study data for each participant or unit. Well-designed forms use clear labels, controlled response options, and required fields to reduce entry errors. Forms should match the data dictionary and protocol so that information flows consistently from collection through analysis.

Question 19

What is a unique identifier, and what rules should govern its use?

Accepted Answer

A unique identifier is a code that distinguishes each record from every other record in a dataset. Identifiers should be assigned systematically, never reused, and kept separate from direct personal details whenever the study design allows. Access to any linking key should be restricted to authorized staff and documented under organizational policy.

Question 20

Why is audit trail documentation important for electronic records?

Accepted Answer

An audit trail is a time-stamped log that records who created, changed, or deleted a data entry and when the change occurred. It allows reviewers to reconstruct the history of a record and verify that edits were authorized and justified. Without it, organizations struggle to demonstrate data integrity during monitoring, inspections, or reviews.

Question 21

What is missing data, and why does the reason for missingness matter?

Accepted Answer

Missing data are values that were expected but not recorded for a unit. The reason matters because data missing at random behave differently from data missing for systematic reasons, such as participants with the worst outcomes leaving a study. Teams should track missingness patterns, document reasons when known, and plan analysis methods accordingly.

Question 22

What is data cleaning, and what distinguishes it from data analysis?

Accepted Answer

Data cleaning is the process of detecting and correcting errors, inconsistencies, and invalid entries in a dataset before analysis begins. Its goal is accuracy and consistency, not answering the research question. Cleaning decisions, such as how outliers or duplicates were handled, should be documented so that every change can be traced and explained.

Question 23

Why should data transfers between systems be validated?

Accepted Answer

Each transfer between systems can introduce changes such as truncated text, shifted dates, dropped records, or altered codes. Validation compares record counts, checksums, and sample values before and after transfer to confirm that content survived intact. Documenting the validation steps and results gives reviewers evidence that the data they analyze match the source.

Question 24

What is a source document, and how does it support verification?

Accepted Answer

A source document is the original record of an observation or event, such as a visit note or instrument printout, from which study data are drawn. Verification compares entries in the dataset against these originals to confirm accuracy and completeness. Linking entries to sources eases monitoring and reduces disputes about which value is correct.

Question 25

What is a statistical analysis plan, and when should it be finalized?

Accepted Answer

A statistical analysis plan is a detailed document describing the planned analyses, variables, populations, handling of missing data, and criteria for decisions. It should be finalized before outcome data are unblinded or examined for group comparisons. Writing it early discourages changes made after seeing results, which could otherwise undermine credibility.

Question 26

Why should analysis be scripted rather than performed by point and click?

Accepted Answer

Scripted analysis records every step in a file that can be rerun, reviewed, and shared. This makes results reproducible, lets colleagues check logic, and creates a clear audit trail when numbers change. Manual steps in spreadsheets are harder to verify and easier to alter accidentally, which increases the risk of undetected errors.

Question 27

What is double programming, and where is it used?

Accepted Answer

Double programming means two people independently write code to produce the same output, then compare the results to detect discrepancies. It is commonly applied to key tables, derived variables, and primary analyses where errors would be consequential. Disagreements are investigated and resolved by documented discussion rather than by quietly adopting one version.

Question 28

What is a data monitoring committee, and what does it review?

Accepted Answer

A data monitoring committee is an independent group that periodically reviews accumulating study data to protect participants and the integrity of the study. Its members are kept separate from the study team so that interim results do not influence conduct. The committee advises whether the study should continue, change, or stop.

Question 29

How can summary tables help detect data errors before analysis?

Accepted Answer

Summary tables show frequencies, ranges, and means for each variable, making unusual patterns easy to spot. A value far outside the expected range, an unexpected category, or a site with abrupt shifts often signals a recording or coding problem. Reviewing these tables routinely helps teams fix problems early and document each correction.

Question 30

What is a prespecified sensitivity analysis, and why is it planned in advance?

