This was a single-centre cohort study conducted in the Department of Neurosurgery of Bouaké University Hospital, Côte d’Ivoire. It combined a retrospective phase, from 1 January 2017 to 31 December 2021, and a prospective phase, from 1 January 2025 to 30 June 2026, reported according to the STROBE recommendations [16]. The 2022-2024 period was not covered by the database built for this study: this period corresponded to the transition to a new department and to a new patient-record archiving system, resulting in disorganisation and loss of records for that period. The participant flow diagram is presented in Figure 1.

Figure 1. Participant flow diagram (STROBE): records screened and excluded, stratification of the cohort by lesion completeness at admission, mortality and numbers analysed for functional outcome.

All patients hospitalised in the department for an acute traumatic spinal injury during the two study periods were eligible. A traumatic spinal injury was retained when a history of recent trauma was associated with a spinal column injury (fracture, dislocation or disco-ligamentous injury) and/or signs of spinal cord involvement on imaging—plain radiographs, computed tomography (CT) and, when available, magnetic resonance imaging (MRI)—whatever the neurological status at admission. Exclusion criteria were non-traumatic spinal conditions (disc herniation, myelomeningocele, tumours and metastases, infectious spondylodiscitis), duplicate records, and inconclusive records, defined as records in which the traumatic origin of the lesion or the admission neurological grade could not be established from the available documentation. Of the 383 records screened, 17 were excluded—9 for a non-traumatic condition, 5 duplicates and 3 inconclusive records—leaving 366 patients for analysis (Figure 1). No patient was excluded because of missing data on the analysed variables.

The main exposure of this study was complete injury, defined by an ASIA grade A at admission (n = 147). The comparator group, referred to as non-complete injuries (n = 219), comprised 176 patients with an incomplete spinal cord injury (ASIA B, n = 51; C, n = 60; D, n = 65) and 43 patients (11.7%) admitted for a traumatic spinal column injury without neurological deficit (ASIA E). ASIA E patients were retained because the study population was defined by admission for traumatic spinal injury to a neurosurgical unit, and not by the presence of a deficit: the ASIA A versus B - E contrast corresponds to the question raised at admission across the whole case-mix of the unit. Because ASIA E patients do not have a spinal cord injury in the strict sense and could dilute the comparator, the comparisons of in-hospital mortality and complications were repeated with ASIA B - D as the comparator (sensitivity analysis), and ASIA E patients were excluded from the analysis of neurological recovery, as they cannot gain an ASIA grade.

The neurological grade was established according to the ASIA/ISNCSCI scale [5]. The first ISNCSCI examination was performed by the neurosurgical team on admission to the department, i.e., a median of 2 days (interquartile range 1 - 3) after the trauma, with most patients having been referred from another facility. Grade A defines complete injury: the absence of motor and sensory function in the sacral segments S4 - S5. Grades B to D define incomplete injuries of decreasing severity, and grade E the absence of deficit. The ASIA grade recorded at admission was considered the baseline neurological grade; ASIA A therefore denotes the neurological status at admission, not definitive completeness. Neurological reassessment was not systematically performed at a predefined time after the resolution of spinal shock, and the last documented grade was used for the analysis of neurological recovery. Spinal shock was recorded when the admitting clinician documented a flaccid, areflexic paralysis below the level of injury; the bulbocavernosus reflex and the timing of its return were not systematically recorded. Bony lesions were classified on admission imaging (CT, or plain radiographs when CT was unavailable) according to the AOSpine subaxial cervical classification for subaxial cervical injuries (C3 - C7) [17] and the AOSpine thoracolumbar classification for thoracic, lumbar and sacral injuries [18]; in both systems, type C denotes a translational injury (displacement or dislocation in any plane). Upper cervical (C1 - C2) injuries were included in the cervical level for topographical analyses but were not assigned an AOSpine subaxial morphological type, as the subaxial classification applies to C3 - C7 injuries. For multi-level injuries, the most severe classifiable morphological lesion was retained. Classification was performed by the treating neurosurgical team as part of routine care and was therefore not blinded to the neurological examination. Dislocation, recorded separately, was defined as a partial or complete loss of the normal anatomical relationships between two adjacent vertebrae on imaging, with abnormal translation or rotation, with or without facet disengagement; a partial or rotational dislocation not reaching the characterised translation of type C was not classified as type C. Spinal shock, respiratory distress, haemodynamic instability, spinal topography and the elementary lesions on imaging—spinal cord compression and contusion, dislocation, instability—were documented at admission. The outcomes were hierarchised: the primary outcome was in-hospital mortality; the secondary outcomes were in-hospital complications (respiratory, neurological, cutaneous, urinary, thromboembolic), neurological recovery (gain of at least one ASIA grade between admission and the last documented assessment), functional walking and sphincter recovery, the latter two assessed among survivors.

