Study design and study population
We conducted a prospective population-based study using routine clinical data from digital medical records and study specific data of the study site. The study site is the largest Swiss facility for SCI care with annually between 600 and 700 hospitalizations due to SHCs. In this study we included all persons admitted within 24 hours after first contact with the clinic (emergency admissions) from 01.01.2017 to 30.06.2018 (18 months). Inclusion criteria were: persons with SCI due to traumatic or non-traumatic cause and having an acute SHC with “the need for emergency admission within 24 hours after first contact”. Excluded were persons without a diagnosis of SCI (for example: amyotrophic lateral sclerosis or cerebral palsy), persons admitted outside the defined time frame of 24 hours after first contact, or persons with direct referral to the intensive care unit.
LOS
LOS was defined as the length of the inpatient period of care, calculated from time from the day of hospitalization until the day of discharge, thus basically reflecting the number of overnight stays. For persons with admission and discharge on the same day (n = 2 persons) LOS was set at 0.5.
Demographic variables and lesion characteristics
Person-characteristics, including age (at admittance), sex and TSI, were categorized according to the recommendations of the International Spinal Cord Injury Core Data Sets (14), while collapsing adjacent groups in case of sparse data. Age groups for analysis thus included persons aged 15–29, 30–44, 45–59, and 60 years or older. TSI groups included less than 6, 6–15, 16–25, and 26 years or more. Level and severity of SCI was reported according to the International Standards for Neurological Classification of Spinal Cord Injury (ISNCSCI) (13). SCI etiology was classified as traumatic and non-traumatic SCI, according to the definition of the International Spinal Cord Society (ISCoS) Core Data Set (13). Discharge destination was classified as private home, nursing home, death, or unspecified.
Diagnosis of the SHC
We defined, according to the definition of Jensen, Molton (5), the following groups of SHC as admission causes: 1. pressure ulcers, 2. UTI, 3. respiratory infections, 4. fractures, 5. other urological causes, 6. autonomic dysregulation (AD), 7. other infections and 8. other acute health conditions (for example bowel problems, baclofen pump defect).
Comorbidities
For our study we chose the most relevant comorbidities, defined as a health condition that is likely to be associated with worse health outcome (especially prolonged LOS) (15). For this study, following comorbidities were assessed: 1. Diabetes mellitus (American Diabetes Association (ADA) Classification 1–4) (16), 2. Osteoporosis (T-Value < 2.5) (17), 3. Arterial hypertension (American Society of Hypertension (ASH) Stage 1–3) (18), 4. Renal failure (KDIGO - Kidney Disease Improving Global Outcomes) (19), 5. Heart failure (definition according to the Heart Failure Society of America) (20), 6. Deep vein thrombosis or pulmonary embolism, 7. Vascular disease (peripheral artery disease), 8. Current treatment for cancer, and 9. Moderate and severe sleep apnea (Apnoe-Hypopnoe-Index > 15/h) (5).
Complications during hospital stay
We comprehensively assessed complications during hospital stay that are likely to influence LOS, including sepsis, UTIs, respiratory infections, severe pain, severe spasticity, autonomic dysregulation, and pressure ulcers. In statistical analysis complications were used as a binary variable, with 0 indicating "not occurring" and 1 indicating "occurring".
Statistical Method
All statistical analyses were performed using the Stata statistical software (Stata/SE version 16.1 for Windows; Stata Corp, College Station, TX, USA).
Descriptive analyses include crude numbers and percentages to depict the population and to evaluate differences in LOS across classes of demographic and lesion characteristic variables. Variation in LOS will be reported using the mean and standard deviation (SD) as well as median and interquartile rage (IQR). Univariable and multivariable linear regression analysis was used to identify main determinants of LOS (days) among predictor variables, including sex, age class, lesion level, completeness of SCI, SCI etiology, cause of admission, medical complications, and pre-existing comorbidities. Prior to analysis, the variable LOS was log-transformed (lnLOS) to achieve normal distribution as confirmed using a Kolmogorov-Smirnov's test. Marginal predictions for LOS from the multivariable model were derived using exponentiation as to back-transform estimates and respective 95% CI to the original scale (days).