Early versus late surgery after cervical spinal cord injury: a Japanese nationwide trauma database study

The institutional ethics committee of the hospital approved the present study. Our analysis did not include personal identifying information, so the requirement for written informed consent was waived.

This was a retrospective cohort study the using Japan Trauma Data Bank (JTDB) database which is a nationwide, multicenter database administered by The Japanese Association for the Surgery of Trauma and The Japanese Association for Acute Medicine [16]. This database contained trauma cases classified as Abbreviated Injury Score (AIS) grade 3 or greater, and 244 hospitals, which included tertiary care emergency hospital (e.g., level I trauma centers) and less specialized emergency care hospitals, participated in the database [16]. Patients with cervical SCI (especially isolated cervical SCI cases without multiple trauma) are transported to not only the tertiary trauma centers but also the less specialized hospitals in Japan. The patients were registered in the JTDB at the last hospital if they were transferred to another hospital. The database contained information regarding patients’ age, sex, vital signs on scene, vital signs in the emergency room, mechanism of injury, diagnosis, AIS score [17], Injury Severity Score (ISS) [18], medical cares, ward, length of ICU stay, and length of hospital stay and mortality. Diagnosis and coding of the AIS were performed by doctors in charge of each institution. This database followed the patients till they discharged of the hospitals, transferred to other hospitals, or died in the hospitals.

We used the information on trauma patients enrolled in the JTDB between 2004 and 2015. We defined cervical SCI and graded the severity of cervical SCI using the AIS 2005 update 2008 code. For cervical SCI, AIS scores were coded from 1 to 6; a bigger number indicated a more severe neurological status. AIS 6 meant currently untreatable [17].

Inclusion criteria were as follows: (1) adult (15 years and above) patients, (2) having blunt isolated cervical SCI, (3) having undergone surgery within 7 days after injury, and (4) being admitted to the ICU. The exclusion criteria were as follows: (1) patients coded AIS 6 (currently untreatable patients), (2) absence of information about mortality, and (3) having died within 7 days after admission. We excluded patients who died within 7 days to avoid immortal time bias [19].

We grouped the participants into two groups: the early group who underwent surgery within 24 h after injury, and the late group who underwent surgery between 25 and 168 h (7 days) after injury. There is no standard definition for “early” surgical intervention in the literature regarding surgery for SCI with the range of “early” surgery being from within 4 h of injury to within 72 h of injury. The more recent trend is to define surgery within 24 h of injury as “early” surgery, and therefore, we defined early surgery as decompression or stabilization within 24 h of injury in the present study [7, 20, 21].

The primary outcome was the length of ICU stay. The secondary endpoint was in-hospital mortality.

We grouped cervical cord injury patients who underwent surgical management into two groups: the early group and the late group according to timing of surgery. To compare the two groups, we compared categorical variables of the patients’ baseline characteristics using a chi-square or Fisher’s test. We analyzed continuous variables using the Student t test or Mann–Whitney U test, appropriately. We further performed a sub-group analysis and evaluated the relationship between the timing of surgery (day 1 to 7) and the outcomes, using Mantel–Haenszel trend test. We then performed multiple imputation, by which each missing value was replaced with a set of five substitute plausible values, to decrease the bias caused by incomplete data. A multivariable regression model was constructed for each imputed data set, and a single model was created by statistical inference about the results of the five imputed datasets [22, 23]. Next, the propensity score was estimated using a logistic regression model adjusted for the patient characteristics, cause of injury, and vital signs at the scene and on admission, as these variables have been shown to be prognostic predictors in previous studies (age, gender, prehospital vital signs, vital signs on hospital arrival, body temperature on arrival, GCS on arrival, ISS, and cervical AIS) [16, 21, 24]. We then performed a linear regression analysis for the length of the ICU stay for cervical SCI to compare different surgery timings. Lastly, we used the Cox proportional hazard regression analysis with variables that were the timing of operation (24 h vs. 7 days), the aim of operation (stabilization and decompression), and the propensity score to evaluate the secondary outcome, in-hospital mortality. The statistical significance threshold was a p value of less than 0.05. All analyses were carried out using SPSS software (IBM Corp., Armork, NY, USA, version 23).

There were some limitations in our study. First, the JTDB does not contain information about neurological status before and after surgery. For example, the database did not include Frankel Grade or American Spinal Injury Association Impairment Scale, so we could not evaluate the relevance of the timing of surgery for cervical SCI and neurological improvements, that is, the efficacy of surgery for SCI. We did assess the severity of cervical SCI using the AIS; however, many previous studies have used the levels of injury, in terms of AO Spine Trauma Classification or Frankel Grade. Therefore, we could not directly compare the present study with many previous studies. Second, we could not obtain information about the indication for surgery, the reason behind the timing of surgery, the comorbidities of the patients, the complications of the operation, or the cause of death. Consequently, we could not analyze the direct effect of timing of surgery on the length of the ICU stay. Third, this database does not include information about comorbidities, number of days on ventilator, complications, and neurological status; therefore, whether there were some causes that had made the hospital stay longer was not clear in our study. Fourth, regarding the AIS coding rule about the spine, paralysis should be coded according to its status at 24 h post-injury [17]. It may be possible that the current study included the patients whose AIS could have changed within 24 h as a result of operation itself or in the natural course (i.e., unrelated to operation) in the early group.