The probability of pre-test and post-test probability (alternate pretest and posttest probabilities) is the probability of a condition (such as a disease) before and after the diagnostic test, respectively. Post-test probabilities , in turn, can be either positive or negative , depending on whether the test falls as a positive test or a negative test, respectively. In some cases, it is used for the possibility of developing interesting conditions in the future.
The test, in this sense, can refer to medical tests (but usually in the sense of diagnostic tests), and in a broad sense also includes questions and even assumptions (such as the assumption that the target individual is female or male). The ability to make the difference between pre and post-test probabilities of various conditions is a major factor in the indication of medical tests.
Video Pre- and post-test probability
Probabilitas pre-test
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 selected if no other characteristics are known to the individual, or can be selected for ease of calculation even if other characteristics are known even though such omissions may lead to inaccurate results
- Post-time probability of conditions resulting from one or more previous tests
- Rough estimates, which may have to be selected if a more systematic approach is not possible or efficient
Maps Pre- and post-test probability
Post-test probability estimate
In clinical practice, post-test probabilities are often predicted or even estimated. This is usually acceptable in finding signs or symptoms of
In fact, however, the subjective probability of the existence of a condition is never exact 0 or 100%. However, there are several systematic methods for estimating that probability. Such methods are usually based on previously tested reference groups where the presence or absence of such conditions is known (or at least predicted by other tests deemed to be very accurate, such as the "Gold standard"), in order to establish test performance data. These data are then used to interpret the test results of each individual tested by this method. An alternative or complement to the reference group method is to compare test results with previous tests on the same individual, which is more common in tests for monitoring.
The most important systematic systematic group method for estimating post-test probabilities includes those summarized and compared in the following table, and described further in the individual sections below.
With predictive value
Predictive values ​​can be used to estimate the probability of an individual's post-test if individual pre-test probabilities can be assumed to be approximately equal to the prevalence in the reference group in which both test results and knowledge of the presence or absence of conditions (eg disease, as may be determined by "Gold standard") is available.
If the test result is a binary classification into a positive or negative test, then the following table can be created:
The pre-test probability can be calculated from the diagram as follows:
Probabilitas pretest = (True positive False negative)/Total sampel
Also, in this case, positive post-test probability (likely to have a target condition if the test falls positively), numeric equals positive predictive value, and negative post-test probability probability of having a target condition if the test falls negatively) numerically completes the negative predictive value ([negative post-test probability] = 1 - [negative predictive value]), again assuming that the individual tested has no other risk factor producing an individual who has the pre-test probabilities are different from the reference group used to establish the positive and negative predictive value of the test.
In the diagram above, this positive post-test probability , ie the posttest probability of the target condition given the positive test result, is calculated as:
Positive posttest probability = True positive/(True positive Positive error)
Similarly:
The post-test probability of the disease given negative results is calculated as:
Negative posttest probability = False negative/(Negative false Negative true)
The validity of the above equation also depends on that the sample of the population has no substantial sampling bias which makes their group have conditions and those that are substantially disproportionate of the prevalence of the corresponding and "non-prevalent" in the population.. Consequently, the above equation does not apply only to case-control studies that separately collect one group with conditions and one group without it.
With possible ratios
The above method is inappropriate to use if the pretest probability differs from the prevalence in the reference group used to establish, inter alia, the positive predictive value of the test. The difference may occur if another test is preceded, or the person involved in the diagnostic considers that another pretest probability should be used because of knowledge about, for example, a specific complaint, other elements of medical history, signs in physical examination, either by counting on each finding as the test itself with its own sensitivity and specificity, or at least make a rough estimate of individual pre-test probabilities.
In this case, the prevalence in the reference group is not entirely accurate in representing the pre-test probabilities of the individual, and, consequently, the predictive value (whether < i> positive or negative ) is not entirely accurate in representing the post-test probability of an individual with a target condition.
In this case, the probability of posttest can be estimated more accurately by using the likelihood ratio for the test. The likelihood ratio is calculated from the sensitivity and specificity of the test, and thus it does not depend on the prevalence in the reference group, and, likewise, it does not change with the changing pre-test probabilities , different from positive or negative predictive value (which will change). Also, in essence, the validity of the post-test probabilities determined from the probability ratios are not susceptible to the sampling bias with regard to them with and without conditions in the sample population, and can be done as a case. -control studies that separately collect them with and without conditions.
