Introduction

The capacities of any medical department involved in handling patients after a mass casualty incident (MCI) are to a certain extent limited, and planning of resources is crucial to improve performance and outcome [1–8]. Consequently, fast and reliable diagnostic tests are needed under such conditions.

As reported in other studies, initial field triage helps to distribute patients to hospitals with different levels of specialization and facilities according to the injury patterns [2, 9–12]. However, errors in this distribution are likely to some extent, so secondary triage after admission is important to correct categorization and to avoid overuse and blocking of resources [13]. Overtriage will happen in any mass casualty incident and will culminate in increased mortality if not recognized and corrected [3].

Ultrasound is in use as an imaging method for triage and seems adequate because it is widely available, quick, and accurate to a certain level [14]. However, several limitations to its use in major trauma have been described, which demand more sophisticated imaging such as computed tomography (CT) for conclusive patient triage [15].

In a previously published study, a protocol for triage with multidetector CT (MDCT) and fast diagnostic investigation of victims after mass casualty incidents was introduced [16]. Even if the throughput of patients was increased with this protocol, the simulation was very theoretic and did not finally state if the CT workflow would be stable and reproducible during a real event with variable admission rates.

The purpose of this study was to assess the feasibility and reproducibility of the CT triage protocol and the linked workflow in two separate large-scale MCI exercises that also simulated field triage and transport of the patients to the hospital.

Materials and methods

Review board approval

As no human subjects were used in our study, approval of the Institutional Review Board was not needed.

Study supervision

The complete radiology workflow was supervised independently by two of the authors (M.K. and M.M.K.) who were not further involved in the course of the exercise or image reading. The first author (M.K.) recorded all times needed for the further analyses.

Exercise settings

The evaluation of the CT workflow was part of two separate scheduled MCI simulations that were conducted by the city fire department, the regional emergency medical system, and urban hospitals. The time between simulations was 4 months. The data collected for this study derived from a level-I trauma center located in the inner city, about 10 miles away from the site of the incident.

Both exercises were planned on a Saturday, one in the morning and the second at noon. In both events, the simulated scene was a bomb attack at the local soccer stadium with a total of 100 casualties in each, 30 of whom were triaged as “critical” (immediate priority for transport) by field triage teams, and 7 of whom were assigned to be admitted to our hospital.

After the simulated explosion an alarm call was received from the emergency coordination center that informed the responsible staff leader and the hospital about the MCI and the expected number of critical patients to be admitted. The transport of the patients from the site to the emergency department by ambulances was also simulated, but was not part of the evaluation of this study.

Radiology staff

The radiology staff involved was different for each simulation. None of them knew of the simulation before the alarm call. The staff present during the simulations was representative of the staff present for weekends and after-hours (one senior resident, one junior resident, and two general radiology technicians). In addition, the general attending radiologist on call was requested to come to the hospital before the arrival of the first patient for both simulations.

The team was instructed about their specific tasks according to the department’s disaster plan as follows:

  • Technicians: operation of CT and positioning of patients

  • Junior resident: organization of CT workflow and handling of patients

  • Senior resident: reading of images and writing the report

  • Attending: team leader and reading of images

Patient phantoms

Resuscitation simulator dummies were used as phantoms for the CT examinations (Fig. 1). All dummies were equipped with a color-coded triage card and bar-coded tags for identification purposes. Before the exercises all the phantoms were given a history, including clinical presentation at the times of field triage, transport, and admission to the emergency department.

Fig. 1
figure 1

The setup for the simulation of the CT examination. The resuscitation dummy is positioned on the CT table. For simulation of the preparation time, the connection of the ventilator was simulated as shown on the image

Full size image

In-hospital workflow

After admission, the dummies were transferred to the CT room or to the trauma room if the CT was blocked, and placed on the examination table. All patients were assumed to be clinically stable enough to undergo CT. The times needed for the connection of technical equipment such as ventilator, ECG, and power injector for contrast material were simulated. With emphasis on quickly clearing the CT suite, the phantoms were placed on stretchers after the end of the CT examination and moved to the operating room, intensive care unit, or emergency department, depending on their injuries.

In both events, we only simulated the CT workflow. Any other radiological procedures such as radiography and ultrasound were not used or included in the evaluation.

CT examination

All CT examinations were conducted on a four-detector row CT system (Siemens VolumeZoom, Siemens Medical Solutions, Erlangen, Germany) with a previously published dedicated triage CT algorithm [16].

