(L-458) Predicting Ambulation Ability at Discharge from Inpatient Rehabilitation for Individuals Living with Spinal Cord Injury

Introduction:: Sustaining a spinal cord injury (SCI) has a profound impact on an individual's mobility and overall quality of life [1–3], and if and how well an individual will walk versus using a wheelchair (mobility prognosis) is often unclear [4–6]. Predicting mobility prognosis is critical to setting goals and providing appropriate therapeutic interventions [7–10]. Prior studies have evaluated prognosis following SCI [11,12], but most focus on a binary measure of walking or wheeling, rather than the quality of walking, limiting their clinical utility. Further, they primarily focus on outcomes at 1-year [13,14]. Studies are sparse in predicting mobility at discharge from inpatient rehabilitation (IPR), a time when individuals are expected to be returning home and reintegrating into the community [15]. The goal of this study was to predict more meaningful domains of ambulation at discharge from IPR (assistance needed, speed, and endurance in ambulation) and assess if predictive features differed between domains. Additionally, we investigated if daytime physical activity levels, as measured using limb-based accelerometers, improved the model accuracy, as compared to models that only considered demographic and clinical measures. Accelerometer-measured physical activity level is related to functional mobility [16,17] but has not been used for predicting the recovery of ambulation in people with SCI.

Materials and Methods:: We recruited participants with acute, incomplete SCI within the first three days after admission to IPR. Individuals were excluded if they had leg injuries impacting ambulation or cognitive impairments that could inhibit safety with wearing monitors. Participants wore accelerometers (Actigraph GTX9) on the non-dominant wrist and both ankles for the first week of IPR. We determined daily physical activity levels by calculating accumulated time spent at each activity level (see Table 1) and averaging it across 5-days (ignoring the first and last day of data collection)[18,19]. Time spent in moderate to vigorous physical activity(MVPA) level was calculated to indicate high-intensity activities. Additional predictors collected at admission to IPR included participant demographics (age, gender, BMI) and clinical assessments (American Spinal Injury Association impairment scale (AIS), strength, and sensation). As part of the AIS, strength, and sensation were assessed for the L1 to S5 myotomes and dermatomes, and the severity of the SCI was classified from AIS A to E (complete injury to normal function)[20]. Three domains of ambulation were assessed in the last week of IPR, including the Walking Index for SCI (WISCI - need for assistive devices), 10-meter walk test (10mWT - efficiency in ambulation), and 6-minute walk test (6MWT – endurance in ambulation). Performance in each domain was categorized into three levels, as shown in Table 2. A Support Vector Machine (SVM) model was developed to predict each functional ambulation outcome. The models were trained using backward elimination and three-fold cross-validation, and the accuracy of the predictions was reported.


Results, Conclusions, and Discussions:: The study enrolled 43 adults (32 male, 11 female) with AIS scales B: 8, C: 18, D: 17 between the ages of 18 and 82. SVM models demonstrated good performance in all three domains of functional ambulation with accuracies as follows: WISCI (82.9%), 10mWT (84.2%), and 6MWT (85.2%). Furthermore, models that incorporated physical activity features had higher accuracy than those that used only demographic and clinical assessments. (Increased by 7.3% in the WISCI score level, 11.4% in the 10mWT, and 4.8% in the 6MWT).

Each SVM model identified specific subsets of key predictors to predict functional ambulation outcomes, as shown in Table 3. Our models share some common clinical features with predictive models for binary walking/wheeling outcomes [12], including predictors such as strength and sensation at L3 and S1, and age, which are among the most common predictors for outcomes at 1-year. In our models, L3 strength was significant for efficiency and endurance, and S1 strength predicted the need for assistance. Age was selected as a feature for both the WISCI and 6MWT. This indicates the importance of these features for functional ambulation in both the short and long term.

Features of physical activity of the weaker lower limb contributed significantly to all models; specifically, MVPA or vigorous level of physical activity was selected by all models. This suggests that weaker leg activity has a more significant impact on ambulatory ability. The 10mWT model included physical activity levels for both sides, indicating the importance of both lower extremities for the speed of ambulation. For the 6MWT, only vigorous physical activity was selected, suggesting that intense daily activity may support greater endurance in ambulation.

Incorporating physical activity as a predictive factor improves the accuracy of predictions of more complex measures of ambulation, shortly after rehab beyond relying solely on clinical and demographic data. This finding may have implications for developing personalized treatment strategies in an IPR setting. Future research should aim to collect larger sample sizes and incorporate additional relevant predictors.


