Advanced Accident Research System Based on a Medical and Engineering Data in the Metropolitan Area of Florence

Simone Piantini; David Grassi; Marco Mangini; Marco Pierini; Giovanni Zagli; Rosario Spina; Adriano Peris


BMC Emerg Med. 2013;13(3) 

In This Article


The study is based on the direct collaboration between the Department of Mechanics and Industrial Technologies at the University of Florence (Italy) and the ICU of the Emergency Department (Careggi Teaching Hospital, Florence, Italy), and, indirectly, with police forces involved in the road accident detection, the Emergency Medical System (EMS) of Florence and the Emergency Room (ER) of the Careggi Teaching Hospital.[21]

Internal Review Board waived the need of ethical approval due to the nature of the study. And the the aim of this research is to conduct an in-depth investigation into road accidents that have generated severe injuries (major trauma and potential ones) in the metropolitan area of Florence, and to reconstruct the causes and the mechanisms of the injuries. Moreover, the study aims to collect information regarding the disabilities sustained by the injured in order to evaluate their social costs, and also to determine what changes and improvements to vehicle design might mitigate or prevent these injuries in the future. To this purpose, a network of physicians, statisticians and engineers was established to link environmental data acquired on the scene of the accident with crash parameters and clinical information about the injuries.

The study selected all the road accidents where at least one of the persons involved was admitted to the ICU with a diagnosis of major trauma, i.e. an ISS greater than 15. None dead on-scene or in the ER case were collected in this study. The working team, named In-SAFE team, is composed by ICU physicians, engineers and statisticians.

Sampling Area and Representatives

The road accidents analyzed in this study were in the metropolitan area of the city of Florence. This area is made up of nine municipalities, covers a surface of 466 km2, with a population of approximately 604.000 people (Figure 1). The sampling area is mainly composed by urban zones and in small part by extra urban areas.

Figure 1.

Sampling area.

Since 2005, the trauma network of the Tuscan Region has organized the ICU, which works on major trauma through the hub/spoke system. For the Province of Florence, the hub hospital of reference is the Careggi Teaching Hospital which receives all major traumas of patients that are more than 16 years old.

In 2010, Florence was the province with highest number of road accidents and injured in Tuscany (Figure 2). Sixty-five percent of the major traumas in Tuscany were caused by road accidents, and only 3% of these occurred on highways. The access for major trauma to the ER of the Careggi Teaching Hospital confirms the regional trend (Figure 3). Therefore, the metropolitan area selected should ensure that the distribution of the sample is similar to the TTR.

Figure 2.

Number of road accidents and injured in Tuscany for 2010.

Figure 3.

Number of major trauma in Tuscany and at the Careggi University Hospital for 2010.

An In-depth Multidisciplinary Investigation

With the cooperation of the police forces, the In-SAFE team acquires general information about: the crash scene, e.g. point of impact, point of rest; description of the environment, e.g. roadway configuration, traffic control data, weather conditions; the vehicle, e.g. type and model, engine size; and people involved in the crash, e.g. gender, age, type of licence and so on. In the following the main phases of the study are outlined. They are also shown in Figure 4.

Figure 4.

Flow chart of phases and data of the study.

On-site Investigation

The team collects more detailed information, such as skid marks, debris, deposit of liquids, point of rest of the vehicle, line of sight of each vehicle's driver/rider or people involved in a crash, in order to substantiate the exact point of impact.

Vehicles Examination

Each vehicle involved in the accident is carefully examined by the In-SAFE team. All damage (direct or indirect) or contact points are photographed (Figure 5).

Figure 5.

On scene and vehicle damage photography documentation.

Exterior Parts The damage profile is quantified measuring the damage width. The latter is subdivided in six parts (C1-C6), where the dimension of the damage is quantified (CRASH3 method).[22,23] In order to describe the nature and the location of the direct contact on the vehicle in car and van accidents the Collision Deformation Classification (CDC)[24] is used. For accidents involving medium and heavy trucks, and articulated combinations, the Track Deformation Classification (TDC)[25] is used.

The Wraps Around Distance (WAD) measurement for determination of the pedestrian or cyclist interaction with the vehicle is also acquired. Finally, for the PTW, the wheelbase shortening is collected.