The post-test probability estimate of the probability of pre-test and the likelihood ratio occurs as follows:
- Initial probability = (Pretest probability/(1 - Pretest probability)
- Possible endpoint = Pretension opportunity * Possibility ratio
In the above equation, positive post-test probability is calculated using the positivity ratio , and negative post-test probability is calculated using the possible negative .
- Posttest probability = Possible posttest/(Odds final test 1)
This relationship can also be estimated by the so-called Fagan nomogram (shown right) by making a straight line from the given pre-test probability point to a given the likelihood ratio on their scale, which, in turn, estimates the probability of post-test at the point where the straight line cuts the scale.
Post-test probabilities can, in turn, be used as a pre-test probability for additional tests if it continues to be calculated in the same way.
It is possible to calculate the likelihood ratio for testing with continuous values ​​or more than two results similar to those for dichotomous results. For this purpose, a separate likelihood ratio is calculated for each test result level and is called the interval-specific or strata-specific likelihood ratio.
Example
A person is screened by an occult fecal blood test (FOB) to estimate the probability for a person who has a targeted condition of bowel cancer, and it falls positively (blood is detectable in the stool). Prior to the test, the individual had a pre-test probability of having bowel cancer, for example, 3% (0.03), as might be expected by evaluation, eg, previous medical history, examination and test of the individual..
The sensitivity, specificity etc of the FOB test were established with a sample population of 203 persons (without such offspring), and fell as follows:
From this, the possible ratio of tests can be set:
- Positive probability ratio = sensitivity/(1 - specificity) = 66.67%/(1 - 91%) = 7.4
- Negative possibility ratio = (1 - sensitivity)/specificity = (1 - 66.67%)/91% = 0.37
- Pretest probability (in this example) = 0.03
- Initial probability = 0.03/(1 - 0.03) = 0.0309
- Odds positive posttest = 0.0309 * 7.4 = 0.229
- Positive posttest probability = 0.229/(0.229 1) = 0.186 or 18.6%
Thus, the individual had a post-test probability (or "post-test risk") of 18.6% having bowel cancer.
Prevalence in population samples is calculated as:
- Prevalence = (2 1)/203 = 0.0148 or 1.48%
The probability of individual pre-test is more than twice that of one population sample, although the probability of individual post-test is less than twice that of one population sample (estimated with a positive predictive value of 10% test) as opposed to what would be produced by a lesser method accurate by simply multiplying the relative risk.
Source of specific inaccuracy
The sources of special inaccuracies when using the likelihood ratio for determining post test probabilities include interference with prior determinants or tests or overlapping test targets, as described below:
Interference with test
The post-test probabilities , as predicted from the pre-test probabilities with the likelihood ratio , should be handled with caution in individuals with other determinants (eg as a risk factor) than the general population, as well as in individuals who have had previous tests, because the determinant or test may also affect the test itself in unpredictable ways, still causing inaccurate results. An example with obesity risk factors is that additional abdominal fat may make it difficult to palpate abdominal organs and reduce abdominal ultrasound resolution, and similarly, residual barium contrast from previous radiographs may impair subsequent abdominal examinations, which essentially reduces the sensitivity and specificity of subsequent tests. On the other hand, the interference effects have the potential to increase the efficacy of subsequent tests compared with use in the reference group, as some abdominal examinations become easier when performed on thin people.
Overlap testing
Furthermore, the validity of the calculations on each pre-test probability derived from the previous test depends on that the two tests do not significantly overlap in terms of the tested target parameters, such as the blood test of one substance and the same crazy metabolic pathway. An extreme example of such overlap is where sensitivity and specificity have been assigned to blood tests that detect "substance X", and also to one detect "substance Y". If, in fact, "substance X" and "substance Y" are one and the same substance, then, making two successive tests of one and the same substance may have no diagnostic value at all, even though the calculations seem to indicate a difference. In contrast to the disruption described above, increased overlap tests only reduce their efficacy. In medical settings, diagnostic validity increases by combining different modality tests to avoid substantial overlap, for example in making a combination of blood tests, biopsy and radiography.
Method to resolve inaccuracies
To avoid the source of such inaccuracies by using the likelihood ratio, the optimal method is to collect a large reference group of equivalent individuals, to set a separate predictive value for the use of the test on that individual. However, with more knowledge about individual medical history, physical examination and previous tests etc. That the individual becomes more differentiated, with the increasing difficulty of finding reference groups to set customized prediction values, making probability estimates post test with invalid predictive value.
