Hubbry Logo
Pre- and post-test probabilityPre- and post-test probabilityMain
Open search
Pre- and post-test probability
Community hub
Pre- and post-test probability
logo
7 pages, 0 posts
0 subscribers
Be the first to start a discussion here.
Be the first to start a discussion here.
Pre- and post-test probability
Pre- and post-test probability
Pre- and post-test probability
View on Wikipedia
from Wikipedia

Pre-test probability and post-test probability (alternatively spelled pretest and posttest probability) are the probabilities of the presence of a condition (such as a disease) before and after a diagnostic test, respectively. Post-test probability, in turn, can be positive or negative, depending on whether the test falls out as a positive test or a negative test, respectively. In some cases, it is used for the probability of developing the condition of interest in the future.

Test, in this sense, can refer to any medical test (but usually in the sense of diagnostic tests), and in a broad sense also including questions and even assumptions (such as assuming that the target individual is a female or male). The ability to make a difference between pre- and post-test probabilities of various conditions is a major factor in the indication of medical tests.

Pre-test probability

[edit]

The pre-test probability of an individual can be chosen as one of the following:

  • The prevalence of the disease, which may have to be chosen if no other characteristic is known for the individual, or it can be chosen for ease of calculation even if other characteristics are known although such omission may cause inaccurate results
  • The post-test probability of the condition resulting from one or more preceding tests
  • A rough estimation, which may have to be chosen if more systematic approaches are not possible or efficient

Estimation of post-test probability

[edit]

In clinical practice, post-test probabilities are often just estimated or even guessed. This is usually acceptable in the finding of a pathognomonic sign or symptom, in which case it is almost certain that the target condition is present; or in the absence of finding a sine qua non sign or symptom, in which case it is almost certain that the target condition is absent.

In reality, however, the subjective probability of the presence of a condition is never exactly 0 or 100%. Yet, there are several systematic methods to estimate that probability. Such methods are usually based on previously having performed the test on a reference group in which the presence or absence on the condition is known (or at least estimated by another test that is considered highly accurate, such as by "Gold standard"), in order to establish data of test performance. These data are subsequently used to interpret the test result of any individual tested by the method. An alternative or complement to reference group-based methods is comparing a test result to a previous test on the same individual, which is more common in tests for monitoring.

The most important systematic reference group-based methods to estimate post-test probability includes the ones summarized and compared in the following table, and further described in individual sections below.

Method Establishment of performance data Method of individual interpretation Ability to accurately interpret subsequent tests Additional advantages
By predictive values Direct quotients from reference group Most straightforward: Predictive value equals probability Usually low: Separate reference group required for every subsequent pre-test state Available both for binary and continuous values
By likelihood ratio Derived from sensitivity and specificity Post-test odds given by multiplying pretest odds with the ratio Theoretically limitless Pre-test state (and thus the pre-test probability) does not have to be same as in reference group
By relative risk Quotient of risk among exposed and risk among unexposed Pre-test probability multiplied by the relative risk Low, unless subsequent relative risks are derived from same multivariate regression analysis Relatively intuitive to use
By diagnostic criteria and clinical prediction rules Variable, but usually most tedious Variable Usually excellent for all test included in criteria Usually most preferable if available

By predictive values

[edit]

Predictive values can be used to estimate the post-test probability of an individual if the pre-test probability of the individual can be assumed roughly equal to the prevalence in a reference group on which both test results and knowledge on the presence or absence of the condition (for example a disease, such as may determined by "Gold standard") are available.

If the test result is of a binary classification into either positive or negative tests, then the following table can be made:

Condition
(as determined by "Gold standard")
Positive Negative
Test
outcome 		Positive True Positive False Positive
(Type I error) 		→ Positive predictive value
Negative False Negative
(Type II error) 		True Negative → Negative predictive value
↓
Sensitivity 		↓
Specificity 		↘
Accuracy

Pre-test probability can be calculated from the diagram as follows:

Pretest probability = (True positive + False negative) / Total sample

Also, in this case, the positive post-test probability (the probability of having the target condition if the test falls out positive), is numerically equal to the positive predictive value, and the negative post-test probability (the probability of having the target condition if the test falls out negative) is numerically complementary to the negative predictive value ([negative post-test probability] = 1 - [negative predictive value]),[1] again assuming that the individual being tested does not have any other risk factors that result in that individual having a different pre-test probability than the reference group used to establish the positive and negative predictive values of the test.

In the diagram above, this positive post-test probability, that is, the posttest probability of a target condition given a positive test result, is calculated as:

Positive posttest probability = True positives / (True positives + False positives)

Similarly:

The post-test probability of disease given a negative result is calculated as:

Negative posttest probability = 1 - (True negatives / (False negatives + True negatives))

The validity of the equations above also depend on that the sample from the population does not have substantial sampling bias that make the groups of those who have the condition and those who do not substantially disproportionate from corresponding prevalence and "non-prevalence" in the population. In effect, the equations above are not valid with merely a case-control study that separately collects one group with the condition and one group without it.

By likelihood ratio

[edit]

The above methods are inappropriate to use if the pretest probability differs from the prevalence in the reference group used to establish, among others, the positive predictive value of the test. Such difference can occur if another test preceded, or the person involved in the diagnostics considers that another pretest probability must be used because of knowledge of, for example, specific complaints, other elements of a medical history, signs in a physical examination, either by calculating on each finding as a test in itself with its own sensitivity and specificity, or at least making a rough estimation of the individual pre-test probability.

In these cases, the prevalence in the reference group is not completely accurate in representing the pre-test probability of the individual, and, consequently, the predictive value (whether positive or negative) is not completely accurate in representing the post-test probability of the individual of having the target condition.

In these cases, a posttest probability can be estimated more accurately by using a likelihood ratio for the test. Likelihood ratio is calculated from sensitivity and specificity of the test, and thereby it does not depend on prevalence in the reference group,[2] and, likewise, it does not change with changed pre-test probability, in contrast to positive or negative predictive values (which would change). Also, in effect, the validity of post-test probability determined from likelihood ratio is not vulnerable to sampling bias in regard to those with and without the condition in the population sample, and can be done as a case-control study that separately gathers those with and without the condition.

Estimation of post-test probability from pre-test probability and likelihood ratio goes as follows:[2]

  • Pretest odds = Pretest probability / (1 - Pretest probability)
  • Posttest odds = Pretest odds * Likelihood ratio

In equation above, positive post-test probability is calculated using the likelihood ratio positive, and the negative post-test probability is calculated using the likelihood ratio negative.

  • Posttest probability = Posttest odds / (Posttest odds + 1)
Fagan nomogram[3]

The relation can also be estimated by a so-called Fagan nomogram (shown at right) by making a straight line from the point of the given pre-test probability to the given likelihood ratio in their scales, which, in turn, estimates the post-test probability at the point where that straight line crosses its scale.

