Epidemiology and Public Health Statistics

Why This Topic Matters for the CSP Exam

Epidemiology and public health statistics sit inside the BCSP Certified Safety Professional (CSP) blueprint under the Occupational Health and Applied Science domain. The reason BCSP tests this material is simple: a safety professional cannot manage what cannot be measured, and cannot defend a control decision without evidence. Epidemiology gives you the tools to determine whether an exposure (noise, silica, repetitive motion, shift work, a new chemical) is actually associated with an outcome (hearing loss, silicosis, carpal tunnel syndrome, fatigue-related injuries). Statistics gives you the tools to decide whether a difference you observe is real or simply random variation.

On the exam you will see this content in several forms: calculating rates and risk ratios, interpreting study results, choosing the correct study design for a question, recognizing bias and confounding, reading sensitivity and specificity of a screening test, and applying descriptive statistics to injury or exposure data. Because many questions are quantitative, this is one of the areas where disciplined practice pays off most directly in points. It also overlaps heavily with other domains: OSHA recordkeeping rates (TRIR, DART), industrial hygiene sampling statistics, risk assessment, and program evaluation all rely on the same foundation.

What Epidemiology Is

Epidemiology is the study of the distribution and determinants of health-related states or events in specified populations, and the application of that study to control health problems. The key words are:

Distribution – who gets the disease or injury, where, and when (person, place, time).
Determinants – the factors (exposures, behaviors, environments) that influence the occurrence of the outcome.
Populations – epidemiology is about groups, not individuals. A clinician asks why this worker got sick; an epidemiologist asks why this workforce has more cases than expected.

Occupational epidemiology applies these methods to the workplace. Classic examples the exam may reference include Percival Pott linking chimney soot to scrotal cancer in chimney sweeps (1775), Alice Hamilton's investigations of lead and industrial poisons, Irving Selikoff's asbestos studies among insulation workers, and the vinyl chloride and angiosarcoma of the liver cluster that led to the OSHA vinyl chloride standard.

Core Measures of Disease and Injury Frequency

Counts – The raw number of cases. Useful for resource planning but poor for comparison because they ignore population size.

Ratio – One quantity divided by another where the numerator is not necessarily part of the denominator (e.g., male-to-female cases, 3:1).

Proportion – A ratio in which the numerator is included in the denominator, often expressed as a percentage (e.g., 15 of 300 workers = 5%).

Rate – A measure of how fast events occur in a population over time. A true rate includes a time dimension in the denominator.

Prevalence – The proportion of a population that has a condition at a given point (point prevalence) or during a period (period prevalence). Formula: Prevalence = Existing cases / Total population at that time. Prevalence is a snapshot; it is influenced by both how many new cases occur and how long people remain cases. Prevalence is approximately Incidence × Duration when conditions are stable.

Incidence – The occurrence of new cases over a period of time. Two forms:
• Cumulative incidence (incidence proportion or risk) = New cases during period / Population at risk at start of period. Example: 12 new cases of dermatitis among 400 workers in one year = 3% annual risk.
• Incidence rate (incidence density) = New cases / Total person-time at risk. Example: 12 new cases over 3,800 person-years = 3.16 per 1,000 person-years. Person-time accounts for workers who join, leave, or are followed for different durations.

Exam distinction: if a question says new cases, think incidence; if it says existing cases or how many currently have, think prevalence. A chronic, long-lasting disease (like noise-induced hearing loss) can have high prevalence even with modest incidence because cases accumulate. A rapidly fatal or quickly cured condition can have high incidence but low prevalence.

Mortality rate – Deaths per population per time. Case fatality rate – Deaths among those who have the disease divided by number of cases, expressed as a percentage; it measures severity, not frequency. Attack rate – Cumulative incidence used in outbreak settings, often for acute events (e.g., a foodborne illness at a plant cafeteria).

Standardization – When comparing populations with different age or sex structures, crude rates mislead. Direct standardization applies the study population's age-specific rates to a standard population. Indirect standardization produces the Standardized Mortality Ratio (SMR) = Observed deaths / Expected deaths × 100. An SMR of 150 means 50% more deaths than expected. SMRs are common in occupational cohort studies. Be aware of the healthy worker effect: employed populations are generally healthier than the general population, so SMRs below 100 are common even when a real occupational hazard exists. The healthy worker effect is a form of selection bias that tends to underestimate occupational risk.

