Clinical Prediction Rules for Ambulation After SCI

Jordan Bosch, Jen Buczkowski, Seth Conway, Tyler Lacy, Taylor Pfiefer

Clinical Question: Is there conclusive evidence validating the use of CPRs in predicting ambulation after a traumatic spinal cord injury?


Screen Shot 2017-04-12 at 3.48.24 PM.png

Spinal Cord Injuries

  • Spinal Cord: is the main pathway of communication between the brain and the rest of the body. It serves as a connection for our central nervous system and functions by nerves that carry impulses to and from the brain to the rest of the body.
  • Spinal Cord Injury (SCI): physical damage through trauma, like a sudden blow to the spine, that fractures or dislocates vertebrae and results in a loss/impaired function.

Incidence Rate
  • Current population size of 314 million people in the U.S., annual incidence of spinal cord injury (SCI) is approximately 54 cases per million population or approximately 17,000 new SCI cases each year.
  • Number of people alive in 2016 from SCI has been estimated to be approximately 282,000 persons (range: 243,000-347,000)
  • Males - 80% of new SCI cases
  • Age - 42 years old

American Spinal Cord Injury Association (ASIA): standard system of examinations to classify the level of SCI severity based on a thorough neurological examination. The examination includes:
Screen Shot 2017-04-12 at 4.18.27 PM.png
  • Sensory and motor levels
  • The completeness of the injury
  • ASIA impairment scale
  • The zone of partial preservation for complete injuries

Clinical Prediction Rules (CPRs)

Clinical Prediction Rules (CPRs): a tool designed to improve decision making in clinical practice by assisting practitioners in making a particular diagnosis, establishing a prognosis, or matching patients to optimal interventions based on a subset of predictor variables.
  • Provides practitioners with powerful diagnostic information (history and physical examination) that may help as an accurate decision-making tool in comparison to a more expensive diagnostic test.
  • Physical therapy - subgroup patients into specific classifications to help guide treatment approaches

Example: Ottawa Ankle Rules (100% sensitivity, 40% specificity)
Conventional radiographs should be ordered if:
  • Pain in the malleolar area AND
  • Tenderness at posterior aspect or tip of lateral malleolus, OR
  • Tenderness at posterior aspect or tip of medial malleolus, OR
  • Inability to WB immediately and in emergency department

CPRs are developed and validated according to rigorous methodological standards.
McGinn et al. have suggested a 3-step process:
Screen Shot 2017-04-20 at 10.11.03 AM.png
  • 1. Creating the Clinical Prediction Rule
    • Predictive Factors
    • Sample Size
  • 2. Validating the Clinical Prediction Rule - “Test Set”
    • Doesn’t occur by chance
    • Unique to particular population
    • Training to accurately apply predictor factors
  • 3. Conducting an Impact Analysis
    • Only useful if it can improve clinically relevant outcomes, increase patient satisfaction, and decrease costs once it is implemented

Levels of Evidence Hierarchy: determine whether the CPR is appropriate in the decision-making process
Screen Shot 2017-04-19 at 3.26.54 PM.png

A clinical prediction rule for ambulation outcomes after traumatic spinal cord injury: a longitudinal cohort study. - by van Middendorp et al.

Background - van Middendorp et al.

  • Before this study, there was no CPR for patients with spinal cord injuries to predict their ability to walk independently.
  • This was a longitudinal cohort study of patients with SCI to create a CPR to predict ambulation.
  • A simple CPR was created, using factors from the international scale for spinal cord classification and the patient’s age at time of injury as prognostic variables.

Methods - van Middendorp et al.

  • The investigators gathered data from 19 locations across Europe, and of 1442 potential candidates with spinal cord injuries, 492 had the outcome data the researchers were looking for.
  • Functional status and neurological data were recorded within 15 days of the SCI, and follow-up measurements were taken at 1 month, 3 months, 6 months, and 12 months post-injury.
  • The treatments each patient underwent were not recorded, as there is no proven effective treatment for SCI.
  • Validation for the created CPR was done on 99 subjects that met inclusion criteria after the CPR was created.
  • Methods - van Middendorp et al.

Prognostic Variables
  1. Age (Two Groups, 0-64, 65+)
  2. Motor Score Testing
    • Adaptation of Medical Research Council scale
    • L3 & S1 (quad/gastrosoleus)
  3. Light Touch Sensory
    • L3 & S1
  4. Pinprick Sensation (removed from initial CPR)
  5. Sacral Sparing Scores (removed from initial CPR)
    • Anal contraction and sensation

Clinical Prediction Rule Variables - Original Model

Screen Shot 2017-04-18 at 1.26.13 PM.png

Functional Outcome Measure of Interest: Spinal Cord Independence Measure (SCIM): Item 12 (mobility indoors <10ft)

Functional Outcome Measure of Interest: Spinal Cord Independence Measure (SCIM): Item 12 (mobility indoors <10ft)
Screen Shot 2017-04-11 at 3.39.26 PM.png

Results - van Middendorp et al.

