Background
Systems for pre-hospital critical care exist worldwide, but emergency medical service (EMS) systems differ in resources, organizational and operational models; from simple systems providing basic life support to sophisticated systems providing critical care [
1‐
10].
Treatment and diagnostic options in pre-hospital care are increasing and several in-hospital techniques are currently being applied in the pre-hospital setting [
10‐
14]. To enable more point-of-care diagnostics and increase in advanced interventions, some EMS systems, especially in Europe and Australasia, have introduced helicopters and rapid-response cars staffed with specially trained physicians [
10,
15]. The effect of physician-staffed EMS (p-EMS) is debated and studies report contradicting results [
16‐
28]. A substantial challenge to assess quality of health care is lack of uniform documentation, this is also pertinent to p-EMS [
29,
30].
The concept of consensus-developed condition-specific datasets has proven useful for research and quality assessment in several areas of critical care [
30‐
33]. To evaluate the effect and efficiency of p-EMS, a template for uniform reporting of data from p-EMS was published [
34]. However, to implement a template for documentation, feasibility of the template to collect the requested data in the context intended should be demonstrated [
29,
35‐
39].
In Scandinavia, p-EMS is well established, and services are relatively similar, thus joint research efforts may be valuable [
5,
34]. Finland is currently the only country where the template for documenting and reporting from p-EMS is implemented, thus the only country able to provide routinely collected template data. To evaluate template feasibility, we wanted to compare two different data collection methods in two similar systems. P-EMS in Finland and Norway employ the same operational and medical concept and differences between services are mainly seen in time variables, patient volume and service area [
5]. We considered comparison of Finland and Norway to be feasible; thus, we decided to include these two countries for the present study.
The aim of the present study was to evaluate the feasibility of pre-hospital physicians to collect patient and system level data by using the template for uniform reporting of data from p-EMS [
34], comparing data collection from a standard to a focused data collection method.
Discussion
When efforts are optimized, p-EMS can achieve high completeness rates in collecting prospective data using a template. Motivation and focus on documentation, rather than operational context, seems to affect data completeness rates most.
Lack of documentation is often highlighted as a limitation for research in emergency medicine, especially for retrospective registry studies [
5,
35,
43‐
45]. Putting attention to increase the quality of routinely collected data may enable such data to be an important and effective source to monitor and compare services. As such, strategies to increase data capture should be sought [
30,
43,
45‐
48]. Training programs may increase data capture, most likely by increasing attention to documentation [
43,
48]. In our study the effect of motivation was evident, where significantly more data were registered with a focused data collection method than with a standard data collection method. Feedback on how high-quality research or quality assurance will benefit from complete data registration can make physicians more aware of the importance of data registration, thereby increasing data capture.
Echoing our results, several studies have found physiological variables to be the least documented variables [
44,
45,
49]. Laudermilch et al. [
44] found that 28% of patient records had missing physiological data and Bergrath et al. [
45] reports vital parameters necessary to document Mainz Emergency Evaluation Score (MEES) to be present at two time points in only 31.08% of patients. Gravel et al. reports from the paediatric population that high rates of vital signs data are missing [
50]. With a standard and a focused data collection method, 48 and 85% of physiological variables were registered, respectively, indicating that high completeness rates are achievable. However, physiological data were not complete, even with a focused data collection method. Good clinical assessment depends on correct evaluation of vital signs; thus, documentation of physiological variables is important [
48,
50] and strategies for improvement of reporting should be sought.
Physiological data change according to patient state and repeated registrations of the same variable capture trends and reveal changes in patient condition and the effect of treatment [
30‐
32,
46,
47,
51]. The p-EMS template requests documentation of physiological variables at two time points. For all repeated parameters we found the first value to be more complete than the last value, thereby complicating intervention comparison and comparison of changes in patient state. This is comparable with the findings of Bergrath et al. [
45]. Medical directors should emphasize the statutory requirement for temporal documentation of physiological parameters and that this also pertains to p-EMS [
52].
Ideally data capture in pre-hospital critical care should be simple, accurate and fast. For both cohorts, clinical data are registered on paper during the mission and are later digitally registered. This process is time-consuming, inexpedient and carries a risk for recall-bias and documentation fatigue. Automated data capture from monitors may increase completeness rates and is widely used in anaesthetic services documenting every change in the patient state [
53]. Implementation of these readily available concepts to the pre-hospital environment is increasing [
54]. Although there are still challenges, automated data capture may reduce administrative workload, improve patient focus and transferal of patient documentation to the next level of care [
54‐
58].
Laudermilch et al. [
44] suggests that datasets are less complete for more severely injured patients and that increased workload reduce data capture. This is in contrast to our findings, where data capture was increased or remained equal for patients with a critical condition (Table
3). Corresponding with our findings, Bergrath et al. report calculability of MEES to improve with increasing medical severity [
45]. Patients with minor complaints might be considered to require less attention and thereby an increased amount of missing data occurs [
49,
59,
60]. However, with a focused data collection method, we found no differences in data completeness for less critical patients.
Time available for data capture may affect completeness rates. We found missions with less than 20 min of patient encounter were associated with lower completeness rates than missions lasting more than 20 min. This may reflect increased workload. For children below 10 years of age, we found lower completeness rates of vital parameters than for patients above 10 years of age. For less severely ill or injured children, measuring blood pressure can be uncomfortable and doctors may be reluctant to perform the measurements, resulting in lower completeness rates.