Acknowledgements (Optional): :

References (Optional): : [1] Hiremath SV, Hogaboom NS, Roscher MR, Worobey LA, Oyster ML, Boninger ML. Longitudinal Prediction of Quality-of-Life Scores and Locomotion in Individuals With Traumatic Spinal Cord Injury. Archives of physical medicine and rehabilitation 2017; 98: 2385–2392.

[2] Spinal Cord Injury Facts and Figures at a Glance. The journal of spinal cord medicine 2014; 37.

[3] National Spinal Cord Injury Statistical Center. 2020 Annual Statistical Report for the Spinal Cord Injury Model Systems; 2020 2020.

[4] Malla R. External validation study of a clinical prediction rule for ambulation outcomes after traumatic spinal cord injury. Texas Medical Center Dissertations 2013.

[5] Engel-Haber E, Zeilig G, Haber S, Worobey L, Kirshblum S. The effect of age and injury severity on clinical prediction rules for ambulation among individuals with spinal cord injury. The spine journal : official journal of the North American Spine Society 2020; 20: 1666–1675.

[6] Rigot S, Worobey L, Boninger ML. Gait Training in Acute Spinal Cord Injury Rehabilitation-Utilization and Outcomes Among Nonambulatory Individuals: Findings From the SCIRehab Project. Archives of physical medicine and rehabilitation 2018; 99: 1591–1598.

[7] Kirshblum S, Snider B, Eren F, Guest J. Characterizing Natural Recovery after Traumatic Spinal Cord Injury. Journal of neurotrauma 2021; 38: 1267–1284.

[8] Chay W, Kirshblum S. Predicting Outcomes After Spinal Cord Injury. Physical medicine and rehabilitation clinics of North America 2020; 31: 331–343.

[9] Simpson LA, Eng JJ, Hsieh JTC, Wolfe DL. The health and life priorities of individuals with spinal cord injury: a systematic review. Journal of neurotrauma 2012; 29: 1548–1555.

[10] Ditunno PL, Patrick M, Stineman M, Ditunno JF. Who wants to walk? Preferences for recovery after SCI: a longitudinal and cross-sectional study. Spinal cord 2008; 46: 500–506.

[11] DeVries Z, Hoda M, Rivers CS, et al. Development of an unsupervised machine learning algorithm for the prognostication of walking ability in spinal cord injury patients. The spine journal : official journal of the North American Spine Society 2020; 20: 213–224.

[12] van Middendorp JJ, Hosman AJF, Donders ART, et al. A clinical prediction rule for ambulation outcomes after traumatic spinal cord injury: a longitudinal cohort study. The Lancet 2011; 377: 1004–1010.

[13] Rigot SK, Boninger ML, Ding D, et al. Toward Improving the Prediction of Functional Ambulation After Spinal Cord Injury Through the Inclusion of Limb Accelerations During Sleep and Personal Factors. Archives of physical medicine and rehabilitation 2022; 103: 676-687.e6.

[14] Belliveau T, Jette AM, Seetharama S, et al. Developing Artificial Neural Network Models to Predict Functioning One Year After Traumatic Spinal Cord Injury. Archives of physical medicine and rehabilitation 2016; 97: 1663-1668.e3.

[15] Frontera WR, DeLisa JA, Gans BM, Robinson LR. DeLisa’s Physical Medicine and Rehabilitation: Principles and Practice: Lippincott Williams & Wilkins 2019.

[16] Levins SM, Redenbach DM, Dyck I. Individual and Societal Influences on Participation in Physical Activity Following Spinal Cord Injury: A Qualitative Study. Physical therapy 2004; 84: 496–509.

[17] Stevens SL, Caputo JL, Fuller DK, Morgan DW. Physical activity and quality of life in adults with spinal cord injury. The Journal of Spinal Cord Medicine 2008; 31: 373–378.

[18] Troiano RP, Berrigan D, Dodd KW, Mâsse LC, Tilert T, McDowell M. Physical activity in the United States measured by accelerometer. Medicine and science in sports and exercise 2008; 40: 181–188.

[19] Warms CA, Belza BL. Actigraphy as a measure of physical activity for wheelchair users with spinal cord injury. Nursing research 2004; 53: 136–143.

[20] International standards for neurological classification of spinal cord injury (Revised 2011). The journal of spinal cord medicine.