Neurocrit Care

https://doi.org/10.1007/s12028-023-01845-8

NEUROPROGNOSTICATION

Guidelines for neuroprognostication

in adults with traumatic spinal cord injury
Dea Mahanes1, Susanne Muehlschlegel2, Katja E. Wartenberg3, Venkatakrishna Rajajee4, Sheila A. Alexander5,
Katharina M. Busl6, Claire J. Creutzfeldt7, Gabriel V. Fontaine8, Sara E. Hocker9, David Y. Hwang10, Keri S. Kim11,
Dominik Madzar12, Shraddha Mainali13, Juergen Meixensberger14, Panayiotis N. Varelas15, Christian Weimar16,17,
Thomas Westermaier18 and Oliver W. Sakowitz19*

© 2023 The Author(s)

Abstract
Background: Traumatic spinal cord injury (tSCI) impacts patients and their families acutely and often for the long
term. The ability of clinicians to share prognostic information about mortality and functional outcomes allows patients
and their surrogates to engage in decision-making and plan for the future. These guidelines provide recommenda-
tions on the reliability of acute-phase clinical predictors to inform neuroprognostication and guide clinicians in coun-
seling adult patients with tSCI or their surrogates.
Methods: A narrative systematic review was completed using Grading of Recommendations Assessment, Develop-
ment, and Evaluation methodology. Candidate predictors, including clinical variables and prediction models, were
selected based on clinical relevance and presence of an appropriate body of evidence. The Population/Intervention/
Comparator/Outcome/Timing/Setting question was framed as “When counseling patients or surrogates of critically ill
patients with traumatic spinal cord injury, should < predictor, with time of assessment if appropriate > be considered a
reliable predictor of < outcome, with time frame of assessment >?” Additional full-text screening criteria were used to
exclude small and lower quality studies. Following construction of an evidence profile and summary of findings, rec-
ommendations were based on four Grading of Recommendations Assessment, Development, and Evaluation criteria:
quality of evidence, balance of desirable and undesirable consequences, values and preferences, and resource use.
Good practice recommendations addressed essential principles of neuroprognostication that could not be framed in
the Population/Intervention/Comparator/Outcome/Timing/Setting format. Throughout the guideline development
process, an individual living with tSCI provided perspective on patient-centered priorities.
Results: Six candidate clinical variables and one prediction model were selected. Out of 11,132 articles screened,
369 met inclusion criteria for full-text review and 35 articles met eligibility criteria to guide recommendations. We
recommend pathologic findings on magnetic resonance imaging, neurological level of injury, and severity of injury as
moderately reliable predictors of American Spinal Cord Injury Impairment Scale improvement and the Dutch Clinical
Prediction Rule as a moderately reliable prediction model of independent ambulation at 1 year after injury. No other
reliable or moderately reliable predictors of mortality or functional outcome were identified. Good practice recom-
mendations include considering the complete clinical condition as opposed to a single variable and communicating
the challenges of likely functional deficits as well as potential for improvement and for long-term quality of life with
SCI-related deficits to patients and surrogates.

*Correspondence: oliver.sakowitz@uni-heidelberg.de
19
Department of Neurosurgery, Neurosurgery Center Ludwigsburg-
Heilbronn, Ludwigsburg, Germany
Full list of author information is available at the end of the article
 Conclusions: These guidelines provide recommendations about the reliability of acute-phase predictors of mor-
tality, functional outcome, American Spinal Injury Association Impairment Scale grade conversion, and recovery of
independent ambulation for consideration when counseling patients with tSCI or their surrogates and suggest broad
principles of neuroprognostication in this context.
Keywords: Spinal cord injuries, Trauma, Prognosis, Functional status, Outcome

Introduction [7]. The North American Clinical Trials Network, a net-

According to the 2016 Global Burden of Disease study, work of ten US and Canadian centers, was established
the age-standardized incidence rate of traumatic spinal in 2005 to promote advancement in tSCI management
cord injury (tSCI) is estimated to be 13 cases per 100,000 through research. Now affiliated with the Christopher
people, or a total of 930,000 new cases worldwide. The and Dana Reeve Foundation, the North American Clini-
global prevalence of tSCI in 2016 was approximately cal Trials Network maintains an SCI registry focused on
27 million [1]. Average age at the time of injury has the natural history of recovery in the context of current
increased, primarily due to an increase in injuries among treatments, information that can then be used for com-
individuals aged 65 and older [2]. Deficits vary based on parison in future interventional studies [8]. In Canada,
level of injury to the spinal cord. Based on data from the more than 30 centers contribute information to the Rick
United States, incomplete tetraplegia is most common, Hansen Spinal Cord Injury Registry (RHSCIR) using a
impacting approximately 47% of patients, followed by prospectively established standardized data set. Although
complete and incomplete paraplegia [3]. The morbidity initially focused on tSCI, the RHSCIR has expanded to
associated with tSCI is substantial and out of proportion include nontraumatic SCI. Similar to many other con-
to the incidence. Although traumatic brain injury (TBI) sortiums or networks, centers that belong to the RHSCIR
occurs approximately 30 times more often and the preva- commit to the implementation of best practices for SCI
lence of patients with TBI sequelae is twice as high, tSCI care [9]. Outside of North America, the European Mul-
is associated with more years lived with disability [1, 4]. ticenter Study about Spinal Cord Injury (EMSCI) was
Prognostication following tSCI focuses primarily on founded in 2004 as a collaborative effort between sev-
functional outcomes such as ambulation and the ability eral leading spinal cord injury centers across Europe.
to perform activities of daily living. Short-term and long- The initial aim of the organization was to establish a
term survival is also a consideration, especially for older, comprehensive database of clinical and demographic
medically complex patients or patients with respiratory information on patients with SCI to improve the quality
compromise. The ability to provide prognostic informa- of care for these individuals. Since its inception, EMSCI
tion may help patients and family with both short-term has grown to include more than 20 member centers in
decision-making and long-term care planning. Accurate countries including Austria, Belgium, Denmark, France,
prognostication supports realistic goals and enables cli- Germany, Italy, the Netherlands, Norway, Spain, Sweden,
nicians to better focus rehabilitation on either return of Switzerland, and the United Kingdom. The organization
function or adaptation. In 1999, the Consortium for Spi- has also expanded its research focus beyond data collec-
nal Cord Medicine published a comprehensive practice tion to include the design and implementation of clinical
guideline focused on outcomes after spinal cord injury, trials and observational studies aimed at improving the
with support from the Paralyzed Veterans of America [5]. treatment and outcomes of tSCI. Over the years, EMSCI
Much of the research into outcomes after tSCI is drawn has been instrumental in promoting collaboration and
from observational data submitted to large registries knowledge sharing among researchers and health care
or from centralized consortiums and networks. In the professionals across Europe and beyond [10].
United States, the Spinal Cord Injury Model Systems
(SCIMS) program includes centers that provide com-
prehensive care to individuals with SCI across the con- Scope, Purpose, and Target Audience
tinuum of treatment [6]. The SCIMS program began in The scope of these Grading of Recommendations Assess-
1970 with a single center [6] and currently includes 14 ment, Development and Evaluation (GRADE) guidelines
institutions [7]. In addition to providing services to peo- is the prognostication of neurological outcome in criti-
ple with SCI and participating in research, each SCIMS cally ill adult patients with tSCI. The purpose of these
center contributes data to a central database. The SCIMS guidelines is to provide evidence-based recommenda-
database is managed through the National Spinal Cord tions on the reliability of predictors of neurological
Injury Statistical Center and is available to researchers outcome in critically ill adult patients with tSCI, to aid
clinicians in formulating a prognosis. The target audience purposes of these guidelines and cannot be formally rec-
consists of clinicians responsible for such counseling. ommended for prognostication on their own. Variables
deemed not reliable, however, may be a component of
How to Use These Guidelines reliable or moderately reliable prediction models.
These guidelines provide recommendations on the reli-
ability of select demographic and clinical variables as well Methods
as prediction models when counseling patients, families, An in-depth description of the methodology used
and surrogates of individuals with tSCI. We categorized in these guidelines is available in Supplementary
these predictors as reliable, moderately reliable, or not Appendix 1.
reliable. We based this categorization on a GRADE-based
assessment of certainty in the body of evidence, as well Selection of Guideline Questions
as effect size (quantification of predictor accuracy) across Candidate predictors were selected based on clinical rel-
published studies, as detailed in Supplementary Appen- evance and the presence of an appropriate body of litera-
dix 1 and Table 1. Reliable predictors and prediction ture. Candidate predictors and prediction models were
models, for the purposes of these guidelines, may be used considered “clinically relevant” if the predictor or com-
to formulate a prognosis when the appropriate clinical ponents of the model were (1) accessible to clinicians,
context is present in the absence of potential confound- although universal availability was not required, and (2)
ers. These are predictors with clear, actionable thresh- likely to be considered by clinicians when formulating a
olds or clinical/radiographic definitions and a low rate of neurological prognosis for critically ill adult patients with
error in prediction of outcomes, with at least moderate tSCI. An appropriate body of literature was considered
certainty in the body of evidence. When prognosis is for- present for any clinical variable that was evaluated in at
mulated on the basis of one or more reliable predictors, least two studies that included a minimum of 50 study
the clinician may describe the outcome as “very likely” participants and established as an independent predic-
during counseling. Given the inherent limitations in neu- tor in a multivariate analysis that incorporated age and
roprognostication research, the clinician must neverthe- measures of injury severity (level and/or completeness
less acknowledge the presence of uncertainty—even if of injury). For clinical prediction models, an appropri-
low—in the prognosis during counseling. Moderately ate body of literature was considered present for mod-
reliable individual predictors may be used for prognos- els with at least one external validation study of at least
tication only when additional reliable or moderately reli- 50 patients in addition to the initial report on develop-
able individual predictors are present, in addition to the ment of the model (also with a minimum of 50 patients).
appropriate clinical context. These are also predictors Because of the relatively low incidence of tSCI, studies of
with clear, actionable thresholds or clinical/radiographic at least 50 study participants were included to broaden
definitions and a low rate of error in prediction of out- the available literature. The panel’s goal was to delineate
comes, but with lower certainty in the body of evidence, factors for prognosis of the natural course following tSCI
frequently as a result of smaller studies that result in based on acute-phase assessments in the critical care
imprecision or other risk of bias, often rooted in meth- environment. Treatment of patients with trauma is heter-
odology. When the prognosis is formulated on the basis ogeneous, but the impact of therapeutic interventions on
of multiple moderately reliable predictors, the clinician outcome was outside the scope of these prognostication
may describe the outcome as “likely” during counseling guidelines. Based on these criteria, the following candi-
but must acknowledge “substantial” uncertainty in the date predictors were selected.
prognosis. Moderately reliable clinical prediction models
that generate predicted probabilities of outcomes, in con- Clinical Variables and Description
trast, may be used for prognostication during counseling 1. Age at time of injury. Age at the time of injury may be
of patients, families, and surrogates of individuals with used as a continuous variable or, more often, divided
tSCI in the absence of other reliable or moderately reli- into categories. These categories vary widely across
able predictors. However, it is recommended that the cli- studies and may be impacted by changes in the epi-
nician describe the predicted probability of the outcome demiology of SCI, population characteristics, and
as “an objective estimate only, subject to considerable evolving societal perspectives on aging. Although age
uncertainty.” Although the panelists recognize that those is often dichotomized at 65 years, other thresholds,
predictors that do not meet the criteria to be described as such as 60 or even 50 years, have been suggested.
reliable or moderately reliable are often used by clinicians 2. Comorbidities. Comorbid conditions such as cardiac,
in formulating their subjective impressions of prognosis, pulmonary, or renal disease may be present at the
they have nevertheless been deemed not reliable for the time of tSCI. These conditions may complicate man-
Table 1 Predictor Characterization and Use—Reliable and Moderately Reliable Predictors
Category GRADE criteria Point Use dur- Presence Suggested language
of predictor/ estimates ing counseling of additional during counseling of patients or
model of accuracy of patients or specific surrogates
in the body surrogates? reliable or
Risk of Bias Inconsistency Imprecision Indirectness Quality of of evidence moderately Likelihood Disclaimer
Evidence- predictors of outcome of Uncertainty
Overall required during coun-
for use dur- seling
ing counseling