The post-test probability can, in turn, be used as pre-test probability for additional tests if it continues to be calculated in the same manner.[2]

  • Diagram relating pre- and post-test probabilities, with the green curve (upper left half) representing a positive test, and the red curve (lower right half) representing a negative test, for the case of 90% sensitivity and 90% specificity, corresponding to a likelihood ratio positive of 9, and a likelihood ratio negative of 0.111. The length of the green arrows represent the change in absolute (rather than relative) probability given a positive test, and the red arrows represent the change in absolute probability given a negative test. It can be seen from the length of the arrows that, at low pre-test probabilities, a positive test gives a greater change in absolute probability than a negative test (a property that is generally valid as long as the specificity isn't much higher than the sensitivity). Similarly, at high pre-test probabilities, a negative test gives a greater change in absolute probability than a positive test (a property that is generally valid as long as the sensitivity isn't much higher than the specificity).
    Diagram relating pre- and post-test probabilities, with the green curve (upper left half) representing a positive test, and the red curve (lower right half) representing a negative test, for the case of 90% sensitivity and 90% specificity, corresponding to a likelihood ratio positive of 9, and a likelihood ratio negative of 0.111. The length of the green arrows represent the change in absolute (rather than relative) probability given a positive test, and the red arrows represent the change in absolute probability given a negative test.
    It can be seen from the length of the arrows that, at low pre-test probabilities, a positive test gives a greater change in absolute probability than a negative test (a property that is generally valid as long as the specificity isn't much higher than the sensitivity). Similarly, at high pre-test probabilities, a negative test gives a greater change in absolute probability than a positive test (a property that is generally valid as long as the sensitivity isn't much higher than the specificity).
  • Relation between pre-and post-test probabilities for various likelihood ratio positives (upper left half) and various likelihood ratio negatives (lower right half).
    Relation between pre-and post-test probabilities for various likelihood ratio positives (upper left half) and various likelihood ratio negatives (lower right half).

It is possible to do a calculation of likelihood ratios for tests with continuous values or more than two outcomes which is similar to the calculation for dichotomous outcomes. For this purpose, a separate likelihood ratio is calculated for every level of test result and is called interval or stratum specific likelihood ratios.[4]

Example

[edit]

An individual was screened with the test of fecal occult blood (FOB) to estimate the probability for that person having the target condition of bowel cancer, and it fell out positive (blood were detected in stool). Before the test, that individual had a pre-test probability of having bowel cancer of, for example, 3% (0.03), as could have been estimated by evaluation of, for example, the medical history, examination and previous tests of that individual.

The sensitivity, specificity etc. of the FOB test were established with a population sample of 203 people (without such heredity), and fell out as follows:

Patients with bowel cancer
(as confirmed on endoscopy)
Positive Negative
Fecal
occult
blood
screen
test
outcome 		Positive TP = 2 FP = 18 → Positive predictive value
= TP / (TP + FP)
= 2 / (2 + 18)
= 2 / 20
= 10%
Negative FN = 1 TN = 182 → Negative predictive value
= TN / (FN + TN)
= 182 / (1 + 182)
= 182 / 183
≈ 99.5%
↓
Sensitivity
= TP / (TP + FN)
= 2 / (2 + 1)
= 2 / 3
≈ 66.67% 		↓
Specificity
= TN / (FP + TN)
= 182 / (18 + 182)
= 182 / 200
= 91% 		↘
Accuracy
= (TP + TN) / Total
= (2 + 182) / 203
= 184 / 203
= 90.64%

From this, the likelihood ratios of the test can be established:[2]

  1. Likelihood ratio positive = sensitivity / (1 − specificity) = 66.67% / (1 − 91%) = 7.4
  2. Likelihood ratio negative = (1 − sensitivity) / specificity = (1 − 66.67%) / 91% = 0.37
  • Pretest probability (in this example) = 0.03
  • Pretest odds = 0.03 / (1 - 0.03) = 0.0309
  • Positive posttest odds = 0.0309 * 7.4 = 0.229
  • Positive posttest probability = 0.229 / (0.229 + 1) = 0.186 or 18.6%

Thus, that individual has a post-test probability (or "post-test risk") of 18.6% of having bowel cancer.

The prevalence in the population sample is calculated to be:

  • Prevalence = (2 + 1) / 203 = 0.0148 or 1.48%

The individual's pre-test probability was more than twice that of the population sample, although the individual's post-test probability was less than twice that of the population sample (which is estimated by the positive predictive value of the test of 10%), opposite to what would result by a less accurate method of simply multiplying relative risks.

Specific sources of inaccuracy

[edit]

Specific sources of inaccuracy when using likelihood ratio to determine a post-test probability include interference with determinants or previous tests or overlap of test targets, as explained below:

Interference with test
[edit]

Post-test probability, as estimated from the pre-test probability with likelihood ratio, should be handled with caution in individuals with other determinants (such as risk factors) than the general population, as well as in individuals that have undergone previous tests, because such determinants or tests may also influence the test itself in unpredictive ways, still causing inaccurate results. An example with the risk factor of obesity is that additional abdominal fat can make it difficult to palpate abdominal organs and decrease the resolution of abdominal ultrasonography, and similarly, remnant barium contrast from a previous radiography can interfere with subsequent abdominal examinations,[5] in effect decreasing the sensitivities and specificities of such subsequent tests. On the other hand, the effect of interference can potentially improve the efficacy of subsequent tests as compared to usage in the reference group, such as some abdominal examinations being easier when performed on underweight people.

Overlap of tests
[edit]

Furthermore, the validity of calculations upon any pre-test probability that itself is derived from a previous test depend on that the two tests do not significantly overlap in regard to the target parameter being tested, such as blood tests of substances belonging to one and the same deranged metabolic pathway. An example of the extreme of such an overlap is where the sensitivity and specificity has been established for a blood test detecting "substance X", and likewise for one detecting "substance Y". If, in fact, "substance X" and "substance Y" are one and the same substance, then, making a two consecutive tests of one and the same substance may not have any diagnostic value at all, although the calculation appears to show a difference. In contrast to interference as described above, increasing overlap of tests only decreases their efficacy. In the medical setting, diagnostic validity is increased by combining tests of different modalities to avoid substantial overlap, for example in making a combination of a blood test, a biopsy and radiograph.

Methods to overcome inaccuracy
[edit]

To avoid such sources of inaccuracy by using likelihood ratios, the optimal method would be to gather a large reference group of equivalent individuals, in order to establish separate predictive values for use of the test in such individuals. However, with more knowledge of an individual's medical history, physical examination and previous test etc. that individual becomes more differentiated, with increasing difficulty to find a reference group to establish tailored predictive values, making an estimation of post-test probability by predictive values invalid.