Interior Parts Vehicle interiors are thoroughly investigated for evidence of occupant contacts, and to quantify the intrusions. These data are then stored using the Passenger Compartment Classification (PCC) developed by Standardization of Accident and Injury Registration Systems (STAIRS) project.[26] Special attention is given to the usage of the seatbelt, activation of the pretensioner, and airbag activation (Figure 6).

Figure 6.

Interior vehicle photography documentation.

Accident Reconstruction Methodologies

From the previously collected data, the accident is reconstructed to evaluate the accident dynamics and the main physical parameters concerning the crash phase, as well as pre-crash phase manoeuvres, such as avoidance actions.

The post-crash velocity of each vehicle involved in a crash is evaluated by means of the analysis of the post-crash motion. The deformation energy and the velocity variation (ΔV) are estimated through Crash3 method.[27] The impact velocity of each vehicle is assessed by applying the principles of energy balance and the momentum analysis. By the use of crash simulation software (PC-Crash 8.3 and Virtual CRASH 2.2) all the previous data are verified and validated, and other parameters, such as the PDOF and the impact angle between the vehicles, are also evaluated.

In order to assess the range of uncertainty of the analysis, the Finite Difference Method (FDM)[28] is used. This is a numeric approach to partial differentiation of the equation used. The method consists in the calculation of the uncertainty range around the nominal value.

Injury and Physiological Derangement Evaluation

The medical data collected in the database are selected to provide a clear correlation between the trauma's dynamic and the injury's localization and severity.

The main information coming from the EMS (e.g. Glasgow Coma Scale, blood pressure, and intubation) and ER/ICU (e.g. diagnostics), the AIS and ISS scores, the EMTRAS and the Computed Tomography information, are the scores and data chosen for the previous aims.

The AIS was developed by the Automotive Committee On Medical Aspects of Automotive Safety in 1971.[29] The last revision of the score is the AIS 2005, updated in 2008. Because the different AIS versions are not always compatible, injury severity scoring tools using the new AIS should be compared to those using previous versions in terms of score and predictive performance:[30] Carroll et al. show a reduction in traumatic brain injury (TBI) AIS when recorded using the 2005 revision versus the 1998 one.[31] For this reason, the In-SAFE database includes the AIS 2005 and AIS 1998 codifications, in order to asses differences in trauma severity classifications, and to allow the comparison with other databases using both revisions of the AIS. The ISS was introduced by Baker in 1974 to classify the severity of traumas involving lesions in more than one of the AIS regions. The score is calculated summing the square of the three highest AIS of three different body regions. No more than one AIS can be taken from a single region.[30,31] [30,31] If a lesion is graded as 6, the ISS is automatically calculated as 75. This choice put greater attention on the multiplicity of trauma injury but at the same moment it can overlook multiple lesions suffered by the same part of the body. For this reason in 1997 Osler et al. developed the NISS, which is calculated summing the square of the 3 highest AIS, without any regard to the body region.[32,33] The authors affirm the superiority of the NISS to the ISS to predict the outcome of the trauma patient, and this conclusion is supported by Lavoie et al..[34] In addition, for research purposes, the EMTRAS score, a new trauma score developed in Germany in 2009 that is calculated by using the age of the patient, the on-scene GCS, the Base excess, and the Prothrombin Time at the ER,[35] has been added to the In-SAFE database. Drug and alcohol abuse are a major cause of loss of life, threatening injury in motor vehicle accidents, both in the US and in Europe.[36,37] Drugs test includes ethanol, cannabis, cocaine, amphetamine, benzodiazepine, barbiturate, and opioids dosage, collected upon admission in the ER, and recorded in In-SAFE. Ethanol was measured with head-space gas chromatography/mass spectrometry, whereas cannabis, cocaine, amphetamine, benzodiazepine, barbiturate ,opioids were dosed using Enzyme-Linked ImmunoSorbent Assay). To avoid false positives, daily internal control dosage are performed, and in case of a patient with elevated concentration of a substance (in absence of a known addiction), the analysis is repeated Moreover, on scene drugs are recorded, as well as first aid medical treatments.