Another method of dealing with inaccuracies is to evaluate the test results in the context of the diagnostic criteria, as described in the next section.
With relative risk
Post-test probabilities can sometimes be estimated by multiplying the probability of pre-test with the relative risks given by the test. In clinical practice, this is usually applied in the evaluation of an individual's medical history, where "tests" are usually questions (or even assumptions) about various risk factors, eg sex, smoking or weight, but can potentially be substantial tests such as putting individuals on a weighing scale. When using relative risk, the resulting probability is usually somewhat related to the individual developing the condition over a period of time (similar to the occurrence in the population), rather than the individual probability of having a condition in the present, but indirectly becoming the last estimate.
The use of hazard ratio can be used somewhat similar to relative risk.
One risk factor
To establish relative risk, the risks in the exposed group are divided by risks in the unexposed group.
If only one individual risk factor is taken into account, post-test probabilities can be estimated by multiplying the relative risk with the risk in the control group. The control group usually represents an unexposed population, but if very low fractions of the population are exposed, then prevalence in the general population can often be assumed to be the same as the prevalence in the control group. In such cases, post-test probabilities can be estimated by multiplying the relative risk with risk in the general population.
For example, the incidence of breast cancer in a woman in England from age 55 to 59 is estimated to be about 280 cases per 100,000 per year, and risk factors have been exposed to high dose ionizing radiation to the chest (for example, as a treatment for other cancers) giving a relative risk of breast cancer between 2.1 and 4.0, compared with not exposed. Because a small proportion of the population is exposed, prevalence in unexposed populations can be assumed to be the same as prevalence in the general population. Furthermore, it can be estimated that a woman in England between the ages of 55 and 59 and who has been exposed to high doses of ionizing radiation should have a risk of developing breast cancer for a period of one year between 588 and 1.120. 100,000 (ie, between 0.6% and 1.1%).
Multiple risk factors
Theoretically, the total risk with the existence of some risk factors can be approximated roughly by multiplying with each relative risk, but generally much more inaccurate than using the likelihood ratio, and is usually done simply because it is much easier to do when only relative risk is given, by, for example, converting source data to sensitivity and specificity and calculating with the likelihood ratio. Likewise, relative risk is often given rather than the ratio of possibilities in the literature because the former is more intuitive. The source of inaccuracy doubles the relative risks involved:
- The relative risk is affected by the prevalence of the conditions in the reference group (in contrast to the likelihood ratio, which is not), and this problem results in that the probability validity of the post test becomes less valid with the increasing difference between the prevalence in the reference group and the probability of pre-test for each individual. Any known risk factor or prior test of an individual almost always confers the difference, reducing the validity of using relative risk in estimating the total effect of some risk factors or tests. Most physicians do not accurately consider differences in prevalence when interpreting test results, which can lead to unnecessary diagnostic testing and error.
- The inaccurate source of inaccuracies to multiply some relative risk, considering only positive tests, is likely to exaggerate total risk compared to using the likelihood ratio. This overestimation can be explained by the inability of the method to compensate for the fact that the total risk should not 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 age younger than 40 years in women in the UK can be estimated at about 2%. Also, studies on Ashkenazi Jews have shown that mutations in BRCA1 provide a relative risk of 21.6 developing breast cancer in women under 40, and mutations in BRCA2 provide a relative risk of 3.3 developing breast cancer in women under 40. years of age. From this data, it can be estimated that a woman with BRCA1 mutation will have about 40% of breast cancer risk developing at younger than 40 years, and women with BRCA2 mutations will have a risk of about 6%. However, in situations that are somewhat improbable to have both , BRCA1 and BRCA2 mutations, simply multiplying with both the relative risk will result in the risk of more than 140% of breast cancers developing before age 40, which may not be accurate in reality.
The (latter) effect of being too high can be compensated by converting risk into opportunity, and risk relative to the odds ratio. However, this does not offset the (previously mentioned) effects of differences between the pre-test probabilities of individuals and the prevalence in reference groups.
The method to compensate both sources for inaccuracy is to establish relative risk with multivariate regression analysis. However, in order to maintain its validity, such a set relative risk must be multiplied by all other risk factors in the same regression analysis, and without the addition of other factors outside the regression analysis.
