Healthcare Quality Analytics and Decision-Making

Introduction

Healthcare quality management increasingly depends on the ability to convert operational data into defensible decisions. Hospitals and health systems generate data through electronic health records, laboratory systems, patient-experience surveys, claims, incident reports, staffing systems, registries, and remote-monitoring devices. Data alone, however, do not constitute evidence of quality. Managers must distinguish signal from noise, compare like with like, identify inequities, assess risk, and select interventions whose benefits outweigh their costs and unintended consequences.

Quality analytics is the disciplined use of data, statistical reasoning, and visual communication to understand healthcare performance and guide improvement. It links measurement with action. Descriptive statistics summarize what has happened; inferential statistics help determine whether observed differences are likely to reflect more than random variation; control charts examine performance over time; risk adjustment supports fair comparisons; and dashboards translate complex evidence into usable management information.

Analytics should not be treated as a purely technical activity. The choice of denominator, case definition, time period, comparison group, and presentation format embeds values and assumptions. A technically precise measure can still produce poor decisions if it encourages gaming, ignores patient experience, or conceals disparities. Effective healthcare quality analytics therefore combines statistical competence, clinical understanding, operational knowledge, ethical governance, and managerial judgment.

Learning Objectives

By the end of this chapter, the learner should be able to:

1. Distinguish descriptive and inferential approaches to healthcare quality analysis.
2. Calculate and interpret common measures of central tendency, variation, rates, ratios, and proportions.
3. Select and interpret appropriate run charts and control charts.
4. Explain risk adjustment and its importance in comparing providers and populations.
5. Diagnose common and special causes of variation.
6. Design dashboards that support timely, equitable, and actionable decisions.
7. Evaluate data limitations, ethical risks, and the appropriate role of artificial intelligence in quality analytics.
8. Apply an integrated analytics framework to a healthcare quality problem.

1. From Measurement to Decision

A quality indicator is a defined measure of a structure, process, outcome, experience, or balancing result. Examples include nurse staffing hours, medication reconciliation completion, mortality, readmission, waiting time, falls, pressure injuries, and patient-reported communication quality.

A useful indicator has a clear numerator, denominator, population, observation period, data source, inclusion criteria, exclusion criteria, and reporting frequency. For example, “door-to-antibiotic time” is incomplete unless the organization specifies which patients are included, when the clock starts and stops, how transfers are handled, and whether the measure is a mean, median, or percentile.

The measurement-to-decision pathway is:

Question → operational definition → data acquisition → data validation → analysis → interpretation → decision → intervention → reassessment.

The central managerial question is not simply “Is the rate high or low?” It is “What decision should this evidence support, for whom, and with what degree of confidence?”

Types of quality analytics

• Descriptive analytics: summarizes observed data, such as monthly infection rates.
• Diagnostic analytics: investigates why performance differs, such as stratifying falls by ward and shift.
• Inferential analytics: estimates uncertainty and tests hypotheses about a wider population.
• Predictive analytics: estimates future risk, such as deterioration within 24 hours.
• Prescriptive analytics: compares possible actions and their expected consequences.

These categories overlap. A dashboard may be descriptive but can support diagnostic inquiry when it permits drill-down by unit, patient group, or time.

2. Descriptive Statistics for Quality Management

2.1 Counts, proportions, rates, and ratios

A count is the number of events. A proportion is the number of events divided by the relevant total, with the numerator included in the denominator. If 18 of 240 eligible patients receive a documented medication reconciliation, the proportion is:

Here, x is the number meeting the criterion and n is the number eligible. The result is dimensionless; multiplying by 100 expresses it as a percentage.

A rate incorporates a time or exposure denominator. For example, patient falls may be expressed as falls per 1,000 patient-days:

where k is a scaling constant, such as 1,000. If 12 falls occur during 4,800 patient-days, the rate is 2.5 falls per 1,000 patient-days. Rates are useful when exposure differs between periods or units. A ratio compares two quantities, which need not be part of the same whole, such as nurses per occupied bed.

A common error is to compare numerators without comparing denominators. A larger hospital may have more falls but a lower fall rate. Another error is to use an inappropriate exposure measure: outpatient visits, occupied bed-days, procedures, and device-days answer different questions.

2.2 Central tendency

The mean is:

where xᵢ is observation i and n is the number of observations. The mean uses every value but is sensitive to extreme observations.