The impact of road accident dynamics and lesions on the outcome are studied by recording length of stay, mortality at 6 months, and the follow-up program at 6 months on the ICU database. As an indicator of the quality of life recovered at 6 months after the event (follow-up at 6 months) the Glasgow Outcome Scale (GOS)[38] is used, as well as the questionnaire EuroQol5 EQ5-D with scale EQ5-D-VAS,[39] which includes a medical examination. In case a patient cannot sustain a medical visit, a telephone interview is performed. Patient pre-accident drug treatment and pre-existing medical conditions seem to correlate with worse outcome, in terms of complication, ICU and Hospital length of stay, and lower functional outcome.[40–43] For this reason these data are recorded in a dedicated section of the database that includes the type and number of pre-existing medical conditions, and the type and dosage of each drug (ethanol, cannabis, cocaine, amphetamine, benzodiazepine, barbiturate, opioids). Despite some limitations due to risk related to ionizing radiation, CT remains the most sensitive imaging exam to assess trauma lesions:[44–47] for this reason for head, neck, face, chest and abdomen CT slices are chosen.

In addition to the encoding of each lesion using the AIS code, these are identified by means of a three-dimensional localization tool that uses a discretization of the human body based on a set of CT slices equipped with an active matrix (Figure 7).

Figure 7.

Graphical method for the active injuries' localization.

This was done by dividing a human body not affected by clinical pathologies through cross sections of CT scan made at regular intervals in the sagittal plane (z axis). Each slice (or plane) is divided into a point's matrix. In this way, each point has its Cartesian coordinate (x, y, z) fixed, where x and y are read in the transverse plane (CT slice) while the z coordinate is the height of the CT slice, with zero value at top of the head. The matrix dimension depends on of the size of the section. The body regions head-face, neck, thorax, and abdomen are divided, respectively, into 8, 3, 15 and 13 slices. For the facial bones, vertebrae, rib cage, pelvis, and limbs, an active matrix built on the anatomical atlas figure is used to localize lesions with more sensitivity.

This type of localization of the lesions, for example, provides a means to compare the distribution of the damage (in terms of extent of the lesion) among different people, or even to realize the frequency distributions of the damage (mean and standard deviation) relative to a certain region of the body. More generally, it provides the possibility of correlating the area of damage with other types of information (i.e. impact velocity or direction, type of crash).

Injury Correlation Phase

This phase is the heart of the study but also the most complex and subjected to errors. In this stage, the kinematics and dynamics of vehicles and people involved and the injuries are correlated. The injury information is assessed mainly by CT scan performed at the admission in the ER; other imaging exams (i.e. vascular CT Scan, Magnetic Resonance Imaging) can be added to CT to identify specific lesions.

The dynamic and kinematic information of the vehicles and people involved are assessed through physical principles and software. Once the injuries and dynamics are clearly identified, a meeting between intensive care physicians and engineers is organized in order to correlate each injury to its cause. By merging the data previously gathered and using state-of-the-art biomechanics of impact, it is possible to understand cause and mechanism of injuries.

In the end, for each association, the definition of a level of reliability of the correlation process (β), in percentage, indicates the quality of the data produced.

The reliability is defined as

β = 1 − a

where α is the uncertainty that we have about the association (injury vs. cause).

During the data analysis phase, a threshold value, fixed in β = 60%, is used for the selection of the most significant associations (Table 1).

Data Stored System

All the data collected are stored in a relational database (In-SAFE), where the variables are coded in accordance with the state-of-the-art techniques. The standardized protocols taken as reference are the Common International Methodology for in-depth accident investigation (OECD)[48,49] and STAIRS project.[26] The In-SAFE database contains about 700 variables divided in three main groups: environment, vehicles and people. The people group contains both demographic and medical information (Figure 8).

Figure 8.

Database In-SAFE – Main clustering of data collected.

Correlation Analysis between Injuries and Dynamics: A Case Study

This accident, which occurred on an urban road, involved a 26 year old rider of a moped (scooter style) in a head-on collision against a road sign (single vehicle accident). Informed consent to publish this case and any accompanying images was obtained from the next of kin of the patient. The road was straight and divided into two roadways separated with a curb indicated by the road sign, as seen in Figure 9.