Another method to overcome such inaccuracies is by evaluating the test result in the context of diagnostic criteria, as described in the next section.

By relative risk

[edit]

Post-test probability can sometimes be estimated by multiplying the pre-test probability with a relative risk given by the test. In clinical practice, this is usually applied in evaluation of a medical history of an individual, where the "test" usually is a question (or even assumption) regarding various risk factors, for example, sex, tobacco smoking or weight, but it can potentially be a substantial test such as putting the individual on a weighing scale. When using relative risks, the resultant probability is usually rather related to the individual developing the condition over a period of time (similarly to the incidence in a population), instead of being the probability of an individual of having the condition in the present, but can indirectly be an estimation of the latter.

Usage of hazard ratio can be used somewhat similarly to relative risk.

One risk factor

[edit]

To establish a relative risk, the risk in an exposed group is divided by the risk in an unexposed group.

If only one risk factor of an individual is taken into account, the post-test probability can be estimated by multiplying the relative risk with the risk in the control group. The control group usually represents the unexposed population, but if a very low fraction of the population is exposed, then the prevalence in the general population can often be assumed equal to the prevalence in the control group. In such cases, the post-test probability can be estimated by multiplying the relative risk with the risk in the general population.

For example, the incidence of breast cancer in a woman in the United Kingdom at age 55 to 59 is estimated at 280 cases per 100.000 per year,[6] and the risk factor of having been exposed to high-dose ionizing radiation to the chest (for example, as treatments for other cancers) confers a relative risk of breast cancer between 2.1 and 4.0,[7] compared to unexposed. Because a low fraction of the population is exposed, the prevalence in the unexposed population can be assumed equal to the prevalence in the general population. Subsequently, it can be estimated that a woman in the United Kingdom that is aged between 55 and 59 and that has been exposed to high-dose ionizing radiation should have a risk of developing breast cancer over a period of one year of between 588 and 1.120 in 100.000 (that is, between 0,6% and 1.1%).

Multiple risk factors

[edit]

Theoretically, the total risk in the presence of multiple risk factors can be estimated by multiplying with each relative risk, but is generally much less accurate than using likelihood ratios, and is usually done only because it is much easier to perform when only relative risks are given, compared to, for example, converting the source data to sensitivities and specificities and calculate by likelihood ratios. Likewise, relative risks are often given instead of likelihood ratios in the literature because the former is more intuitive. Sources of inaccuracy of multiplying relative risks include:

  • Relative risks are affected by the prevalence of the condition in the reference group (in contrast to likelihood ratios, which are not), and this issue results in that the validity of post-test probabilities become less valid with increasing difference between the prevalence in the reference group and the pre-test probability for any individual. Any known risk factor or previous test of an individual almost always confers such a difference, decreasing the validity of using relative risks in estimating the total effect of multiple risk factors or tests. Most physicians do not appropriately take such differences in prevalence into account when interpreting test results, which may cause unnecessary testing and diagnostic errors.[8]
  • A separate source of inaccuracy of multiplying several relative risks, considering only positive tests, is that it tends to overestimate the total risk as compared to using likelihood ratios. This overestimation can be explained by the inability of the method to compensate for the fact that the total risk cannot be more than 100%. This overestimation is rather small for small risks, but becomes higher for higher values. For example, the risk of developing breast cancer at an age younger than 40 years in women in the United Kingdom can be estimated at 2%.[9] Also, studies on Ashkenazi Jews has indicated that a mutation in BRCA1 confers a relative risk of 21.6 of developing breast cancer in women under 40 years of age, and a mutation in BRCA2 confers a relative risk of 3.3 of developing breast cancer in women under 40 years of age.[10] From these data, it may be estimated that a woman with a BRCA1 mutation would have a risk of approximately 40% of developing breast cancer at an age younger than 40 years, and woman with a BRCA2 mutation would have a risk of approximately 6%. However, in the rather improbable situation of having both a BRCA1 and a BRCA2 mutation, simply multiplying with both relative risks would result in a risk of over 140% of developing breast cancer before 40 years of age, which can not possibly be accurate in reality.

The (latter mentioned) effect of overestimation can be compensated for by converting risks to odds, and relative risks to odds ratios. However, this does not compensate for (former mentioned) effect of any difference between pre-test probability of an individual and the prevalence in the reference group.

A method to compensate for both sources of inaccuracy above is to establish the relative risks by multivariate regression analysis. However, to retain its validity, relative risks established as such must be multiplied with all the other risk factors in the same regression analysis, and without any addition of other factors beyond the regression analysis.

In addition, multiplying multiple relative risks has the same risk of missing important overlaps of the included risk factors, similarly to when using likelihood ratios. Also, different risk factors can act in synergy, with the result that, for example, two factors that both individually have a relative risk of 2 have a total relative risk of 6 when both are present, or can inhibit each other, somewhat similarly to the interference described for using likelihood ratios.

By diagnostic criteria and clinical prediction rules

[edit]

Most major diseases have established diagnostic criteria and/or clinical prediction rules. The establishment of diagnostic criteria or clinical prediction rules consists of a comprehensive evaluation of many tests that are considered important in estimating the probability of a condition of interest, sometimes also including how to divide it into subgroups, and when and how to treat the condition. Such establishment can include usage of predictive values, likelihood ratios as well as relative risks.

For example, the ACR criteria for systemic lupus erythematosus defines the diagnosis as presence of at least 4 out of 11 findings, each of which can be regarded as a target value of a test with its own sensitivity and specificity. In this case, there has been evaluation of the tests for these target parameters when used in combination in regard to, for example, interference between them and overlap of target parameters, thereby striving to avoid inaccuracies that could otherwise arise if attempting to calculate the probability of the disease using likelihood ratios of the individual tests. Therefore, if diagnostic criteria have been established for a condition, it is generally most appropriate to interpret any post-test probability for that condition in the context of these criteria.

Also, there are risk assessment tools for estimating the combined risk of several risk factors, such as the online tool [1] from the Framingham Heart Study for estimating the risk for coronary heart disease outcomes using multiple risk factors, including age, gender, blood lipids, blood pressure and smoking, being much more accurate than multiplying the individual relative risks of each risk factor.

Still, an experienced physician may estimate the post-test probability (and the actions it motivates) by a broad consideration including criteria and rules in addition to other methods described previously, including both individual risk factors and the performances of tests that have been carried out.

Clinical use of pre- and post-test probabilities

[edit]

A clinically useful parameter is the absolute (rather than relative, and not negative) difference between pre- and post-test probability, calculated as:

Absolute difference = | (pre-test probability) - (post-test probability) |

A major factor for such an absolute difference is the power of the test itself, such as can be described in terms of, for example, sensitivity and specificity or likelihood ratio. Another factor is the pre-test probability, with a lower pre-test probability resulting in a lower absolute difference, with the consequence that even very powerful tests achieve a low absolute difference for very unlikely conditions in an individual (such as rare diseases in the absence of any other indicating sign), but on the other hand, that even tests with low power can make a great difference for highly suspected conditions.