The median is the middle ordered value. It is often preferable for length of stay, waiting time, or cost because these variables are commonly right-skewed. The mode is the most frequent value and is useful for categorical or discrete data.

Managers should state which summary is used. A mean waiting time of 42 minutes and a median of 25 minutes describe different operational realities; the difference indicates a long right tail affecting some patients.

2.3 Variation and distribution

The range is maximum minus minimum. The interquartile range (IQR) is:

where Q₁ and Q₃ are the 25th and 75th percentiles. The IQR contains the middle 50% of observations and is relatively resistant to outliers.

The sample variance is:

and the sample standard deviation is:

The denominator n − 1 is used when estimating population variability from a sample. Standard deviation is expressed in the same units as the original measure. The coefficient of variation is:

It is useful for comparing relative variability across measures with different scales, provided the mean is positive and meaningfully different from zero.

Worked example: waiting time

Suppose five patients wait 10, 12, 15, 18, and 45 minutes. The mean is 20 minutes, while the median is 15 minutes. The range is 35 minutes. The mean is pulled upward by the 45-minute wait; therefore, reporting only the mean would conceal the experience of most patients. A manager should examine the cause of the extreme delay while also reporting a percentile or median for typical performance.

2.4 Stratification and cross-tabulation

An aggregate rate can conceal important differences. Performance should commonly be stratified by ward, diagnosis, age, sex, ethnicity, language, disability, socioeconomic status, payer, shift, clinician, and risk category when ethically and statistically appropriate. Stratification can reveal inequity, Simpson’s paradox, or a concentration of events in a small subgroup.

Small denominators require caution. A change from 1/10 to 2/10 is a 10-percentage-point increase but is based on very limited evidence. Dashboards should display denominators and, where appropriate, suppress or flag estimates that could identify individuals.

3. Inferential Statistics and Uncertainty

Descriptive results concern the observed data. Inferential statistics address uncertainty when a sample is used to learn about a broader population or when observed groups are compared.

3.1 Confidence intervals

For a sample mean, an approximate confidence interval is:

where t* is a critical value determined by the desired confidence level and degrees of freedom, s is the sample standard deviation, and n is sample size. This interval assumes independent observations and, for small samples, an approximately normal sampling distribution or a suitable method of inference.

For a proportion, a simple large-sample approximation is:

where p̂ = x/n and z* is the standard-normal critical value. This approximation can be inaccurate with small samples or proportions near 0 or 1; Wilson or exact intervals are preferable in such cases.

A 95% confidence interval is not the probability that a fixed parameter lies in one already calculated interval. Rather, a method producing intervals in repeated samples would capture the parameter approximately 95% of the time under its assumptions.

3.2 Hypothesis tests and practical significance

A hypothesis test compares a null hypothesis with observed data. A p-value is the probability, under the null model and its assumptions, of obtaining results at least as extreme as those observed. It is not the probability that the null hypothesis is true, nor is it a measure of clinical importance.

Statistical significance can result from a very large sample and a trivial effect. Conversely, a clinically important effect may fail to reach conventional significance in a small sample. Managers should examine effect size, confidence interval, clinical meaning, cost, feasibility, and potential harm.

For two independent means, a common comparison is a t-test; for categorical proportions, a chi-square or exact test may be used. These tests require appropriate independence and distributional assumptions. Repeated observations from the same patient, unit, or clinician violate simple independence and may require paired methods, mixed-effects models, or cluster-robust analysis.

3.3 Association is not causation

A correlation or association does not establish that one variable caused another. Confounding, selection bias, reverse causation, measurement bias, and time trends can produce misleading relationships. Quality leaders should ask whether the data arise from an experiment, a quasi-experiment, or an observational study, and whether the analytic design addresses plausible alternative explanations.

Interrupted time-series analysis, difference-in-differences, matched cohorts, and controlled before-and-after designs can strengthen evaluation of improvement interventions, but each has assumptions that must be examined.

4. Run Charts and Control Charts

4.1 Run charts

A run chart plots data in time order, usually with a median reference line. It helps identify shifts, trends, cycles, and unusually long sequences above or below the median. Time order is essential because randomly reordered observations can destroy evidence of process change.

A run is a sequence of points on one side of the median; points exactly on the median are generally omitted when counting runs. Formal run rules depend on the number of observations and the chosen method. A run chart is especially useful early in an improvement project when the process baseline is not yet stable.