The study was conducted in accordance with the Declaration of Helsinki. Bouaké University Hospital had no constituted institutional ethics committee during the retrospective period; data collection was initially authorised by the Department of Neurosurgery and the hospital management, and the study protocol, covering both components, was subsequently approved by the Institutional Ethics Committee of Bouaké University Hospital (approval no. CEI/CHU-BKE/129/2026). For the retrospective component (2017-2021), data were extracted from routine medical records and anonymised before analysis, and individual consent was waived with the approval of the committee. For the prospective component (January 2025-June 2026), informed consent to the use of clinical data was obtained from each patient or, when the patient was unable to consent, from a relative.

Qualitative variables are expressed as counts and percentages, quantitative variables as median (interquartile range). Comparisons between complete and non-complete injuries used the χ² test, Fisher’s exact test or the Mann-Whitney test, according to the nature of the variables and the expected counts. Two multivariable logistic regression models were built: the first identified the factors independently associated with complete injury (dependent variable: ASIA A), the second assessed the factors associated with in-hospital mortality, including complete injury as the main exposure. Variables were selected by combining their clinical relevance, the data from the literature and their association in bivariate analysis (p < 0.20), while respecting a sufficient number of events per variable (at least ten). The database built for this study contained no missing data on the model variables, which allowed a complete-case analysis. The absence of important collinearity was verified by calculating the variance inflation factor (VIF), all below 1.2. Discrimination was assessed by the area under the ROC curve (AUC) and calibration by the Hosmer-Lemeshow test. The crude and adjusted odds ratios (OR) are presented with their 95% confidence interval (95% CI). In a sensitivity analysis, the comparisons of in-hospital mortality and complications were repeated after exclusion of the 43 ASIA E patients (comparator ASIA B - D, n = 176). Because complete injury and several outcomes were frequent, odds ratios were not interpreted as relative risks. The significance threshold was set at 0.05. Analyses were performed with Python 3.12 (pandas, statsmodels, scikit-learn). As this is an observational study, the reported associations should be interpreted as prognostic and not as causal relationships.

In the model of the admission characteristics associated with complete injury, selection favoured variables representing distinct dimensions and clinically antecedent to completeness: bony morphology (type C per the AOSpine classification), acute physiological severity (spinal shock, respiratory distress), associated traumatic burden (polytrauma), topography (cervical level), the access pathway (non-medicalised transport) and age. Several variables significant in bivariate analysis were deliberately not retained. Sphincter disorders partly overlap with the very definition of grade A, based on the abolition of sacral S4 - S5 function, and their inclusion would have introduced circularity with the modelled outcome. Spinal cord compression and contusion are radiological expressions of the cord injury of which completeness constitutes the clinical endpoint: located on the lesional pathway rather than upstream, they would have conditioned the model on intermediate variables. Dislocation, a morphological marker largely captured by the AOSpine type C category (translation injuries), was set aside for this same redundancy. Finally, the thoracic level constitutes, within a single topographical variable, the complement of the cervical level: the two could not be introduced jointly without collinearity, and the cervical level was retained as the reference contrast, because of its prognostic relevance and its association with mortality and respiratory morbidity.

Of the 366 patients in the cohort, 147 (40.2%) had a complete spinal cord injury (ASIA A). This was the most frequent ASIA grade at admission, followed by grades D (65; 17.8%), C (60; 16.4%), B (51; 13.9%) and E (43; 11.7%) (Figure 2(a)). These patients were predominantly male (81.6%), with a median age of 35 years (28 - 49), with no significant difference in age or sex from patients with a non-complete injury. Thoracic involvement was as frequent as cervical involvement within complete injuries (42.9% and 41.5% respectively), whereas cervical involvement clearly predominated among non-complete injuries (55.7%; p = 0.008 for cervical, p = 0.003 for thoracic; Figure 2(b)). Traumatic circumstances and the admission pathway were broadly comparable between the two groups (Table 1).

Figure 2. Burden of complete injuries. (a) Distribution of ASIA grades at admission; (b) Distribution of spinal segments by lesion completeness (complete, ASIA A; non-complete, ASIA B - E).

Table 1. Characteristics at admission by lesion completeness.