The probabilities in this sense may also need to be considered in context of conditions that are not primary targets of the test, such as profile-relative probabilities in a differential diagnostic procedure.

The absolute difference can be put in relation to the benefit for an individual that a medical test achieves, such as can roughly be estimated as:

, where:

  • bn is the net benefit of performing a medical test
  • Λp is the absolute difference between pre- and posttest probability of conditions (such as diseases) that the test is expected to achieve.
  • ri is the rate of how much probability differences are expected to result in changes in interventions (such as a change from "no treatment" to "administration of low-dose medical treatment").
  • bi is the benefit of changes in interventions for the individual
  • hi is the harm of changes in interventions for the individual, such as side effects of medical treatment
  • ht is the harm caused by the test itself

In this formula, what constitutes benefit or harm largely varies by personal and cultural values, but general conclusions can still be drawn. For example, if the only expected effect of a medical test is to make one disease more likely than another, but the two diseases have the same treatment (or neither can be treated), then ri = 0 and the test is essentially without any benefit for the individual.

Additional factors that influence a decision whether a medical test should be performed or not include: cost of the test, availability of additional tests, potential interference with subsequent test (such as an abdominal palpation potentially inducing intestinal activity whose sounds interfere with a subsequent abdominal auscultation), time taken for the test or other practical or administrative aspects. Also, even if not beneficial for the individual being tested, the results may be useful for the establishment of statistics in order to improve health care for other individuals.

Subjectivity

[edit]

Pre- and post-test probabilities are subjective based on the fact that, in reality, an individual either has the condition or not (with the probability always being 100%), so pre- and post-test probabilities for individuals can rather be regarded as psychological phenomena in the minds of those involved in the diagnostics at hand.

See also

[edit]

References

[edit]
Revisions and contributorsEdit on WikipediaRead on Wikipedia

Pre- and post-test probability

View on Grokipedia
from Grokipedia
Pre-test probability is the estimated likelihood that a has a specific or condition prior to undergoing a diagnostic test, typically derived from factors such as prevalence, , symptoms, and factors. Post-test probability represents the updated likelihood of the after incorporating the results of the diagnostic test, accounting for the test's performance characteristics like . These concepts, grounded in , enable clinicians to quantify diagnostic uncertainty and interpret test results in a probabilistic framework rather than relying solely on binary outcomes. The transition from pre-test to post-test probability is calculated using , which mathematically updates the (pre-test) with evidence from the test result. Specifically, post-test odds are obtained by multiplying pre-test by the likelihood (LR) of the test: post-test odds = pre-test odds × LR, where pre-test odds = pre-test probability / (1 - pre-test probability), and LR for a positive test = sensitivity / (1 - specificity). This approach transforms raw test data into actionable probabilities; for instance, a likelihood ratio greater than 1 increases the probability of disease, while one less than 1 decreases it. In , pre- and post-test probabilities are essential for deciding the utility of diagnostic tests, as tests are most informative when pre-test probability is intermediate (neither very low nor very high), avoiding unnecessary testing or . Clinical thresholds—such as treatment thresholds (e.g., 5-10% for certain conditions like )—further guide decisions by indicating when post-test probability justifies intervention without additional testing. Graphical tools like Fagan's simplify these calculations by allowing users to plot pre-test probability and likelihood ratios to directly read off post-test probabilities, enhancing bedside . Recent advancements extend the Bayesian pre-test/post-test framework by integrating to explicitly consider costs, benefits, and risks alongside probabilities, addressing limitations in traditional models that focus solely on diagnostic accuracy. For example, simplified decision boundaries can help clinicians weigh treatment options without predefined cost utilities, as demonstrated in reanalyses of clinical scenarios like asymptomatic bacteriuria. These developments underscore the framework's evolving role in and diagnostic stewardship.

Fundamentals

Definitions

Pre-test probability is defined as the clinician's subjective estimate of the likelihood that a has a specific before any diagnostic test is performed. This probability is expressed on a scale from 0 to 1, where 0 indicates no likelihood of and 1 indicates , or equivalently as a ranging from 0% to 100%. Post-test probability represents the revised estimate of the 's likelihood after incorporating the results of a diagnostic test into the assessment. It is calculated separately for positive test results (positive post-test probability, reflecting the chance of true presence given a positive outcome) and negative test results (negative post-test probability, reflecting the chance of disease absence given a negative outcome). Pre-test probability acts as the foundational prior in diagnostic reasoning, serving as the initial input for Bayesian processes that integrate new from results to yield the post-test probability. Fundamentally, pre-test probability draws from aggregated prior clinical data, including symptoms, history, , and epidemiological factors such as disease ; post-test probability then refines this estimate by accounting for the diagnostic provided by the .

Bayesian Foundation

The Bayesian foundation for pre- and post-test probability in medical diagnostics is rooted in , which provides a mathematical framework for updating the probability of a disease based on test results. Formulated by the English mathematician and Presbyterian minister (c. 1701–1761), the theorem was published posthumously in 1763 as part of his essay "An Essay towards solving a Problem in the Doctrine of Chances" in the Philosophical Transactions of the Royal Society. This theorem enables the transition from pre-test probability—the initial estimate of disease likelihood before testing—to post-test probability by incorporating the test's performance characteristics. In medical contexts, is expressed as the of given a test result: P(D∣T)=P(T∣D)×P(D)P(T),P(D|T) = \frac{P(T|D) \times P(D)}{P(T)}, where P(D)P(D) is the pre-test probability of , P(T∣D)P(T|D) is the sensitivity of the (probability of a positive given ), and P(T)P(T) is the total probability of a positive test result. The denominator P(T)P(T) is derived from the : P(T)=P(T∣D)×P(D)+P(T∣¬D)×P(¬D)P(T) = P(T|D) \times P(D) + P(T|\neg D) \times P(\neg D), where P(T∣¬D)P(T|\neg D) is the false-positive rate (1 - specificity) and P(¬D)=1−P(D)P(\neg D) = 1 - P(D). This formulation ensures that the resulting post-test probability remains bounded between 0 and 1, preventing overestimation or invalid outcomes by normalizing against all possible scenarios. A more intuitive derivation begins with the odds form of the theorem, which simplifies calculations in clinical settings. are defined as the of probability to its complement: pre-test = P(D)/(1−P(D))P(D) / (1 - P(D)). The post-test are then obtained by multiplying the pre-test by the likelihood (LR) of the result: post-test = pre-test × LR. For a positive , LR⁺ = sensitivity / (1 - specificity); for a negative , LR⁻ = (1 - sensitivity) / specificity. To derive this, start from the definition of : P(D∣T)=P(D∩T)/P(T)P(D|T) = P(D \cap T) / P(T). Substitute P(D∩T)=P(T∣D)×P(D)P(D \cap T) = P(T|D) \times P(D) and divide both numerator and denominator by P(¬D∩¬T)P(\neg D \cap \neg T), yielding the form: P(D∣T)P(¬D∣T)=P(T∣D)P(T∣¬D)×P(D)P(¬D).\frac{P(D|T)}{P(\neg D|T)} = \frac{P(T|D)}{P(T|\neg D)} \times \frac{P(D)}{P(\neg D)}. Converting back to probability gives P(D∣T)=post-test [odds](/page/Odds)1+post-test [odds](/page/Odds)P(D|T) = \frac{\text{post-test [odds](/page/Odds)}}{1 + \text{post-test [odds](/page/Odds)}}, maintaining the 0–1 bound through normalization. The application of to medicine gained prominence in the , particularly for evidence-based diagnostics. Early uses included Jerome Cornfield's 1951 analysis of smoking and risk, which employed Bayesian updating to estimate disease rates from clinical data. By the , it underpinned the Rational Clinical Examination series, launched in 1992, which systematically applied the theorem to evaluate physical exam findings and tests, integrating likelihood ratios with pre-test probabilities to guide diagnostic decision-making.