Reliable One Downgrade Downgrade Downgrade Moderate or High. Predic- Yes Preferred, but “Very likely” or Present, but low
downgrade NOT NOT NOT High tion models not absolutely for clinical
permitted permitted permitted permitted require required prediction
AUC > 0.8, models use
no evidence predicted
of miscali- probability of
bration in outcome
external
validation
studies
Moderately One Downgrade One One Any High Yes Yes “Likely” Substantial
reliable downgrade NOT downgrade downgrade
individual permitted permitted permitted permitted
predictors
Moderately One Downgrade One One Any High. Yes No Use predicted “The predicted
reliable downgrade NOT downgrade downgrade Prediction probability of probability is an
clinical permitted permitted permitted permitted models outcome objective esti-
prediction require mate, subject
models AUC > 0.7, to considerable
some uncertainty”
miscalibra-
tion allowed
in external
validation
studies
Not reliable Downgrade Downgrade Downgrade Downgrade Any Any Noa Not applicable Not applicable Not applicable
permitted permitted permitted permitted

AUC​, area under the curve, GRADE, Grading of recommendations assessment, development, and evaluation
a
Many predictors designated “not reliable” are practically used by clinicians in formulating and communicating real-world subjective impressions of prognosis. The purpose of these guidelines is to identify predictors, if
any, that meet reliable or moderately reliable criteria
 agement and potentially impact mortality. Comor- with complete injuries may show signs of partial sen-
bidities at the time of acute hospitalization are often sory or motor innervation below the NLI, described
reported in large-scale registries or can be extracted as zones of partial preservation (ZPP). It has been
from the electronic health record. Common meth- suggested that occurrence and segmental width of
ods for measuring this predictor include use of the ZPPs may predict future functional improvements.
Charlson Comorbidity Index (CCI) (Supplementary Timing of assessment is of critical importance for
Table 1) or a count of the number of comorbid condi- this predictor, but significant variability exists in
tions reported. the literature. For the purposes of prognostication,
3. Concomitant injury. The presence of concomitant determination of NLI should be deferred for 72 h.
brain injury or other multisystem trauma may impact The initial examination may be impacted by spi-
outcomes. Concomitant TBI is typically classified by nal shock and by other factors such as concomitant
its presence, the Glasgow Coma Scale (GCS) score injuries, medications, and ongoing resuscitation. An
following initial resuscitation, or by the three tra- examination performed at 72 h after injury is likely
ditional gradations of severity: mild (GCS 13–15), more predictive of long-term functional outcomes
moderate (GCS 9–12), and severe (GCS 3–8). Injury than earlier assessments [14–16]. Previous guidelines
scoring systems such as the Abbreviated Injury Scale for prediction of outcomes after tSCI recommend
and the Injury Severity Score (ISS) provide informa- comprehensive assessment 3 to 7 days after injury
tion about the severity of trauma and body regions [5]. One caveat is that for the purpose of scientific
impacted. The Abbreviated Injury Scale, developed analysis, assessments performed at later time inter-
in the 1960s in response to an increase in automo- vals are more likely to reflect treatment effects than
bile accidents, assigns an injury severity score by assessments done at the time of initial presentation.
body region (head/neck, face, chest, abdomen/pel- For the purposes of this systematic review, the NLI
vic organs, extremities/pelvic girdle, and external). was defined as the first reported ISNCSCI scale or a
Each region is scored from 0 to 6 according to injury similar scoring system. In most cases, the first NLI
severity (0 none, 1 minor, 2 moderate, 3 serious, 4 was recorded at admission or within 72 h, although
severe, 5 critical, and 6 maximal/untreatable). The this information was not reported in all studies. Stud-
ISS, used by trauma systems internationally, is the ies were not excluded if the time of determination of
sum of the squares of Abbreviated Injury Scale scores NLI was not specified, provided the NLI was deter-
for the three body regions with the highest severity mined during the initial postinjury hospitalization.
scores. The resulting sum score for injured patients 5. Magnetic resonance imaging (MRI) findings. Radio-
ranges from 1 to 75 [11, 12]. For the purposes of logical evaluation by computed tomography is stand-
these guidelines, we did not limit our search to a spe- ard in trauma systems. MRI diagnostics have evolved
cific definition of concomitant injury to avoid exclud- as state-of-the-art tools in the management of tSCI
ing potentially relevant predictors. but may be limited outside of core working hours in
4. Neurological level of injury (NLI). NLI reflects the some areas. Costs, staffing issues, safety concerns,
most caudal segment of the spine with normal sen- and sometimes logistics of patient care may be pro-
sation and at least antigravity motor strength, pro- hibitive. However, MRI provides good soft tissue
vided that rostral movement and sensation is nor- contrast and the ability to evaluate details such as
mal. The International Standards for Neurological neurovascular injury, ongoing compression, extent
Classification of Spinal Cord Injury (ISNCSCI) scale, of secondary injuries, and resolution of injury (signs
commonly called the “American Spinal Injury Asso- of cord compression, lesion size, edema) following
ciation (ASIA) Scale,” provides a structured approach definitive treatment. MRI findings with a potential
to assessment and determination of both NLI and role in prognostication include spinal canal com-
severity (completeness) of injury [13]. The ISNCSCI promise, intramedullary signal change (e.g., hemor-
scale may be impacted by spinal shock and requires rhage, ischemia), maximum spinal cord compres-
training to perform accurately, so it is not reported sion, and extension of edema. MRI protocols differ
in all studies of early prognostication after SCI. Some between hospital systems and the specific protocol
studies identify the specific NLI by spinal segment, used needs to be carefully considered in prognostic
whereas others categorize injury by region (cervical, studies. Advanced microstructural, biochemical, or
thoracic, lumbar, thoracolumbar, conus, or other). functional imaging techniques of the spinal cord such
The broadest differentiation based on level of injury as magnetic resonance spectroscopy, diffusion tensor
is tetraplegia (cervical and cervicothoracic SCI) or imaging, positron-emission of single-photon-emis-
paraplegia (thoracic or thoracolumbar SCI). Patients sion tomography, or functional MRI are in develop-
 ment. For the purposes of these guidelines, specific level and completeness of injury before and after signifi-
MRI findings were not defined prior to the literature cant treatment interventions.
search to avoid excluding potentially relevant predic-
tors.
Clinical Prediction Models
6. Severity of injury to the spinal cord (complete or
incomplete). Following tSCI, loss of motor and The Dutch Clinical Prediction Rule (DCPR), derived from
sensory function may be complete or incomplete. the EMSCI data set, was introduced in 2011 to aid in
Incomplete injury indicates conduction of electrical early prediction of independent ambulation among adult
impulses past the area of cord injury. Completeness of patients after tSCI. Variables in the final model were age
injury is most often reported using the ASIA Impair- dichotomized at 65 years, motor scores of the quadriceps
ment Scale (AIS) score (Supplementary Table 2), femoris (L3) and gastrocnemius (S1) muscles, and light
although earlier research may report the Frankel touch sensation in the corresponding dermatomes [18].
grade, a predecessor to AIS grading. A patient with Neurological assessment was performed within 15 days
a complete injury (AIS A) has no sensation or move- of injury. International (external) validation studies have
ment below the lesion. Incomplete injury reflects been conducted in North America (USA and Canada)
varying degrees of sensory or motor preservation and and Australia [19–22]. The DCPR generates a total score
must include evidence of preserved S4-5 function, of − 10 to 40 that can be used to predict probability of
which can be motor (voluntary anal contraction), ambulation at 1 year. Table 2 provides an overview and
sensory (light touch, pin prick sensation, or deep anal examples of DCPR score calculation, and Table 3 indi-
pressure sensation), or both. Preservation of sensa- cates the predicted probability of independent ambula-
tion with complete loss of motor function below the tion at 1 year based on selected DCPR scores.
NLI is categorized as AIS B, sensory incomplete. The
classifications AIS C and AIS D denote motor incom- Guideline Questions
plete injuries. If less than half of key muscle groups The Population/Intervention/Comparator/Outcome/
below the NLI have at least antigravity strength, Time frame/Setting question was framed for the specific
the injury is classified as AIS C; if at least half of the candidate predictors as follows:
key muscles below the NLI have at least antigravity
strength, the injury is classified as AIS D. A grade of When counseling patients, family members, and/
AIS E indicates resolution of SCI-related deficits [13]. or surrogates of adults with acute traumatic spinal
Antigravity muscle strength (muscle function grade cord injury, should <predictor, with time of assess-
of 3 on a 5-point scale) is the minimum necessary for ment if appropriate> be considered a reliable pre-
recovery of functional activities. Similar to NLI, the dictor of <outcome, with time frame of assessment>?
timing of assessment is important. For the purposes
of prognostication, determination of completeness of Selection of Outcomes
injury (AIS grade) should be deferred for 72 h, with Outcomes were selected using the GRADE 1–9 scale
the previously noted considerations and limitations with input from experts on the writing panel and the
in the available literature. patient representative. The outcomes rated “critical”
were mortality at discharge from the acute care hospital
Note on early surgical decompression as a predictor of or later (average rating 7.67), functional outcome at dis-
outcome: Treatment of tSCI has been driven by medi- charge from rehabilitation or later (average rating 8.67),
cal management, surgical restoration of spinal column improvement in ASIA Impairment Scale (AIS grade con-
stability, and surgical decompression of the spinal canal version) at discharge from rehabilitation or later (average
to counter existing or developing pressure on the spinal rating 7.0), independent ambulation at discharge from
cord. Although the benefit of early surgical decompres- rehabilitation or later (average rating 8.0), and bowel
sion has been a topic of debate, there is consensus that and bladder control (average rating 8.0). Outcomes were
it is safe and may improve neurological outcomes [17]. defined as follows:
The impact of therapeutic interventions was outside the
scope of these guidelines. However, clinicians should •  Mortality. Because the factors impacting mortality
be aware of the potential beneficial impact of treatment related to tSCI vary based on time from injury, the
approaches, specifically early surgical decompression, reliability of predictors of acute in-hospital mor-
on neurological outcomes. Information about treatment tality (occurring during the initial acute hospitali-
approaches should be incorporated into research into zation after injury) has been described separately
prognosis after tSCI and should include measurement of from predictors of long-term mortality (cumulative
Table 2 Dutch clinical prediction rule (DCPR) variables Table 3 Dutch Clinical Prediction Rule (DCPR): predicted
and scoring probability of ambulation by selected scores
Variable Range Weighted Minimum Maximum DCPR Score Predicted probability of
of test coefficients score score independent ambulation
scores at 1 year (%)

