Ask a simple question
Although it sounds simple, asking the wrong question – one that has little interest or utility to others or one that is not clear and focused – is one of the most common errors in research. Simply put, every research question should pass the “So what?” test and be clear and focused. Although a study can involve some subtopics or smaller questions related to the main question, the researcher must be able to state clearly in one or two sentences the main study question. Asking colleagues or a research mentor to review the question before starting the study will help ensure that the question is important, clear, and understandable and may also lead to additional ideas for conducting the project.
It is important that the researcher ask an important and focused question first and then design an approach to answering it, rather than the other way around. Searching through a mound of data (e.g. numbers generated by a quality improvement project or dumped out of a computer-aided dispatch system) for a question that could be addressed can result in either misleading or uninteresting results. Large databases are tempting to explore because it seems there must be a study lurking in the numbers, but care must be taken to ask a clear question before analyzing the data.
Write a hypothesis or a clear, simple objective
Once a research question has been formulated, a hypothesis or a clear objective must be composed. A hypothesis is a declaration to be proved or disproved. There are several types of hypotheses, depending on the phrasing. A null hypothesis states that no difference exists between two (or more) groups being studied. An example from a published study states, “Our null hypothesis is no difference in survival exists between an EMS system using targeted response and one using a uniform or all advanced life support response model” [4]. The alternative hypothesis states that a difference between groups will exist. This difference can be directional, specifying a better/worse or higher/lower effect (e.g. “We hypothesized that 4-hour survival would be greater among patients randomized to load distributing band CPR compared with those randomized to manual CPR” [5]). It also can be non-directional, where no direction is specified for the sought-after difference.
It is important to recognize that not all studies can have traditional hypotheses, but every study should have a clear focus. For example, the following is a focused and clear objective: “The objective of the current study, the OPALS Respiratory Distress Study, was to assess the incremental benefit with respect to morbidity and mortality that results from the implementation of an advanced-life-support program for the evaluation and management of respiratory distress before patients arrive at the hospital” [6]. Many observational EMS studies observe patterns and operational features of systems rather than compare groups, and a formal hypothesis may not be appropriate. Such studies can refine a question and explore areas in need of future work rather than define an association.
Review the current literature
Reviewing the available literature on the topic to be studied is not glamorous or fun, but it is crucial. A review will confirm that the question is appropriate and important. If the question has already been studied, there may be no need for a trial. This is unlikely for a good clinical question – even if previous work exists, a confirmatory study or a study conducted in a different type of EMS system may be of value. More likely, a researcher will find work on a similar but different question. A review may also help identify problems that may be encountered or methods that may be better suited to the study question than those originally planned.
To ensure a comprehensive review of the literature, perform a computer search using at least two different databases. For the EMS literature, the MEDLINE and CINAHL (Cumulative Index to Nursing and Allied Health) databases are of particular value. MEDLINE is produced by the National Library of Medicine, indexes approximately 5,600 journals, and is generally considered to be the authoritative biomedical index. CINAHL indexes more than 5,400 journals and includes “emergency services” among the 17 allied health fields on which it focuses. On-line abstracts are given for the majority of citations, and these can help screen for articles that need to be located and reviewed in their entirety. Online Google Scholar searches may also assist in identifying relevant published research. No matter which search engine is used, computerized searches will miss some papers [7], making a manual search of cited references from all relevant articles found in the computer search essential, although frequently tedious.
After completing the literature review, the investigator should be an expert on the subject and have a clear idea of the direction the study is going to take. In fact, many investigators write review papers from this effort or are able to frame the introduction, methods, and even parts of the discussion section of the eventual study manuscript before beginning the study based on this gained expertise.
Select a study design
A study design is a general plan for testing the study hypothesis or approaching the objective. There are many designs, and the best choice will depend on the study question and available resources. The researcher must select a design that balances costs, time, feasibility, and ethics and will produce useful results.
Although often perceived as the “gold standard” from an evidentiary strength view, a randomized controlled trial (RCT) is not always possible or even desirable. For example, some outcomes may occur so far in the future that it is not practical to wait for them to occur; similarly, an RCT may be unethical if clear benefit, harm, or a lack of supporting data for a particular therapy exists. New questions may require background work to define the magnitude of the problem or identify potential solutions, usually best accomplished using non-RCT designs.
When choosing a study design, researchers must ask three questions.
- Can they follow participants over time?
- Can they intervene with participants (instead of simply observing)?
- Will they look at events that have already occurred or as they occur?
The first question determines if the study will be cross-sectional or longitudinal. Cross-sectional studies are those that measure all the study variables at the same point in time (or during a brief interval, perhaps a week or a month), providing a “snapshot” of data. Most surveys are cross-sectional, in that they gather data regarding the state of a particular problem or variable across a number of respondents, all at approximately the same time. A survey of all 9-1-1 communication centers in the state of North Carolina regarding dispatcher training and protocols for stroke and myocardial infarction is an example [8].
In contrast, longitudinal studies examine variables over time by following patients. Longitudinal studies are challenging in EMS, where patients generally do not maintain contact with the EMS system after their encounters. Although many out-of-hospital cardiac arrest studies follow survivors out to 1 year [9], very few studies follow patients through multiple EMS encounters over time. Some studies seem to have characteristics of both types, as with a study of changes in out-of-hospital cardiac arrest survival rates over an 11-year period. In this example, the investigators assembled sequential 1-year snapshots (rather than individual patients being followed over time) over more than a decade [10].
The second question determines if the study will be observational or interventional. An observational study monitors what is happening but makes no attempt to influence outcome or otherwise intervene in the events being studied. By comparison, an interventional design imposes a change or perturbation and studies the effects. Contrasting examples are a study by Hostler and colleagues examining blood samples from out-of-hospital cardiac arrest patients for the presence of thrombin/antithrombin complexes (an observational trial) [11], and a randomized controlled trial by Rickard and colleagues comparing intranasal fentanyl to intravenous morphine for prehospital pain management (an interventional design) [12].
The third question determines if the study will be prospective or retrospective. In a prospective study, the events of interest have not yet occurred when the patient is identified. For example, in the analgesia study by Rickard and colleagues, patients were enrolled before any pain medication was given. After enrollment, each patient randomly received one of the two study drugs, which was then administered, followed by outcome assessment (the amount of change in the patient’s verbal rating score of the degree of pain). In a retrospective study, the events of interest have already occurred at the time of patient identification. A retrospective study might have reviewed the records of patients to see what pain medications they had received and compared outcomes.
In general, longitudinal, interventional, and prospective studies are of greater value in assessing causal relationships – ”this affects/changes that” – than are cross-sectional, observational, or retrospective studies. This benefit comes at an expense: the more powerful designs are generally harder to perform, are often more expensive, and are usually more time consuming.