Pre-test Probability

Estimation Methods

Estimating pre-test probability begins with establishing a baseline using prevalence derived from epidemiological studies or clinical databases, which provides an initial estimate of the likelihood of in a relevant . For instance, prevalence data from cross-sectional studies can anchor the probability for patients presenting with specific symptoms, such as the <10% initial probability of in patients with suspected . This approach ensures the estimate reflects evidence-based rates rather than unsubstantiated assumptions. To individualize the estimate, adjustments are made for patient-specific factors including age, sex, symptoms, and clinical context, transforming a population-level probability into one tailored to the individual. For example, in assessing risk, a baseline might be elevated from under 10% to 55% for an elderly female patient post-surgery with symptoms like calf tenderness and , using structured adjustments via clinical prediction rules. These rules, such as the Wells score, incorporate weighted factors to refine the probability quantitatively. The process emphasizes that pre-test probability must be patient-centered, accounting for local and clinician judgment to avoid over-reliance on generic . A step-by-step estimation process typically starts with the base as the pre-test probability. Next, apply adjustments using , tables, or rules to integrate factors, yielding an updated probability. If preparing for Bayesian updating, convert this probability to by dividing it by (1 - probability); for a 20% probability, the odds are 0.25:1. Tools like the Fagan can visualize this logic by plotting pre-test probability against likelihood ratios, though the core relies on the sequential integration of and details. This methodical approach enhances diagnostic accuracy by grounding decisions in probabilistic reasoning.

Sources of Data

Primary sources for deriving pre-test probability estimates include systematic reviews and meta-analyses, such as those published by the Cochrane Collaboration, which aggregate data from multiple diagnostic test accuracy studies to inform prevalence and risk assessments. Disease registries provide representative population-level data on incidence and , offering robust foundations for estimating disease likelihood in specific cohorts. Large cohort studies, exemplified by the , yield long-term epidemiological data on cardiovascular risk factors and event rates, enabling precise pre-test probability calculations for conditions like . Secondary sources encompass clinical guidelines that compile prevalence tables derived from synthesized evidence, such as those from the National Institute for Health and Care Excellence (), which outline pre-test probabilities for stable based on age, sex, and symptom profiles. Similarly, the U.S. Preventive Services Task Force (USPSTF) incorporates baseline prevalence data into screening recommendations, adjusting for population risks in conditions like or cancer. Electronic health records (EHRs) serve as valuable local data repositories, allowing for real-time estimation of pre-test probabilities tailored to institutional or regional patient demographics through of historical diagnoses and testing patterns. For rare diseases, Bayesian priors can be drawn from global databases like Orphanet, which systematically surveys literature to estimate and incidence, providing essential starting points for low-probability scenarios. These estimates must be adjusted for regional variations, as seen in where is markedly higher in endemic areas like and compared to low-burden regions, influencing pre-test probabilities in diagnostic algorithms. Key challenges in using these sources include outdated data, such as pre-2020 estimates for respiratory diseases that fail to account for COVID-19's disruption of baseline prevalence and testing behaviors, potentially leading to inaccurate risk assessments. in studies and registries further complicates reliability, as overrepresentation of certain demographics can skew prevalence figures away from true population risks.

Post-test Probability Estimation

Predictive Values Approach

The predictive values approach to estimating post-test probability involves calculating the positive predictive value (PPV) and negative predictive value (NPV) of a diagnostic test, which directly provide the probability of disease presence or absence given a positive or negative test result, respectively. The PPV is defined as the probability that a patient has the disease given a positive test result, while the NPV is the probability that a patient does not have the disease given a negative test result. These values are derived from but are practically computed using the test's along with the pre-test probability () of disease in the population. The formula for PPV is: PPV=sensitivity×pre-test probability(sensitivity×pre-test probability)+((1−specificity)×(1−pre-test probability))\text{PPV} = \frac{\text{sensitivity} \times \text{pre-test probability}}{(\text{sensitivity} \times \text{pre-test probability}) + ((1 - \text{specificity}) \times (1 - \text{pre-test probability}))} The NPV is calculated similarly: NPV=specificity×(1−pre-test probability)(specificity×(1−pre-test probability))+((1−sensitivity)×pre-test probability)\text{NPV} = \frac{\text{specificity} \times (1 - \text{pre-test probability})}{(\text{specificity} \times (1 - \text{pre-test probability})) + ((1 - \text{sensitivity}) \times \text{pre-test probability})} In application, these formulas can be derived directly from a 2x2 , which categorizes test outcomes as true positives (TP), false positives (FP), true negatives (TN), and false negatives (FN). Here, PPV = TP / (TP + FP) and NPV = TN / (TN + FN), with the table populated based on sensitivity, specificity, and pre-test probability to reflect population counts. A key characteristic of this approach is that predictive values depend heavily on the pre-test probability, unlike which are intrinsic to the test; low pre-test probabilities can substantially reduce PPV even for highly accurate tests. For instance, a test with 90% sensitivity and 90% specificity yields a PPV of approximately 50% when the pre-test probability is 10%, highlighting the need for adequate disease prevalence to achieve clinically useful positive predictions.