IQR: interquartile range; RTA: road traffic accident; AOSpine: AOSpine classification.

In bivariate analysis, complete injury was significantly associated with spinal shock (29.9% vs 10.0%; p < 0.001), type C bony lesion per the AOSpine classification (51.0% vs 38.4%; p = 0.017), dislocation (64.6% vs 52.5%; p = 0.022), spinal cord compression (82.3% vs 65.3%; p < 0.001), spinal cord contusion (61.2% vs 50.2%; p = 0.038) and sphincter disorders (84.4% vs 47.0%; p < 0.001) (Table 1). Cervical involvement was, on the contrary, less frequent among complete injuries (41.5% vs 55.7%; p = 0.008), as was non-medicalised transport (47.6% vs 63.0%; p = 0.004).

In multivariable analysis (VIF all below 1.2), the admission characteristics independently associated with complete injury were spinal shock (adjusted OR: 3.46; 95% CI: 1.90 - 6.31) and type C bony lesion per the AOSpine classification (adjusted OR: 1.66; 95% CI: 1.06 - 2.61), whereas non-medicalised transport was inversely associated with it (adjusted OR: 0.55; 95% CI: 0.34 - 0.88) (Table 2; Figure 3). An adjusted OR of 3.46 means that, all else being equal, the odds of a complete injury were about three and a half times higher in the presence of spinal shock; because complete injury affected 40.2% of the cohort, this odds ratio overstates the corresponding ratio of probabilities and should not be read as a relative risk. Age, introduced as a continuous variable per 10-year band, was not significantly associated with lesion completeness (adjusted OR: 1.02; 95% CI: 0.92 - 1.13).

In-hospital mortality was very strongly linked to completeness: 35.4% of ASIA A patients died, versus 5.5% of ASIA B-E patients (crude OR: 9.44; p < 0.001) (Figure 4). When the 43 ASIA E patients, 3 of whom died, were excluded from the comparator, mortality was 5.1% (9/176) in ASIA B - D patients and the association was unchanged (crude OR: 10.16; 95% CI: 4.79 - 21.53; p < 0.001). In multivariable analysis, complete injury remained the factor most strongly associated with

Figure 3. Admission characteristics independently associated with complete injury (adjusted odds ratios; the vertical dashed line corresponds to OR = 1).

Table 2. Factors independently associated with complete injury (ASIA A).

n = 366; events (ASIA A) = 147; 7 terms in the model; EPV = 21; VIF all < 1.2. OR: odds ratio; 95% CI: 95% confidence interval.

Figure 4. Vital and functional prognosis and complications by lesion completeness. Mortality and complications are calculated among all patients; the functional indicators (recovery ≥ 1 ASIA grade, functional walking) are calculated among survivors, with ASIA B-D survivors as comparator for recovery ≥ 1 ASIA grade.

mortality (adjusted OR: 10.5; 95% CI: 5.2 - 21.0) after adjustment for cervical level, respiratory distress, polytrauma and age, with good discrimination (AUC: 0.79) and no major lack of calibration detected (Hosmer-Lemeshow p = 0.54) (Table 3). Within complete injuries themselves, mortality was highest for cervical involvement (41.0%; 95% CI: 29.5 - 53.5), followed by thoracic (34.9%; 95% CI: 24.3 - 47.2) and lumbar involvement (22.2%; 95% CI: 9.0 - 45.2). Sacral or multi-level injuries remained few (20.0%; 1/5), which underlines the aggravating role of a high lesional level. No significant interaction was found between completeness and cervical level (p = 0.52), suggesting that the effect of complete injury on mortality was homogeneous regardless of level.

Table 3. Vital and functional prognosis and complications by lesion completeness.

*Adjusted OR for cervical level, respiratory distress, polytrauma and age (AUC: 0.79; Hosmer-Lemeshow p = 0.54). Survivors: ASIA A: n = 95; ASIA B - E: n = 207. Recovery of at least one ASIA grade excludes the 40 ASIA E survivors (comparator ASIA B - D: n = 167). Sensitivity analysis (comparator ASIA B - D: n = 176): in-hospital mortality 9 (5.1%), crude OR: 10.16 (95% CI: 4.79 - 21.53), p < 0.001; any complication 98 (55.7%), crude OR: 1.99 (1.25 - 3.17), p = 0.005; respiratory complication 27 (15.3%), crude OR: 2.59 (1.52 - 4.44), p < 0.001.