Age > 65 years 0–1 − 10 − 10 0 − 10 <1

Motor score L3 0–5 2 0 10 −5 1
Motor score S1 0–5 2 0 10 0 4
Light touch score 0–2 5 0 10 2 6
L3 5 13
Light touch score 0–2 5 0 10 10 35
S1
15 68
Total − 10 40
20 89
van Middendorp JJ, Hosman AJF, Donders ART, Pouw MH, Ditunno
JF, Curt A, et al. A clinical prediction rule for ambulation outcomes 25 97
after traumatic spinal cord injury: a longitudinal cohort study. Lancet. 28 99
2011;377:1004–10 30 > 99
Scoring of motor and sensory scores is in accordance with the international 35 100
standards for neurological classification of spinal cord injuries (ISNCSCI), from
40 100
the American Spine Injury Association (ASIA) and the International Spinal Cord
Society (ISCoS). Assessment is performed within 15 days following injury. The Predicted probability of independent ambulation corresponding to various
best motor and sensory score (left vs. right) is used DCPR scores. See Table 2 for method of calculation of DCPR score. Predicted
The motor score is calculated as follows: 0 = Total paralysis; 1 = Palpable or probabilities are calculated based on the following model
visible contraction; 2 = Active movement with gravity eliminated; 3 = Active Predicted probability = EXP(− 3.273 + 0.267 × DCPR − score)/
movement against gravity; 4 = Active movement against some resistance; (1 + EXP[− 3.279 + 0.267 × DCPR − score])
5 = Active movement against full resistance; NT = Not testable. Motor function is
Point estimates for example cases A and B from Table 2 are given
tested with knee extension at L3 and plantar flexion at S1
van Middendorp JJ, Hosman AJ, Donders AR, Pouw MH, Ditunno JF, Jr., Curt A,
The sensory score is calculated as follows: 0 = Absent; 1 = Altered; 2 = Normal;
et al. A clinical prediction rule for ambulation outcomes after traumatic spinal
NT = Not Testable. Sensation is typically tested at the medial femoral condyle for
cord injury: a longitudinal cohort study. Lancet. 2011;377(9770):1004–10
the L3 dermatome and the lateral aspect of the heel for the S1 dermatome
The age, motor L3, motor S1, light touch L3 and light touch S1 scores are
multiplied by the respective weight coefficients. The sum of these numbers
is then the final score, which ranges from − 10 (patient 65 years or older with
complete paralysis and absent sensation at both L3 and S1) to 40 (patient FIM motor score, constructed from 13 subscales,
younger than 65 years with active movement against full resistance and normal therefore ranges from 13 to 91. Functional inde-
sensation at both L3 and S1)
pendence is assumed once scores of 6 or higher are
For example, a 65-year-old patient (A) with a 1 out of 5 strength in knee
extension (L3) and 0 out of 5 strength in plantarflexion (S1) and scant sensation
reached on all subscales. The SCIM assesses three
in both dermatomes will obtain a score of 2 (− 10 + 2 + 0 + 5 + 5), while a domains: self-care, respiration and sphincter man-
50-year-old patient (B) with 2 out of 5 strength in L3 and 2 out of 5 strength in S1 agement, and mobility. Because FIM and SCIM
and unimpaired sensation in the corresponding dermatomes will obtain a score
of 28 (0 + 4 + 4 + 10 + 10) scores are expected to vary based on NLI, no spe-
cific score was used to define good versus poor out-
comes. Scores are expected to improve in the first 3
mortality measured at any time point after hospital months and plateau within 6–9 months with con-
discharge). tinued rehabilitation. In clinical studies, functional
•  Functional outcome. Assessment of functional out- outcome assessments based on the FIM score are
come after tSCI is most often performed with the typically performed at 1 year. When 1-year out-
Functional Independence Measure (FIM) [23–25] comes are unavailable, 6 month-scores are accept-
or Spinal Cord Independence Measure (SCIM) able [18, 27].
[26] (Supplementary Tables 3 and 4). Both FIM and •  AIS grade conversion. Studies reporting only change
SCIM are validated tools that measure the patient’s in motor score on the ISNCSCI scale or its prede-
ability to perform tasks important to daily living. cessors were excluded because an isolated change
The FIM can be used with a wide range of patients in motor score is difficult to interpret and may
and is divided into a motor subscale (self-care, not reflect improved function. A number of stud-
sphincter control, transfers, and locomotion) and ies reported AIS grade conversion as a surrogate
a cognition subscale (communication and social marker of improved function. These studies were
cognition). Scores range from 1 (totally depend- included because AIS grade provides broad cat-
ent) to 7 (device- and helper-independent). The egories of function for discussion with patients and
families. In addition, improvement in AIS grade may
 reflect increased potential for future participation in that did not report model discrimination. Studies of lab-
self-care. As with other functional outcome meas- oratory biomarkers were included only if the biomarker
ures (see above), most improvement is seen within 3 was considered clinically relevant and had been evaluated
months following injury with a plateau 6–9 months in more than one published study that met other crite-
following injury. This measure is widely used clini- ria. Spinal cord injury unrelated to trauma (for example,
cally during acute care and at follow-up. ischemic injury) was excluded.
•  Independent ambulation. Independent ambulation A summary of individual studies of predictors is in
was broadly defined as the ability to walk for short Supplementary Appendix 3. The GRADE evidence pro-
distances with or without assistive devices. Most file and summary of findings table is in Table 4.
studies report ambulation using a subscale of either
the FIM or SCIM, with scores dichotomized into two Evidence to Recommendation Criteria
groups (independent ambulation/walkers or no inde- 1. Quality of evidence/certainty in the evidence and
pendent ambulation/nonwalkers) for analysis. The effect size: For the purpose of these guidelines, pre-
specific definition of independent ambulation var- dictors described as “reliable” have both a higher
ies slightly by the scale used (FIM or SCIM) and the overall certainty in the evidence and greater effect
version used (SCIM I, II, or III). Studies that utilize size than “moderately reliable” predictors (Table 1).
the FIM typically define independent ambulation as For “reliable” predictors and prediction models, one
a locomotion (walking) score of 6 or 7, representing downgrade was permitted for risk of bias, but none
modified or complete independence [21, 22]. Alter- for inconsistency, imprecision or indirectness, and
natively, independent ambulation has been defined by the overall quality of evidence had to be high or mod-
a locomotion-walking score of 5 (supervision only) or erate. “Reliable” prediction models were required to
higher with at least 50 m of unassisted walking [28]. demonstrate an area under the receiver operating
All versions of SCIM include multiple subscales for curve (AUC) of > 0.8, and no evidence of miscali-
mobility. Independent ambulation is typically defined bration in external validation studies that reported
as an indoor mobility score > 3, reflecting the ability calibration. Single downgrades within each of the
to walk short distances inside without supervision, domains of risk of bias, imprecision, and indirect-
with or without assistance devices such as a walking ness were permitted for “moderately reliable” pre-
frame, crutches, or canes [18]. dictors, but a downgrade for inconsistency was not.
•  Bowel and bladder control. Bowel and bladder con- In addition, “moderately reliable” prediction mod-
trol, broadly defined as the ability to maintain con- els were required to demonstrate an AUC > 0.7, and
tinence with or without the use of adjuncts, was some miscalibration in some external populations
included as an outcome of interest because conti- was allowed, given the lower risk of withdrawal of
nence is a functional outcome that impacts quality life support in this disease. Predictors that did not fit
of life [29, 30]. Although the initial literature search “reliable” or “moderately reliable” criteria were classi-
returned several articles addressing this outcome, fied as “not reliable.”
none of the studies met criteria for inclusion, either 2. Balance of desirable and undesirable consequences:
due to methodological flaws or because they did not Accurate prognostication of functional outcomes
address prognostication based on information avail- after acute tSCI supports the ability of patients, fami-
able during critical care management. lies, and clinicians to plan for future needs and to
better focus rehabilitation activities on recovery of
Systematic Review Methodology function versus adaptation. Inaccurate prediction of
An in-depth description of systematic review methodol- a poor functional outcome may lead to psychological
ogy for these guidelines is in Supplementary Appendix 1. consequences (depression, hopelessness) that nega-
The librarian search string used for this systematic review tively impact engagement in rehabilitation. Among
is in Supplementary Appendix 2 and the PRISMA flow patients requiring early ventilator support, inaccu-
diagram is in Fig. 1. Full-text screening was performed rate prediction of a poor functional outcome could
with the following exclusion criteria: sample size less than become a self-fulfilling prophecy, with mortality due
50, focuses on a highly selected subgroup (such as pene- to withdrawal of life-sustaining treatment. Although
trating trauma or central cord syndrome in patients with the patient’s ability to engage in self-determination is
underlying degenerative cervical myelopathy), studies of a desirable outcome, early withdrawal of life-sustain-
predictors not established as independent with multivari- ing treatment may be an undesirable consequence
ate analysis, studies focused on a genetic polymorphism when based on imperfect prognostic information.
as a predictor, and studies of clinical prediction models The panel and patient representative considered early
Fig. 1 PRISMA flow diagram—guidelines for neuroprognostication: traumatic spinal cord injury

withdrawal of life-sustaining treatments in patients sion with providers. They agreed that most individu-
with tSCI to be less likely than other populations, als with tSCI, along with their families and surro-
such as those with severe TBI or comatose cardiac gates, would likely value information about predicted
arrest survivors, but still relevant for consideration. functional outcomes even when that information is
Although clinical variables may be associated with incomplete or uncertain because it supports planning
higher in-hospital and long-term mortality, the panel for the future. The patient representative stressed the
exercised caution when making recommendations importance of providing balanced information when
about the use of these variables as predictors of mor- discussing prognosis, specifically addressing both the
tality due to concerns about early withdrawal of life- potential for a good quality of life and the burdens
sustaining treatment in patients with the potential associated with functional losses.
for recovery or an acceptable quality of life with SCI- 4. Resource use: The predictors and models that
related disability. have been tested in prognostication of spinal cord
3. Values and preferences: The panel and patient rep- injury reflect information already collected as part
resentative noted that many patients with isolated of routine care. Clinical neurological assessment is
tSCI are conscious and able to participate in discus- required for care and therefore requires no addi-
Table 4 Evidence profile/summary of findings
Outcome Predictor or model Quality of evidence Summary of findings
(narrative of effect size)
RoB Incon- Indirect- Impreci-sion QoE- Summary
sist- ness
ency

Mortality, acute in-hospital Age ↓ – – ↓ Low ↑ with age, but ≥ 75% survive
to discharge in most studies
even among older patients
Comorbidities ↓ ↓ – – Low ↑ with 2 or ≥ 3 vs. no comor-
bidities (HR 1.73–2.19; OR
2.2–2.7), ≥ 75% survive to
discharge. Inconsistent find-
ings in smaller studies
Concomitant Injury ↓ – – ↓ Low ↑ with multisystem injury
(HR 1.4–1.85), higher Injury
Severity Score and TBI.
Overall survival > 80%
NLI ↓ ↓ – ↓ Very low ↑ with high vs. low cervical
injury (OR 2.14–5.24),
Results vary for other levels
Severity of SCI ↓ – – ↓ Low ↑ with complete injury, espe-
cially cervical, but > 75%
survive even with complete
cervical injury
Mortality, long-term Age ↓ – – ↓ Low ↑ with age at time of injury,
but age categories vary
Comorbidities ↓ ↓ – – Low Variable results and limited
data
Concomitant Injury ↓ – – – Moderate No significant ↑ in long-term
mortality with concomitant
injury
NLI ↓ ↓ – ↓ Very low Variable results. ↑ for C1–4 vs.
other levels in largest study
(HR 1.6), survival in C1–4
group ~ 70%
Severity of SCI ↓ – – ↓ Low Study size/methodology
varies. HR of 0.3–0.6 for
incomplete vs. complete
injury in largest study, with
overall mortality of 16.2%
Functional outcome, score- Age ↓ – – – Moderate ↓ motor recovery (FIM motor
based score) with age ≥ 65 years.
AIS grades B/C patients at
younger age fared better
NLI ↓ – – ↓ Low ↑ future motor outcome for
thoracolumbar injuries vs.
cervical. Substratified by
neurological structures
assessed by MRI, cauda
equina > conus medulla-
ris > thoracic myelon
Severity of SCI ↓ – – – Moderate ↑ future motor outcome with
↓ injury severity (AIS)
Table 4 (continued)
Outcome Predictor or model Quality of evidence Summary of findings
(narrative of effect size)
RoB Incon- Indirect- Impreci-sion QoE- Summary
sist- ness
ency