Likelihood Ratios Approach

The likelihood ratio positive (LR+) is defined as the ratio of the probability of a positive test result in patients with the disease to the probability of a positive test result in patients without the disease, mathematically expressed as LR+=sensitivity1−specificity.LR^+ = \frac{\text{sensitivity}}{1 - \text{specificity}}. Similarly, the likelihood ratio negative (LR-) is the ratio of the probability of a negative test result in patients with the disease to the probability of a negative test result in patients without the disease, given by LR−=1−sensitivityspecificity.LR^- = \frac{1 - \text{sensitivity}}{\text{specificity}}. These ratios quantify how much a test result changes the odds of disease presence; values of LR+ greater than 1 increase the odds, while LR- less than 1 decreases them. To update pre-test probability to post-test probability using likelihood ratios, first convert the pre-test probability to pre-test odds by dividing the probability by (1 minus the probability). The post-test odds are then calculated as pre-test odds multiplied by the appropriate likelihood ratio (LR+ for positive test, LR- for negative). Finally, convert post-test odds back to probability using the formula probability = odds / (1 + odds). This Bayesian updating process allows modular application independent of disease prevalence. Consider a presenting with in the , where the pre-test probability of (MI) is estimated at 20%, corresponding to pre-test of 0.2 / 0.8 = 0.25 (or 1:4). Suppose an electrocardiogram (ECG) shows new ST-segment , with an LR+ of 5. The post-test are then 0.25 × 5 = 1.25 (or 5:4). Converting to probability yields 1.25 / (1 + 1.25) ≈ 0.56, or approximately 56% post-test probability of MI—effectively doubling the initial suspicion and often warranting urgent intervention. This example illustrates how even moderate LR values can substantially shift clinical suspicion. Unlike positive predictive values and negative predictive values, which depend on disease prevalence and thus vary by setting, likelihood ratios are inherent properties of the test itself, enabling their pooling across studies via similar to risk ratios. For continuous or multicategory tests, interval likelihood ratios can be derived for specific result thresholds, preserving more information than dichotomizing outcomes. Likelihood ratios can be inaccurate due to spectrum bias, where test performance () varies across patient populations differing in disease severity or spectrum, leading to over- or underestimation in dissimilar clinical settings. Additionally, may inflate reported LR values, as studies with statistically significant or favorable results are more likely to be published.

Advanced Estimation Techniques

Diagnostic Criteria

Standardized diagnostic criteria provide structured frameworks for quantifying pre-test probabilities by using symptom checklists that assign incremental risk based on clinical features met. For instance, the criteria for disorders, such as , require meeting at least five of nine symptoms over a two-week period to establish a presumptive , with partial fulfillment informing a lower pre-test probability of the condition. Similarly, the Rome IV criteria for gastrointestinal disorders, like , define the disorder through recurrent (at least 1 day per week in the last 3 months) associated with two or more changes related to defecation, stool frequency, or form, where fewer symptoms suggest a reduced pre-test likelihood of functional gastrointestinal issues. These checklists enable clinicians to estimate disease probability before confirmatory testing by tallying met criteria against established thresholds. In the process, each met criterion contributes to an overall score, often with weighted points for more indicative features, yielding a pre-test probability that stratifies risk levels. The total score categorizes patients into low, moderate, or high pre-test probability groups, which then guide the selection and interpretation of diagnostic tests to refine the estimate post-test. For example, the Wells criteria for deep vein thrombosis assign points for factors like active cancer (+1 point), immobilization (+1 point), and calf swelling >3 cm (+1 point), among others, with the aggregate informing the initial probability before imaging or blood tests. This weighted approach ensures that more salient symptoms disproportionately influence the pre-test assessment, promoting consistent probability estimation across clinical settings. These criteria emerged from consensus-driven processes in the late 20th and early 21st centuries, as medical communities sought to standardize disparate diagnostic practices through expert panels and iterative refinements. A seminal example is the Wells criteria for DVT, where a score greater than 2 corresponds to a 75% pre-test probability of the condition in validation cohorts. Such scoring systems have since become integral to evidence-based diagnostics, balancing simplicity with prognostic accuracy. However, these criteria may introduce subjectivity in symptom assessment and are generally less precise than statistically derived clinical prediction rules. Integration with Bayesian reasoning allows these pre-test probabilities to be updated post-test; for instance, a positive result (likelihood ratio ~3-5) applied to a Wells score-derived moderate (17%) pre-test probability elevates the post-test probability to approximately 40-60%, often warranting further imaging. This method extends to more statistically validated clinical rules, enhancing precision in probability adjustments.

Clinical Prediction Rules

Clinical prediction rules (CPRs) are statistically derived tools that estimate the probability of a specific clinical outcome, such as the presence of a , using multiple predictor variables including patient history, findings, and initial test results. These rules are developed through multivariate analysis, typically for binary outcomes, to identify independent predictors and quantify their combined effect on probability. Unlike simpler diagnostic criteria based on expert consensus, CPRs rely on empirical data from large cohorts to assign weights or points to variables, enabling a more precise pre-test probability assessment. Development of CPRs begins with prospective observational studies in representative populations, where potential predictors are evaluated for their association with the target outcome. Researchers use multivariable to select variables with significant independent predictive value, often applying techniques like backward elimination to refine the model and avoid . The resulting model is simplified into a scoring system, where predictors are assigned points based on their regression coefficients (e.g., rounded to integers for clinical ). A seminal example is the HEART score for predicting major adverse cardiac events in patients with suspected (ACS). In this rule, derived from a cohort of over 2,000 patients, variables are scored as follows: (0–2 points), ECG changes (0–2), Age ≥65 years (0–2), Risk factors (0–2), and elevation (0–2). A total score of 0–3 indicates low risk (approximately 1–2% probability of ACS within 6 weeks), 4–6 moderate risk (12–17%), and ≥7 high risk (50–65%). This point-based system facilitates rapid bedside calculation without needing computational tools. Validation is essential to ensure CPRs perform reliably beyond the derivation cohort, assessing metrics like calibration (agreement between predicted and observed probabilities) and discrimination (ability to distinguish outcomes, often via the C-statistic). Internal validation uses techniques such as on the same dataset, while external validation applies the rule to independent populations, which may involve temporal (same setting, different time), geographical (different locations), or domain (different patient groups) testing. However, external validation remains uncommon; a of clinical prediction models found that only about 16% undergo external validation after initial development, highlighting a gap in rigorous evaluation during the and beyond. For the HEART score, prospective external validations in diverse cohorts, including U.S. emergency departments, have confirmed good (C-statistic ≈0.76) and , supporting its use across settings. In clinical practice, CPRs provide a direct estimate of pre-test probability, standardizing subjective judgments and stratifying patients into categories to guide initial . These probabilities can then be updated to post-test values by incorporating results from additional diagnostic tests using likelihood ratios, integrating seamlessly with Bayesian approaches for refined .