The CT examination consisted of a nonenhanced helical head CT (4 × 2.5-mm collimation, slice thickness 5.0 mm, no gantry angulation), followed by a contrast-enhanced combined single helical CT of the trunk (4 × 2.5-mm collimation, slice thickness 5.0 mm) including the cervical spine, chest, abdomen, and pelvic to the ischiac tuberosities (Table 1). The injection of intravenous contrast material at 4.0 ml/s was simulated with a delay of 45 s before initiating CT data acquisition. Only transverse sections were acquired by default. The raw data of the examinations were digitally stored for later generation of secondary reformations.

Table 1 Triage CT protocol
Full size table

Image reading and reporting

The attending radiologist and the senior resident on call in consensus read the images on the PACS system (IMPAX, Release 4.1, AGFA-Gevaert, Mortsel, Belgium). To simulate the reading and reporting process, stored multisystem trauma cases containing images from a previous CT examination were assigned to each phantom (Table 2). The mean injury severity score (ISS) of these patients was 27, ranging from 16 to 59. Prior to the simulation, the authors defined the injuries that were potentially life-threatening.

Table 2 Trauma-related injuries that were present in the cases used for simulation of the image reading
Full size table

The examinations were loaded to the PACS cache before the simulation by the supervisors. The reading team was instructed to start viewing the images when the PACS transfer of the generated phantom images was completed to simulate the image transfer time as well. Because the dedicated triage CT protocol initially consists of transverse images alone, the readers were instructed to read these images of the CT examinations exclusively. All relevant and potentially fatal injuries had to be included in the final handwritten report. These reports were retrospectively compared with the existing written reports of the CT examinations for missed injuries. Missed injuries were grouped in critical (life-threatening injuries) and not critical (any other injury) (Table 2).

Although the images generated from the CT examinations of the dummies were not needed for image reading, they were sent to the PACS archive over the hospital’s 100-MBit local area network (LAN). The time needed for image transfer was recorded for later evaluation.

Evaluation of the workflow

The following times were recorded for all subjects: arrival in the CT examination room, start of the scout, start of the head CT, end of the head CT, end of the trunk CT, start of final image reconstruction, end of final image reconstruction, end of complete transfer of images to PACS, start of reading, end of writing the report, and time of patient leaving the CT examination room.

From the times recorded, the following time frames were defined:

  1. 1.

    Transfer and preparation of patients (time from arrival in the scanner room to start of the scout)

  2. 2.

    Examination time (time from start of the scout to end of trunk CT)

  3. 3.

    Time for reconstruction of final transverse images (time from start to end of final image reconstruction)

  4. 4.

    Total time in CT examination room (time from arrival to patient leaving the room)

  5. 5.

    Time for transfer of images to PACS (time from end of final image reconstruction to end of complete transfer of images to the PACS)

  6. 6.

    Time for reading of images and writing of reports (time from end of complete image transfer to PACS to end of writing of the report)

Mean capacities per hour were calculated to better display the results. The capacities were calculated as follows: CT examinations completed (60 min/mean total time in scanner room), studies transferred to PACS (60 min/mean time for PACS transfer), reports completed (60 min/mean time for reading and writing of reports), and patients admitted (60 min/[time from first to last patient’s arrival/7]).

The results from both studies were compared using the Mann–Whitney U test. Data were analyzed with a dedicated software package (SPSS 14.0, SPSS Inc., Chicago, USA).

Results

Detailed data are presented in Tables 3–5.

Table 3 Results from two mass casualty incident exercises
Full size table
Table 4 Results for numbers of images and PACS transfer from both exercises
Full size table
Table 5 Capacities calculated from the time frames from both exercises
Full size table

Mass casualty incident exercise I (MCI I)

The time interval between the alarm call from the emergency coordination service and the arrival of the first patient was 72 min. The attending radiologist on call arrived 34 min after the alarm call.

The mean time for patient transport from the entry of the CT room on to the examination table, patient positioning, and preparation for the scan was 2.3 min. The duration of the CT examination from the scout to the end of the chest and abdominal CT acquisition, including the delay for injection of contrast material, was 2.7 min. The mean time needed for reconstruction of the final images was 2.6 min. The mean total time in the room from entering to leaving after completed CT examination was 6.4 min. The resulting patient throughput per hour was 9.4 (CT examinations completed per hour).

Mass casualty incident exercise II (MCI II)

The time interval between the alarm call and the arrival of the first patient was 89 min. The attending radiologist on call arrived 23 min after the alarm call.