Accepted Answer

A sensitivity analysis checks whether conclusions hold under reasonable alternative assumptions, such as different handling of missing data or outliers. Planning it in advance shows that the team did not select the version that looked best after seeing results. Consistent conclusions across reasonable approaches strengthen confidence, while sharp disagreement signals fragile findings.

Question 31

What is reproducibility in analysis, and how is it supported?

Accepted Answer

Reproducibility means that someone else can rerun the same data and procedures and obtain the same results. It is supported by clear documentation, scripted workflows, version-controlled files, and a record of software versions and random seeds. Reproducibility is a minimum standard for trust, distinct from whether the conclusions are ultimately correct.

Question 32

What does a quality control checklist accomplish for an analysis?

Accepted Answer

A quality control checklist turns review expectations into specific, verifiable steps, such as confirming sample counts, checking variable definitions, and reviewing output labels. Using a checklist reduces reliance on memory and makes reviews consistent across team members and projects. Signed, dated checklists also provide evidence that required checks actually took place.

Question 33

What should a results report state about the population analyzed?

Accepted Answer

A results report should describe how many units were approached, enrolled or included, analyzed, and excluded, with reasons at each step. This accounting, often shown as a flow diagram, lets readers judge whether the analyzed group represents the target population. Omitting exclusions can make results look more favorable than the full data justify.

Question 34

What does it mean to report results with transparency about uncertainty?

Accepted Answer

Transparent reporting presents point estimates together with intervals or measures of variability, and states the limitations of the data and methods. It avoids presenting only statistically significant findings and discloses analyses that were exploratory. Readers should be able to tell which conclusions were planned and which emerged later.

Question 35

What is the role of a data use agreement?

Accepted Answer

A data use agreement is a contract that defines how a dataset may be accessed, used, shared, stored, and destroyed by the receiving party. It typically specifies permitted purposes, security obligations, restrictions on re-identification, and responsibilities for reporting breaches. Such agreements protect both parties by setting clear expectations before transfer.

Question 36

What is an institutional review board, and what is its function?

Accepted Answer

An institutional review board is a committee that reviews research involving human participants to confirm that it meets ethical standards and regulatory requirements. Its work includes reviewing protocols, consent materials, risk and benefit balance, and privacy safeguards, and it may require changes. Biostatistics staff contribute by explaining sample size justification and data handling plans.

Question 37

What do privacy protections such as de-identification mean for records?

Accepted Answer

De-identification removes or masks direct identifiers, such as names and contact details, and lowers the chance of linkage. It does not guarantee anonymity, because combinations of dates, locations, and rare characteristics can still re-identify individuals. Records managers should assess re-identification risk and limit access to any linking keys.

Question 38

What is overinterpretation, and how can a reporter avoid it?

Accepted Answer

Overinterpretation occurs when a report claims more than the design, sample, or analysis can support, such as implying causation from an observational association. Reporters avoid it by matching language to design, stating limitations, and distinguishing associations from causal claims. Peer review and internal checks help catch overstatement before publication or release.

Question 39

What is a data governance policy, and who typically is responsible for it?

Accepted Answer

A data governance policy defines who owns data, who may access it, how quality is maintained, and how decisions about data use are made. Responsibility is usually shared among senior leaders, data stewards, system owners, and statisticians, with clear roles written down. Good governance makes accountability visible and supports consistent handling across projects.

Question 40

What should an organization do when a data breach or quality error is discovered?

Accepted Answer

The organization should contain the problem, assess its scope and affected records, and notify responsible parties according to applicable policy and law. It should document the timeline, root cause, corrective actions, and who approved each step. Lessons learned should feed back into training, procedures, and system controls to prevent recurrence.

Biostatistics Management Flashcards

1Foundations of Biostatistics Management

2Study Design and Protocols

3Data Collection and Records

4Analysis Planning and Quality Control

5Reporting, Ethics, and Governance

Learning Resources