Patients with a complete injury accumulated a heavier in-hospital morbidity: at least one complication was documented in 71.4% of them, versus 51.1% of non-complete injuries (crude OR: 2.39; p < 0.001). This difference was mainly driven by respiratory (32.0% vs 13.7%; crude OR: 2.96; p < 0.001) and thromboembolic complications (6.8% vs 1.8%; crude OR: 3.92; p = 0.023). Neurological, cutaneous (pressure ulcers) and urinary complications were numerically more frequent in ASIA A patients, without reaching the significance threshold (Table 3).

After exclusion of ASIA E patients (comparator ASIA B - D, n = 176), at least one complication was documented in 98 patients (55.7%) with ASIA B - D injuries (crude OR: 1.99; 95% CI: 1.25 - 3.17; p = 0.005), and a respiratory complication in 27 patients (15.3%) (crude OR: 2.59; 95% CI: 1.52 - 4.44; p < 0.001); the differences with ASIA A patients therefore persisted.

Among the 95 survivors of a complete injury, neurological recovery remained limited. A conversion to a less severe grade, a gain of at least one ASIA grade, was observed in 31.6% of them, versus 53.9% of survivors of an incomplete injury of ASIA grade B - D—survivors with ASIA E, unable to gain, being excluded from this analysis (crude OR: 0.39; 95% CI: 0.23 - 0.67; p < 0.001). This conversion was most often by a single grade (23 A → B conversions, 4 A → C, 3 A → D), with conversion to the functional grade E remaining nil (Figure 5). Since grade D is based on a threshold of manually tested motor strength and not on a walking criterion, its conversion does not imply resumption of walking: indeed, no ASIA A survivor, including among the 3 patients who progressed to grade D, regained functional walking, whether independent or assisted, versus 57.0% of survivors of a non-complete injury (p < 0.001); all ASIA A survivors remained without resumption of walking, in a wheelchair, with difficult standing or an absence of recovery. Sphincter recovery, in contrast, was not significantly different between the two groups (40.0% vs 36.2%; p = 0.53). These results illustrate a clear dissociation between neurological recovery—a gain of one ASIA grade may be observed—and functional recovery, resumption of walking requiring a level of motor recovery that was not reached by any survivor of a complete injury.

The proportion of complete injuries (40.2%) is high compared with high-income country series, often below 30%, and close to African cohorts (36.5% in Ethiopia; Table 4). A notable finding is that thoracic involvement was as frequent as cervical involvement among complete injuries, whereas cervical involvement dominated non-complete injuries. The relative narrowness of the thoracic canal and its watershed vascularisation expose the thoracic cord to complete injuries, whereas cervical injuries are distributed across the whole spectrum of severity. Spinal shock and type C lesion per the AOSpine classification, independently associated with completeness, are consistent with a high energy transferred to the cord. Spinal shock, however, is also a transient clinical state that can mimic a complete injury, so that part of this association may reflect grading during shock rather than definitive completeness (see Strengths and Limitations). Mechanistically, the primary injury is followed by a cascade of secondary lesions—ischaemia, oedema, inflammation, programmed cell death—which extends tissue destruction and provides the rationale for strategies limiting these lesions. Since cervical level was not significantly associated with completeness after adjustment, neurological severity and lesional level appear as two distinct dimensions. The inverse association with non-medicalised transport probably reflects the triage of the most severe presentations towards medicalised transport, and not a protective effect.

Table 4. Comparison with sub-Saharan African spinal cord injury cohorts.

NR: not reported; hosp.: in-hospital mortality; level of evidence per OCEBM (cohort studies). Definitions of mortality (in-hospital, 30-day) and follow-up windows differ, which limits direct comparisons.

Complete injury was the main factor associated with in-hospital mortality (adjusted OR: 10.5; AUC: 0.79), with a mortality of 35.4% versus 5.5%, consistent with the Ethiopian cohort, where complete injury was strongly associated with mortality (adjusted OR: 6.5) [19]. This result reflects the expected pathophysiological role: a severe anatomical and functional interruption of the cord exposes to dysautonomia, haemodynamic instability and respiratory failure, particularly in the case of cervical involvement, where mortality reached 41.0% within complete injuries. This association was unchanged when patients without neurological deficit were excluded from the comparator (crude OR: 10.2 versus ASIA B - D), and is therefore not an artefact of the inclusion of ASIA E patients. The respiratory excess mortality—complications more than twice as frequent (32.0% vs 13.7%)—probably constitutes one of the main mechanisms and underlines the role of critical care, as decisive as the surgical procedure in the in-hospital outcome.