Improvement in AIS Age ↓ – – – Moderate Improvement in AIS up to

12–24 months is independ-
ent of age at time of injury
and age*gender interaction
MRI ↓ – – ↓ Low ↑ in AIS less likely with signa-
tures of tSCI such as maxi-
mum spinal cord compres-
sion (> 50%) or extension of
spinal cord edema
NLI ↓ – – – Moderate ↑ in AIS at 6–12 months is
1.5–fourfold more likely
with tSCI to conus level or
below compared with cervi-
cal or thoracic injuries
Severity of SCI ↓ – – ↓ Low ↑ in AIS is 4–eightfold more
likely with AIS B–D com-
pared with AIS A
Independent ambulation MODEL: Dutch Clinical ↓ – – – Moderate DCPR allows for accurate early
Prediction Rule prediction of an individual
(DCPR) patient’s ability to walk at 1
year post injury (AUC > 0.95);
simplifying the rule by leav-
ing out age or one motor
score and one sensory score
may yield similar results
(AUC 0.87–0.94); predic-
tion accuracy may vary by
initial injury severity (AIS
A + D > B + C)
Long-term mortality reflects cumulative mortality measured at any time point following hospital discharge
AIS, American Spinal Injury Association Impairment Scale, AUC​, area under the curve, DCPR, Dutch clinical prediction rule, FIM, Functional independence measure, HR,
hazard ratio, MRI, magnetic resonance imaging, NLI, neurological level of injury, OR, odds ratio, QoE, quality of evidence, RoB, risk of bias, TBI, traumatic brain injury,
tSCI, traumatic spinal cord injury