In addition, multiplying some relative risk poses the same risk as the significant overlap loss of the included risk factors, similar to when using the likelihood ratio. Also, different risk factors can act synergistically, with the result that, for example, two factors that both individually have a relative risk 2 have a total relative risk 6 when both exist, or can inhibit each other, somewhat similar to the disorder described using the likelihood ratio.
With diagnostic criteria and clinical prediction rules
Most diseases have established diagnostic criteria and/or clinical prediction rules. The establishment of diagnostic criteria or clinical prediction rules comprises a comprehensive evaluation of many of the tests considered important in estimating the probability of an attractive condition, sometimes including how to divide it into subgroups, and when and how to treat the condition. Such formation may include the use of predictive value, probability ratio and relative risk.
For example, the ACR criteria for systemic lupus erythematosus define the diagnosis as the presence of at least 4 of the 11 findings, each of which can be regarded as the target value of the test with its own sensitivity and specificity. In this case, there has been a test evaluation for this target parameter when used in combination in terms of, for example, interference between them and overlapping the target parameters, thus attempting to avoid the inaccuracies that may arise if trying to calculate the probability of the disease using the likelihood ratio of the individual tests. Therefore, if diagnostic criteria have been established for a condition, it is generally best to interpret each post-test probability for the condition in the context of this criterion.
Also, there is a risk assessment tool to estimate the combined risks of several risk factors, such as the online tool [1] of the Framingham Heart Study to estimate the risk of coronary heart disease results using various risk factors including age, sex, blood lipid, blood pressure and smoking , is much more accurate than multiplying the relative risk of each of each risk factor.
However, experienced physicians can estimate the probability of post-test (and motivated action) by extensive considerations including criteria and rules in addition to other previously described methods, including individual risk factors and performance tests that have been performed..
Clinical use of pre and post-test probabilities
Clinically useful parameters are absolute (not relative, and not negative) differences between pre and post-test probabilities, calculated as:
Absolute difference = | (pre-test probability) - (post-test probability) |
The main factor for such absolute differences is the strength of the test itself, as can be explained in terms of, for example, sensitivity and specificity or the likelihood ratio. Another factor is the pre-test probability, with lower pre-test probabilities resulting in lower absolute differences, with the consequence that even very strong tests achieve a low absolute difference for conditions that are highly unlikely in individuals (such as rare diseases in no the presence of other indicating signs), but on the other hand, that even a low power test can make a big difference to a highly suspect condition.
Probability in this sense may also need to be considered in the context of conditions that are not the primary target of the test, such as relative-profile probabilities in differential diagnostic procedures.
The absolute difference can be attributed to the benefits for a person whose medical test reaches, as approximately can be estimated as:
, di mana:
- b n adalah manfaat bersih dari melakukan tes medis
- ? p adalah perbedaan mutlak before probable pra-dan pasca-uji dari kondisi (seperti penyakit) yang ingin dicapai tes.
- r i adalah tingkat berapa banyak perbedaan probabilitas yang diharapkan menghasilkan perubahan dalam intervensi (seperti perubahan dari " tidak ada perawatan "menjadi" administrasi perawatan medis dosis rendah ").
- b i adalah manfaat dari perubahan dalam intervensi untuk individu
- h i adalah bahaya perubahan dalam intervensi bagi individu, seperti efek samping dari perawatan medis
- h t adalah kerugian yang disebabkan oleh tes itu sendiri
In this formula, what constitutes benefits or dangers varies greatly according to personal and cultural values, but general conclusions can still be drawn. For example, if the only effect expected from a medical test is to make one disease more likely than another, but both diseases have the same (or untreatable) treatment, then r i = 0 and testing basically does not benefit the individual.
Additional factors that influence the decision whether medical tests should be performed or excluded: test costs, availability of additional tests, potential interference with subsequent tests (such as potentially intestinal palpation inducing bowel activity that disturbs subsequent auscultation of auscultation), time taken for the exam or aspect practical or other administrative. Also, even if it is not beneficial to the individual being tested, the results may be useful for statistical establishment to improve health care for other individuals.
Subjectivity
The pre- and post-test probabilities are subjective based on the fact that, in fact, an individual has a condition or not (with probability always being 100%), so the pre and post-test probabilities for the individual is better regarded as a psychological phenomenon in their minds involved in the diagnosis at hand.
See also
- Interpretation of diagnostic tests, including common sources of inaccuracy and inaccuracy
References
Source of the article : Wikipedia
0 Comments