Methods
The Strengthening the Reporting of Observational Studies in Epidemiology guideline was used to ensure proper reporting of methods, results, and discussion .10
Study design
This was a multicenter, nationwide, retrospective cohort study.
Data source
The Japan Trauma Databank (JTDB) is a multicenter, nationwide trauma registry in Japan that was established in 2003 by the Japanese Association for the Surgery of Trauma and the Japanese Association for Acute Medicine to improve the quality of trauma care in Japan. In principle, trauma patients who are transported to the participating hospitals and have an Abbreviated Injury Scale (AIS) score ≥3 are registered in the JTDB. Two hundred and ninety-two acute care hospitals that provide trauma care throughout Japan were the participating centers of the JTDB as of December 31, 2021. Data are entered into the JTDB by physicians or medical assistants who are trained in AIS coding and include various items such as prehospital care, initial treatment, diagnosis, in-hospital treatment, and clinical outcomes (eg, mortality and LOS). In this study, we used the JTDB data that were released in 2021 and included trauma patients treated between January 1, 2004 and May 31, 2019.
Participants
The inclusion criteria were as follows: (1) age ≥18 years, (2) intentional fall from a height, and (3) LOS ≥72 hours. Patients who had an AIS score ≥5 in the head region, those who died after admission, or those whose Injury Severity Score (ISS) data were missing were excluded from this study. However, those who died after admission were included in the sensitivity analysis with a competing risk model, which is discussed in the Statistical analysis section.
Data collection
Data related to patient and hospital information were obtained from the JTDB, which included information on the demographics; prehospital, ED, and in-hospital treatments; AIS scores; ISS; and clinical outcomes.
Exposure
NRF was developed based on our previous study7 which investigated the differences in trauma injury patterns and severity between intentional and accidental falls from a height. In the intentional fall group, the trauma severity increased in the lower extremities and pelvic region as the ISS increased, whereas in the accidental fall group the trauma severity in the head region increased. The intentional fall group with ISS <16 showed more fractures of the lower extremity, pelvis, and spine (lumbar) compared with the accidental fall group. NRF reflected these characteristics.
Definitions of the variables
The AIS provides an internationally accepted tool for ranking injury severity and is an anatomically based global severity scoring system that classifies an individual injury by body region according to its relative severity on a 6-point scale (1=minor; 6=lethal). In this study, AIS 98 was used.10 Moreover, the AIS provides the basic framework for the ISS, which is a recognized tool for assessment of overall injury severity. The ISS is the sum of the squares of the highest AIS code in each of the three most severely injured ISS body regions.10 11 The medical history of each patient was identified by the data registered in the JTDB. The presence or absence of a medical history of psychiatric disorders was recorded. However, psychiatric diseases were not descriptively recorded; therefore, data on the types of psychiatric diseases, such as depression and schizophrenia, were unavailable. Trauma-specialized hospitals were defined by the number of severe trauma patients with ISS ≥16 who were registered in the JTDB since January 1, 2004, in each participating hospital; hospitals where the number of these patients was in the top half of all participating hospitals were defined as trauma-specialized hospitals. The LOS included only the duration of the hospital stay during which the patient was registered in the JTDB. The distribution of fractures was identified by the AIS code registered for each patient. The regions in the lower extremity included the femur, patella, tibia, fibula, and foot. The NRF was calculated and ranged from 0 to 3.
Outcomes
The primary endpoint was the association between the NRF and the proportion of patients with LOS ≥30 days. The secondary endpoint was the association between the NRF and the proportion of patients with LOS ≥60 days. These were determined based on the clinical significance from the viewpoint of psychiatrists. Patients who intentionally fall from a height often have a medical history of psychiatric diseases such as mood disorders and schizophrenia. Antidepressant therapy and modified electroconvulsive therapy require approximately 1 month between their initiation and evaluation. In other words, knowing that the patient is likely to be hospitalized for more than a certain period would help in the decision-making for initiating psychiatric treatment.
Statistical analysis
Continuous variables with normal distribution are expressed as mean±SD and as median (IQR) for non-parametric variables.12 Categorical variables are presented as numbers and percentages. Continuous variables were compared using the Mann-Whitney U test, and categorical variables were compared using the χ2 test.
The association between the NRF and the LOS was assessed in a multivariable model. Considering clinical relevance and prior studies,1 3–9 the covariates for the logistic regression model were determined as follows: age, sex (male), medical history of psychiatric diseases, cerebrovascular diseases, chronic obstructive pulmonary disease, chronic heart failure (CHF), or chronic kidney dysfunction requiring hemodialysis, admission to trauma-specialized hospitals, transferred from other hospitals, systolic blood pressure (BP), respiratory rate, heart rate, body temperature, Glasgow Coma Scale (GCS) score, ISS, intubation in the ED, and admission to non-intensive care unit (ICU) wards. Vital signs were obtained from those measured in the ED. The height of falls and the condition of the ground were considered to be appropriate covariates, but these were unavailable in the JTDB. Instead, we used the ISS because it is associated with the height of falls.9 Moreover, we assessed the association between the NRF and the LOS in a Kaplan-Meier analysis.
To address the issue of missing data, we used multiple imputations by a chained equation.13 Ten complete data sets were created based on a multiple regression model that included variables with potential associations and were available in the data sets, and estimates were combined using Rubin’s rules.14 The number of imputed data sets was set above the percentage of patients with any missing data.13
We performed a predetermined sensitivity analysis to confirm the robustness of the primary analysis: a complete case analysis and a competing risk model (the Fine and Gray model).15 In the competing risk model, patients who died after admission were included because death after admission is a competing risk event for hospital discharge (figure 1). The complete case analysis included the patients who were included in the primary analysis. The covariates were the same variables as those used in the primary analysis.
Figure 1Participant enrollment flow chart. AIS, Abbreviated Injury Scale; ISS, Injury Severity Score; JTDB, Japan Trauma Databank.
All analyses were performed using SPSS V.28 (IBM, Armonk, NY) and R V.3.6.3 (R Foundation for Statistical Computing, Vienna, Austria) with the “mice” package. OR, HR, and 95% CI were calculated as appropriate. Statistical tests were two-sided, with p<0.05 set to indicate statistical significance.