tional expenditure. Although MRI has cost asso- Good practice statement #1: We recommend that prog-
ciated, it is considered a basic assessment tool for nostication should be performed with consideration of
tSCI in most centers. Conversely, the costs associ- the complete clinical condition and never based on a sin-
ated with the long-term sequelae of SCI are sub- gle variable (strong recommendation, evidence cannot be
stantial and having information about prognosis can graded).
assist with planning [3]. Rationale: Neuroprognostication in tSCI is complicated
by heterogeneity in injury patterns, management strat-
egies, and study methodology. Clinicians must there-
Good Practice Statements fore use caution when formulating a prognosis with the
In accordance with recommendations of the GRADE predictors addressed in these guidelines. In addition,
network these statements were considered by the panel the overall body of literature on tSCI outcomes is poor
to be actionable, supported by indirect evidence where quality. Most studies performed retrospective analysis
appropriate, and essential to guide the practice of neuro- of registry databases that, even with prospective enroll-
prognostication. The good clinical practices reflected in ment, may exclude important variables. Studies based
these statements lacked a meaningful body of direct sup- on retrospective chart review have additional limita-
porting evidence—typically because of insufficient clini- tions, especially for determination of outcomes. Given
cal equipoise—but were considered by the panel to be these limitations, prognostication can be improved
unequivocally beneficial [31]. through consideration of multiple predictors. Decisions
 should always be made with attention to individual urement, confounding, statistical analysis and report-
patient characteristics, as well as preferences and val- ing, and self-fulfilling prophecy. Many studies had
ues. high risk of bias in the domain of confounding because
treatment effect was not considered, and in statistical
Good practice statement #2: We suggest that when analysis and reporting because the studies were under-
discussing prognosis with patients and surrogates in the powered, or the analysis was not well described. Most
immediate postinjury period, clinicians provide informa- studies also had moderate to high risk of bias in the
tion about the significant challenges of likely functional domain of self-fulfilling prophecy because death was
deficits as well as the potential for improvement and for either preceded by withdrawal of life-sustaining treat-
long-term quality of life with SCI-related deficits (strong ment or this information was not reported. The qual-
recommendation, evidence cannot be graded). ity of evidence was not limited by inconsistency or
Rationale: Good communication is essential when indirectness but was limited by imprecision. Only one
conveying information about the short-term and long- study found no significant impact of age on acute in-
term impacts of tSCI. Research specific to tSCI is lim- hospital mortality, and this study had methodologic
ited, although two small studies emphasize hearing ini- limitations, primarily related to study participation and
tial information from a physician and communication statistical reporting [34]. While the body of evidence
that is truthful but supports hope [32, 33]. Unlike some was consistent in identifying older age as a predictor
neurological diagnoses that impact cognitive function, of acute in-hospital mortality in patients with cervi-
the patient with tSCI is often able to actively participate cal injury [35–38] and when patients with all levels of
in discussion about prognosis. Because the impact of injury were included [39–41], the panel could not rec-
tSCI on family members and other caregivers is sub- ommend the use of age alone as a reliable or moder-
stantial, they should be included in communication ately reliable predictor because, irrespective of age, the
with the patient’s consent, when appropriate. While majority of individuals survive to hospital discharge. In
clinical considerations are often the initial focus, the one of the largest studies (n=3,389 patients with all lev-
financial impact of tSCI is significant and may include els and severity of injury), 528/668 patients (79%) ≥ 65
loss of income as well as the costs for equipment and years survived to discharge [40].
caregivers. By devoting time to ongoing discussion and
providing information about the range of possible func- Question: When counseling patients, family mem-
tional outcomes, clinicians can best support patients bers, and/or surrogates of adults with acute tSCI, should
with tSCI and their families as they plan for the future. comorbidities alone be considered a reliable predictor of
The patient’s NLI and AIS Grade provide information acute in-hospital mortality?
about anticipated functional deficits, although uncer-
tainty must be acknowledged. Early involvement of Recommendation: When counseling patients, family
rehabilitation professionals may be helpful in discus- members, and/or surrogates of adults with acute tSCI,
sion of the long-term impacts of tSCI. we suggest that comorbidities alone not be considered
a reliable predictor of acute in-hospital mortality (weak
recommendation, low quality evidence).
Recommendations: Clinical Variables as Predictors Rationale: The body of evidence was downgraded for
Outcome: Mortality (Acute, in‑Hospital) risk of bias, with various studies demonstrating poten-
Question: When counseling patients, family members, tial bias in the QUIPS domains of study participation,
and/or surrogates of adults with acute tSCI, should older prognostic factor measurement (assessment of comor-
age at the time of injury alone be considered a reliable bidities through retrospective chart review), con-
predictor of acute in-hospital mortality? founding (potential impact of treatment factors), and
self-fulfilling prophecy (withdrawal of life-sustaining
Recommendation: When counseling patients, family treatment not considered or not reported). The qual-
members, and/or surrogates of adults with acute tSCI, ity of evidence was not limited by imprecision or indi-
we suggest that older age alone not be considered a reli- rectness, but inconsistency did downgrade the quality
able predictor of acute in-hospital mortality (weak rec- of evidence. The presence of two or more comorbid
ommendation; low quality evidence). conditions compared to no comorbidities was associ-
Rationale: The body of evidence was downgraded for ated with increased mortality in two large studies that
risk of bias, with various studies demonstrating poten- included a total of more than 6,000 patients [39, 40].
tial bias in the Quality in Prognostic Studies (QUIPS) However, even with ≥ 3 comorbidities, most patients
domains of study participation, prognostic factor meas- in these two studies survived to hospital discharge
 (in-hospital mortality rate 16.9–22.8%). In one smaller Question: When counseling patients, family members,
study (n = 297) that utilized both the CCI and number and/or surrogates of adults with acute tSCI, should NLI
of comorbidities, the number of comorbidities did not alone be considered a reliable predictor of acute in-hos-
impact in-hospital mortality, although the study popu- pital mortality?
lation included a large proportion (60.9%) of patients
with AIS D SCI [42]. Across studies, the body of litera- Recommendation: When counseling patients, fam-
ture demonstrated substantial variation in study popu- ily members, and/or surrogates of patients with acute
lation and in prognostic factor measurement. The panel tSCI, we suggest that NLI alone not be considered a
was unable to recommend the use of comorbidities as reliable predictor of acute in-hospital mortality. (weak
a reliable or moderately reliable predictor of mortality recommendation; very low quality evidence)
because most patients will survive despite the presence Rationale: The body of evidence was downgraded for
of multiple comorbidities. risk of bias, with various studies demonstrating poten-
tial bias in the QUIPS domains of study participation,
Question: When counseling patients, family members, prognostic factor measurement, study confounding,
and/or surrogates of adults with acute tSCI, should con- statistical analysis and reporting, and self-fulfilling
comitant injury alone be considered a reliable predictor prophecy. In addition, several studies included limited
of acute in-hospital mortality? or no information about the timing of prognostic fac-
tor measurement and decisions about withdrawal of
Recommendation: When counseling patients, family life-sustaining treatment. The quality of evidence was
members, and/or surrogates of adults with acute tSCI, not limited by indirectness but was limited by incon-
we suggest that concomitant injury alone not be con- sistency and imprecision. Most studies included only
sidered a reliable predictor of acute in-hospital mortal- patients with cervical injuries. In the only study that
ity (weak recommendation; low quality evidence). included patients with all levels of injury, the majority
Rationale: The body of evidence was downgraded for of the population (83.5% of 6,827 patients) had cervi-
risk of bias, with various studies demonstrating poten- cal injury [41]. There were no significant differences in
tial bias in the QUIPS domains of prognostic factor mortality based on NLI when comparing cervical, tho-
measurement, confounding, statistical analysis and racic, or cauda equina to lumbar levels of injury [41].
reporting, and self-fulfilling prophecy. Prognostic fac- Among studies focused on cervical injury, mortality
tor measurement varied between studies, including a was higher in patients with upper cervical injury, which
count of body systems injured [40], total ISS [39, 41], was defined as C4 or higher in two studies [34, 35] and
and measures of the severity of associated TBI [35, 39, C5 or higher in the largest study [43]. In contrast, Mar-
41]. Most studies did not describe or control for the tin and colleagues found no significant difference in
potential impact of treatment variables, descriptions of mortality between patients with C1–4 and C5–8 inju-
statistical analysis lacked detail, and the impact of with- ries [36]. Across all studies, the vast majority of patients
drawal of life-sustaining treatment was not addressed. (approximately 80–90%) survived to hospital discharge.
The quality of evidence was not limited by inconsist- Additional limitations that favor not considering the
ency or indirectness but was decreased by impreci- predictor include the potential for withdrawal bias
sion. Studies found increased risk of in-hospital mor- among patients with cervical injury and the likelihood
tality in patients with multisystem injury (hazard ratio that patients with complete injury at C1 or C2 do not
1.46–1.85) [40] and patients with higher total ISS [39, survive to reach the hospital.
41]. One study used the Trauma Score and ISS (TRISS)
and found no significant association with acute in- Question: When counseling patients, family members,
hospital mortality [36]. Concomitant TBI was associ- and/or surrogates of adults with acute tSCI, should sever-
ated with higher acute in-hospital mortality, although ity of injury (complete versus incomplete) alone be con-
TBI was defined differently across studies [35, 38, 39, sidered a reliable predictor of acute in-hospital mortality?
41]. However, in those studies in which mortality rates
were separately reported, more than 80% of patients Recommendation: When counseling patients, fam-
with TBI or other multisystem trauma survived to hos- ily members, and/or surrogates of patients with acute
pital discharge [40, 41]. The panel was unable to recom- tSCI, we suggest that severity of injury (complete ver-
mend the use of concomitant injury alone as a reliable sus incomplete) alone not be considered a reliable pre-
or moderately reliable predictor of mortality because dictor of acute in-hospital mortality (weak recommen-
most patients will survive despite the presence of con- dation; low quality evidence).
comitant injuries.
 Rationale: The body of evidence was downgraded for tSCI) increases with age [44–48], the age at which this
risk of bias, with various studies demonstrating poten- increase is seen varies substantially between studies,
tial bias in the QUIPS domains of study participation, follow-up periods vary, confidence intervals are wide,
study attrition, prognostic factor measurement, study and comparator groups vary. Survival at a year or more
confounding, statistical analysis and reporting, and postinjury remains > 50% for almost all age groups [44,
self-fulfilling prophecy. In several studies, the popula- 45, 47], although one study [45] reported a five-year
tion was either selective or was poorly described, and survival of 40.4% among those individuals aged 75 or
some studies did not consider treatment effect, the older. With one exception [46], acute in-hospital mor-
impact of polytrauma, or withdrawal of life-sustaining tality was included in the cumulative mortality rates
treatments. Limited descriptions of statistical analy- reported at later intervals.
sis also contributed to risk of bias. The quality of evi-
dence was not limited by inconsistency or indirectness, Question: When counseling patients, family mem-
but was impacted by imprecision with wide confidence bers, and/or surrogates of adults with acute tSCI, should
intervals in one large study [43]. Mortality risk appears comorbidities alone be considered a reliable predictor of
increased with complete injury when compared to long-term mortality (cumulative mortality measured at
incomplete injury, especially at the cervical level, but any time point after hospital discharge)?
variability in comparison groups and statistical analysis
make the risk difficult to quantify [34, 36, 38–41, 43]. Recommendation: When counseling patients, fam-
Even among patients with complete cervical injury, ily members, and/or surrogates of patients with acute
early survival is consistently reported to exceed 75% tSCI, we suggest that comorbidities alone not be con-
[34, 38, 39, 43]. sidered a reliable predictor of long-term mortality
(cumulative mortality measured at any time point after
There was insufficient evidence to provide a recom- hospital discharge; weak recommendation; low quality
mendation on the use of MRI findings as a predictor of evidence).
in-hospital mortality. Rationale: Three studies [45, 46, 49] evaluated the pre-
dictive value of comorbid conditions at the time of
Outcome: Long‑Term Mortality (Cumulative, Measured acute hospitalization for long-term cumulative mortal-
After Discharge) ity. Moderate risk of bias was present, with potential
Question: When counseling patients, family members, bias in at least one study for all QUIPS domains. The
and/or surrogates of adults with acute tSCI, should older quality of evidence was not limited by imprecision or
age at the time of injury alone be considered a reliable indirectness but was limited by inconsistency. One
predictor of long-term mortality (cumulative mortality study assessed all-cause mortality 5 years post injury
measured at any time point after hospital discharge)? in patients with all levels and varying severity of tSCI,
and found no correlation with CCI during acute hospi-
Recommendation: When counseling patients, fam- talization [45]. Similarly, Esmoris-Arijón and colleagues
ily members, and/or surrogates of patients with acute (2021) found no association between CCI during acute
tSCI, we suggest increased age at time of injury alone hospitalization and death at 1 year [49]. The third
not be considered a reliable predictor of long-term and largest (n=2,685) study evaluated the association
mortality (cumulative mortality measured at any time between number of comorbidities and mortality after
point after hospital discharge). (weak recommendation; discharge from acute hospitalization. This study fol-
low quality evidence) lowed patients for variable periods (longest follow-up
Rationale: The body of evidence was downgraded for > 10 years) and found that patients with two or more
risk of bias, with various studies demonstrating poten- comorbid conditions were at increased risk for death.
tial bias in the QUIPS domains of study participation, However, 84% of patients remained alive at last follow-
outcome measurement, confounding, statistical analy- up [46].
sis and reporting, and self-fulfilling prophecy. The
quality of evidence was not limited by inconsistency Question: When counseling patients, family members,
or indirectness, but was limited by imprecision with and/or surrogates of adults with acute tSCI, should con-
wide confidence intervals throughout. While there comitant injury alone be considered a reliable predictor
is low quality evidence that suggests that the risk of of long-term mortality (cumulative mortality measured
longer-term all-cause mortality (at least 1 year after at any time point after hospital discharge)?
 Recommendation: When counseling patients, fam- or above, approximately 60–70% were still alive 8 to 10
ily members, and/or surrogates of patients with acute years post injury [44, 46], limiting the utility of NLI as a
tSCI, we suggest that concomitant injury alone not be predictor of long-term mortality.
considered a reliable predictor of long-term mortality
(cumulative mortality measured at any time point after Question: When counseling patients, family mem-
hospital discharge; weak recommendation; moderate bers, and/or surrogates of adults with acute tSCI, should
quality evidence). severity of injury alone be considered a reliable predictor
Rationale: The evidence was downgraded for risk of of long-term mortality (cumulative mortality measured
bias in the QUIPS domains of study participation, at any time point after hospital discharge)?
prognostic factor measurement, confounding, and self-
fulfilling prophecy. The quality of evidence was not Recommendation: When counseling patients, fam-
limited by inconsistency, imprecision, or indirectness. ily members, and/or surrogates of patients with acute
All studies reported cumulative mortality, encompass- tSCI, we suggest that severity of injury alone (defined
ing in-hospital and post-discharge mortality. Concomi- by complete versus incomplete injury or by ASIA
tant injury was defined by chest trauma [44], ISS [45], motor score) not be considered a reliable predictor of
or GCS [45, 47], and patient population also varied by long-term mortality (cumulative mortality measured
level and severity of injury. None found a significant at any time point after hospital discharge; weak recom-
impact of concomitant injury on long-term survival mendation; low quality evidence).
after tSCI. Rationale: The body of evidence was downgraded
for risk of bias, with various studies demonstrating
Question: When counseling patients, family members, potential bias in the QUIPS domains of study partici-
and/or surrogates of adults with acute tSCI, should NLI pation, prognostic factor assessment, study confound-
alone be considered a reliable predictor of long-term ing, statistical analysis and reporting, and self-fulfilling
mortality (cumulative mortality measured at any time prophecy. The quality of evidence was not limited by
point after hospital discharge)? inconsistency or indirectness but was impacted by
imprecision. Although increased severity of injury
Recommendation: When counseling patients, fam- (defined by lower ASIA motor score or worse AIS/
ily members, and/or surrogates of patients with acute Frankel grade) appeared to be associated with increased
tSCI, we suggest that NLI alone not be considered a mortality risk, the evidence was limited by variation in
reliable predictor of long-term mortality (cumulative study methodology, and lack of information regarding
mortality measured at any time point after hospital dis- the timing of assessment of injury severity. The larg-
charge). (weak recommendation; very low quality evi- est study reported a hazard ratio of 0.3–0.6 for incom-
dence) plete injury (Frankel grades B-E) versus complete
Rationale: Four studies addressed the impact of NLI on injury (Frankel grade A) with overall mortality (across
long-term outcomes, with follow-up intervals varying all grades) of 16.2% [46]. Although follow-up intervals
from 1 year to almost 12 years. The body of evidence varied and severity of injury did impact mortality, sur-
was downgraded for risk of bias, with various studies vival rates at final follow-up remained above 60 percent
demonstrating potential bias in the QUIPS domains of across studies even for patients with the most severe
study participation, prognostic factor measurement, injuries. Casper et al. (2018) grouped patients by ASIA
study confounding, statistical analysis and reporting, motor score and found that lower motor scores were
and self-fulfilling prophecy. The quality of evidence was associated with higher mortality; however, 5-year sur-
not limited by indirectness but was limited by incon- vival was 68.2% even among the group with the most