Measures of Association

Most occupational epidemiology questions on the CSP exam revolve around a 2×2 table. Learn this layout cold:

Exposed with disease = a
Exposed without disease = b
Unexposed with disease = c
Unexposed without disease = d

Relative Risk (Risk Ratio, RR) – Used in cohort studies and experimental studies where incidence can be calculated.
RR = Incidence in exposed / Incidence in unexposed = [a/(a+b)] / [c/(c+d)].
Interpretation: RR = 1 means no association; RR greater than 1 means exposure increases risk; RR less than 1 means exposure is protective. RR = 2.5 means exposed workers have 2.5 times the risk, or 150% higher risk.

Odds Ratio (OR) – Used in case-control studies, where incidence cannot be calculated because the investigator selects by disease status.
OR = (a × d) / (b × c).
Interpretation parallels RR. When the disease is rare (roughly under 10% in the population), the OR approximates the RR.

Attributable Risk (Risk Difference, AR) – The absolute excess risk due to exposure.
AR = Incidence in exposed − Incidence in unexposed.
Tells you how many cases per unit population would be prevented if exposure were eliminated.

Attributable Risk Percent (Attributable Fraction among the exposed) = (RR − 1) / RR × 100, or (Incidence exposed − Incidence unexposed) / Incidence exposed × 100. The percentage of cases among the exposed that is due to the exposure.

Population Attributable Risk (PAR) = Incidence in total population − Incidence in unexposed. The excess in the whole population attributable to exposure; depends on how common the exposure is.

Worked Example
A cohort of 500 welders and 500 non-welding fabricators is followed for five years. Forty welders and 10 fabricators develop a respiratory condition.
Incidence exposed = 40/500 = 0.08 (8%).
Incidence unexposed = 10/500 = 0.02 (2%).
RR = 0.08 / 0.02 = 4.0. Welders have four times the risk.
AR = 0.08 − 0.02 = 0.06, or 60 excess cases per 1,000 welders over five years.
AR% = (4 − 1)/4 = 75%. Three-quarters of the welders' cases are attributable to welding exposure.
If this were a case-control study with the same cell values (a=40, b=460, c=10, d=490), OR = (40×490)/(460×10) = 19,600/4,600 = 4.26, close to the RR because the outcome is relatively uncommon.

Study Designs You Must Recognize

Descriptive studies – Case reports, case series, and ecological (correlational) studies. Ecological studies compare group-level data (e.g., cancer rates by county versus industrial emissions by county). They generate hypotheses but suffer from the ecological fallacy: an association at the group level may not hold at the individual level.

Cross-sectional (prevalence) study – Exposure and outcome are measured at the same time in a population. Fast and inexpensive, good for prevalence, but cannot establish temporal sequence (did the exposure precede the disease?). Common in workplace health surveys.

Case-control study – Start with people who have the disease (cases) and people who do not (controls), then look backward at exposure history. Efficient for rare diseases and diseases with long latency; relatively quick and inexpensive; can examine many exposures. Weaknesses: recall bias, difficulty selecting appropriate controls, cannot directly calculate incidence or RR, so you use the OR.

Cohort study – Start with exposed and unexposed groups who are free of the disease and follow them forward for outcomes. Prospective cohorts follow from the present into the future; retrospective (historical) cohorts use existing records to reconstruct exposure and follow-up that have already occurred, which is very common in occupational settings using employment and death records. Strengths: establishes temporality, calculates incidence and RR, can study multiple outcomes from one exposure. Weaknesses: expensive, time-consuming, loss to follow-up, inefficient for rare diseases.

Experimental studies (randomized controlled trials, intervention studies) – Investigator assigns exposure or intervention, ideally at random. Strongest design for causal inference because randomization balances known and unknown confounders. Rarely used for harmful occupational exposures for ethical reasons, but used to evaluate interventions such as training programs, ergonomic redesigns, or hearing protection strategies.

Hierarchy of evidence (strongest to weakest for causation): randomized controlled trial, prospective cohort, retrospective cohort, case-control, cross-sectional, ecological, case series, case report, expert opinion. Systematic reviews and meta-analyses pool results across multiple studies and are often placed at the top.

Validity: Bias, Confounding, and Chance

When an association is observed, there are four possible explanations: it is causal, it is due to bias, it is due to confounding, or it is due to chance. The exam frequently tests your ability to name the problem.