Screen Shot 2017-04-18 at 1.29.11 PM.png
  • Probabilities of walking independently (SCIM Item 12 score 4-8) 1 year post-injury
  • Shaded area is 95% confidence interval

Screen Shot 2017-04-18 at 1.37.35 PM.png
  • Left is derivation group
  • Right is validation group

Discussion - van Middendorp et al.

  • It was found that the CPR created has a significantly higher accuracy of predicting ambulation outcomes 1-year post-SCI than the ASIA Standard Neurological Scale
  • CPR - 0.956 accuracy of predicting outcomes [95% CI 0·940–0·978] (p<0·0001)
  • ASIA - 0.898 accuracy of predicting outcomes [95% CI 0·867–0·928, (p<0·0001)

American Spinal Injury Association (ASIA) / International Spinal Cord Society Neurological Standard Scale

Screen Shot 2017-04-11 at 3.38.51 PM.png

Screen Shot 2017-04-18 at 5.06.27 PM.png
  • It was also found that the CPR had a greater accuracy of predicting outcomes than each of the ASIA Grades individually.

Clinical Application

  • This created CPR is reliable to predict 1-year ambulatory outcomes in SCI patients
  • Less time-consuming than ASIA test
  • Higher accuracy than ASIA test
  • A con - the physician performing the 2 motor and 2 sensory tests must be experienced with examining SCI patients

Validation of the Dutch clinical prediction rule for ambulation outcomes in an inpatient setting following traumatic spinal cord injury - by van Silfhout et al.

Background - van Silfhout et al.

  • Purpose:
    • A retrospective study to further assess the accuracy and validity of the proposed Dutch CPR used in the Middendorp et al. study.
    • To replicate the Dutch CPR study as closely as possible in clinical practice and evaluate the efficacy of those findings.
    • Use the CPR previously generated to interpret the results

Methods - van Silfout et al.

  • Study Design:
    • Inclusion Factors
      • Traumatic SCI patients above the age of 18
    • Exclusion Factors
      • Non-traumatic SCI
      • No neurological assessment within 15 days post injury, or mobility assessment within 6 months post injury

  • Patient Population:
    • Austin Hospital SCI Unit: 2006-2012
    • Total Participants:184
    • Average days in ICU: 8.2 days
    • 1- Year Follow up Measure:116 subjects
    • * Almost double the amount of subjects used compared to the original study: 184/99
  • Patient Characteristics:
      • Patient Characteristics.png
  • Prognostic Variables
    1. Age
    2. Motor Score Testing
    3. Light Touch Sensory
    4. Pinprick Sensory
    5. Sacral Sparing

Results - van Silfhout et al.

Walking Vs. Non-Walking Factors.png
  • Significant Ambulation outcome factors:
  1. Age at Injury
  2. Severity of Injury
  3. Time spent in ICU
  4. Time in Hospital

An ROC curve was utilized to compare the clinical results recorded with the Middendorp et al. study.
Area Under the Curve.png
  • Deviation into the top left corner compared to baseline shows that our CPR has highly significant predictive value.
  • AUC: 0.939, 95% CI

Discussion - van Silfhout et al.

  • Results obtained from this study were not significantly different than those seen by Middendorp et al.
    • Original Study (Middendorp et al.)
      • AUC: (0.956, 95% CI (0.936, 0.976))
      • AUC: (0.939, 95% CI (0.892, 0.986))
  • Replication of the Dutch CPR study in a clinical setting looks consistent with previous findings.

Conclusion - van Silfhout et al.

  • Replication of the Dutch CPR study found no statistical differences when compared to the results of this study.
  • Having no significant differences between studies, shows significant validation for the use of this CPR in a practical clinical setting.
  • The Dutch CPR generated by Middendorp et al. seems to work as intended.

A clinical prediction model for long-term functional outcome after traumatic spinal cord injury based on acute clinical and imaging factors - Wilson et al.

Background - Wilson et al.

“There is a pressing unmet need to accurately prognosticate, early after SCI, a patient's functional outcome.”
  • Physicians have little to scientifically guide their prediction of outcome following SC
  • The few studies that do exist, employ variables collected outside of the initial injury period (more than 3 days post-SCI)
    • Therefore they are less useful as acute clinical prediction rules
  • The purpose of this study is to improve the ability of clinicians to predict long-term outcomes in the acute clinical setting (1-3 days)
    • Aim to create a CPR that relates clinical and imaging findings to functional outcome at 1 year following SCI

Methods - Wilson et al.