4.2 Control charts

A control chart adds statistically derived control limits to a time-ordered process measure. The center line estimates the process average, while upper and lower control limits indicate the expected range of common-cause variation under a stable process.

For individual measurements with a stable process, limits are often estimated using the moving range. If MR̄ is the average of successive absolute differences, an estimate of standard deviation is:

where d₂ is a control-chart constant dependent on subgroup or moving-range size; for a two-observation moving range, d₂ is approximately 1.128. Limits are then:

For proportions, a p-chart uses:

and, for subgroup i,

Limits below 0 are set to 0. Here xᵢ is the number of events, nᵢ is the number eligible, and p̄ is the overall proportion. These limits assume independent Bernoulli observations within a suitable process framework. Overdispersion, autocorrelation, changing definitions, or very small denominators may make the conventional chart inappropriate.

Other charts include c-charts for counts with constant opportunity, u-charts for counts per unit with varying opportunity, and X-bar/R or X-bar/S charts for rational subgroups of continuous measurements. Chart selection must reflect the data type and sampling structure.

4.3 Common and special causes

Common-cause variation is produced by the system: staffing patterns, workflow design, equipment, policies, and ordinary environmental conditions. It calls for system-level redesign. Special-cause variation is attributable to a specific, unusual, or identifiable influence, such as a temporary closure, data-entry error, outbreak, or new technology. It calls for investigation and targeted correction.

Overreacting to common-cause variation creates tampering: managers repeatedly alter a stable process, increasing instability. Conversely, ignoring a special cause allows a correctable problem to persist. A control chart should therefore be interpreted with process knowledge, not mechanically.

5. Risk Adjustment and Fair Comparison

Raw outcome comparisons can be misleading when patient populations differ. A tertiary referral center may treat more severely ill patients than a community facility. Risk adjustment attempts to account for differences in baseline risk so that comparisons more closely reflect quality rather than case mix.

A logistic model for a binary outcome can be expressed as:

where p is the probability of the outcome, β₀ is the intercept, βⱼ are coefficients, and Xⱼ are predictor variables such as age, comorbidity, physiological status, or diagnosis. The model assumes an appropriate functional form, adequate data quality, independent or appropriately clustered observations, and predictors measured before the outcome where causal interpretation is intended.

A commonly used summary is the standardized mortality ratio:

An SMR of 1 indicates observed deaths equal expected deaths under the reference model; values above or below 1 indicate more or fewer deaths than expected. The SMR is not automatically a causal measure of hospital quality. Its interpretation depends on the reference population, model calibration, coding practices, and uncertainty.

Risk adjustment can use direct standardization, indirect standardization, regression, matching, or hierarchical models. Models should avoid adjusting for variables that occur after the care process being evaluated, because doing so may remove part of the quality signal. For example, adjusting for a complication caused by treatment could unfairly hide poor performance.

Risk adjustment does not solve inequity. If social deprivation, language barriers, disability, or structural access barriers are omitted, organizations serving disadvantaged populations may be penalized. If they are included without careful interpretation, legitimate disparities may be normalized. Analysts should report both adjusted and unadjusted results where useful, explain the model, assess calibration and discrimination, and examine subgroup performance.

6. Variation Analysis and Root-Cause Inquiry

Variation analysis combines statistical evidence with process knowledge. A practical sequence is:

1. Define the outcome and operational process.
2. Verify data completeness, accuracy, timeliness, and consistency.
3. Plot performance over time.
4. Identify changes, clusters, trends, and outliers.
5. Stratify by relevant clinical and operational factors.
6. Develop hypotheses about causes.
7. Test hypotheses using chart review, observation, interviews, or further analysis.
8. Implement a proportionate intervention and reassess.

Useful tools include Pareto charts, process maps, fishbone diagrams, scatterplots, failure mode and effects analysis, and statistical process control. A Pareto display may show that a small number of failure categories account for most delays, but the 80/20 pattern is a heuristic, not a law.

A cause should not be accepted merely because it is plausible. Analysts should seek triangulation: agreement among administrative data, clinical records, direct observation, and staff or patient accounts. The investigation should distinguish proximate causes from latent system conditions such as poor interface design, unclear accountability, workload, training gaps, or incompatible incentives.

