Back to EPIDEMIOLOGYEveryday, medical practitioners are involved with situations, in which a diagnostic process is needed. This process is extremely complicated, and includes semiology, clinical knowledge, previous experience, and mainly, suppositions. For example, when doctors are in emergency rooms, they are in the front line of medical care. In this situation, they are supposed to reach a diagnosis and, in case it is possible, a treatment. In order to achieve this, since a patient enters the room the doctor is observing “glues” that will help later in diagnosis. The main purpose of interview and physical examination is to find out and exclude as much symptoms and signals of disease as possible, in order to obtain support for clinical speculation. During and after clinical examination, lots of patient’s characteristics are discovered, such as fever, diarrhea, fourth heart sound and many others. Each one leads us to think about some diseases that typically manifests itself in this special way.
These manifestations are the basis of the diagnostic process, and when associated to a large clinical knowledge, probably will result in a correct clinical policy. However, many of these prominent “glues” cannot be categorized in only one disease, but in a lot of pathologies. Consequently, further study is often necessary, such as laboratorial tests, radiological images, biopsies and etc. Choosing these exams is pivotal in clinical diagnosis, and is based on the most important hypothesis and on the properties of each exam. The same statement is used in clinical examination, history and tests, in which the positive and negative result of any information can guide differential diagnosis. The value of an exam, information or clinical presentation can be expressed by accuracy, sensitivity, specificity and predictive values. These epidemiological data can express the quality of certain positive or negative result in the determination or exclusion of diagnostic hypothesis and help to reach a correct diagnosis. Most useful epidemiological characteristics of a test are described below and represented in table 1.
The gold-standard is a test that is considered to be the most accurate among all the known tests. All the others should be compared with this test, in order to indicate whether they are reliable. This comparison is useful in most cases, because most gold-standards are expensive or invasive, so that less accurate tests are preferred, unless their accuracy is not acceptable in relation to important results of misdiagnosis. The gold-standard can be a test ( PCR ), a proceeding ( biopsy ) or even follow-up.
The accuracy of a test expresses exactly the quality of the test, and includes all the times that this test resulted in a correct result according to a gold-standard. It represents true positive and negative results among all the results of the test.
The sensitivity of a test is the proportion of positive results among all the patients that have certain disease. A sensitive test is the one that rarely loose a case ( a sick patient ). Whereas, the specificity of a test expresses the proportion of negative results among all the disease-free patients. A specific test rarely categorizes healthy persons as sick ones. The use of these tests depends on doctor’s purpose, because if there are a lot of differential diagnosis it would be expensive and maybe impossible to do specific test to all of them. In this case, the best policy is starting with sensitive tests, which will guide investigation, and when there are only a few hypothesis, specific tests can be helpful. High-sensitive tests are extremely useful in screening, such as ELISA to determine anti-HIV presence in blood donors. In this case the sensitivity is essential because if the test do not diagnose HIV donors, the consequences may be catastrophic. However, this test is not diagnostic of HIV presence, and further examination with specific tests is always necessary. For example, Western blot or other specific techniques are needed to determine if the patient that had a positive ELISA test is really HIV+.
The relation between specificity and sensitivity is characteristic. While one of them is increasing the other is decreasing. It is extremely important because when there are different values to consider a test as normal or abnormal ( cut-points ), you should trade-off the one that gives the necessary sensitivity and specificity.
Considering sensitivity and specificity you can choose what test is necessary or helpful, but when you have the results in hands, the most important information is predictive value. Results of a test can be positive or negative. In case the test is positive or abnormal, it is necessary to know some important information about the disease, and then calculate the positive predictive value, which expresses how many times the positive result of the test really represents disease. This value is better achieved with specific tests and depends on prevalence, sensitivity and specificity according to the equation:
Positive Sensitivity X Prevalence
predictive = ______________________________________________
value
(Sensitivity X Prevalence) + ( 1 - Specificity) X ( 1 - Prevalence)
On the other hand, negative predictive value is the probability of a negative result really correlates to a disease-free person. This result is better correlated with sensitive tests, and depends on prevalence.
This important information is extremely used in clinical diagnosis, because it can indicate that certain disease is common or uncommon, and then doctors can make their clinical investigation, starting from the most to the least probable hypothesis. Each data from clinical exam, history and other tests leads to certain probability of a disease. An information must be added to other in order to obtain a diagnose by making differential diagnosis. For example, a 55 years-old man, heavy smoker and alcoholic, with progressive dysphagia and weight loss in the past 2 years. The doctor requests laboratory tests and a CT scan. The CT scan shows an esophageal tumor. It is well known that the most common carcinoma of esophagus is the squamous cell carcinoma, and there are others, such as adenocarcinoma ( less common ) and some rare cancers. According to this prevalence data, the most probable cancer is squamous cell carcinoma. Then a biopsy was taken and this prevalence statement was confirmed. The history and clinical data leaded to some hypothesis with high prevalence among the patients with those clinical manifestations. The complementary exams gave a great probability.
The example showed how prevalence can be used in clinical practice, because some characteristics are associated with a greater prevalence of certain disease, such as smoking & cancer, dysphagia & esophageal disorders, age & cancer, age & heart disease, among others. This specific clinical situations are associated with a greater incidence and prevalence of certain disease and can be useful in differential diagnosis. Prevalence data can be founded in medical literature, and is always helpful. If a positive result of a test indicates a high-prevalence disease, probably this result is really positive. On the other hand, if the disease is uncommon, it can be a false-positive. The same statement is also useful to interpret negative results, in which a high-prevalence disease makes a false-negative probable, and a low-prevalence disease make this result probably true. Another important comment is that some care must be taken with reference hospitals and its produced literature in relation to prevalence, because in this cases, prevalences tend to be overestimated.
Characteristics of a test in relation to real disease (gold standard). The false-positive is represented by “b” and false negative result is represented by “c”. The formulas to calculate all the aforementioned test characteristics are also showed.
Disease |
|||
|---|---|---|---|
yes |
no |
||
Test |
positive |
a |
b |
negative |
c |
d |
|
The use of epidemiological approach is extremely important in clinical decision making, and the use of test properties to choose the right test and to interpret its results is essential. This properties are of great interest in diagnosis, because it can help practitioners to reach a diagnosis without spending time and money. Every medical student or practitioner should be aware of some important matters. For example, in the beginning of an investigation or in screening, the proper test is the one with a great sensitivity. At this time specificity is not so important, because you cannot loose the disease and you do not need to reach a final diagnosis. After the first results, doctors must balance pros and cons of each differential diagnose, and then choose specific tests to reach a final conclusion ( if it is possible ). When you have the result of a test, you should know the predictive values, in order to interpret this and confirm the previous hypothesis. This statements are useful in clinical examination and history, because the same characteristics of tests can be applied to physical exam, symptoms and etc. The use of this approach is extremely important because makes this process easier and systematic, and can simplify medical decision making.
Back to EPIDEMIOLOGY
Back to SPECIALTIES
Back to MEDSTUDENTS HOMEPAGE