sistency and imprecision. One study reported 5-year severe injury (ASIA motor score 0–20) [45].
mortality for 426 patients with cervical, thoracic, and
lumbar injury of variable severity and found no sig- There was insufficient evidence to provide a recom-
nificant impact of NLI on multivariate analysis [45]. mendation on the use of MRI findings as a predictor of
Another found no significant difference in cumulative long-term mortality.
mortality at 1 year for cervical versus thoracolumbar
injury [47]. The remaining two studies found increased Outcome: Functional Outcome (at Discharge
mortality in patients with upper cervical injury (C1- from Rehabilitation or Beyond)
C4); the comparison group was patients with lower cer- Question: When counseling patients, family members,
vical injury in one study [44] and all levels of injury in and/or surrogates of adults with acute tSCI, should
the other [46]. Even among patients with injuries at C4 older age at time of injury alone be considered a reliable
predictor of worse functional outcome at discharge and self-fulfilling prophecy. Imprecision was present.
from rehabilitation or beyond? In a multivariate analysis thoracolumbar injuries inde-
pendently predicted better functional outcome than
Recommendation: When counseling patients, fam- cervical injuries [51]. The study was limited however
ily members, and/or surrogates of patients with acute by a relatively low proportion (<10%) of thoracic and
tSCI, we suggest that age alone not be considered a lumbar SCI. In thoracic SCI the NLI based on the sen-
reliable predictor of functional outcome at 1-year fol- sory exam was also an independent predictor of func-
low-up (weak recommendation; moderate quality evi- tional outcome. The analysis of 400 patients enrolled in
dence). the SCIMS database revealed better FIM motor scores
Rationale: Functional outcome was assessed with the at 1-year follow-up in patients with low (T10–T12)
FIM score in studies that met our criteria. Evidence as opposed to high (T2–T9) NLI [28]. This predictor
was downgraded for the overall risk of bias with poten- could not be recommended as a reliable (or moderately
tial bias in the QUIPS domains of participation, study reliable) predictor in isolation because a substantial
attrition, confounding, statistical analysis and report- number of patients with higher NLI with mitigating
ing and self-fulfilling prophecy. The quality of evidence factors such as incomplete injury achieve functional
was not limited by inconsistency, indirectness, or independence.
imprecision. Available data from four large prospec-
tive databases suggested that age at time of the injury Question: When counseling patients, family mem-
was independently and inversely associated with func- bers, and/or surrogates of adults with acute tSCI,
tional outcome at 1 year. Age could not be considered should the initial severity of injury as measured by the
a reliable, or moderately reliable predictor, because ASIA impairment score alone be considered a reli-
these effects were weak [28, 47, 50, 51]—albeit statis- able predictor of functional outcome at discharge from
tically significant—and an age threshold that reliably rehabilitation or beyond?
predicted poor outcome could not be identified. In
one study, advanced age impacted functional outcome Recommendation: When counseling patients, fam-
and recovery after tSCI as assessed by the motor score ily members, and/or surrogates of patients with acute
of the FIM. This effect varied across the spectrum of tSCI, we suggest that initial severity of injury as meas-
injury severity. Potential for improvement generally ured by the ASIA impairment score alone not be con-
diminished with completeness or incompleteness of the sidered a reliable predictor for future motor outcome
injury. Age was particularly relevant for AIS grade B/C at 1-year follow-up. However, patients and surrogates
patients, with patients < 65 years demonstrating better should be counseled that motor incomplete injury
functional outcomes [51]. is associated with a higher probability of functional
improvement (weak recommendation; moderate qual-
Question: When counseling patients, family mem- ity evidence).
bers, and/or surrogates of adults with acute tSCI, Rationale: The body of evidence was downgraded for
should the NLI alone be considered a reliable predictor risk of bias, with various studies demonstrating poten-
of functional outcome at discharge from rehabilitation tial bias in the QUIPS domains of study attrition, con-
or beyond? founding, statistical analysis and reporting, and self-
fulfilling prophecy. The quality of evidence was not
Recommendation: When counseling patients, fam- limited by inconsistency, indirectness, or imprecision.
ily members, and/or surrogates of patients with acute Evidence was from prospective registries and study
tSCI, we suggest that NLI alone not be considered a data sets revealed a strong relationship between initial
reliable predictor of poor functional outcome at 1-year ASIA grade and functional outcome as measured by
follow-up. However, patients and surrogates should be the motor score of the FIM. As independent predictors
counseled that a higher level of injury is associated with of functional outcome, the AIS and NLI are most con-
more functional deficits and greater dependence (weak sistently included in clinical prediction models. Patients
recommendation; low quality evidence). with motor complete injuries (AIS A and B) were least
Rationale: The body of evidence was downgraded likely to achieve functional independence, compared
for risk of bias, with various studies demonstrating with patients with AIS C and D injuries [50, 51]. As
potential bias in the QUIPS domains of study attri- previously outlined, these studies primarily included
tion, confounding, statistical analysis and reporting, patients with cervical tSCI, about half with motor com-
 plete injuries. However, an analysis by Lee et al. (2016) predictor of AIS conversion at discharge from rehabilita-
confirmed these results in a population with thoracic tion or beyond?
tSCI, in which 90% of patients had motor complete
injuries [28]. This predictor could not be recommended Recommendation: When counseling patients, fam-
as a reliable (or moderately reliable) predictor in iso- ily members, and/or surrogates of patients with acute
lation because a substantial number of patients with tSCI, we suggest that the absence of pathological find-
motor complete injuries demonstrate AIS grade con- ings on MRI be considered a moderately reliable pre-
version and variable functional independence at long- dictor of AIS conversion at 6–12 months follow-up
term follow-up. (weak recommendation; low quality evidence).
Rationale: The body of evidence was downgraded for
There was insufficient evidence to provide a recom- risk of bias, with various studies demonstrating at least
mendation on comorbidities as a predictor of functional moderate bias in the QUIPS domains of study partici-
outcome. Data from individual studies of concomitant pation and attrition, prognostic factor measurement,
injuries and MRI as predictors of functional outcome confounding, and self-fulfilling prophecy. Imprecision
were also insufficient to support recommendations. A was present. In one study of 86 patients with cervi-
summary of these studies is available in Supplementary cal tSCI (AIS grades B-D), with improvement of at
Appendix 4. least one AIS grade in 77% of all patients at 12-month
follow-up, intramedullary edema >36 mm and facet
dislocation on MRI were independent predictors of
Outcome: AIS Improvement (Conversion) at Discharge the absence of AIS conversion regardless of initial AIS
from Rehabilitation or Beyond grade [55]. In another study, stepwise multiple logis-
Question: When counseling patients, family members, tic regression analysis demonstrated a 5% reduction in
and/or surrogates of adults with acute tSCI, should age at the probability of AIS conversion with each mm incre-
the time of injury alone be considered a reliable predictor ment in intramedullary lesion length [56]. In a group
of AIS improvement (conversion) assessed at discharge of 55 patients who received definitive treatment for
from rehabilitation or beyond? subaxial cervical fracture dislocations, AIS conversion
was observed in 54% at follow-up. Maximal spinal cord
Recommendation: When counseling patients, fam- compression >55.8% on MRI was an independent pre-
ily members, and/or surrogates of patients with acute dictor of the absence of AIS conversion in this study
tSCI, we suggest age at the time of injury alone not [57].
be considered a reliable predictor of AIS conversion
assessed at 12–24 months follow-up (weak recommen- Question: When counseling patients, family members,
dation; moderate quality evidence). and/or surrogates of adults with acute tSCI, should the
Rationale: The body of evidence was downgraded for NLI be considered a reliable predictor of AIS conversion
risk of bias, with various studies demonstrating at least at discharge from rehabilitation or beyond?
moderate bias in the QUIPS domains of study partici-
pation, study attrition, prognostic factor assessment, Recommendation: When counseling patients, fam-
outcome measurement, study confounding, statisti- ily members, and/or surrogates of patients with acute
cal analysis and reporting, and self-fulfilling prophecy. tSCI, we suggest that NLI be considered a moderately
Three studies met inclusion criteria. In two studies, reliable predictor of AIS conversion at 6–12 months
no effect of age at the time of injury on the rate of AIS follow-up. Patients and surrogates should be counse-
conversion was observed. In a study of 14,433 patients led that an improvement in AIS is more likely following
with tSCI from the SCIMS database, neither age alone, lumbar and thoracolumbar injury than cervical or tho-
nor the interaction of gender and age, was a significant racic injury (weak recommendation; moderate quality
predictor of AIS conversion [52, 53]. In one study of evidence).
57 patients with cervical tSCI who underwent surgi- Rationale: The body of evidence was downgraded for
cal management, the median age was 37 years in the risk of bias, with various studies demonstrating at least
group that demonstrated AIS conversion, compared to moderate bias in the QUIPS domains of study partici-
a median age of 64 years in the group that did not [54]. pation and attrition, prognostic factor and outcome
measurement, confounding, statistical analysis and
Question: When counseling patients, family mem- reporting, and self-fulfilling prophecy. In a retrospec-
bers, and/or surrogates of adults with acute tSCI, should tive study of 931 patients with motor complete (AIS
pathological findings on MRI be considered a reliable A/B) tSCI NLI was an independent predictor of AIS
 improvement at 6–12 months. Patients with injury at was present in more than half of these patients at the
the conus or below (lumbar) were more likely to dem- time of initial evaluation [55]. In a study of 58 patients
onstrate an improvement in AIS than patients with operated for thoracic or thoracolumbar tSCI, AIS con-
cervical or thoracic tSCI. [58]. In a study of 95 patients version at 1-year follow-up occurred in 35% of patients
with thoracic and thoracolumbar injuries, an AIS con- with initial AIS A, 82% of patients with AIS B, 80% of
version rate of 92.9% was observed following lum- patients with AIS C and 90% of patients with AIS D
bar tSCI (conus, L1-S5) versus 22.4% for thoracic and [61]. In a study of 86 patients with tSCI who underwent
thoracolumbar tSCI [59]. In a multivariate analysis of surgery only AIS B (OR 4.3, 95%CI 1.2–15.4) and AIS D
86 surgically treated patients, compared with patients (OR 5.2, 1.2–21.6) were independent predictors of AIS
with cervical tSCI, patients with thoracolumbar and improvement [60].
lumbar injury were significantly more likely to demon-
strate AIS improvement (odds ratio [OR] 3.9), whereas There was insufficient evidence to provide a recom-
patients with thoracic injury were less likely to improve mendation on concomitant injury as a predictor of AIS
(OR 0.5) [60]. conversion. Data from individual studies of the predic-
tive value of comorbidities for AIS conversion were
Question: When counseling patients, family members, also insufficient to support recommendations. A sum-
and/or surrogates of adults with acute tSCI, should the mary of these studies is available in Supplementary
initial severity of injury as measured by the ASIA impair- Appendix 4.
ment score be considered a reliable predictor of AIS con-
version at discharge from rehabilitation or beyond? Outcome: Independent Ambulation (at Discharge
from Rehabilitation or Beyond)
Recommendation: When counseling patients, fam- The body of evidence for individual predictors of inde-
ily members, and/or surrogates of patients with acute pendent ambulation was insufficient to support recom-
tSCI, we suggest that severity of injury as assessed by mendations. A summary of individual studies is available
the initial ASIA impairment score be considered a in Supplementary Appendix 4.
moderately reliable predictor of AIS conversion at 6–12
months follow-up. Patients and surrogates should be Recommendations: Clinical Prediction Models
counseled that an improvement in AIS is more likely in One model, the DCPR for ambulation, met criteria for
the presence of incomplete injury (weak recommenda- inclusion. The DCPR predicts the probability of inde-
tion; low quality evidence). pendent ambulation at 1 year post injury using informa-
Rationale: The body of evidence was downgraded tion available during early management of tSCI. Tables 2
for risk of bias, with various studies demonstrating at and 3 provide more information about DCPR score cal-
least moderate potential bias in the QUIPS domains culation and use.
of study participation and attrition, prognostic fac-
tor and outcome measurement, confounding, statisti- Outcome: Independent Ambulation (at Discharge
cal analysis and reporting, and self-fulfilling prophecy. from Rehabilitation or Beyond)
Imprecision was present. The body of evidence was Question: When counseling patients, family members,
consistent in demonstrating an independent associa- and/or surrogates of adults with acute tSCI, should the
tion between severity of SCI as assessed by the initial DCPR for ambulation, with neurological assessment per-
ASIA impairment score and AIS conversion. Less than formed within 15 days following injury, be considered a
one third of patients with AIS A injuries demonstrate reliable predictor of the ability to walk independently at
improvement at 1 year. AIS conversion is 4 to 8-fold discharge from rehabilitation or beyond?
more likely with AIS B–D injuries compared with AIS
A injuries [53, 54, 58]. In a study of 931 patients with Recommendation: When counseling patients, fam-
motor complete tSCI (AIS A/B) 21% of all patients ily members, and/or surrogates of patients with acute
with initial AIS A showed neurological improvement tSCI, we suggest the Dutch Clinical Prediction Rule for
at 1-year follow-up compared to 69% with initial AIS ambulation (DCPR), with neurological assessment per-
B [58]. Among 86 patients with incomplete tSCI (AIS formed within 15 days following injury, be considered
B–D), AIS improved by one or more grades in nearly a moderately reliable predictor of the ability to walk
75%. Although incomplete injury was an independ- 1-year following injury (weak recommendation; mod-
ent predictor of AIS improvement in this study, AIS D erate quality evidence).
 Rationale: The body of evidence was downgraded for This system is constantly reviewed and refined. New
risk of bias, with various studies demonstrating bias in items, such as ZPP, may substantially add to future
the Prediction model Risk Of Bias Assessmnent Tool research endeavors. The impact of consistent use of the
(PROBAST) domains of analysis and self-fulfilling ISNCSCI scoring system on the scientific value of the
prophecy. published evidence cannot be overemphasized because
The DCPR was introduced in 2011 and is based on a it provides a standardized nomenclature and thus the
patient sample set from the EMSCI. It was first vali- ability to compare similar populations across studies.
dated internally and then externally in various interna- The impact of timing of assessment on the accuracy of
tional cohorts [19–22]. Excellent discrimination (AUC outcome prediction requires additional research.
>0.95) was reported in most studies [18, 20]. Simplify- Most of the evidence used as the basis for recom-
ing the rule by leaving out age (AUC 0.94) or leaving mendations within these guidelines was derived from
out one motor score (S1) and one sensory score (L3) multicenter registries and data sets. Statistical value
(AUC 0.87) may yield acceptable discrimination [21]. increases with larger samples, and accurate, detailed
Some recent studies suggest that prediction accuracy descriptions of the population. The use of predefined
may vary based on injury severity (ASIA A+D > B+C) common data elements (CDEs) in published studies is
[22, 62]. Clinical utility may be greatest in patients crucial [64].
with AIS B and C injuries since improvement to func- Modern imaging technology (MRI, tractography,
tional independence is inherently unlikely with AIS spectroscopy) is increasingly a standard in tSCI care,
A injuries and inherently likely with AIS D injuries. and will ultimately enhance multimodal prognostica-
Other studies suggest superior model performance tion. Advanced neuromonitoring of parameters such as
with a lower age threshold than 65 years [63]. In a vali- intraspinal pressure, perfusion and oxygenation is cur-
dation study using the SCIMS database dichotomiza- rently not in routine clinical use but may contribute to
tion of age at 50 years yielded superior model perfor- both management and prognostication in the future.
mance, with both the original DCPR and a simplified Similarly, novel biomarkers from serum and cerebro-
(“3-variable”) version [22]. spinal fluid may assist with prognostication and provide
insights into novel therapeutic options in the future. Fur-
Future Directions ther preclinical and clinical biomarker research is there-
While individual clinical variables may independently fore necessary.
predict an outcome in multivariate analysis, they rarely While these guidelines focus on a limited range of
achieve the predictive accuracy necessary to serve as outcomes, such as long-term independent ambula-
the sole basis for clinical neuroprognostication. A mul- tion, patients and families value a wide range of bod-
timodal approach is therefore essential. The interac- ily functions and self-care activities. Nonspecific
tion of age and comorbidities, probably best described outcomes such as AIS grade conversion do not trans-
as “frailty,” will need better scientific exploration and late linearly to improved function and have limited
consensus on descriptors (“frailty indices”) used. In the meaning to patients and families. In a survey of more
setting of tSCI it is also essential to consider multidi- than 600 patients, recovery of arm and hand function
mensional (motor, sensory, cognitive, socioeconomic, was most important to quadriplegics, while recovery of
etc.), patient-centered functional outcomes, as well as sexual function was of highest priority to paraplegics.
the potential interaction between social determinants Both ranked improvement of bowel and bladder func-
of health and outcomes. Clinical prediction models tion equally high [65]. The body of evidence for predic-
incorporate multiple independent predictors to improve tors of recovery of these functions is especially scarce.
prognostic accuracy, but only a limited number of mod- Future research should focus on these additional
els exist. While models such as the DCPR may benefit patient-centered outcomes.
from further modifications and validation to improve Therapeutic interventions were outside the scope of
predictive accuracy, they provide a reasonable esti- these guidelines, and our recommendations assume pro-
mate of the likelihood of meaningful outcomes, such as vision of the standard of care (best medical practice).
a return to ambulation. These estimates allow patients However, the treatment of tSCI is heterogeneous, and
and families to set realistic expectations and plan for the best medical practice is constantly evolving. As an exam-
future. ple, early surgical decompression of the spinal cord has
The core of tSCI research is a standardized approach received greater emphasis more recently, in addition to
to clinical examination and documentation through restoration of spinal stability. Future prognostication
consistent utilization of the ISNCSCI scoring system. studies should consistently include critical aspects of
Table 5 Summary of recommendations
Outcomes (all outcomes except mortality measured at discharge from rehabilitation or beyond)
Mortality, Mortality, Functional AIS conversion Ambulation Bladder
acute measured outcome (SCIM/ and bowel
in-hospital after discharge, FIM) function
cumulative