The functional outcome of survivors was very poor: conversion to a less severe grade remained limited (31.6%), most often by a single grade, and no survivor regained functional walking. This result highlights an essential distinction: a measurable neurological improvement, for example, an ASIA A to B conversion reflecting partial sensory recovery, does not equate to a functional improvement. Recovery of walking integrates residual motor strength, postural control, balance and endurance. It is a patient-centred outcome, which conditions autonomy, return home and social participation. Its absence in all ASIA A survivors illustrates a lasting disability, beyond neurological impairment alone, and imposes realistic rehabilitation objectives: wheelchair independence, prevention of complications, bladder-sphincter management. Our results thus distinguish barely modifiable factors—initial severity, level, AOSpine—from potentially modifiable factors—delays, intensive care, prevention of complications, rehabilitation. The latter constitute windows of intervention, even in the most severe patients.

Our results are consistent with the large international and African cohorts, in which initial neurological severity remains the main prognostic factor for mortality and recovery [19][21][22]. The frequency of complete injuries and the mortality observed lie within the range reported in sub-Saharan Africa [10] and underline the specificities of resource-limited systems. The differences from high-income countries probably relate as much to the organisation of care—delays, transport, intensive care, rehabilitation—as to the universal pathophysiology of spinal cord trauma. By documenting this subgroup in an Ivorian referral centre, this study reduces the under-representation of West African data and fits within the objectives of global neurosurgery [23][24].

What the study shows: The high frequency of complete injuries, their independent association with spinal shock and type C lesion per the AOSpine classification (and an inverse association with non-medicalised transport), marked in-hospital excess mortality and respiratory morbidity, and the absence of functional walking in survivors.

What itsuggests: The value of early prognostic stratification based on ASIA assessment and of optimised organisation of care.

What it does not demonstrate: Neither a causal effect of the factors studied, nor the comparative effectiveness of surgical or rehabilitation strategies [25][26], nor the external validity of the associations, which require prospective multicentre confirmation.

Figure 6 summarises this conceptual framework: anatomical severity conditions initial neurological severity, which determines complications, mortality and, among survivors, neurological then functional recovery. This is an observed prognostic trajectory, and not a formal therapeutic recommendation.

Figure 6. Conceptual framework of the prognostic trajectory after traumatic spinal cord injury, illustrated by the cohort data for complete injuries (ASIA A).

The main strengths are the size of the cohort, the internal quality control that led to a database without missing data on the model variables, the use of standardised classifications (ASIA grade, AOSpine classification), the joint analysis of burden, associated admission characteristics and prognosis, and multivariable analyses with verification of collinearity and calibration. Several limitations must be taken into account. The single-centre nature and recruitment in a referral centre expose to a selection bias liable to over-represent severe forms. The frequency of complete injuries should therefore not be interpreted as a national incidence. The partly retrospective component exposes to an information bias and to residual confounding: comorbidities, nutritional status, socio-economic conditions, rehabilitation adherence. The prognostic model was not externally validated, and the absence of reliable dates prevented any temporal survival analysis. The ASIA grade was assessed at admission and at a median of 2 days after the trauma. Because neurological reassessment was not systematically performed at a predefined time after resolution of spinal shock, some patients initially classified as ASIA A may have been transiently misclassified during spinal shock; this would inflate the frequency of complete injuries and bias the association between spinal shock and completeness away from the null. Consequently, part of the observed A-to-B/C/D conversions may represent resolution of spinal shock rather than true neurological recovery. The AOSpine classification was performed by clinicians aware of the neurological status, which may have favoured a concordant classification of the most severe cases. The comparator group included 43 patients without neurological deficit; the sensitivity analysis restricted to ASIA B - D did not modify the mortality or complication results. Finally, MRI was not systematic, which prevented the analysis of intramedullary contusion or haemorrhage. Functional walking was based on a clinical categorisation, not on a validated scale (SCIM, WISCI). Data on rehabilitation and long-term quality of life were not available.

In the short term, the priorities concern the improvement of care pathways, the standardisation of ASIA assessment and the strengthening of complication prevention. In the medium and long term, a prospective multicentre West African registry, based on common definitions, would make it possible to validate the associations identified, to integrate standardised and patient-reported functional outcomes—quality of life, participation, return to work—to assess the impact of delays and rehabilitation, and to develop and validate prognostic models adapted to African populations [22], in line with the updated international recommendations on the management of acute SCI [27]. A specific model of functional walking constitutes the most natural extension of this work.