Selection bias – Systematic error in how participants are chosen or retained, so the study population does not represent the target population. Examples: the healthy worker effect; volunteer bias; differential loss to follow-up; choosing hospital controls that differ systematically from cases.

Information (measurement) bias – Systematic error in how exposure or outcome is measured. Examples: recall bias (cases remember exposures more thoroughly than controls); interviewer bias (the interviewer probes exposed subjects more aggressively); misclassification (assigning workers to the wrong exposure category). Non-differential misclassification generally biases results toward the null (makes the association look weaker). Differential misclassification can bias in either direction.

Confounding – A third variable that is associated with both the exposure and the outcome and is not on the causal pathway distorts the apparent relationship. The classic occupational example: smoking confounds the relationship between a dusty job and lung cancer if smoking is more common among those workers. Confounding is controlled in design by randomization, restriction, or matching, and in analysis by stratification or multivariable regression (adjusted rates). If adjusting for a variable substantially changes the RR, that variable was a confounder.

Effect modification (interaction) – Different from confounding. The effect of the exposure differs by level of a third variable. Example: asbestos and smoking interact so that the lung cancer risk in smokers exposed to asbestos is multiplicatively greater than either alone. Effect modification is a real finding to report, not a nuisance to remove.

Chance – Random error, addressed through statistical testing and confidence intervals, discussed below.

Causation: The Bradford Hill Criteria

Association is not causation. Sir Austin Bradford Hill proposed considerations for judging causality, and the exam may ask you to identify or apply them:

1. Strength – Larger RR or OR is more convincing.
2. Consistency – Repeated findings across different populations, places, and investigators.
3. Specificity – One exposure leads to one particular outcome (weakest criterion; many exposures cause multiple diseases).
4. Temporality – Exposure must precede the outcome. This is the only absolutely required criterion.
5. Biological gradient (dose-response) – More exposure yields more disease.
6. Plausibility – A biological mechanism makes sense.
7. Coherence – Fits with known natural history and biology of the disease.
8. Experiment – Removing exposure reduces disease.
9. Analogy – Similar exposures cause similar outcomes.

Screening and Diagnostic Test Characteristics

Medical surveillance programs (audiometry, spirometry, blood lead monitoring) depend on screening tests. You need another 2×2 table, this time comparing test result to true disease status:

True positive (TP) = disease present, test positive
False positive (FP) = disease absent, test positive
False negative (FN) = disease present, test negative
True negative (TN) = disease absent, test negative

Sensitivity = TP / (TP + FN). The proportion of people with the disease who test positive. A highly sensitive test rarely misses disease, so a negative result on a sensitive test effectively rules disease out. Memory aid: SnNout.

Specificity = TN / (TN + FP). The proportion of people without the disease who test negative. A highly specific test rarely mislabels healthy people, so a positive result on a specific test effectively rules disease in. Memory aid: SpPin.

Positive Predictive Value (PPV) = TP / (TP + FP). Given a positive test, the probability the person truly has the disease. PPV rises as prevalence rises.

Negative Predictive Value (NPV) = TN / (TN + FN). Given a negative test, the probability the person is truly disease-free. NPV falls as prevalence rises.

Key exam concept: sensitivity and specificity are properties of the test and do not change with prevalence; predictive values depend heavily on prevalence. Screening a low-prevalence workforce with even a good test yields many false positives and a low PPV. This is why initial screening often uses a sensitive test and positives are confirmed with a more specific test.

Worked Example
A screening test for a condition is given to 1,000 workers; 50 truly have the condition. The test identifies 45 of them (TP=45, FN=5) and falsely flags 95 healthy workers (FP=95, TN=855).
Sensitivity = 45/50 = 90%.
Specificity = 855/950 = 90%.
PPV = 45/140 = 32%. Only one in three positives actually has the condition.
NPV = 855/860 = 99.4%.

Workplace Injury and Illness Rates

These are applied epidemiology and are heavily tested. OSHA rates are normalized to 200,000 hours, representing 100 full-time employees working 40 hours per week for 50 weeks.

Total Recordable Incident Rate (TRIR or TCIR) = (Number of OSHA recordable cases × 200,000) / Total employee hours worked.

DART Rate = (Number of cases with Days Away, Restricted duty, or job Transfer × 200,000) / Total hours worked.

Lost Time Case Rate (LTCR / DAFWII) = (Cases with days away from work × 200,000) / Total hours worked.

