The most important finding of this study was that multislice computed tomography missed at least one diagnosis in 25%, and at least one potentially fatal diagnosis in 4.1%, as determined by autopsy in trauma patients not surviving the first 48 h after admission. 67% of the missed diagnoses and 100% of the potentially fatal diagnoses were located in the thoracic region. Total numbers of MD and MPFD were significantly lower for the newer CT generation (64 MSCT, N = 11), compared to older one (4 MSCT, N = 26). The time period of 48 h until decease of the patient as inclusion criterion for the study was chosen to exclude secondary potentially fatal complications, such as pneumonia, SIRS, sepsis, organ dysfunction or MODS as possible confounding factors.
Our demographic data were in accordance with the previous reported cohorts. Fung Kon Jin et al. quote a number of 16 patients in one year (trauma, autopsy, at least partial diagnostic assessment but variable times of death after admission) [11]. Sharma et al. included 160 patients in four years (trauma, autopsy, death within 24 h, at least partial diagnostics but also patients admitted under CPR or already dead) [7]. The average age (53.16 years) and distribution of gender (63% male) of our cohort also matches the values of comparable studies [4, 11, 12].
In our cohort, 4.4% of all trauma patients admitted over the ER did not survive the first 48 h in hospital. Compared to previous studies, this is a quite low value. Matthes et al. found an early lethality within the first 24 h of 8.7% in severely injured patients [13]. Other authors found an early lethality of 46 through 61% within the first 48 h [4, 7, 10, 12]. Differences in definitions used for “trauma” among the various studies consequently lead to different selection criteria for eligible patients.
The incidence of all missed diagnoses in our study was 25%. In other studies, the incidence ranges from 0% (Forsythe et al. [10]) to 57.6% (Sharma et al. [7]). Different autopsy rates in many studies, differences in quality, regulations of accuracy and requested exploration may be reasons for the discrepancies. Furthermore, the definition of a missed diagnosis varies. Many authors only included severe MD or MPFD [7, 10]. The MPFD rate in this study was 4.1%. Reported rates of MPFD range from 0 to 14% [4, 7, 8, 12].
Injuries of the thorax represented the area with the majority of MD (66.6%), followed by head/neck (18.2%), and abdomen/pelvis/extremities (15.2%). This corresponds well to quotes in prior studies [4, 12].
Secondary complications and comorbidities were not counted as MD. During autopsy, venous thrombosis of the sinus was found in five patients as a diagnosis not known before. This diagnosis was valued as secondary complication and thus not counted as a MD. Nevertheless, diagnoses like bronchopneumonia and severe haemorrhage are dedicated to be the most common MD in trauma patients [4, 7].
Reasons for missing a diagnosis by means of MSCT scan are numerous, including a variety of human and technical errors. While human errors appear to be more or less constant and independent from the resolution of the CT scanner, technical errors appear to be reducible by sufficient image thickness and image intervals. In the particular setting of hemodynamically unstable patients, insufficient circulation patterns may compromise the distribution of injected contrast medium and accordingly the detection of active bleeding injuries.
As for the survey of the ISS, severities of traumas were reflected in the identified values. After autopsy, we did not find a significant change of the ISS (median ISS 24 before, and 25 after autopsy). Hodgson et al. reports an increase from 30 to 43 in trauma patients after autopsy [8]. Other authors describe an increased ISS after autopsy in 7–69% of all patients [12, 14].
The question whether MSCT is accurate enough to replace autopsy is of particular interest. In 2007, Molina et al. conducted a study on 113 trauma patients who obtained a CT scan within 24 h before death and who finally underwent autopsy [15]. They found unacceptably low rates of sensitivity and specificity as well as positive and negative predictive values for MSCT. This study distinctly challenges the idea of post mortem CT diagnostics, particularly in the forensic area. However, it has to be considered that this study was conducted between 2002 and 2005, using CT scanners with image thickness of 5–10 mm at 5-to-10-mm intervals for the skull, 3-mm image thickness at a 2-mm interval for the cervical spine, and 5-mm image thickness at a 5-mm interval for the chest, abdomen, and pelvis. This makes their findings comparable to our group 1 and illustrates the impact of high-resolution CT scans with sufficient slice thickness and slice intervals on diagnostic accuracy. Our results suggest that the rate of technical errors can be reduced by using a 64 MSCT scanner compared to the older 4 MSCT scanner.
This study has several limitations. First, only missed diagnoses offering a direct relation to the causative trauma were considered. Second, MSCT scan was done in some patients with insufficient circulation patterns, which made it impossible to exactly evaluate the distribution of injected contrast medium and accordingly to detect all active bleeding injuries. Third, due to the retrospective study design and partially insufficient documentation—e.g., imaging from external referring hospitals or lost records, we were not able to fully report all missed diagnoses on all patients. 32 patients had to be excluded for this reason.
