Structured clinical question.

Any knowledge gap should be formulated as a structured clinical question. Its components are: patients, intervention, comparison and outcome.

…will never have answers. My biochemistry’s teacher taught me this almost two lives ago, when I was a medicine freshman. I don’t remember what else she taught me but I have this etched into my memory because, I don’t want to remember how many years later, it still remains valid.

And it turns out that the wheel of evidence-based medicine starts spinning with a question. Of course, the problem is that, in medicine, we do not always get an answer for a lot of questions, and according to some, in four out of five times we will not get a satisfactory answer, no matter how well we look for it.

Types of clinical question

We physicians, let’s face it, are pretty ignorant, and anyone who thinks otherwise is because he doesn’t know how ignorant he’s, which is much worse and more dangerous. We are often challenged by gaps in our knowledge that we want to fill with the available information. It has been estimated that, at Primary Care level, we ask two questions for every 10 patients we receive, increasing this number to five for each patient admitted to Hospital Care. It is easy to understand that we cannot do a bibliography search every time we have a question, so we will have to set priorities.

At our beginnings, when we are very, very ignorant, the questions are quite general. These are called background questions, seeking information on general aspects of diseases and treatments. They are usually composed of a root with a word like how, how much, when or something similar, and a verb followed by the disease or whatever we are dealing with. Questions of this kind are, for example, “what germ causes risperidiosis?” or “how do we treat a dander attack?”

In general, the answer to background questions can be found in textbooks or review articles. There are digital sources of reviews on general topics, such as the one that is undoubtedly one of the most worshiped: UpToDate. We will all meet some uptodater, who are people easily recognizable because, in the first hour of the morning, they already have the latest information obtained from UpToDate, so they give you the answer even before you have asked yourself the question.

But, as we become wiser, the questions that we ask start to involve specific aspects of treatment, prognosis or whatever of a disease in a given patient or population. These advanced or foreground questions often have characteristics that differ qualitatively from that of the background questions: they are usually asked as part of the clinical decision making when we are seeking for information about any problem we are interested in.

Structured clinical question

Therefore, it’s essential to set them properly and formulate them clearly because if not, they won’t serve to plan the search strategy and to make the right decision that we’re looking for. They are formed by what is known as a structured clinical question, also known in the jargon of evidence-based medicine as PICO questions, after the initial of its components, as we can see below.

P stands for patient, but also for the problem of interest or the clinical description of the situation that we are studying. We must define very well the most relevant characteristics of the group of patients or the population that originated the question, trying not to restrict too much the characteristics of the group, because it may happen that later we find nothing that answers the question. It is often preferable to select the population more generally and, if the search is unspecific (we have many results), we can always restrict it later.

I represents the main intervention, which can be a treatment, a diagnostic test, a risk factor or exposure, etc. C refers to the comparison with which we contrast the intervention, and may be another treatment, placebo or, sometimes, do nothing. This component is not mandatory in the structure of the question, so we can avoid it in cases that we do not need it.

Finally, O represents the outcome of clinical interest in our question, whether in terms of symptoms, complications, quality of life, morbidity, mortality, or any other outcome variable we choose. Thus, it is important to emphasize that the result that we choose should have importance from the clinical point of view, especially importance from the point of view of the patient. For example, in a study to prevent coronary disease, we can measure the effect by decreasing troponin, but the patient will certainly appreciate it more if we estimate the decrease in mortality from myocardial infarction.

Sometimes, as I have already said, it’s not relevant to do any comparison with anything, so PICO becomes PIO. Some people add a fifth parameter, the time, and PICO becomes PICOt. You can also see it as PECO or PECOt if you prefer to say exposure rather than intervention. But, no matter what letters you use, the important thing is to divide the question into its components, because these elements will be the ones that will determine the keywords for the search of information and the type of study design that we’ll need to find the answer (some people add the type of study design as a fifth or sixth letter to PICO).

The difficult balance between amplitude and precision

It’s very important to find a good balance between the scope and accuracy of the question. For instance, the question “in infants with cranial traumatism, do treatment with corticoids improve the prognosis?” may be too general to be of any use. In addition, “in 3-6 month-old infants who fall from the crib from 20 centimeters high and suffer a left side of his forehead traumatism against a carpeted floor, can we improve the prognosis using methylprednisolone at a dose of 2 mg/kg/day during five days?” seems to me as too specific to be used in the search strategy or to be useful for the clinical decision making. A better way of structure the question would be something like “in infants with minor cranial traumatism (minor trauma’s criteria must be previously stablished) does steroid treatment improve the prognosis?” P would be the infants who suffer the trauma, I the treatment with corticosteroids, C would be, in this case, not to give steroids and, finally, O would be the prognosis (which could be replaced by something more specific such as the probability of hospital admission, time until discharge, death, etc).

Let’s see another example: in (P) infants with bronchiolitis, (I) the use of intravenous corticosteroids, instead of inhaled (C), decreases the risk of hospital admission (O)?. Or this one: in (P) infants with otitis, does the use of antibiotics (I) shorten the duration of illness (O)?.

Types of structured clinical question

Depending on the type of answer that they are looking for, clinical questions can be classified into four basic types: diagnosis, treatment, prognosis and etiology/harm. Diagnostic questions are about how to select and interpret diagnostic tests. Treatment questions have to do with the treatment we can choose to provide more benefits than risks and with lower economic cost and resources. Prognosis questions give us the probability of a certain clinical course and anticipate complications. Finally, etiology/harm questions are those that serve to identify the causes of diseases, including iatrogenic.

The type of question is important because it will define the type of study design that most likely will answer to our question. Thus, diagnostic questions are best answered with studies with a design that is specific for the evaluation of diagnostic tests. Treatment or harm questions can be answered with clinical trials (ideally) or with observational studies. However, prognostic questions usually require observational studies to find the answer. Finally, just to mention that there’re other types of clinical questions besides the four basic ones such as the frequency (which will be answered using systematic reviews and observational studies) or cost-benefit questions (who need economic evaluation studies).

We’re leaving…

A well-structured clinical question can help us to solve a clinical problem but it also often serves to make more questions, which with we can fill the gaps of our knowledge and become a little less ignorant. In addition, if we don’t structure our question in its different components, it will be practically impossible to find useful information. Those of you that don’t believe what I’m saying, just write “asthma” in the search field of PubMed or any other search engine and see the number of results. Some browsers, such us Trip Database, even allow search using the PICO structure of the clinically structure question. But unfortunately, in most cases we’ll must find synonyms of each component and find the right descriptor for the database where we are doing the search, usually using advances search techniques. But that’s another story…