Clinical variables Age not reliable not reliable not reliable not reliable insufficient data no data
Comorbidities not reliable not reliable no data insufficient data no data no data
Concomitant not reliable not reliable insufficient data no data no data no data
injury
MRI no data no data insufficient data moderately no data no data
reliable
NLI not reliable not reliable not reliable moderately insufficient data no data
reliable
Severity of SCI not reliable not reliable not reliable moderately insufficient data no data
reliable
Predictive model DCPR not applicable not applicable not applicable not applicable moderately not applicable
reliable
No data: No studies were identified that met criteria for inclusion
Insufficient data: One study met criteria for inclusion; body of literature inadequate to make a recommendation based on the a priori methodological agreements of
the panel
DCPR, Dutch Clinical Prediction Rule, FIM, Functional Independence Measure, MRI, magnetic resonance imaging, NLI, neurological level of injury, SCIM, Spinal Cord
Independence Measure, SCI, spinal cord injury

tSCI treatment, such as early surgical decompression, as of life-sustaining treatment will decrease potential
variables in multivariate analysis and, where appropriate, bias associated with self-fulfilling prophecy.
in prediction models. 6. Consistent use of validated tools for measurement of
Based on the most common study limitations identified both prognostic factor and outcomes will substan-
in our systematic review, future studies should consider tially improve the quality of evidence for neuroprog-
the following general principles: nostication after tSCI. Use of the measurement tools
outlined in the CDEs for spinal cord injury is strongly
1. Clear description of the study population and a pre- encouraged [64].
determined time point of outcome assessment is nec- 7. In addition to standardized study design as outlined
essary to limit study bias related to patient participa- above, power calculations and appropriate statistical
tion and attrition. analysis should be performed.
2. Specification of the timing of neurological assess-
ment is essential for studies evaluating the predictive Conclusions
accuracy of NLI and AIS grade.
3. As a consequence of the heterogeneity of tSCI, only These guidelines provide recommendations on the use
the largest registries provide a sufficient sample size of predictors of mortality as well as functional outcome
to analyze predictors across all levels and severity in the context of counseling adult patients with tSCI and
of injury. Participation in multicenter tSCI registries their surrogates. Recommendations are summarized
should be encouraged. in Table 5. A suggested approach to neuroprognostica-
4. The use of logical patient groupings (for example, tion after tSCI is summarized in Fig. 2. Three predictors
separating AIS A, AIS B/C and AIS D, and cervical/ (absence of pathologic findings on MRI, NLI, and injury
high thoracic NLI from low thoracic/lumbar NLI for severity) were considered moderately reliable for the pre-
analysis), will allow researchers to generate clinically diction of AIS conversion at 1-year follow-up. A clinical
meaningful information. prediction model, the DCPR, was considered moderately
5. Withdrawal of life-sustaining treatment is thought to reliable for the prediction of independent ambulation at
be uncommon following tSCI, but the true frequency 1-year follow-up. Future development of additional bio-
is unknown because most studies do not report this markers and models will help enhance the field of neuro-
information. Including information about withdrawal prognostication in tSCI.
 Fig. 2 Suggested approach to neuroprognostication after traumatic spinal cord injury (tSCI). AIS, American Spinal Injury Association Impairment
Scale, FIM, Functional Independence Measure, ISNCSCI, International Standards for the Neurological Classification of Spinal Cord Injury, NLI, Neuro-
logical Level of Injury