Severity Rate = (Total days away or restricted × 200,000) / Total hours worked. Some organizations use 1,000,000 hours (the older ANSI Z16.1 convention); always use the constant the question gives you.

Worked Example: A plant with 350 employees worked 712,000 hours last year and recorded 9 recordable injuries, 4 of which involved days away or restriction totaling 86 days. TRIR = (9 × 200,000)/712,000 = 2.53. DART = (4 × 200,000)/712,000 = 1.12. Severity = (86 × 200,000)/712,000 = 24.2 days per 100 full-time workers.

Remember that these are lagging indicators and, for small populations, are highly unstable; a single injury in a 20-person shop can swing the TRIR by several points. Statistical thinking tells you to compare with caution and look at confidence intervals or multi-year averages.

Descriptive Statistics

Measures of central tendency:
• Mean – arithmetic average; sensitive to outliers.
• Median – middle value when data are ordered; robust to outliers; preferred for skewed data such as lost workdays or exposure concentrations.
• Mode – most frequent value.
In a right-skewed distribution (long tail to the high side, typical of exposure and claim-cost data), the mean exceeds the median, which exceeds the mode. In a perfectly normal distribution all three coincide.

Measures of dispersion:
• Range – maximum minus minimum.
• Variance – average of squared deviations from the mean.
• Standard deviation (SD) – square root of variance; same units as the data.
• Coefficient of variation (CV) = SD / Mean, often as a percentage; lets you compare variability across different scales. Used in industrial hygiene to describe sampling and analytical error.
• Geometric mean and geometric standard deviation (GSD) – appropriate for lognormal data such as airborne contaminant concentrations. A GSD near 1 indicates little variability; GSD above about 3 indicates highly variable exposures.

The normal distribution: symmetrical and bell-shaped. Approximately 68% of values fall within ±1 SD of the mean, 95% within ±2 SD (more precisely 1.96), and 99.7% within ±3 SD. A z-score = (value − mean)/SD tells you how many standard deviations a value lies from the mean. Many exam questions ask you to find the percentage of workers above or below a threshold using these rules.

Example: Noise exposures across a department are normally distributed with a mean of 84 dBA and SD of 3 dBA. What proportion exceed 90 dBA? z = (90−84)/3 = 2.0. About 2.5% lie above +2 SD, so roughly 2.5% of workers exceed the PEL.

Inferential Statistics

Population vs. sample – You measure a sample to infer about a population. The standard error of the mean = SD / square root of n describes how much sample means vary; larger samples give more precise estimates.

Confidence interval (CI) – A range that, with a stated probability (usually 95%), contains the true population value. For a mean: Mean ± 1.96 × (SD/√n). For an RR or OR, if the 95% CI includes 1.0, the association is not statistically significant at the 0.05 level. If the CI is entirely above 1.0 (e.g., 1.4 to 3.2), the exposure is significantly associated with increased risk. A narrow CI indicates precision; a wide CI indicates a small sample or high variability.

Hypothesis testing – The null hypothesis (H0) states there is no difference or no association. The alternative hypothesis (H1) states there is. The p-value is the probability of observing data at least as extreme as what you found if the null hypothesis were true. If p is less than the chosen alpha level (commonly 0.05), you reject the null and call the result statistically significant. A p-value does not measure the size or importance of an effect, and statistical significance is not the same as practical significance.

Type I error (alpha) – Rejecting a true null hypothesis; a false positive; concluding an intervention works when it does not.
Type II error (beta) – Failing to reject a false null hypothesis; a false negative; missing a real effect.
Power = 1 − beta; the probability of detecting a true effect. Power increases with larger sample size, larger true effect size, and higher alpha. Many small workplace studies are underpowered, which is why a non-significant finding does not prove safety.

Common tests and when they apply:
• t-test – compares means of two groups (independent samples) or before-and-after measurements on the same subjects (paired).
• ANOVA – compares means across three or more groups.
• Chi-square test – compares proportions or frequencies among categories; the natural test for a 2×2 table (e.g., injured vs. not injured by shift).
• Correlation coefficient (r) – measures the strength and direction of a linear relationship between two continuous variables, ranging from −1 to +1. r² (coefficient of determination) is the proportion of variance explained. Correlation never proves causation.
• Regression – models an outcome as a function of one or more predictors; logistic regression is used for binary outcomes and yields adjusted odds ratios.
• Nonparametric tests (Mann-Whitney, Wilcoxon, Kruskal-Wallis) – used when data are not normally distributed or are ordinal.