The mean times recorded for this exercise were transfer and preparation 1.7 min, CT examination 2.8 min, image reconstruction 2.5 min, and total time in CT examination room 6.1 min. The corresponding patient throughput per hour was 9.8. There were no significant differences between the times from MCI I and MCI II.

Images and transfer to PACS

The mean number of images generated was 224 for MCI I and 211 for MCI II. The times needed for complete transfer of these images to the digital PACS archive were 3.9 min for MCI I and 3.4 min for MCI II. Both measures were significantly lower in MCI II (Table 4). The mean transfer rates (mean number of images/mean duration of transfer) were 57 and 62 images/min, respectively. The PACS capacities for sending studies to the archive per hour (60 min/mean duration of transfer) were 15.4 (MCI I) and 17.7 (MCI II).

Image reading and report writing

For MCI I, image reading and writing of the reports were completed in a mean time of 7.3 min, resulting in a mean film reading capacity of 8.2 reports per hour. For MCI II, the reading and reporting were done in a mean time of 8.3 min, which represents a mean capacity of 7.2 reports per hour.

The written reports were evaluated after the exercises. In both simulations, all trauma-related injuries that were classified as potentially life-threatening were identified (Table 2). Of 97 injuries 93 (96%) were correctly described in the reports; 4 of 97 (4%) injuries were missed in two patients, specifically the fractures of the transverse process of two thoracic and one lumbar vertebra were missed in one patient and lung contusions to one lung were missed in another patient.

Admission of casualties

The number of critically ill patients admitted to the emergency department was 7 for both exercises. In MCI I these patients arrived during a period of 76.8 min, which means one patient arrived every 10.9 min. The computed patient admission rate per hour was 5.5.

In MCI II, the seven patients were admitted within 56.5 minutes, which means one patient was admitted every 8.1 min. The resulting admission rate per hour was 7.4.

Discussion

In 2006 a dedicated four-detector-row CT protocol (Triage MSCT) that was suited to the triage of a large number of casualties mainly because of the reduced average length of the patients’ stay in the CT examination room was published [16]. The patient throughput per hour was increased to 6.7; however, the study was only a theoretical simulation, and there were still some problems to be solved, such as time for tube cooling. Whether the capacity for patients would be sufficient for a real event was not stated.

The results from our study have shown no significant differences for both exercises in the calculated time frames for transfer of patients, CT examination, image reconstruction, total time in the CT examination room, and reading of images (Tables 3 and 4). We were able to prove that the dedicated triage CT protocol was suitable for quick imaging of larger numbers of patients, and that the results were stable and reproducible with different staff. In both events, the maximum patient throughput per hour was even increased in comparison to previously published results [16].

When we compared the calculated capacities there were some differences, although the number of completed CT examinations per hour was constant in both events (Table 5). The number of completed reports per hour was lower in MCI II. In both simulations the image reading capacity was lower than the capacity for acquiring CT data. When the number of patients examined is higher than the reading capacity, delays for further treatment may be expected in a real event. However, in the first event the number of patients admitted per hour was reasonably lower than the reading capacity (5.5 compared with 8.2), which in this case would not have led to appreciable delays.

In the second simulation, the admission rate was slightly higher than the reading capacity (7.4 compared with 7.2) but still delays would be only short. However, this hypothesis is valid only if reading time and admission rate of patients were constant. In our experience from both exercises, admission rate can be variable. While at some points two or more patients arrived simultaneously, we also noted periods of 10 min or more when no patient was admitted, and the CT equipment was not in use because of the variable admissions of patients.

The time for the digital transfer of the images to the PACS must also be considered. In the current study, the image count in the second event was significantly lower as was the time needed for PACS transfer. In both events, the transfer time was considerably lower than previously reported [16]. Performance of the sending process is dependent on technical specifications and network load. Our hospital uses a 100-MBit LAN connection, which is shared with other departments. As the two simulations reported in this study were conducted at weekends, the network load was expected to be lower than during the main working week. If a mass casualty incident occurred during the week, transfer time could be quite longer, and could seriously limit the overall performance of the radiology department. Solutions to this problem may be faster or dedicated departmental networks, or direct connection of the CT scanner and the viewing console.