Clinical Applications

Decision-Making Processes

In evidence-based medicine (EBM), pre- and post-test probabilities play a central role in guiding clinical decisions about whether to perform a diagnostic test or initiate treatment. These probabilities help clinicians determine if the potential post-test probability after testing would exceed a treatment threshold, beyond which the benefits of intervention outweigh the risks; for instance, if the post-test probability remains below this threshold, unnecessary testing or treatment can be avoided to prevent harm and reduce costs. Furthermore, they integrate with patient utilities—such as preferences for quality of life and risk tolerance—within decision tree models, allowing for a structured evaluation of expected outcomes under uncertainty. Clinical decision-making processes incorporating these probabilities often emphasize shared decision-making, where clinicians discuss pre- and post-test estimates with patients to align choices with individual values and circumstances. This approach facilitates by framing diagnostic uncertainty in accessible terms, such as the relative costs of over- versus under-treatment. Studies indicate that using probability-based reasoning improves diagnostic accuracy by enhancing the synthesis of clinical evidence, with one analysis suggesting reductions in overestimation of disease likelihood that could otherwise lead to inappropriate actions. A representative example is the evaluation of suspected (PE), where pre-test probability, assessed via tools like the Wells score, determines the need for imaging. If the pre-test probability exceeds 15% (indicating moderate to high risk), guidelines recommend proceeding to computed tomography pulmonary angiography; the test result then updates this to a post-test probability, which may rule in or out PE and guide anticoagulation decisions. The use of pre- and post-test probabilities in clinical decision-making was popularized in the late 1980s through David Sackett's foundational work on the rational clinical examination, which emphasized evidence-based appraisal of history and physical findings to refine diagnostic probabilities.

Test Thresholds

Test thresholds represent the pretest probability of disease at which the expected benefit of performing a diagnostic equals the expected harm, guiding clinicians on whether to proceed with testing based on harm-benefit trade-offs. This concept, introduced in the threshold approach to clinical decision making, defines two key probabilities: the threshold, below which no testing is warranted as the disease is unlikely, and the treatment threshold, above which intervention is justified without additional testing. For low-risk tests with minimal patient discomfort or side effects, such as routine blood draws, the threshold is typically in the range of 5-10%, reflecting the low harm relative to potential diagnostic gains. In contrast, tests involving greater risks, such as with its associated , may have adjusted thresholds around 1% to account for the incremental cancer risk from , estimated at approximately 125 cases per 100,000 women screened annually from ages 40 to 74. The calculation of the test threshold incorporates the relative harms and benefits, often approximated as the ratio of the harm from a false-positive test result to the sum of the test's benefit and the harm from a false-negative result: Threshold=test false-positive harmtest benefit+false-negative harm\text{Threshold} = \frac{\text{test false-positive harm}}{\text{test benefit} + \text{false-negative harm}} This formula balances the costs of unnecessary testing (e.g., anxiety, follow-up procedures, or direct harms like radiation) against the value of confirming or ruling out disease to avoid missed diagnoses. The exact value varies by test characteristics and clinical scenario; for instance, tests with high specificity minimize false-positive harms, lowering the threshold, while those with significant risks raise it. Similarly, the treatment threshold is derived analogously, as the point where treatment benefits outweigh harms, commonly expressed as Treatment threshold=treatment harmtreatment benefit+treatment harm\text{Treatment threshold} = \frac{\text{treatment harm}}{\text{treatment benefit} + \text{treatment harm}}, often yielding values like 16.7% when treatment reduces mortality by 5% but carries a 1% risk of adverse events. In application, if the pretest probability exceeds the test threshold but falls below the treatment threshold, the diagnostic test is performed to refine the probability; a post-test probability above the treatment threshold then prompts intervention. This sequential approach ensures testing is only pursued when it can meaningfully alter management. For example, in evaluating , clinicians might set a test threshold around 5-10% (e.g., for chest ) and a treatment threshold near 20-40%, based on empirical estimates where unnecessary antibiotics pose moderate harm but missing carries high risk; studies show physicians' implicit thresholds average about 9.5% for testing and 43% for treatment in acute cough scenarios.

Limitations

Subjectivity Issues

Clinician bias introduces significant subjectivity into pre-test probability estimation, often leading to systematic errors in diagnostic reasoning. One prominent example is the , where clinicians overestimate the likelihood of diseases that are more vivid or recently encountered in their practice, such as dramatic cases of following media reports, while underestimating less memorable conditions. This bias stems from reliance on personal recall rather than epidemiological data, distorting probability assessments and contributing to inconsistent across providers. Inter-observer variability further exacerbates these issues, with studies demonstrating substantial disagreement among clinicians when estimating pre-test probabilities for the same scenarios. For instance, in assessments of suspected , independent physicians categorized clinical probability differently in approximately 25-31% of cases, yielding moderate agreement levels with weighted values of 0.54 to 0.60. Similarly, surveys of practicing clinicians reveal wide ranges in estimates, from 5% to 100% for common conditions like deep vein thrombosis, with only 6.7-12% of responses falling within 20 percentage points of evidence-based values. Notably, clinical experience does not mitigate this variability; specialists often exhibit even greater dispersion in their estimates compared to residents, with standard deviations up to 21% across scenarios. Such subjectivity has profound impacts on , frequently resulting in unnecessary testing when fear or overestimation overrides low pre-test probabilities, thereby increasing burden, , and costs without improving outcomes. For example, practitioners commonly overestimate disease likelihoods—such as 80% for versus evidence-based 25-42%—prompting redundant diagnostics in low-risk cases. Overconfidence in high pre-test scenarios compounds this, as meta-analyses indicate clinicians routinely exceed actual accuracy in probability judgments, leading to diagnostic errors. Mitigation strategies focus on structured approaches to enhance reliability, including in probabilistic thinking and the of decision aids like clinical prediction rules. These interventions correct for biases by anchoring estimates to objective data, such as and factors, and have been shown to improve agreement; for instance, using quantitative pre-test tools reduced unnecessary in evaluations by 24%, from 33% to 25% of low-risk patients. Compared to subjective estimates, guideline-based methods like simplified clinical models increase inter-observer from 0.23 (unaided) to 0.60-0.66, substantially narrowing variability and promoting more consistent application of evidence.