7. Dashboards and Data Visualization

A dashboard is a visual interface that presents selected measures for a defined audience and decision cycle. A board dashboard, executive dashboard, service-line dashboard, and frontline huddle board have different purposes and levels of detail.

A good dashboard has:

• a named decision owner;
• a clear purpose and action threshold;
• stable definitions and visible denominators;
• appropriate time granularity;
• trend information, not only a single score;
• targets and control limits distinguished from one another;
• drill-down capability without overwhelming the user;
• stratification for equity and operational learning;
• data refresh and validation information;
• accessibility for users with visual or cognitive impairments.

Use line charts for trends, bars for category comparisons, dot plots for distributions, and scatterplots for relationships. Avoid three-dimensional charts, unnecessary decoration, truncated axes that exaggerate differences, and color schemes that rely only on red-green distinctions. A red status should not mean merely “below target” if the process is statistically stable; a target is a managerial aspiration, whereas a control limit describes process behavior.

Dashboard design should begin with decisions rather than available data. For example, a nurse manager deciding whether to redesign discharge workflow needs daily throughput, queue time, staffing, rework, patient experience, and balancing measures—not a collection of unrelated metrics.

Leading, lagging, and balancing measures

Lagging measures such as mortality or readmission are important but may respond slowly. Leading measures, such as completion of medication reconciliation or timely escalation, may indicate whether a process is changing. Balancing measures protect against unintended effects; reducing length of stay, for example, should be monitored alongside emergency returns, patient understanding, and post-discharge harm.

8. An Integrated Decision Framework

A disciplined quality decision can be organized around five questions:

1. What is happening? Use valid definitions, descriptive summaries, visualization, and stratification.
2. Is the change real? Examine confidence intervals, run rules, control limits, and data-quality changes.
3. Why might it be happening? Investigate process, context, case mix, and competing explanations.
4. What should be done? Select an intervention linked to a hypothesized mechanism and assess feasibility, cost, equity, and risk.
5. Did it work without harm? Use outcome, process, and balancing measures with sufficient follow-up.

A useful decision matrix combines statistical signal with operational impact. A statistically unusual result with low practical importance may require monitoring. A clinically serious event with little statistical power may require immediate action despite uncertainty. Statistical thresholds should never replace patient-safety escalation rules.

9. Illustrative Case: Reducing Emergency Department Length of Stay

Background

This fictional case illustrates the application of healthcare quality analytics. A 350-bed hospital reports a median emergency department (ED) length of stay of 238 minutes. Senior leaders want to reduce delays without increasing left-without-being-seen visits, diagnostic errors, or unplanned returns.

Problem definition

The team defines ED length of stay as minutes from registration to physical departure for discharged patients. Patients admitted to inpatient units are analyzed separately because bed availability creates a distinct process. The team also specifies inclusion criteria, excludes cancelled encounters, and validates timestamps against a sample of records.

Analysis

Monthly medians alone show modest improvement, but a run chart reveals a sustained shift after a new triage process. A control chart for individual daily median values shows two points above the upper limit during a period of laboratory instrument failure. The team does not treat those points as evidence that the entire process has deteriorated; it investigates the special cause.

Stratification shows that delays are concentrated among patients awaiting imaging and among admitted patients awaiting inpatient beds. A raw comparison by physician suggests one clinician has longer times, but the apparent difference disappears after stratifying by high-acuity patients and time of arrival. This demonstrates why crude comparisons can mislead.

The team reports the following baseline results:

Approach and solution

A cross-functional team maps registration, triage, diagnostics, consultation, and discharge. It introduces an early senior clinical assessment for selected patients, a radiology escalation protocol, and a predicted-bed-status communication process. The intervention is tested first on weekday evening shifts using a Plan–Do–Study–Act cycle.

The dashboard displays daily median and 90th-percentile stay, left-without-being-seen rate, unplanned returns, diagnostic turnaround time, and staffing. Data are shown with control limits and annotated with intervention dates. Patient comments are reviewed alongside numerical measures.

Outcome

After twelve weeks, the median falls to 205 minutes and the 90th percentile to 365 minutes. The left-without-being-seen rate remains 2.0%, while seven-day returns fluctuate within the previous range. The control chart suggests a sustained shift rather than a single favorable observation, but the team continues monitoring because seasonal demand and staffing changes may affect results.

