Now that we have specified our research question, we have to plan out how we may approach the research question in order to answer it. Research methods are the procedures that researchers employ to answer a research question. The research question will dictate what methods should be used; in other words, certain types of research methods are best used to answer certain types of research questions.

The three broad approaches to research are quantitative, qualitative, and mixed methods. Each broad approach uses specific methods, or in the case of mixed methods a complementary combination of quantitative and qualitative.

Quantitative methods in general focus on describing phenomena or differences numerically. Using our earlier example, we may compare pain scores between those that received a traction splint and those that did not. Quantitative methods are the most commonly used approach in medical sciences, and although useful for describing “what,” it does not often explain “why.”

That is where a qualitative approach may be best. Qualitative methods generally focus on exploring phenomena in-depth. Using our earlier example, we may want to assess not just a patient’s numerical pain score, but how a traction splint made them feel. The difficulty of adequately measuring suffering via a single numerical value, as well as the characteristics of catastrophizing and other pain modifiers, has been discussed elsewhere, and has great relevance to the field of WEMS.⁶ We may also want to explore barriers to the use of a traction splint in the WEMS setting from a provider’s perspective. Both of these research questions might be better answered using qualitative methods.

Finally, a mixed-methods approach may use both quantitative and qualitative methods to more comprehensively answer a research question. Using our example, to fully understand the implications of using a traction splint in the WEMS environment, the best study may first assess whether there was a quantitative difference in pain scores between those that received a traction splint and those that did not, then explore why this difference may or may not have occurred from the patients’ and providers’ perspectives. We will go into detail on all three of these approaches later in the chapter.

Once a study is complete, and often published, the results should be translated into action. This is sometimes called knowledge translation (KT). Action could mean that it changes personal practice, or in the case of those that work for a WEMS system that has standing protocols or procedures, perhaps results in a change to these policies. It is very important that users of research take into account how well the researchers answered their research question (in other words the rigor) before translating it into practice. As we have previously mentioned, rigor incorporates the ideas of using the best research methods (in balance with feasibility) to answer the research question, and applying these methods correctly. Another important aspect of applying research into practice is the idea that the setting of the study is sufficiently similar to your own setting to apply the research findings (we will cover this in more detail later in the chapter).

Finally, in some instances there may be conflicting conclusions from multiple studies. An important component of KT is getting the evidence straight, which means summarizing all known studies on a topic with due consideration to the rigor of these studies, and in combination with your own experience, translate the findings into practice.⁷

A randomized controlled trial (RCT) is considered by many quantitative researchers to be the gold standard design for studies of therapies or interventions. If we recall the section on rigor, randomly allocating an intervention to two or more groups of individuals is the closest thing we can get to the “time machine” approach.

In the “time machine” approach, we could put the traction splint on a patient, assess their pain, then travel back in time and put on a traditional splint. Remember, the reason this is so powerful is because everything about the patient is the same, and the only thing that differs is the splint. Therefore, any differences we see in outcome must be attributable to the splint. This is a concept called causality; the traction splint is causing a change in pain level. (Causality is important to differentiate from “correlation”—we will cover this later in the chapter.)

An RCT is an extension of this idea, except that two (or more) groups are created, one that gets the intervention and the other that does not. The key feature is that allocation of who gets what intervention is done randomly—in other words by chance. The reason why we want to do this randomly is that in theory the only systematic difference between the two groups should be the intervention (similar to the “time machine” example). All other known (or unknown) factors that may influence the outcome should be nearly the same. So if we apply this to our example of a traction splint, we could create a research study where one group of patients randomly receives a traction splint and the other group does not. These groups should theoretically only differ by the type of splint they received. This type of study design is an example of an experimental study design, because the researcher is controlling who gets the intervention and who does not.

So why would we not allow responders to decide who gets a traction splint and who does not? Why is it better to randomly decide? The reason is that we all have biases. It may be that responders believe that traction reduces pain, and so they may place a traction splint on those patients who appear to be genuinely in a lot of pain over those that appear to have less pain. This may result in the postsplinting pain scores between the two groups being similar, not because the traction splint is not effective, but because patients in the traction splint group had a higher pain score to start off with. You can see that bias is an important factor, but can often be addressed in the design or analysis phases of a research study (more on this later).

We now know why we need to randomize patients, but is randomization enough to prevent bias? Not entirely, we also ideally need to “blind” patients, providers, and outcome assessors
on what intervention they are receiving. This is to prevent changes in data collection, assessment, treatment, or perception of treatment based on knowledge of the intervention. In some RCTs this is possible, like a drug trial where two drugs can be made to look identical. For trials involving a traction splint, this is obviously impossible. In trials where it is not possible to blind the intervention (called unblinded or open trials), it is important to record other factors that may influence the outcome to make sure these other factors are not influencing the perceived or real effect of the intervention. For example, with our traction splint study, we may want to record presplinting pain scores to ensure that the group that received traction splints had the same initial score as those that did not.

One may come across numerous terminologies to describe who was blinded in a study. Some may suggest that a “single blind” study is when the patients involved in the study do not know what intervention they are being given, but the researchers do. A study may be “double-blind” when both the patients and the researchers who are following the patients do not know what intervention has been provided, although there is no universal agreement on these terms. It is important to note that there is often inconsistency of the meaning of “single blind” and “double blind,” and caution in your interpretation is needed.^8,9 Although these terms are still used, a more modern approach describes who was blinded instead of using this terminology, making it easier for the reader to interpret.⁸

Thus far, we have learned about how randomly assigning groups of patients to receive an intervention (an RCT) is the most rigorous approach we can use, without using a “time machine.” It is rigorous because we can infer causality between an intervention and an outcome (eg, traction and pain). It is important to realize though that there are many situations where using an RCT approach is either not feasible or is unethical.

An RCT is expensive and takes a tremendous amount of time and resources to implement. Imagine if we were to implement a study that randomly assigned WEMS patients to receive traction or no traction. First, we would likely need many different systems to participate in the study so that we could enroll enough patients to achieve an appropriate sample size (more on this later). Second, we would have to train enough WEMS providers on the study methods so that they could identify patients that meet inclusion criteria for the study, consent the patient to be part of the study, and then apply the intervention. Third, we would have to create a random allocation process, so that patients could be randomly assigned to a group. This would involve working with operations managers of the WEMS systems to ensure that crews had the right equipment on hand on the scene (ie, both types of splints), and that they were not aware of what intervention they were providing until after they consented the patient to participate (to reduce bias). In addition to this, we would need to have an extensive research team that may include epidemiologists, biostatisticians, clinicians, WEMS operations managers and responders, and perhaps other specialized academics, such as health economists, geographical information system (GIS) specialists, etc.

In addition to feasibility, there is also an ethical responsibility to not conduct an RCT if the alternative intervention is not thought to be clinically equivalent. For example, would it be ethical to conduct an RCT where one group of patients gets direct pressure for a severely bleeding wound, and another group gets no treatment to directly control bleeding? Definitely not, as direct pressure is the gold standard for controlling bleeding, and doing nothing would be unethical! This is a clear example of an unethical RCT, but sometimes it is not that clear. What about traction on a femur fracture? Would it be ethical to carry out an RCT where some patients received traction and others did not? That may depend on the balance of existing evidence around the benefit and harm of the intervention. Some RCTs are required to have a data monitoring committee, that ensures participants are not being harmed, or to establish futility (ie, when the trial reaches a point that it cannot hope to achieve its objectives no matter if all future patients had a certain outcome).