Supplementary Information Practice, University of Illinois at Chicago, Chicago, IL, USA. 12 Department
The online version contains supplementary material available at https://​doi.​ of Neurology, University of Erlangen-Nuremberg, Erlangen, Germany.
13
org/​10.​1007/​s12028-​023-​01845-8. Department of Neurology, Virginia Commonwealth University, Richmond,
VA, USA. 14 Department of Neurosurgery, University of Leipzig, Leipzig, Ger-
many. 15 Department of Neurology, Albany Medical College, Albany, NY, USA.
16
Author details Institute of Medical Informatics, Biometry and Epidemiology, University Hos-
1
Departments of Neurology and Neurosurgery, UVA Health, University pital Essen, Essen, Germany. 17 BDH-Clinic Elzach, Elzach, Germany. 18 Depart-
of Virginia, Charlottesville, VA, USA. 2 Departments of Neurology, Anesthesiol- ment of Neurosurgery, Helios Amper-Klinikum Dachau, Dachau, Germany.
19
ogy and Surgery, University of Massachusetts Chan Medical School, Worcester, Department of Neurosurgery, Neurosurgery Center Ludwigsburg-Heilbronn,
MA, USA. 3 Department of Neurology, University of Leipzig, Leipzig, Germany. Ludwigsburg, Germany.
4
Departments of Neurology and Neurosurgery, University of Michigan, Ann
Arbor, MI, USA. 5 School of Nursing, University of Pittsburgh, Pittsburgh, PA, Acknowledgements
USA. 6 Departments of Neurology and Neurosurgery, College of Medicine, The authors acknowledge Patricia Guerra, our patient representative, for her
University of Florida, Gainesville, FL, USA. 7 Department of Neurology, Univer- contributions to the development of these guidelines and for her feedback on
sity of Washington, Seattle, WA, USA. 8 Departments of Pharmacy and Neu- this manuscript. We would also like to thank Geoffrey R. Smith, MD, FAAPMR,
rosciences, Intermountain Health, Salt Lake City, UT, USA. 9 Department Department of Physical Medicine and Rehabilitation, Mayo Clinic and and
of Neurology, Mayo Clinic, Rochester, MN, USA. 10 Department of Neurology, Deborah M. Stein, MD, MPH, FACS, FCCM, Professor of Surgery, University
University of North Carolina, Chapel Hill, NC, USA. 11 Department of Pharmacy of Maryland School of Medicine and Director Adult Critical Care Services,
University of Maryland Medical Center, who served as expert reviewers for this 5. Consortium for spinal cord medicine. Outcomes following traumatic
guideline. spinal cord injury: clinical practice guidelines for health-care profession-
als. Consortium for Spinal Cord Medicine; 1999.
Endorsements and affirmations 6. Chen Y, DeVivo MJ, Richards JS, SanAgustin TB. Spinal cord injury model
The Society of Critical Care Medicine (SCCM) and German Society of Neuro- systems: review of program and national database from 1970 to 2015.
surgery (DGNC) hav endorsed these guidelines. The American Association Arch Phys Med Rehabil. 2016;97:1797–804.
of Neurological Surgeons/Congress of Neurological Surgeons (AANS / CNS) 7. Model Systems Knowledge Translation Center. Spinal Cord Injury (SCI)
affirms the educational benefit of this document. Model System [Internet]. Spinal Cord Injury (SCI) Model System. 2022
[cited 2022 Jul 6]. https://​msktc.​org/​about-​model-​syste​ms/​sci
Author Contributions 8. Christopher & Dana Reeve Foundation. Our clinical trials network: Reeve
DM and OWS conceptualized the literature search, completed acquisition, Foundation North American Trials Network [Internet]. Our clinical trials
analysis, and interpretation of the data, drafted the initial manuscript, and network: Reeve Foundation North American Trials Network. 2022 [cited
revised it subsequently. All other authors participated in data interpretation 2022 Jul 8]. https://​www.​chris​tophe​rreeve.​org/​resea​rch/​nactn
and critical revision of the manuscript for intellectual content. All authors 9. Praxis Spinal Cord Institute. National SCI Registry (RHSCIR) [Internet].
reviewed and approved the final version of the manuscript. National SCI Registry (RHSCIR). 2022 [cited 2022 Jul 6]. https://​praxi​sinst​
itute.​org/​resea​rch-​care/​key-​initi​atives/​natio​nal-​sci-​regis​try/
Source of support 10. European Multicenter Study about Spinal Cord Injury Website [Internet].
Funded by the Neurocritical Care Society; these are official guide- 2022. https://​www.​emsci.​org/​index.​php
lines of the Neurocritical Care Society and Deutsche Gesellschaft für 11. Baker SP, O’Neill B, Haddon W, Long WB. The injury severity score: a
Neurointensivmedizin. method for describing patients with multiple injuries and evaluating
emergency care. J Trauma. 1974;14:187–96.
Conflict of interest 12. Copes WS, Champion HR, Sacco WJ, Lawnick MM, Keast SL, Bain LW. The
None of the authors have any conflicts to disclose relevant to this article. injury severity score revisited. J Trauma. 1988;28:69–77.
13. American Spinal Injury Association and International Spinal Cord Society.
Ethical Approval International Standards for Neurological Classification of SCI (ISNCSCI)
We confirm adherence to ethical guidelines. Institutional review board Worksheet [Internet]. 2019 [cited 2022 Jul 7]. https://​asia-​spina​linju​r y.​org/​
approval was not required, as data collection was based on available literature inter​natio​nal-​stand​ards-​neuro​logic​al-​class​ifica​tion-​sci-​isncs​ci-​works​heet/
review. 14. Maynard FM, Reynolds GG, Fountain S, Wilmot C, Hamilton R. Neuro-
logical prognosis after traumatic quadriplegia. three-year experience
of California regional spinal cord injury care system. J Neurosurg.
Open Access 1979;50:611–6.
This article is licensed under a Creative Commons Attribution 4.0 International 15. Brown PJ, Marino RJ, Herbison GJ, Ditunno JF. The 72-hour examination
License, which permits use, sharing, adaptation, distribution and reproduction as a predictor of recovery in motor complete quadriplegia. Arch Phys
in any medium or format, as long as you give appropriate credit to the original Med Rehabil. 1991;72:546–8.
author(s) and the source, provide a link to the Creative Commons licence, and 16. Herbison GJ, Zerby SA, Cohen ME, Marino RJ, Ditunno JF. Motor power
indicate if changes were made. The images or other third party material in this differences within the first two weeks post-SCI in cervical spinal cord-
article are included in the article’s Creative Commons licence, unless indicated injured quadriplegic subjects. J Neurotrauma. 1992;9:373–80.
otherwise in a credit line to the material. If material is not included in the 17. Fehlings MG, Vaccaro A, Wilson JR, Singh A, Cadotte WD, Harrop JS,
article’s Creative Commons licence and your intended use is not permitted et al. Early versus delayed decompression for traumatic cervical spinal
by statutory regulation or exceeds the permitted use, you will need to obtain cord injury: results of the surgical timing in acute spinal cord injury
permission directly from the copyright holder. To view a copy of this licence, study (STASCIS). PLoS ONE. 2012;7:e32037.
visit http://creativecommons.org/licenses/by/4.0/. 18. van Middendorp JJ, Hosman AJF, Donders ART, Pouw MH, Ditunno
JF, Curt A, et al. A clinical prediction rule for ambulation outcomes
after traumatic spinal cord injury: a longitudinal cohort study. Lancet.
Publisher’s Note 2011;377:1004–10.
Springer Nature remains neutral with regard to jurisdictional claims in pub- 19. Malla R. External validation study of a clinical prediction rule for ambu-
lished maps and institutional affiliations. lation [Internet]. Texas Medical Center; 2013 [cited 2022 Jun 4]. https://​
digit​alcom​mons.​libra​r y.​tmc.​edu/​disse​r tati​ons/​AAI15​41012/
20. van Silfhout L, Peters AEJ, Graco M, Schembri R, Nunn AK, Berlowitz
DJ. Validation of the Dutch clinical prediction rule for ambulation
Received: 17 July 2023 Accepted: 17 August 2023 outcomes in an inpatient setting following traumatic spinal cord injury.
Spinal Cord. 2016;54:614–8.
21. Hicks KE, Zhao Y, Fallah N, Rivers CS, Noonan VK, Plashkes T, et al. A
simplified clinical prediction rule for prognosticating independent
walking after spinal cord injury: a prospective study from a Canadian
References multicenter spinal cord injury registry. Spine J. 2017;17:1383–92.
1. GBD 2016 Traumatic Brain Injury and Spinal Cord Injury Collaborators. 22. Engel-Haber E, Zeilig G, Haber S, Worobey L, Kirshblum S. The effect
Global, regional, and national burden of traumatic brain injury and spinal of age and injury severity on clinical prediction rules for ambulation
cord injury, 1990–2016: a systematic analysis for the Global Burden of among individuals with spinal cord injury. Spine J. 2020;20:1666–75.
Disease Study 2016. Lancet Neurol. 2019;18:56–87. 23. Keith RA, Granger CV, Hamilton BB, Sherwin FS. The functional inde-
2. Jain NB, Ayers GD, Peterson EN, Harris MB, Morse L, O’Connor KC, et al. pendence measure: a new tool for rehabilitation. Adv Clin Rehabil.
Traumatic spinal cord injury in the United States, 1993–2012. JAMA. 1987;1:6–18.
2015;313:2236–43. 24. Dodds TA, Martin DP, Stolov WC, Deyo RA. A validation of the functional
3. National spinal cord injury statistical center. Traumatic spinal cord injury independence measurement and its performance among rehabilitation
facts and figures at a glance [Internet]. Birmingham, AL: University of inpatients. Arch Phys Med Rehabil. 1993;74:531–6.
Alabama at Birmingham; 2022. https://​msktc.​org/​sites/​defau​lt/​files/​SCI-​ 25. Hamilton BB, Laughlin JA, Fiedler RC, Granger CV. Interrater reliability of
Facts-​Figs-​2022-​Eng-​508.​pdf the 7-level functional independence measure (FIM). Scand J Rehabil Med.
4. Badhiwala JH, Wilson JR, Fehlings MG. Global burden of traumatic brain 1994;26:115–9.
and spinal cord injury. Lancet Neurol. 2019;18:24–5. 26. Itzkovich M, Gelernter I, Biering-Sorensen F, Weeks C, Laramee MT,
Craven BC, et al. The spinal cord independence measure (SCIM) version
 III: reliability and validity in a multi-center international study. Disabil 47. Furlan JC, Fehlings M. The impact of age on mortality, impairment and
Rehabil. 2007;29:1926–33. disability among adults with acute traumatic spinal cord injury. J Neuro-
27. EM-SCI Study Group, van Middendorp JJ, Hosman AJF, Pouw MH, Van trauma. 2009;26(10):1707–17.
de Meent H. Is determination between complete and incomplete 48. Higashi T, Eguchi H, Wakayama Y, Sumi M, Saito T. Risk factors associated
traumatic spinal cord injury clinically relevant? Validation of the ASIA with mortality after traumatic cervical spinal cord injury. OTA Int Open
sacral sparing criteria in a prospective cohort of 432 patients. Spinal Cord. Access J Orthop Trauma. 2018;1: e003.
2009;47:809–16. 49. Esmorís-Arijón I, Galeiras R, Salvador de la Barrera S, Fariña MM, Díaz SP.
28. Lee BA, Leiby BE, Marino RJ. Neurological and functional recovery after Characteristics and survival of patients with acute traumatic spinal cord
thoracic spinal cord injury. J Spinal Cord Med. 2016;39:67–76. injury above T6 with prolonged intensive care unit stays. World Neuro-
29. Hicken BL, Putzke JD, Richards JS. Bladder management and quality of life surg. 2021;152:e721–8.
after spinal cord injury. Am J Phys Med Rehabil. 2001;80:916–22. 50. Wilson JR, Grossman RG, Frankowski RF, Kiss A, Davis AM, Kulkarni AV,
30. Liu C-W, Huang C-C, Yang Y-H, Chen S-C, Weng M-C, Huang M-H. et al. A clinical prediction model for long-term functional outcome after
Relationship between neurogenic bowel dysfunction and health- traumatic spinal cord injury based on acute clinical and imaging factors. J
related quality of life in persons with spinal cord injury. J Rehabil Med. Neurotrauma. 2012;29:2263–71.
2009;41:35–40. 51. Wilson JR, Davis AM, Kulkarni AV, Kiss A, Frankowski RF, Grossman RG,
31. Guyatt GH, Schünemann HJ, Djulbegovic B, Akl EA. Guideline pan- et al. Defining age-related differences in outcome after traumatic
els should not GRADE good practice statements. J Clin Epidemiol. spinal cord injury: analysis of a combined, multicenter dataset. Spine J.
2015;68:597–600. 2014;14:1192–8.
32. Kirshblum SC, Botticello AL, DeSipio GB, Fichtenbaum J, Shah A, Scelza 52. Sipski ML, Jackson AB, Gómez-Marín O, Estores I, Stein A. Effects of gender
W. Breaking the news: a pilot study on patient perspectives of discuss- on neurologic and functional recovery after spinal cord injury. Arch Phys
ing prognosis after traumatic spinal cord injury. J Spinal Cord Med. Med Rehabil. 2004;85:1826–36.
2016;39:155–61. 53. Chen S, Gallagher MJ, Hogg F, Papadopoulos MC, Saadoun S. Visibility
33. Nadeau M, Singh S, Bélanger L, Noonan VK, Hamilton L, Boyd M, et al. graph analysis of intraspinal pressure signal predicts functional outcome
Patient perspective: diagnosis and prognosis of acute spinal cord injuries. in spinal cord injured patients. J Neurotrauma. 2018;35:2947–56.
Spinal Cord. 2021;59:865–73. 54. Dobran M, Iacoangeli M, Nocchi N, Di Rienzo A, di Somma LG, Nasi D,
34. Jin G-X, Li L, Cui S-Q, Duan J-Z, Wang H. Persistent hypoalbumine- et al. Surgical treatment of cervical spine trauma: our experience and
mia is a predictor of outcome in cervical spinal cord injury. Spine J. results. Asian J Neurosurg. 2015;10:207.
2014;14:1902–8. 55. Martínez-Pérez R, Cepeda S, Paredes I, Alen JF, Lagares A. MRI Prognos-
35. Claxton AR, Wong DT, Chung F, Fehlings MG. Predictors of hospital tication factors in the setting of cervical spinal cord injury secondary to
mortality and mechanical ventilation in patients with cervical spinal cord trauma. World Neurosurg. 2017;101:623–32.
injury. Can J Anaesth. 1998;45:144–9. 56. Aarabi B, Sansur CA, Ibrahimi DM, Simard JM, Hersh DS, Le E, et al.
36. Martin ND, Marks JA, Donohue J, Giordano C, Cohen MJ, Weinstein MS. Intramedullary lesion length on postoperative magnetic resonance
The mortality inflection point for age and acute cervical spinal cord imaging is a strong predictor of ASIA impairment scale grade conversion
injury. J Trauma Inj Infect Crit Care. 2011;71:380–6. following decompressive surgery in cervical spinal cord injury. Neurosur-
37. Alsaleh K, Bednar D, Forough F. Acute traumatic quadriplegia in adults: gery. 2017;80:610–20.
predictors of acute in-hospital mortality. Turkish Neurosurgery [Internet]. 57. Cao B-H, Wu Z-M, Liang J-W. Risk factors for poor prognosis of cervical
2016 [cited 2022 Apr 26]; Available from: http://www.turkishneurosur- spinal cord injury with subaxial cervical spine fracture-dislocation after
gery.org.tr/summary_en_doi.php3?doi=https://​doi.​org/​10.​5137/​1019-​ surgical treatment: a consort study. Med Sci Monit. 2019;25:1970–5.
5149.​JTN.​16670-​15.1 58. Failli V, Kopp MA, Gericke C, Martus P, Klingbeil S, Brommer B, et al. Func-
38. Elsamadicy AA, Sandhu MRS, Freedman IG, Reeves BC, Koo AB, Hen- tional neurological recovery after spinal cord injury is impaired in patients
gartner A, et al. Impact of frailty on morbidity and mortality in adult with infections. Brain. 2012;135:3238–50.
patients presenting with an acute traumatic cervical spinal cord injury. 59. Harrop JS, Naroji S, Maltenfort MG, Ratliff JK, Tjoumakaris SI, Frank B, et al.
World Neurosurg. 2021;153:e408–18. Neurologic improvement after thoracic, thoracolumbar, and lumbar
39. Varma A, Hill EG, Nicholas J, Selassie A. Predictors of early mortality after spinal cord (Conus Medullaris) injuries. Spine. 2011;36:21–5.
traumatic spinal cord injury: a population-based study. Spine (Phila Pa 60. Kim M, Hong SK, Jeon SR, Roh SW, Lee S. Early (≤48 hours) versus late
1976). 2010;35:778–83. (>48 hours) surgery in spinal cord injury: treatment outcomes and risk
40. Selassie AW, Varma A, Saunders LL, Welldaregay W. Determinants of in- factors for spinal cord injury. World Neurosurg. 2018;118:e513–25.
hospital death after acute spinal cord injury: a population-based study. 61. Dobran M, Iacoangeli M, Di Somma LG, Rienzo Ad, Colasanti R, Nocchi
Spinal Cord. 2013;51:48–54. N, et al. Neurological outcome in a series of 58 patients operated for
41. Shibahashi K, Nishida M, Okura Y, Hamabe Y. Epidemiological state, pre- traumatic thoracolumbar spinal cord injuries. Surg Neurol Int. 2014;5:329.
dictors of early mortality, and predictive models for traumatic spinal cord 62. Phan P, Budhram B, Zhang Q, Rivers CS, Noonan VK, Plashkes T, et al.
injury: a multicenter nationwide cohort study. Spine. 2019;44:479–87. Highlighting discrepancies in walking prediction accuracy for patients
42. Furlan JC, Kattail D, Fehlings M. The impact of co-morbidities on age- with traumatic spinal cord injury: an evaluation of validated prediction
related differences in mortality after acute traumatic spinal cord injury. J models using a Canadian Multicenter Spinal Cord Injury Registry. The
Neurotrauma. 2009;26(8):1361–7. Spine Journal. 2019;19:703–10.
43. Shao J, Zhu W, Chen X, Jia L, Song D, Zhou X, et al. Factors associated 63. Belliveau T, Jette AM, Seetharama S, Axt J, Rosenblum D, Larose D,
with early mortality after cervical spinal cord injury. J Spinal Cord Med. et al. Developing artificial neural network models to predict function-
2011;34:555–62. ing one year after traumatic spinal cord injury. Arch Phys Med Rehabil.
44. Liu C, Yang X, Meng B, Yang Z, Zhao X, Zhao X, et al. Survival in 222 2016;97:1663-1668.e3.
patients with severe CSCI: an 8-year epidemiologic survey in western 64. Spinal Cord Injury CDE Working Group. NINDS common data elements.
china. Arch Phys Med Rehabil. 2019;100:1872–80. Harmonizing information. Streamlining research. Spinal Cord Injury [Inter-
45. Casper DS, Zmistowski B, Schroeder GD, McKenzie JC, Mangan J, Vatson net]. 2022. https://​www.​commo​ndata​eleme​nts.​ninds.​nih.​gov/​Spinal%​
J, et al. Preinjury patient characteristics and postinjury neurological 20Cord%​20Inj​ury
status are associated with mortality following spinal cord injury. Spine. 65. Anderson KD. Targeting recovery: priorities of the spinal cord-injured
2018;43:895–9. population. J Neurotrauma. 2004;21:1371–83.
46. Cao Y, Krause JS, DiPiro N. Risk factors for mortality after spinal cord injury
in the USA. Spinal Cord. 2013;51:413–8.