Learning points

The case illustrates that: (1) operational definitions precede valid analysis; (2) medians and percentiles reveal different aspects of delay; (3) control charts distinguish a process shift from random fluctuation; (4) stratification prevents misleading attribution; (5) balancing measures protect against harmful improvement; and (6) analytics is most useful when linked to a testable process hypothesis and a decision owner.

10. Governance, Ethics, and Regulation

Quality data are often linked to identifiable health information. Governance should specify lawful use, minimum necessary access, retention, audit trails, role-based permissions, data-sharing agreements, and procedures for correcting errors. Analysts should protect confidentiality in small cells and avoid publishing combinations of variables that permit re-identification.

Data quality has ethical significance. Missingness may be systematically higher for marginalized groups. A dashboard that excludes undocumented, uninsured, or poorly connected patients can produce a falsely favorable picture. Analysts should describe missing data, assess whether missingness differs across groups, and avoid silently replacing missing values with assumptions.

Performance measurement can generate gaming, such as avoiding high-risk patients, reclassifying cases, or documenting completion without meaningful care. Governance should use multiple measures, periodic audits, qualitative review, and patient-centered outcomes. Transparency about definitions and limitations supports trust.

11. Technology and Artificial Intelligence

Modern platforms can automate extraction, natural-language processing, anomaly detection, forecasting, and risk prediction. Artificial intelligence may help identify patients at risk of deterioration or detect documentation patterns associated with harm. However, a model can reproduce historical inequities, perform poorly after workflow changes, or create alert fatigue.

Before deployment, organizations should evaluate discrimination, calibration, discrimination across subgroups, drift, explainability appropriate to the use case, cybersecurity, human oversight, and the consequences of false positives and false negatives. A model used for prioritization should be evaluated not only by area under a receiver-operating-characteristic curve but also by decision consequences, workload, equity, and patient outcomes. Predictive performance does not prove that acting on a prediction improves care.

12. Limitations and Best Practices

Quality analytics is limited by incomplete documentation, inconsistent definitions, changing coding systems, measurement error, selection bias, confounding, small samples, autocorrelation, regression to the mean, and pressure to meet targets. Improvement may appear after an intervention simply because an extreme baseline value naturally moves closer to the average.

Best practices include:

1. Begin with a decision and a clearly defined population.
2. Use a data dictionary and version-controlled measure specifications.
3. Validate numerator, denominator, timestamps, exclusions, and missingness.
4. Display counts and denominators with percentages.
5. Use time-series methods for time-series questions.
6. Distinguish statistical control limits from policy targets.
7. Report uncertainty and practical importance.
8. Stratify for clinical relevance and equity while protecting privacy.
9. Combine quantitative evidence with clinical and patient knowledge.
10. Monitor balancing measures and unintended consequences.
11. Assign ownership for action, review, and escalation.
12. Reassess measures periodically to prevent obsolete or gameable performance systems.

Chapter Summary

Healthcare quality analytics transforms data into evidence for managerial and clinical decisions. Descriptive statistics summarize counts, proportions, rates, central tendency, and variation, while inferential methods quantify uncertainty and compare groups under stated assumptions. Run charts reveal time-based patterns; control charts distinguish common-cause system variation from special causes requiring investigation. Risk adjustment supports more equitable comparisons but depends on valid models, appropriate predictors, and careful interpretation. Stratification exposes hidden variation and disparities, while dashboards organize information around decisions, thresholds, trends, and balancing measures. Effective analytics integrates statistical reasoning with process knowledge, ethical governance, patient perspectives, and continuous reassessment. Artificial intelligence can extend analytic capacity, but it requires validation, equity monitoring, human oversight, and clear accountability.

Key Terms

Analytics: Systematic use of data and methods to generate insight for action.

Balancing measure: Measure used to detect unintended consequences of an intervention.

Confidence interval: Range produced by a statistical procedure to express uncertainty about an estimate.

Control chart: Time-series chart with a center line and statistically derived control limits.

Common-cause variation: Routine variation generated by the system or process.

Dashboard: Visual display of selected measures for a defined audience and decision purpose.

Denominator: Population or exposure used as the basis for a quality measure.

Inferential statistics: Methods for estimating or testing claims beyond the observed sample.

Interquartile range: Difference between the 75th and 25th percentiles.

Median: Middle value of an ordered dataset.

Numerator: Events or observations counted in a quality measure.

P-value: Probability, under a null model, of results at least as extreme as those observed.

Proportion: Part divided by the whole, with the numerator included in the denominator.