Probability basics – Independent events multiply: P(A and B) = P(A) × P(B). Mutually exclusive events add: P(A or B) = P(A) + P(B). Non-mutually exclusive: P(A or B) = P(A) + P(B) − P(A and B). The Poisson distribution models counts of rare events over time (e.g., number of injuries per month) and is the theoretical basis for rare-event rates. The binomial distribution models number of successes in a fixed number of yes/no trials.

Public Health Surveillance Concepts

Surveillance is the ongoing, systematic collection, analysis, interpretation, and dissemination of health data for use in planning, implementing, and evaluating interventions. In the workplace this includes OSHA 300 logs, medical surveillance under specific standards (lead, asbestos, noise, respirators, benzene), hazard surveillance (exposure monitoring), and sentinel health events. Active surveillance seeks cases through outreach and testing; passive surveillance relies on reports submitted to the system. The Sentinel Health Event (Occupational) concept identifies a disease, disability, or death whose occurrence signals a failure of prevention and should trigger investigation.

Levels of prevention:
• Primary – prevent the exposure or disease from occurring (engineering controls, substitution, immunization).
• Secondary – detect early and intervene before symptoms or progression (audiometric testing, blood lead monitoring).
• Tertiary – reduce disability and restore function after disease is established (rehabilitation, return-to-work programs).
Medical surveillance is predominantly secondary prevention. The hierarchy of controls lives in primary prevention.

Latency is the period between first exposure and clinical appearance of disease. Many occupational cancers have latencies of 20 to 40 years, which is why retrospective cohort and case-control designs are needed and why cross-sectional studies of a current workforce can badly underestimate hazards.

How to Approach Exam Questions Step by Step

Step 1: Identify what is being asked. Is it a definition, a calculation, a study-design choice, or an interpretation? Underline the exact quantity requested (incidence, prevalence, RR, OR, sensitivity, PPV, TRIR).

Step 2: Build the table. For any association or screening question, sketch the 2×2 table on your scratch surface and populate a, b, c, d (or TP, FP, FN, TN). Most errors come from putting numbers in the wrong cell, not from the arithmetic.

Step 3: Match the design to the measure. Cohort or experimental → incidence, RR, AR. Case-control → OR only. Cross-sectional → prevalence, prevalence ratio or OR. If the question says cases were selected because they had the disease, it is a case-control study regardless of how it is worded.

Step 4: Check the denominator. Population at risk at the start for cumulative incidence; person-time for incidence rate; total population for prevalence; hours worked for OSHA rates; those with disease for sensitivity; those who tested positive for PPV.

Step 5: Keep units straight. Convert to per 1,000, per 100,000, or per 200,000 hours as the question requires. Check whether a rate is annual or over the full follow-up period.

Step 6: Interpret relative to the null. RR or OR of 1 = no effect. A CI that includes 1 = not significant. A p-value above alpha = fail to reject the null. Distinguish between statistical significance and clinical or practical importance.

Step 7: Sanity-check the answer. Sensitivity, specificity, PPV, and NPV must fall between 0 and 100%. A RR cannot be negative. Prevalence cannot exceed 100%. If your TRIR comes out as 250, you probably forgot to divide by hours.

Exam Tips: Answering Questions on Epidemiology and Public Health Statistics

Tip 1 – Memorize the two 2×2 tables and their formulas. One for association (RR, OR, AR) and one for screening (sensitivity, specificity, PPV, NPV). Write them on your scratch material at the start of the exam so you can refer to them without rethinking under pressure.

Tip 2 – New cases means incidence; existing cases means prevalence. If a question describes a survey conducted on one day, it is prevalence. If it describes following workers over time and counting those who develop the condition, it is incidence.

Tip 3 – Case-control equals odds ratio. You cannot compute incidence or relative risk when you selected people based on having the disease. If the answer choices include RR and OR for a case-control scenario, OR is correct.

Tip 4 – Temporality is the only mandatory Hill criterion. If a question asks which criterion must be satisfied to infer causation, the answer is that exposure precedes disease.

Tip 5 – Prevalence changes predictive values, not sensitivity or specificity. Lower prevalence lowers PPV and raises NPV. A question describing a screening program with many false positives in a healthy workforce is testing this concept.

Tip 6 – SnNout and SpPin. A Sensitive test with a Negative result rules disease out. A Specific test with a Positive result rules disease in. Initial screening favors sensitivity; confirmatory testing favors specificity.

