Data quality of Glasgow Coma Scale and Systolic Blood Pressure in scientific studies involving physician‐staffed emergency medical services: Systematic review

Emergency physicians on‐scene provide highly specialized care to severely sick or injured patients. High‐quality research relies on the quality of data, but no commonly accepted definition of EMS data quality exits. Glasgow Coma Score (GCS) and Systolic Blood Pressure (SBP) are core physiological variables, but little is known about the quality of these data when reported in p‐EMS research. This systematic review aims to describe the quality of pre‐hospital reporting of GCS and SBP data in studies where emergency physicians are present on‐scene.

comparisons and outcomes were carried out by using the data extraction and quality appraisal variables described in methods and depicted in the results section (Figures 2 and 3) and Tables A1 and A2. Data extraction described quality of documentation (accuracy, completeness and capture), study mix, barriers and facilitators of documentation in p-EMS.

| Study selection
The search identified 5530 records after duplicates (435) were removed and 190 full-text articles were assessed for eligibility. Of these, 132 articles were included in the study. In addition, five articles were identified by manual searches and included ( Figure 1). Studies were mainly excluded because SBP or GCS were not reported or because studies did not report data from physicians-staffed units and ordinary EMS separately.

| Study characteristics
Of the included articles, 32 articles reported GCS only, 26 articles reported SBP only whereas 79 articles reported data for both GCS and SBP. Nineteen studies were registry studies and six studies were interventional studies. Nine studies included children only, 60 included adults only, 54 included both children and adults whereas 14 studies did not report age of included patients.
Physicians in the included studies were mostly anaesthesiologists, emergency physicians or a mix of both. A few were registrars from different specialties. For 48 studies the specialty of the physician was unknown.

F I G U R E 1 Flowchart depicting the different stages of the systematic literature review
Most studies (111) were conducted in Europe. Germany (20), United Kingdom (19), France (13), The Netherlands (12), Denmark (11) and Finland (9) conducted three fourths of the studies. Eight studies were conducted in Australia, eight in Japan, two in Brazil, two in Israel and one in USA, Russia and Taiwan respectively. Three studies did not report location.
Sixteen studies reported medical cases, 81 reported trauma cases, one reported neonatal cases and 39 reported a mix of cases.
Fifty-two studies were prospective and 83 were retrospective.
For two studies we could not establish whether the studies were prospective or retrospective. Study design was clearly described for 130 studies.
An ethics committee approved 72 of the studies. For 26 studies it is described that approval was not required and 39 studies did not report information regarding approval.

| Glasgow Coma Scale (111 articles)
Reporting of GCS data are depicted in Figure 2. We found 65  In 56 studies children were included. Of these, one study reported that paediatric GCS 22 was used.
Among studies reporting completeness rates, the lowest completeness rate was 41.5%. For 12 of the studies reporting completeness rates, GCS was a criterion for inclusion and completeness rates were therefore 100%.
Of studies reporting number of excluded patients, exclusion rates ranged from 0 to 64.4%.
Reported facilitators to GCS reporting were the presence of predefined check boxes for reporting GCS and various human factors (motivation, feedback and training of personnel).
Reported barriers to GCS reporting were related to various procedures (sedation, anaesthetic drugs, intubated patients) and difficulties of recording GCS when providing care to critically injured patients due to lack of time. Furthermore, practical challenges (difficulties of recording GCS while providing care to critically injured patients due to lack of time, inadequate documentation tools) and human factors (lack of training, inadequate motivation and inexperience in scoring) were noted as barriers.

| Systolic blood pressure (105 articles)
Reporting of SBP data are depicted in Figure 3

| Quality appraisal
The predefined variables for quality appraisal of the included articles are shown in Table 1. The full quality appraisal of included articles is depicted in Tables A1 and A2. Three articles reported all the items for external validity of SBP whereas two articles reported all the internal validity items. Average amount of reported data was 26% and 45% for external and internal validity data respectively. For either GCS and SBP we found no differences in the reporting rate between prospective and retrospective studies.

