This was a single-center retrospective cohort study conducted at Saiseikai Utsunomiya Hospital, Tochigi, Japan. Recommendation for a whole-body CT was at the discretion of the attending physician, who considered the following indications: severe injury or unknown mechanism, altered mental status, distractingly painful injury, or multiple injuries identified or suspected at physical examination.
The protocol of whole-body CT included a non-contrast CT from the head to the pelvis, an arterial phase from the neck to the pelvis, and a venous phase from the neck to the pelvis. When a limb injury was suspected, the relevant extremity was added to the range of whole-body CT. A 64 detector CT scanner (SOMATOM Definition AS, (Siemens Healthcare, Erlangen, Germany) was used during the study period.
We extracted data on patients who sustained blunt trauma injuries between January 1, 2014 and April 30, 2017. We included patients whose serum D-dimer levels were measured before they underwent whole-body CT with contrast, within 24 h after injury. We excluded patients with systolic blood pressure (sBP) <90 mmHg and those with Glasgow Coma Scale (GCS) score <15 at admission. We also excluded patients with any neurological abnormalities, indicative of injuries that could not be diagnosed using CT, including spinal cord and peripheral nerve injury. Patients who underwent surgery or angiography before whole-body CT were also excluded.
Data collection and definition
Data were extracted from electronic medical records, including information on age, gender, mechanism of injury, vital signs at admission (GCS score, respiratory rate, sBP, and heart rate), abbreviated injury scale (AIS) score, ISS score, Revised Trauma Score, Trauma and Injury Severity Score Probability of Survival, serum D-dimer level, and detailed information obtained from whole-body CT. All images acquired through whole-body CT were re-evaluated by board-certified radiologists not otherwise involved in this study. Disagreements between radiologists were resolved by discussion.
Injury sites were divided into five regions (head/neck, face, chest, abdomen, and limbs/pelvis) according to the AIS coding system. “Isolated” injury was defined as an AIS score ≤ 5 in one of five regions, and an AIS score ≤ 1 in one of other four regions. “Isolated non-severe” injury was defined as an AIS score ≤ 3 in one of five regions and an AIS score ≤ 1 in one of four regions.
Primary outcome was defined as isolated injury; secondary outcome was defined as isolated non-severe injury.
The suitability of using serum D-dimer levels to predict primary and secondary outcomes was assessed by discrimination and reclassification analysis. Receiver operating characteristic (ROC) curves for D-dimer levels, according to isolated injury status, were drawn; the area under the ROC curve (AUROC) was evaluated. Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated for several candidate D-dimer level cutoff values to obtain the lowest possible value that was most likely to eliminate the need for whole-body CT (i.e., a value that predicted an isolated injury).
Sensitivity analyses were performed on a validation cohort of patients with GCS score of 13–14 points at admission, added to the original study population to assess the robustness of the proposed model. Same statistical analyses were applied in sensitivity analyses.
All statistical analyses were performed using EZR version 1.42 (Saitama Medical Center, Jichi Medical University, Saitama, Japan) , which is a graphical user interface for R (The R Foundation for Statistical Computing, Vienna, Austria).