Define sensitivity.

The sensitivity of a test refers to how often a positive result correctly identifies those who have the disease. The higher the sensitivity, the lower the number of results that are falsely negative.

How is sensitivity calculated?

Sensitivity = true positive/(true positive + false negative) = true positive / total diseased

How can SPIN and SNOUT be used to help differentiate between sensitivity and specificity?

SPIN: SPecific tests help to rule IN a disease because the false Positive rate is low. SNOUT: SeNsitive tests help to rule OUT a disease because the false Negative rate is low.

Define specificity.

The specificity of a test determines how often a negative result correctly identifies those who do not have the disease. The higher the specificity, the lower the number of false positive test results (i.e. higher specificity = higher likelihood that a negative result indicates true absence of disease).

How is specificity calculated?

Specificity = true negative / (true negative + false positive) = true negative / total not diseased

Draw a Bayesian foursquare (test results and disease statuses) and use it to demonstrate how to calculate sensitivity.

***Insert Figure 19-1

Draw a Bayesian foursquare (test results and disease statuses) and use it to demonstrate how to calculate specificity.

***Insert Figure 19-1

Draw a Bayesian foursquare (test results and disease statuses) and use it to demonstrate how to calculate positive predictive value.

***Insert Figure 19-1

Draw a Bayesian foursquare (test results and disease statuses) and use it to demonstrate how to calculate negative predictive value.

***Insert Figure 19-1

Define positive predictive value.

The positive predictive value of a diagnostic test is the probability that a patient with a positive test actually has the disease being tested for.

How is positive predictive value calculated?

PPV = true positive / (true positive + false positive) = true positive / all positive tests

What does the positive predictive value help determine?

PPV helps to determine how effective a test is as a screening tool.

Describe the relationship between prevalence of a disease and its effect on sensitivity/specificity vs its effect on positive and negative predictive values.

Sensitivity and specificity are independent of disease prevalence, whereas positive and negative predictive values are influenced by the prevalence of a disease.

Define negative predictive value.

Negative predictive value of a diagnostic test is the probability of not having a disease if the test is negative.

How is negative predictive value calculated?

NPV = true negative / (true negative + false negative) = true negative / all negative tests

Define prevalence.

Prevalence of a disease is the percent of those with the disease in the population being studied.

How is prevalence calculated?

Prevalence = total diseased / total population = (TP + FN) / (TP + FN + TN + FP). ***Note that the Bayesian formula is the same for both prevalence and incidence, but the study setup is different.

How does prevalence affect how one might interpret a given test result?

Prevalence affects the pretest probability associated with a given test.

Define incidence.

The incidence of a disease is the occurrence of new cases of a disease within a specified period of time (it can also be seen as the probability that a person develops that disease during the period of time).

How is incidence calculated?

Incidence = total new cases of disease / total # tested during the given time period = (TP + FN) / (TP + FN + TN + FP). ***Note that the Bayesian formula is the same for both prevalence and incidence, but the study setup is different.

What happens if you increase the threshold of a test?

If you increase the threshold of a test you get more negative results (both true and false negatives). This decreases the sensitivity and increases the specificity.

How can one conceptualize the threshold of a test in relationship with the Bayesian foursquare?

Think of the threshold like the horizontal line of the Bayesian foursquare: as the line moves upward (threshold increases), the total number of negative tests increases; as the line moves downward (threshold decreases), the total number of negative tests decreases. These changes will then be reflected in the sensitivity and specificity calculations due to their affects on the numerators and denominators of those equations.

How is sensitivity affected by changes in the threshold of a test?

As the threshold increases, the sensitivity decreases.

How is specificity affected by changes in the threshold of a test?

As the threshold increases, the specificity increases.

How can one conceptualize the effect of changing disease prevalence/incidence on the false positive and false negative rates of a test using the Bayesian foursquare?

Think of prevalence/incidence as the vertical line. As the line moves left (decreased prevalence/incidence), the number of true positives and false negatives decreases, while the number of false positives and true negatives increases.

What are the two primary categories of study design?

Observational and experimental

What is the purpose of an observational study, and how are they designed?

Observational studies attempt to correlate exposures with outcomes but do not assign patients to one group or another.

What are the two main types of observational study design?

Cohort studies and case-control studies

Describe cohort study design.

Cohort studies are a type of observational study which follow groups of individuals prospectively over time to see which exposures cause disease (i.e. babies exposed to thimerosal vs babies not exposed are followed over time to see if there is a correlation with exposure and developmental disorders later in life).

Describe case-control study design.

Case-control studies are a type of observational study in which people with the disease are compared to those without the disease to identify relevant risk factors (e.g. to see if there is a relationship between acetaminophen use and development of Reye syndrome).

Describe randomized controlled trial study design.

RCTs are a type of experimental study in which the researchers randomly assign patients to different study groups, usually an intervention vs a control.

What is the advantage of a randomized trial?

The strength of RCTs is that the randomization process reduces the risk of confounding variables and biases affecting the data, which makes the results more reliable.

What are the benefits of observational study design?

Observational studies are less expensive than randomized trials and make it possible to study interventions to which researchers cannot ethically randomize subjects (e.g. smoking carcinogens).

What is Type 1 error?

Type 1 error occurs when a study concludes that there is a difference in outcomes when there is not (i.e. rejecting the null hypothesis incorrectly).

What statistical value is used to determine the presence of Type 1 error in a study?

Type 1 errors are exposed by investigating the p value (statistical significance) of the results.

What is Type 2 error?

Type 2 error occurs when a study concludes that there is no difference in outcomes when there actually is (i.e. failing to reject the null hypothesis).

What statistical value is used to determine the presence of Type 2 error in a study?

Type 2 errors are exposed by investigating the power of a study.

What p value is considered statistically significant?

A p value ≤ 0.05 is considered statistically significant, with significance increasing as the p value gets smaller.

What does the p value represent?

The p value is a way of expressing a study’s statistical significance. It represents the probability of finding an association (by chance) when one genuinely does not exist.

What is the power of a study?

The power of a study is the probability that it can detect a treatment effect if it is present (i.e. that the test can correctly reject the null hypothesis).

How would one increase the power of a study?

The only way to increase the power of a study is to increase the number of subjects, continue the study for a longer duration, or both (basically, the more data you are able to collect, the higher your power).

What is considered to be a statistically significant confidence interval?

A CI of 95% is essentially the same as p=0.05. So a consistent result that falls outside the confidence interval of 95% is “statistically significant”.

What is the function of an odds ratio?

The null hypothesis can be expressed as an odds ratio, which is a way of comparing whether the probability of an event is the same for two groups (control vs study). It is a ratio, so if there is no difference between groups, the ratio would be 1 (same odds = no difference)

How is statistical significance represented on confidence interval and/or “forest plot” charts?

The confidence interval is represented by horizontal lines, with the vertical line representing a “no effect” result. Therefore, if the horizontal line crosses the vertical line, the result is not statistically significant.

What is the number needeed to treat?

NNT is the number of people who would need to be treated to prevent one event.

How is number needed to treat calculated?

NNT = 1/ absolute risk reduction = 1/(control event rate - experimental event rate)

What is number needed to harm?

NNH is the number of patients who get a drug or intervention for each patient harmed. The calculation is the same as for number needed to treat, but uses control vs study harm rates rather than rates of positive treatment effects.