Many factors contribute to the inconsistent rate of admission of patients. As reported in analyses of past mass casualty incidents, the admission rate is neither constant nor linear [17, 18]. The first peak of casualties admitted usually consists of walking wounded and noncritical victims who seek care in the hospitals located nearest to the site of the incident. Generally, patients who were triaged as “critical” should be evacuated first. When extrication of these critically injured victims is time-consuming, transport to the hospital may be delayed, which results in a second peak in admissions with more seriously injured patients than in the first peak [19, 20]. Coordination of the emergency medical system and the ambulances involved is also difficult. Einav et al. reported that only 79% of the ambulances were able to reach the site of the mass casualty incidents they analyzed, and only 59% evacuated patients [21].

If the patient distribution to the available hospitals is not carried out properly, delays in transport as well as overcrowding or underutilization of some hospitals can be expected [13, 22]. In our simulation, the distance from the event’s site to our hospital was 10 miles, so consequently delays because of heavy traffic or even breakdowns have to be considered.

We noted problems in correct identification of the patients during both simulations. Although each of the phantoms was marked with a triage tag, which had a unique identification number and barcode, the CT reports were sometimes not correctly assigned. For our simulations this might have been caused by the fact that the images were read with examinations that were not labeled identically as the specimens. In a real event, all images should be labeled with the identification number of the casualty that could help to avoid this error. There are more sophisticated means of labeling patients such as electronic devices, but even with these, mislabeling errors are reported to occur frequently [23–25]. However, a preplanned labeling system is essential for handling multiple casualties.

As our results are based on simulations rather than on real events, this study has some limitations. The slice thickness for the cranial CT was increased compared with our clinical standard in order to reduce the duration of data acquisition and image calculation. Furthermore, only transverse sections at 5-mm slice thickness and with one kernel were calculated for each body part. These procedures contributed to a small image count, a short transfer time to the PACS, and reasonable reading times. Even so it was possible to rule out all present life-threatening injuries sufficiently as shown by our results. Of course we are aware that subtle injuries such as nondislocated fractures or small organ injuries are likely to missed on these images. However, in the initial evaluation of mass casualties, severe injuries that have an impact on patient outcome and further resource planning have to be reported without delay. After the acute situation has been cleared, images at thinner slice thicknesses and other reformations such as MPR can be calculated.

As we did not acquire any actual images, we are not able to finally state if their quality would be sufficient. Exposure had to be reduced because of tube cooling time as observed during a past study, and the quality of images is accordingly reduced, which may lead to possible equivocal findings [16]. In the current study, we did not experience tube overheating, but it still has to be considered if more patients arrive at one time and the frequency of examinations increases.

In both simulations, we concentrated only on the workflow of patients who were categorized as “critical” after field triage, and we did not consider any patients in lower triage categories who arrived. These casualties would also require attention and could block resources needed for the critically injured patients. Our hospital is a campus of dedicated departments separated in different buildings. The study was conducted in the level-I trauma department, which was assigned to accept and treat only critically injured patients. In the case of a mass casualty incident, all other patients are directed to the department of internal medicine, which is located across the street. Even with this precaution we cannot exclude that patients who are not critically injured will overcrowd the emergency department.

For the operation of this second department as well as for radiologic procedures other than CT, further staff is needed. For this reason, our hospital is equipped with an alarming system that automatically gives phone calls to all of the department’s staff to come to the hospital when needed. In both exercises we did not investigate this system and the workflow for patients not triaged as critical because our intention was to evaluate the CT workflow exclusively. Another exercise would be needed to evaluate the arrival times of the staff.

At the time the simulations were conducted, we only had this particular four-detector row scanner installed at our department. We are not able to ultimately state the influence of newer and faster CTs such as 16- and 64-channel CT systems on the workflow, or the impact of having more than one CT unit available. However, CT data acquisition contributed to only about one third of the patient’s total time of stay in the CT examination room (Table 3), which seems to minimize the effect of a faster CT system to improve the number of examinations per hour.

Further investigation is also needed on the effect of using more than one CT system for triage. With the staff available at our simulations, effective operation of more than one CT machine would not have been possible, consequently making utilization of further staff necessary.

Conclusion

With two different mass casualty incident exercises we were able to validate the use of CT for further triage of patients admitted to the hospital. In our experience, the number of CT examinations completed was higher than the admission rate of patients and did not represent a bottleneck in the workflow. The limiting factor in terms of radiology was the image reading capacity, which still nearly equals the admission rate. The results also indicate that regular staff training is necessary to help them recognize and solve problems and to keep them prepared. The workflow described in this study and the suggested CT protocol can be adapted to other CT equipment and should help other departments to practice for mass casualty incidents effectively.