Error Sources

Verification bias arises when only patients with positive test results are referred for confirmatory gold standard testing, leading to overestimation of sensitivity and underestimation of specificity, which in turn distorts likelihood ratios used in pre- and post-test probability calculations. This selective verification often occurs in resource-limited settings where not all negatives are confirmed, resulting in incomplete data on true negatives and false positives. For instance, in studies of tests for cancer, this can lead to inflated estimates of test accuracy by excluding mild or ambiguous cases from verification. Spectrum bias occurs when the study population does not represent the full range of disease severity or patient characteristics encountered in clinical practice, often by including only severe cases or healthy controls, which overestimates diagnostic performance metrics like the positive likelihood ratio (LR+). This mismatch in patient spectrum can substantially inflate LR+ estimates in non-representative cohorts. Reviews have highlighted that such biases are prevalent in diagnostic studies, where real-world patients include comorbidities and atypical presentations not captured in controlled trials. Data-related errors include using outdated prevalence estimates for pre-test probabilities, which can misalign calculations with current epidemiological realities, such as shifts observed post-COVID-19 pandemic where disease incidence varied dramatically by region and time. For example, using outdated rates post-COVID-19 may overestimate post-test probabilities, leading to inappropriate clinical decisions. Common calculation errors in applying or likelihood ratios can lead to over- or underestimation of disease likelihood. In continuous diagnostic tests, such as biomarker assays, errors arise from ignoring the spectrum of consequences— the varying clinical costs of false positives versus false negatives—when selecting thresholds, which can lead to suboptimal probability thresholds that do not balance diagnostic utility with patient outcomes. These inaccuracies in likelihood ratio applications exemplify broader methodological flaws in . To mitigate these errors, sensitivity analyses should be conducted to assess how variations in or bias assumptions affect post-test probabilities, providing a range of plausible outcomes. Robust study designs, including consecutive patient enrollment and full verification protocols, help minimize test-related biases by ensuring representative samples and complete reference standard application.

References

  1. https://www.ncbi.nlm.nih.gov/books/NBK235178/
  2. https://pmc.ncbi.nlm.nih.gov/articles/PMC4744617/
  3. https://bmcmedinformdecismak.biomedcentral.com/articles/10.1186/s12911-024-02610-3
  4. https://pmc.ncbi.nlm.nih.gov/articles/PMC5950618/
  5. https://depts.washington.edu/physdx/eglossary.html
  6. https://hekint.org/2018/01/22/thomas-bayes-bayes-theorem-medicine/
  7. https://pmc.ncbi.nlm.nih.gov/articles/PMC3153801/
  8. https://www.degruyterbrill.com/document/doi/10.12987/9780300175097-010/html
  9. https://pmc.ncbi.nlm.nih.gov/articles/PMC3427763/
  10. https://www.jclinepi.com/article/S0895-4356(21)00346-2/fulltext
  11. https://bmcmedresmethodol.biomedcentral.com/articles/10.1186/s12874-020-00952-w
  12. https://pmc.ncbi.nlm.nih.gov/articles/PMC3673738/
  13. https://www.uspreventiveservicestaskforce.org/uspstf/recommendation/screening-for-prediabetes-and-type-2-diabetes
  14. https://aspe.hhs.gov/reports/feasibility-using-electronic-health-data-research-small-populations-0
  15. https://www.orpha.net/pdfs/orphacom/cahiers/docs/GB/Prevalence_of_rare_diseases_by_alphabetical_list.pdf
  16. https://pmc.ncbi.nlm.nih.gov/articles/PMC7276051/
  17. https://www.ncbi.nlm.nih.gov/books/NBK557491/
  18. https://pmc.ncbi.nlm.nih.gov/articles/PMC10293502/
  19. https://pmc.ncbi.nlm.nih.gov/articles/PMC11633789/
  20. https://pmc.ncbi.nlm.nih.gov/articles/PMC2636062/
  21. https://pmc.ncbi.nlm.nih.gov/articles/PMC6664615/
  22. https://pmc.ncbi.nlm.nih.gov/articles/PMC478236/
  23. https://bmcresnotes.biomedcentral.com/articles/10.1186/1756-0500-5-420
  24. https://www.aafp.org/pubs/afp/issues/2017/0201/p170.html
  25. https://theromefoundation.org/rome-iv/rome-iv-criteria/
  26. https://pmc.ncbi.nlm.nih.gov/articles/PMC5704116/
  27. https://pubmed.ncbi.nlm.nih.gov/9428249/
  28. https://www.mdcalc.com/calc/362/wells-criteria-dvt
  29. https://pmc.ncbi.nlm.nih.gov/articles/PMC4810039/
  30. https://pubmed.ncbi.nlm.nih.gov/15142977/
  31. https://www.sciencedirect.com/science/article/pii/S0049384812003462
  32. https://pmc.ncbi.nlm.nih.gov/articles/PMC6704664/
  33. https://www.sciencedirect.com/science/article/pii/S016752731300315X
  34. https://www.jacc.org/doi/10.1016/j.jacc.2018.07.059
  35. https://www.sciencedirect.com/science/article/pii/S0735109710022355
  36. https://link.springer.com/article/10.1007/s11606-008-0623-z
  37. https://doi.org/10.1186/s12911-024-02610-3
  38. https://www.ahrq.gov/sites/default/files/wysiwyg/patient-safety/reports/issue-briefs/dxsafety-probabilistic-thinking.pdf
  39. https://doi.org/10.1001/jamainternmed.2021.0269
  40. https://pubmed.ncbi.nlm.nih.gov/1573753/
  41. https://sk.sagepub.com/ency/edvol/medical/chpt/testtreatment-threshold
  42. https://www.acpjournals.org/doi/10.7326/M15-1241
  43. https://ebm.bmj.com/content/28/6/392
  44. https://pmc.ncbi.nlm.nih.gov/articles/PMC6193794/
  45. https://www.qualityhealth.org/wp-content/uploads/2021/03/Annals-Avoiding-cognitive-errors.pdf
  46. https://first10em.com/decision-making-in-emergency-medicine-availability-bias/
  47. https://www.jthjournal.org/article/S1538-7836(22)18022-0/fulltext
  48. https://www.mja.com.au/journal/2004/180/9/generating-pre-test-probabilities-neglected-area-clinical-decision-making
  49. https://www.sciencedirect.com/science/article/abs/pii/S0895435605001502
  50. https://www.merckmanuals.com/professional/special-subjects/clinical-decision-making/understanding-medical-tests-and-test-results
  51. https://jamanetwork.com/journals/jamainternalmedicine/fullarticle/2778364
  52. https://www.apa.org/pubs/journals/features/cou-0000105.pdf
  53. https://www.ahajournals.org/doi/10.1161/circimaging.113.001080
  54. https://catalogofbias.org/biases/verification-bias/
  55. https://jamanetwork.com/journals/jama/fullarticle/191668
  56. https://pmc.ncbi.nlm.nih.gov/articles/PMC6738935/
  57. https://catalogofbias.org/biases/spectrum-bias/
  58. https://onlinelibrary.wiley.com/doi/10.1111/acem.12255
  59. https://asm.org/articles/2020/june/why-pretest-and-posttest-probability-matter-in-the
  60. https://www.nature.com/articles/s43856-024-00691-9
  61. https://www.acpjournals.org/doi/10.7326/0003-4819-140-3-200402030-00010
  62. https://bmcmedresmethodol.biomedcentral.com/articles/10.1186/s12874-021-01481-w
Add your contribution
Related Hubs
User Avatar
No comments yet.