Categorization of data may increase data capture compared to registering exact values [
59]. In Finland, where data were collected by a standard method, the template has been modified and pain was reported using a scale from 1 to 10 instead of using the original three-parted scale described in the template; no pain, moderate pain and severe pain. Jennings et al. [
61] recommends the verbal numerical rating scale to measure pain in the pre-hospital setting, corresponding with the FinnHEMS template modification. In Norway, where a focused data collection method was used, pain was reported according to the original template. With the standard data collection method, we found that completeness rates for data on pain were low while with the focused data collection method, data on pain were almost complete, supporting a reduced number of categories to increase data capture. However, fewer categories reduce precision, leading to imprecise estimates, and must be weighed against the need for accuracy.
Outcome comparison often adjust for on-scene time, making low documentation completeness or imprecise registrations of this variable a limitation for research [
62‐
64]. Eckstein et al. [
65] found on-scene time being documented in 70% in a cohort of major trauma patients. In our study, on-scene time was documented in 56% an 91% of the cases with a standard and focused data collection method respectively, indicating that high completeness rates are achievable when attention is directed towards documentation. In Norway, the emergency medical communication centres automatically documented the origin time data whereas the response units registered other time variables on paper or non-portable devices [
59]. Due to weather and operational conditions, paper registration was often not feasible, and variables were often registered in retrospect, increasing the risk for imprecise registrations. Portable devices available for registrations on site could further increase completeness rates and accuracy of data.
In our data Finland report significantly longer median on-scene times than in Norway (22 versus 12 min), still on-scene times are considerable shorter than reported from German (32 min) and Dutch (27.2 min) services [
66]. For trauma patients, the concept of aiming for a pre-hospital time period less than one hour (“The golden hour”) and of keeping on-scene times to not more than 10 min, have been directional for organization of pre-hospital care [
67]. In recent years these concepts have been challenged [
67‐
69]. Harmsen et al. conclude that emphasis should be on making sure the patient receives proper pre-hospital care rather than on getting the patient to hospital as fast as possible [
70]. In our data, physicians in Finland are providing significantly more advanced procedures than in Norway. This may explain the longer on-scene times in Finland. We do not know, for our system, which advanced procedures should be performed by pre-hospital physicians to improve patient care. However, we believe that uniform documentation may enable us, in the future, to identify procedures beneficial in p-EMS.
All process mapping data (procedures performed, and medication administered) are mandatory in Finland, possibly explaining the 100% completeness rates. In Norway, where no data points are mandatory, completeness vary between 26 and 99% for process data.
Two variables showed particularly low completeness rates in Norway: “Diagnostic procedures” and “Breathing – procedures used”. For these two variables there is no option for choosing “none” or “not relevant”, and when no procedure is performed these data fields will appear as missing. We suggest this to be revised in the template.
Comparing data from two countries had some practical challenges. Although both data collection methods collected data according to the same template, the data were registered in different data formats and in different language. To be able to compare the datasets a work-intensive data management job was needed to standardize categories. Thus, to allow rapid and accurate comparison we recommend data to be registered in the same data format. This is achievable and one might suggest a digital template with predefined names and categories to be implemented. This means that data, when transformed into statistical analysis software, must have the same properties, names and limitations to be able to be easily merged into the same database and analyzed. Adaptions where additional variables are included for local purposes can easily be managed within such a digital template without hampering template comparisons. We believe that simplifying the comparison processes by standardizing data entry will generate more multi-centre research.
Limitations
The present study has several limitations. We did not include a formal questionnaire to investigate reasons for missing data, although most physicians in Norway provided informal information regarding this. A questionnaire could have been useful to discover reasons for registration failure of importance to aid revision of the template. In Finland, the physicians were not informed about the study in advance, so the database reflects normal documentation rates in FinnHEMS bases. In Norway, the Hawthorne effect is an obvious and wanted effect, whereas the risk for this in Finland is lower.
The data are from 2013/2014 and this may be considered old. However, documentation method or organisation of p-EMS have not changed in either Finland or Norway since 2014, thus we believe the results still are valid and that newer data would have yielded similar results.
In Norway, 16 physicians participated, and each physician had on average 5 shifts during the data collection period. Each physician registered on average 8 cases, this is on average 1,6 cases per shift. This is a low number if each physician were to be evaluated individually. Because the aim of the study was to evaluate the documentation system, not the individual physician, we find the total number of cases registered in Norway to be acceptable.
The study was conducted in two similar p-EMS settings in two high-income countries and results may not be applicable to all other EMS settings. However, documentation for the study was paper-based, not including expensive equipment. The principles for pre-hospital emergency medical treatment are generally recognized, and international expert consensus on important data to be collected in the field should apply to both low- and high-income EMS systems. The concept of using a template by motivated personnel for data collection may therefore be applicable to other less resource-intensive settings.
Thirteen of the variables are mandatory to register in Finland and electronic patient files cannot be saved unless these variables are registered; completeness rates are therefore 100%. To compare these with Norwegian data will not give an idea of what is possible to collect in an everyday setting or if implementation challenges also apply for this type of data. Finally, the challenge with possible fabricated data to finalize registrations must be addressed.