Rate: Frequency of events relative to a specified exposure or time base.

Risk adjustment: Statistical adjustment for differences in baseline patient or population risk.

Run chart: Time-ordered plot commonly using a median reference line.

Special-cause variation: Unusual, identifiable influence producing process variation.

Standard deviation: Measure of dispersion around the mean, in the original units.

Statistical significance: Evidence against a null model according to a specified statistical criterion.

Variation: Differences in observations across patients, units, or time.

Review Questions

Short Answer Questions

1. Distinguish a proportion, rate, and ratio using one healthcare example of each.
2. Why might the median be preferred to the mean for emergency department waiting time?
3. What is the difference between common-cause and special-cause variation?
4. State two reasons why risk adjustment is used in healthcare comparisons.
5. Identify three characteristics of a decision-useful quality dashboard.

Long Answer Questions

1. Explain how descriptive and inferential statistics complement one another in healthcare quality management.
2. Describe the construction, interpretation, and limitations of a p-chart.
3. Critically discuss the benefits and risks of risk-adjusted provider comparisons.
4. Explain how stratification, time-series analysis, and root-cause inquiry can be integrated in an improvement project.
5. Discuss the ethical and governance requirements for using artificial intelligence in quality analytics.

Analytical / Application Questions

1. A unit records 9 falls over 3,600 patient-days in one month and 12 falls over 4,800 patient-days the next month. Calculate the fall rate per 1,000 patient-days for each month and explain whether the raw counts alone support a conclusion of deterioration.
2. A hospital’s mean length of stay decreases after a discharge intervention, but readmissions increase. Design a dashboard and evaluation approach that would determine whether the intervention represents genuine improvement.
3. A tertiary hospital has higher crude mortality than a community hospital but treats substantially sicker patients. Propose an analysis plan addressing risk adjustment, uncertainty, equity, and the limits of causal interpretation.

Multiple Choice Questions with Answers

Question 1

Which measure is generally most appropriate for a highly right-skewed healthcare waiting-time distribution?

A. Median

B. Mode only

C. Maximum

D. Range only

Correct Answer: A. Median

Question 2

A unit has 12 falls during 4,800 patient-days. What is the rate per 1,000 patient-days?

A. 0.25

B. 2.5

C. 25

D. 250

Correct Answer: B. 2.5

Question 3

What is the primary purpose of a control chart?

A. To rank clinicians by performance

B. To replace clinical judgment

C. To distinguish common-cause from special-cause variation

D. To guarantee that a process meets a target

Correct Answer: C. To distinguish common-cause from special-cause variation

Question 4

In a risk-adjustment model, which variable is generally most problematic to adjust for when evaluating the quality of the treatment process?

A. Age at admission

B. Baseline comorbidity

C. Pre-treatment physiological status

D. A complication caused by the treatment being evaluated

Correct Answer: D. A complication caused by the treatment being evaluated

Question 5

A proportion is calculated as:

A. Events divided by eligible total

B. Exposure divided by events

C. Mean divided by standard deviation

D. Observed deaths divided by expected deaths

Correct Answer: A. Events divided by eligible total

Question 6

What does an SMR of 1 indicate, before considering uncertainty and model validity?

A. No deaths occurred

B. Observed deaths equal expected deaths

C. The hospital has perfect quality

D. The outcome is statistically significant

Correct Answer: B. Observed deaths equal expected deaths

Question 7

Which is a balancing measure when reducing hospital length of stay?

A. Number of discharge planners

B. Number of beds in the hospital

C. Unplanned readmissions

D. Number of dashboard users

Correct Answer: C. Unplanned readmissions

Question 8

A p-value is best described as:

A. The probability that the null hypothesis is true

B. The probability that the intervention is clinically important

C. The percentage of patients receiving the intervention

D. The probability, under a null model, of results at least as extreme as those observed

Correct Answer: D. The probability, under a null model, of results at least as extreme as those observed

Question 9

Which visualization is most suitable for showing performance over successive months?

A. Line chart

B. Pie chart with three-dimensional effects

C. Single numerical card without a time period

D. Alphabetical list

Correct Answer: A. Line chart

Question 10

If the mean waiting time is 20 minutes and the standard deviation is 5 minutes, what is the coefficient of variation?

A. 4%

B. 20%

C. 25%

D. 400%

Correct Answer: B. 20%