Tip 7 – Know the healthy worker effect. It is selection bias that biases occupational SMRs and RRs toward the null or below 1.0. When a study of workers shows lower mortality than the general population, this is the explanation the exam is usually looking for.

Tip 8 – Confounder versus effect modifier. A confounder distorts the association and should be controlled; adjusting for it changes the estimate. An effect modifier means the association genuinely differs between subgroups; it should be reported separately, not adjusted away.

Tip 9 – Recall bias belongs to case-control studies; loss to follow-up belongs to cohort studies. Match the bias to the design and you can eliminate distractors quickly.

Tip 10 – Confidence interval including 1.0 means not significant. For RR or OR. For a difference in means, a CI including 0 means not significant. Narrow intervals indicate precision and usually larger samples.

Tip 11 – Type I is a false alarm; Type II is a missed effect. Alpha is the Type I error rate you set in advance (usually 0.05). Power is the ability to avoid Type II error, and it grows with sample size.

Tip 12 – Mean versus median. When data are skewed (lost workdays, claim costs, exposure concentrations), the median is the more representative measure of center and the mean is pulled toward the tail. For lognormal exposure data, the geometric mean is appropriate.

Tip 13 – Empirical rule numbers. 68-95-99.7 for 1, 2, and 3 standard deviations. For one-sided questions (what percentage exceeds a value), halve the tail: about 16% above +1 SD, 2.5% above +2 SD, 0.15% above +3 SD.

Tip 14 – Use the constant the question gives you. OSHA rates use 200,000 hours, but some questions use 1,000,000 for severity or frequency. Read for the constant and never assume.

Tip 15 – Correlation is not causation, and association is not causation. Any answer choice that claims a correlation coefficient or an RR alone proves cause is wrong. Look for the answer that invokes study design, temporality, confounding, or Hill's criteria.

Tip 16 – Ecological fallacy. When group-level data (county, state, plant-wide) are used to infer about individuals, the exam wants you to name this limitation.

Tip 17 – Rare disease assumption. The OR approximates the RR only when the disease is uncommon. If a question asks when an OR is a good estimate of RR, the answer involves low disease frequency.

Tip 18 – Rate instability in small groups. If a question describes a small department whose TRIR doubled after one injury, the statistical point is that small denominators produce unstable rates and the change may not be meaningful. Multi-year aggregation or confidence intervals are the appropriate response.

Tip 19 – Eliminate by direction. Many distractors are the reciprocal of the right answer or compute the wrong ratio (e.g., FP/(TP+FP) instead of TP/(TP+FP)). Compute your answer, then check whether the distractors represent common mistakes; this confirms your result.

Tip 20 – Manage time on calculations. Epidemiology questions are generally two to four arithmetic steps. If you find yourself more than five steps deep, you have misread the question or mis-assigned a cell. Re-read, rebuild the table, and proceed.

Quick Reference Formula Sheet

Prevalence = Existing cases / Total population
Cumulative incidence = New cases / Population at risk at start
Incidence rate = New cases / Person-time at risk
RR = [a/(a+b)] / [c/(c+d)]
OR = (a × d) / (b × c)
AR = Incidence exposed − Incidence unexposed
AR% = (RR − 1) / RR × 100
SMR = Observed / Expected × 100
Sensitivity = TP / (TP + FN)
Specificity = TN / (TN + FP)
PPV = TP / (TP + FP)
NPV = TN / (TN + FN)
TRIR = Recordables × 200,000 / Hours worked
DART = DART cases × 200,000 / Hours worked
Severity = Lost or restricted days × 200,000 / Hours worked
z = (x − mean) / SD
SE of mean = SD / √n
95% CI for mean = Mean ± 1.96 × SE
CV = SD / Mean × 100%
Power = 1 − beta

Final Perspective

Epidemiology and public health statistics are not academic add-ons to the CSP body of knowledge; they are the logic that connects observation to action. Every time you argue for an engineering control, defend a medical surveillance program, evaluate whether a safety initiative reduced injuries, or interpret a consultant's exposure assessment, you are using these concepts. On exam day, treat each question as a small investigation: identify the population, the exposure, the outcome, the design, and the measure, then compute carefully and interpret against the null. With the formulas internalized and the design-to-measure logic clear, this domain becomes one of the most predictable and reliably scored sections of the exam.