Data Management
What are the three basic tasks of data management discussed in Chapter 12?
Monitor missing data, check for out-of-range values, and monitor study identification numbers.
What are the three measures of central tendency?
Mean, median, and mode.
Approximately how many COVID-19 cases were not reported on time in England's 2020 reporting error?
15,841 positive cases, or nearly 16,000.
What is research dissemination?
Sharing research findings with audiences who can use them to inform practice, policy, or future research.
What is the main purpose of Step 5 of the CDC Program Evaluation Framework?
Generate and support conclusions using evaluation evidence.
What is the difference between a nominal variable and an ordinal variable? Give one example of each.
Nominal variables have categories without an order, such as clinic location. Ordinal variables have ordered categories, such as low, medium, and high risk.
Which graph is most appropriate for continuous numerical data, and which is best for categorical data?
A histogram for continuous data and a bar chart for categorical data.
What technical problem caused the COVID-19 cases to be omitted from reporting?
An Excel spreadsheet reached its row limit, causing records to be cut off during automated data processing.
Name two methods researchers can use to disseminate their findings.
Examples include peer-reviewed journals, conference presentations, research posters, and professional association meetings.
What is the main purpose of Step 6 of the CDC Program Evaluation Framework?
Act on findings by communicating results and using them to improve programs and decisions.
A dataset includes a patient age of 250 years. What type of data error is this, and what should researchers do?
An implausible or out-of-range value. Researchers should check the original record, apply validation rules, and document any correction.
Two communities have the same average blood pressure, but one has a much larger standard deviation. What does this tell us?
The community with the larger standard deviation has more variability in blood pressure, even though both have the same mean.
How did the delayed reporting of thousands of cases affect the appearance of England's COVID-19 trend?
When the missing cases were added together, the data showed an artificial spike that did not represent infections occurring on a single day.
Why should researchers communicate the same findings differently to scientists and policymakers?
Scientists may need detailed methods and statistics, while policymakers often need concise, accessible findings and actionable recommendations.
Why should stakeholder perspectives and program context be considered when interpreting evaluation findings?
They help explain the results, identify limitations, and ensure recommendations are relevant and feasible.
What is the difference between missing completely at random (MCAR) and missing at random (MAR)?
MCAR means missingness is unrelated to observed or unobserved data. MAR means missingness can be explained by observed variables. The distinction affects how missing data should be handled.
A dataset of emergency department waiting times is strongly right-skewed. Should researchers report the mean or median as the typical waiting time, and why?
The median is generally preferable because a few extremely long waiting times can pull the mean upward.
Identify three data validation procedures that could have detected the missing COVID-19 cases earlier.
Reconcile source and destination record counts, automatically flag unexpected drops or truncation, and check for missing IDs or records before publication.
A researcher writes, 'Our intervention proves that this program eliminates health disparities.' Which scientific writing principles are violated?
Avoid absolute claims, support conclusions with evidence, and ensure conclusions do not exceed the study's design and findings.
A community program achieves its participation target but shows little improvement in health outcomes. How should evaluators apply Step 5?
Compare findings with program goals, examine both process and outcome data, consider limitations and context, and develop evidence-supported recommendations.
A health department reports 99% data completeness but has duplicated IDs and systematically missing records from one community. Why might the findings still be biased?
Completeness alone does not guarantee accuracy or representativeness. Duplicates may distort counts, while systematic missingness can underrepresent a community and bias estimates.
Two populations have identical means and standard deviations but different distribution shapes. Why would reporting only these statistics be insufficient?
Mean and SD do not show skewness, outliers, multiple peaks, or the full distribution. Graphs are needed to identify patterns that summary statistics may hide.
If you were leading the response, how would you correct the reporting error while preventing misleading public health conclusions?
Recover and verify omitted cases, update affected reporting dates, disclose the error and its limitations, prioritize delayed tracing, and implement automated reconciliation and monitoring.
A study finds an association between healthcare access and better outcomes. How would you present this finding to policymakers without overstating causality?
State that better access was associated with improved outcomes, explain important limitations, and recommend feasible actions or further evaluation without claiming the association proves causation.
A study finds an association between healthcare access and better outcomes. How would you present this finding to policymakers without overstating causality?
State that better access was associated with improved outcomes, explain important limitations, and recommend feasible actions or further evaluation without claiming the association proves causation.