Figure 9.

Scene of the accident, with point of impact and point of rest of rider and moped.

The rider, with a positive blood alcohol level (2.6 g/l), was riding at night (with road illumination) and heavy rain conditions. The moped was equipped with a windshield. Due to the high blood alcohol level (in this case the primary cause of the accident), the rider failed to keep a straight trajectory and collided with the road sign (1st impact). After a flying phase both the moped and rider impacted with the ground (2nd impact) and continue with a sliding phase before stopping. The total distance covered by the scooter from the point of impact to the point of rest was about 25 m, while the total distance covered by the rider was about 21 m.

Applying the equation of the launched ballistic proposed by Searle[48] it is possible to estimate the impact velocity of the moped (62 ± 5 km/h) and through computer simulation it is possible to reconstruct a 3D scenario of the accident and refine and validate the crash parameters, such as the impact velocity (57 ± 5 km/h) and the delta-V (8 ± 3 km/h).

The moped used for the computer simulation is a generic scooter modelled as a rigid body, resized in terms of mass, wheelbase, and dimension of the wheels. The rider is modelled as a multibody human model available in the software. Comparing the POR of moped and rider obtained with the software and those measured (points 1, 2, 5), as seen in Figure 9, it is possible to see the good quality of the computer simulation performed with the software. The rest position of the rider reconstructed with the software is in good agreement with the actual final position, while the moped one is relatively good but does not perfectly match with the actual position, probably due to the simplified model used to represent the moped and mainly in the modelling of the first impact.

The rider was wearing a demi-jet helmet that became detached after the first impact. For this reason, during the impact against the ground, he sustained serious head injuries and eventually died 47 days after the accident.

The Maximum AIS (MAIS = 4) sustained by the rider is in the head/neck body region and thorax body region, and the ISS score is equal to 33 (Table 2).

In agreement with the on-scene and vehicle investigation and reconstruction, in the first impact the rider crashes with the front-left side of the moped and with his head striking against the yellow part (zone 1) and the blue part of the road sign (zone 2) (Figure 10). After this impact, rider and moped begin a flying phase which ends with landing on the ground and the subsequent slide to the rest position. In this phase, the rider impacts his head and then his thorax on the ground (Figure 11).

Figure 10.

Impact against road sign (1st impact).

Figure 11.

Rider impact on the ground (2nd impact).

As a consequence to the first impact (against the road sign) with the helmet on, the rider sustained the following injuries: left temporal polar lesions (2.5 cm) with millimetric left frontal parietal subdural hemorrhage (Figure 12).

Figure 12.

Head injuries – impact against road sign.

The subdural hematoma (or hemorrhage) is classified as a focal TBI i.e. a coup effect. This is caused by the compressive stresses that are generated when there is a relative motion of the brain with respect to the inner surface of the cranial cavity due to the inertial effects. As a consequence of the detached helmet, the impact against the ground occurs without any protection, causing the most serious head injuries. Ground contact also accounts for the thoracic injuries.

The main head injuries highlighted by CT scan (Figure 13) are: right temporal-parietal-occipital multiple fractures, depressed in the occipital region and diastatic in the mastoid region; diastatic skull base clivus fracture, involving sphenoid bone body and both carotid channel; right temporal styloid process and right tympanic fracture; right petrous fracture with hemotympanum; pneumocephalus bubbles; lacerated and contused right temporal parietal (2.5 cm) lesions; peri mesencephalic subarachnoid haemorrhage, with relative encephalic pons and mesencephalic hypodensity and widespread cerebral oedema.

Figure 13.

Head injuries – impact against the ground.

The depressed skull fractures are caused by the direct contact with the ground that has generated a high deformation of the skull. This is due to the minor lateral strength of the skull with respect to its frontal and rear regions.[50,51]

A right upper lobe lung contusion and bilateral lower lobe lung contusion in the paravertebral area are also sustained in the thoracic region (Figure 14). Both injuries are caused by the compression of the lung at high impact velocity.

Figure 14.

Thorax injuries – impact against the ground.

A summary table with all correlation results and level of reliability in percentage values is shown in Table 1.