The Trip Report: Staffing Models and Scene Times
Reviewed This Month: The Effect of Ambulance Staffing Models in a Metropolitan, Fire-Based EMS System
Authors: Cortez EJ, Panchal AR, Davis JE, Keseg DP.
Published in: Prehosp Disaster Med, 2017; 32(2): 175–9.
This month’s topic is management and operations. While this is a topic that impacts just about every aspect of EMS, it is surprisingly not researched nearly as much as many other topics we’ve discussed.
There are a lot of resources available to EMS leadership that evaluate, compare and discuss management and operations strategies. But most of these are not peer-reviewed scientific papers. Peer review is the process of having subject matter experts evaluate research that was submitted to a medical/scientific journal to determine whether it’s worthy of publication. Granted, there is no shortage of issues with the peer-review process; these issues have been documented in both scientific and nonscientific writings. Nonetheless, peer review is an accepted way of vetting research and attempting to assure high-quality evidence is added to the available literature. Ohio physician Eric Cortez, et al., recently published an interesting peer-reviewed manuscript evaluating ambulance staffing models.
Cortez and company sought to evaluate whether there were benefits to alternative staffing models. They cited about 10 studies published from 2000 to 2015 that highlighted uncertainty in EMS over the “best” model. Some studies indicated that ALS-only models were beneficial because they were more efficient with respect to operations, were less complicated in terms of dispatch and demonstrated some fiscal advantages. On the other hand, some recent literature also indicated that mixed staffing models (ALS and BLS providers on the same ambulance) performed just as well as ALS-only models, and some studies suggested mixed models performed better than ALS-only models when evaluating on-scene times and patient mortality. With no clear consensus in the available literature, there was a need for this research.
The study took place from September to December 2013. It was performed in a large metropolitan area that historically deployed a staffing model with two paramedics on an ambulance. This was a retrospective review of prehospital care reports. Every call during the study period was included, and the principal investigator evaluated all PCRs. The primary outcome of interest was total on-scene time. The authors also evaluated time to ECG, time to IV, IV success and the percentage of protocol violations. They included some examples of what would be considered a protocol violation: not administering aspirin to patients with chest pain; not measuring blood glucose levels in patients with seizures; and not documenting pulse oximetry readings in patients with respiratory complaints.
The study was performed at two stations (station A and station B). Teams with two paramedics on the ambulance were compared to teams with one paramedic and one EMT-Basic. It is important to note that the EMT-Basics were experienced and volunteered for overtime to staff the mixed-model ambulances. The authors reported medians rather than average times. They did this because times in EMS are not and should not be normally distributed. This is just a fancy way of saying that they are and should be short more often than they are long. For times to be normally distributed, there would be a similar proportion of times that are very long and very short. They also used the appropriate statistical tests to detect differences in data that are not normally distributed, namely nonparametric tests.
Both stations replied to a similar number of calls during the study period. Station A had 1,639 calls, station B 1,576. In both stations approximately three-quarters of the calls had an ambulance with two paramedics respond, and approximately one-quarter had an ambulance with one paramedic and one EMT-B respond. The authors found no difference among the patients when evaluating demographics and chief complaints. This is important because it helps put the results into context: If the patients in one station’s area were significantly different than the patients in the other’s, it could introduce bias into the results and limit the usefulness of comparisons.
Interestingly, the authors found that in both stations, on-scene times were shorter for the ambulances with two paramedics, and these results were statistically significant. For station A, the two-paramedic model had a median scene time that was about three minutes less than the mixed-model ambulances. In station B the difference was not as large (approximately one minute) but still statistically significant. There was no statistically significant difference in any of the secondary outcomes assessed.
Statistical vs. Clinical Significance
Now we get to discuss why I really love this paper. Some of you may be thinking, Is a one-minute difference in on-scene time really that important? Some may even wonder whether a three-minute difference is really relevant. The authors did a fantastic job discussing the difference between statistical significance and clinical significance.
Statistical significance is based on a mathematical formula. If the formula returns a number, called a p-value, less than 0.05, then the results are determined to be statistically significant. A lot of things can contribute to a result falling below that 0.05 point. How much data is being analyzed can certainly have an impact. For example, given a large enough data set, you could find that a systolic blood pressure of 122 mmHg is statistically different (p-value less than 0.05) than a systolic blood pressure of 126 mmHg. Now, I’m sure just about everyone reading this wouldn’t say that’s a clinically important difference. Statistical significance alone does not mean the finding is important. It takes subject matter experts to evaluate that.
The authors correctly identified that even though their on-scene results were statistically significant, there were some important reasons why that significance may not be clinical. They identified many reasons, including that the paramedic in the mixed model had to be responsible for all assessments, ALS-level interventions and documentation, whereas in the two-paramedic model, these are typically shared.
The authors also did a good job identifying limitations to this study. For one, they did not assess patient outcomes. Given that the paper was looking at benefits to different staffing models, understanding how they impacted the patient would certainly be an important outcome to assess. However, it is not always easy (as too many EMS providers in the field are aware) to obtain patient outcome information. Further, this study was performed in a fire-based system. It would be interesting to see whether these results differed in other system types.
The authors correctly concluded that mixed staffing models may be beneficial. Again, subject matter expertise is important here. Not only did one of the authors hold a paramedic certification, but it appears all the authors had some EMS affiliation. Had this study been performed by a team without that expertise, they may have erroneously reported that the difference in on-scene time was overly important due to the statistically significant findings. This paper not only added to the available literature on the topic, but it also highlighted the importance of having EMS research performed by EMS professionals.
Research does not have to be intimidating. More involvement by EMS professionals can lead to better research, improved patient care and improved provider performance and safety. Keep reading research and try and get involved when possible.
Antonio R. Fernandez, PhD, NRP, FAHA, is research director at the EMS Performance Improvement Center and an assistant professor in the Department of Emergency Medicine at the University of North Carolina–Chapel Hill. He has been a nationally certified paramedic since 2005 and completed the EMS Research Fellowship at the National Registry of Emergency Medical Technicians.