  • Data was gathered from the combination of two prospective datasets:
    • North American Clinical Trials Network for SCI (NACTN)
      • Enrolls pts with acute traumatic SCI
  • Surgical Timing in Acute Spinal Cord Injury Study (STASCIS)
    • Investigates the role of surgical timing in SCI
  • The data sets were harmonized based on their common data elements to produce a single database.
Screen Shot 2017-04-25 at 11.28.39 PM.png
Predictor Variables
1) Acute ASIA grade
2) Acute ASIA motor score
3) Patient age at injury
4) Intramedullary signal characteristics on spinal
  • All 4 of these predictors have consistently demonstrated prognostic significance in relation to long-term functional outcome after SCI

Functional Outcome Measure of Interest - FIM Motor Score at 1 Year Post-SCI Screen Shot 2017-04-25 at 11.50.28 PM.png

Results - Wilson et al.

  • Better functional outcome when:
    • Less severe ASIA Impairment Scale grade
    • ASIA motor score > 50 at admission
  • Worse functional outcome when:
    • Older age
    • OR MRI signal characteristics with SC edema or hemorrhage
Screen Shot 2017-04-25 at 11.32.48 PM.png
Screen Shot 2017-04-25 at 11.36.09 PM.png

Discussion - Wilson et al.

How effective is this CPR at predicting outcomes?

  • The model predicting FIM motor score demonstrated an R-square of 0.52 in the original dataset and an R square of .52 across a set of 200 bootstraps (random sample).
  • For the logistic model, AUC was 0.93 in the original dataset, and 0.92 (95% CI 0.92, 0.93) across the bootstraps, indicating excellent predictive discrimination.
  • This study produces the first prediction models using acute clinical and radiological data (obtained within the first 3 days of injury), to help predict long term functional outcome after traumatic SCI.

  • This study has created, and internally validated, a prediction model for functional outcome at 1 year based on AIS grade, AMS score, age, and MRI characteristics, with all predictor variables obtained within the first 3 days of injury.
  • Future steps would be to investigate imaging modalities and MRI sequences for predicting outcome for important SCI patient subgroups (incomplete injuies or central cord syndrome).


  • van Middendorp et al. created a CPR to predict ambulation outcomes at 1 year after SCI using 5 measures (age, 2 motor scores, 2 sensation scores), taken within 15 days of SCI with 0.956 accuracy of predicting outcomes.
  • The ASIA scale has a 0.898 accuracy of predicting outcomes and requires many more measurements.
  • van Silfout et al. validated the CPR created by van Middendorp et al. with a clinical setting validation group size twice that of the original study, with an 0.939 accuracy of predicting outcomes.
  • Wilson et al. also created a CPR to predict ambulation outcomes at 1 year after SCI using AISA grade, AMS score, age, and MRI characteristics, taken within 3 days of SCI with 0.93 accuracy of predicting outcomes.


The CPR created by van Middendorp et al. has the highest accuracy of predicting ambulatory outcomes at 1 year after SCI (0.956), gives the largest time frame for taking predictive measures (15 days), and does not require expensive imaging like MRI.


  • Childs, J. D., & Cleland, J. A. (2006). Development and Application of Clinical Prediction Rules to Improve Decision Making in Physical Therapist Practice. Physical Therapy,86(1), 122-131. doi:10.1093/ptj/86.1.122
  • Malla R. External validation study of a clinical prediction rule for ambulation outcomes after traumatic spinal cord injury. Texas Medical Center Dissertations (via ProQuest). 2013:Paper AAI1541012.
  • van Middendorp, J. J., Hosman, a J., Donders, a R., Pouw, M. H., Ditunno Jr., J. F., Curt, A., … Van de Meent, H. (2011). A clinical prediction rule for ambulation outcomes after traumatic spinal cord injury: a longitudinal cohort study. Lancet, 377(9770), 1004–1010. [pii]\r10.1016/S0140-6736(10)62276-3
  • van Silfhout, L., Peters, A. E. J., Graco, M., Schembri, R., Nunn, A. K., & Berlowitz, D. J. (2015). Validation of the Dutch clinical prediction rule for ambulation outcomes in an inpatient setting following traumatic spinal cord injury. Spinal Cord, 54(October), 1–5.
  • Rehab Measures - FIM® instrument (FIM); FIM® is a trademark... (n.d.). Retrieved April 15, 2017, from
  • Wilson, J. R., Grossman, R. G., Frankowski, R. F., Kiss, A., Davis, A. M., Kulkarni, A. V, … Fehlings, M. G. (2012). A clinical prediction model for long-term functional outcome after traumatic spinal cord injury based on acute clinical and imaging factors. Journal of Neurotrauma, 29(13), 2263–71.