| D ISCUSS I ON
In this systematic review, we found a variable rate of accuracy, capture and completeness for reporting of GCS and SBP in p-EMS.
Quality appraisal revealed that most of the predefined variables for assessment of external and internal validity were not reported. High completeness rates are achievable in p-EMS 23 arguing for increased focus on documentation and reporting of data collected. The dynamics of patient physiology can only be captured through repeated measurements. Accurate and complete documentation and reporting are therefore important to identify effects of treatment and changes in patient state. Furthermore, comparison of studies and merging of data is difficult if reporting of data is poorly defined, hampering joint research. 14,24 Uniform documentation promotes comparisons and outcome research of high quality. 25

| Accuracy of reporting
The accuracy of reporting GCS and SBP was low. In most studies timing or method of measurement were not reported, complicating comparisons and evaluation of results.
We found 29 studies reporting GCS as categories. Categorization of GCS originates from neurotrauma research efforts to categorize TBI patient into groups of severe (GCS 3-8), moderate (GCS 9-12) and mild (GCS 13-15) head injury. 13 Among the included studies the categories used were heterogeneous, and we found overall 15 different ways of categorizing GCS. Even for TBI studies, different categorizations were used. The category GCS 3-8 was often used, but there is a clinically significant difference between GCS 3 and GCS 8, and one might question whether categorization into such a heterogeneous group will yield valid conclusions. One study used GCS categories corresponding to the Revised Trauma Score (RTS) categorization. 26 Different categorization may reflect that the use of the scale has expanded to various patient groups, and is no longer used for TBI patients solely, thereby complicating valid comparisons in pre-hospital research. 7,14,24 Furthermore, the categories "severe" (GCS 3-8), "moderate" (GCS 9-12) and "mild" (GCS 13-15) often used in TBI research are not scientifically grounded. The categories were chosen "ad hoc" and the cut-off points are not yet validated. 13 To enable research across different countries and p-EMS systems, we recommend reporting an exact GCS whenever possible. If categories are to be used, agreement of categories and validation of these should be established.
Another obstacle to accuracy of GCS reporting is injuries or illness affecting functions like speech and motor skills. This may interact with the assessment of the GCS components and affect GCS scores. 27 We found no studies reporting how GCS was reported when injuries or illness (eg aphasia, extremity fracture, maxilla-facial trauma and paralysis due to different origins) impaired function. There is no consensus in literature on how to score, for example, aphasic or paralytic patients and strategies vary. 27 Furthermore, p-EMS commonly intubate patients, but 93% of the studies failed to describe how GCS was reported after intubation. Different approaches to GCS reporting for intubated patients are suggested, but still no consensus has been achieved. 13 compared to the full GCS score. 27 However, to enable comparisons, and to increase reliability, a standardized approach is called for. 12,13 Thirty studies specify that the GCS reported is measured before sedation or intubation. Among the studies reporting how GCS was handled after intubation, two studies used the pre-intubation value and three studies used a pseudo score of "3" for all intubated patients. A pseudo score of 3 is different from a true value of 3 and using pseudo scores or conservative coding is not recommended as it does not reflect the situation. 13 It is recommended to report GCS by its three components (EVM) and assign the designation "not testable" (listed with reason) whenever a component is untestable. 13 This will allow imputation methods and provide a more reliable comparisons of patients with illness or injuries that interferes with assessment of the GCS score.
Similar to GCS, the assessment of SBP will be influenced by

| Completeness
Complete documentation and reporting is a quality indicator in p-EMS. 33 Missing data remain a methodologically quality concern in medical sciences 34 and high completeness rates are called for. 7

| Limitations
There is always a danger of selection bias when performing a systematic review, for example, erroneous exclusion or inclusion of studies.
Furthermore, some relevant studies may not have been identified during our database search due to poor indexing or application of imprecise search. Furthermore, including only papers written in English or Scandinavian languages increased the risk of missing relevant stud- ies. The quality appraisal items were designed by the authors in the absence of a universally accepted definition of data quality. Included studies were heterogeneous and information was subjectively interpreted thereby potentially introducing reporting bias.

| CON CLUS IONS
The quality of reporting of GCS and SBP in p-EMS is variable in scientific papers. Uniform documentation and reporting promote comparisons and high-quality outcome research. Given the variable reporting identified in this review, we recommend standardized reporting to enable better comparisons of p-EMS.

ACK N OWLED G EM ENTS
The authors are grateful to the donors of the Norwegian Air Ambulance Foundation who by their contributions funded this study.
We also thank Elisabeth Hundstad Molland (Librarian at Stavanger University Hospital) who helped to design and perform the database searches. The Norwegian Air Ambulance Foundation (NAAF) funded this project. However, the NAAF played no part in study design, data collection, analysis, writing or submitting to publication.

CO N FLI C T O F I NTE R E S T
We declare no conflict of interest.