Carolyn Cheema

Most patients seen in physical therapy (PT) clinics for low back pain (LBP) are treated for chronic low back pain (CLBP), yet PT interventions suggest minimal effectiveness. The Cochrane Back Review Group proposed 'Holy Grail' questions, one being: 'What are the most important (preventable) predictors of chronicity' for patients with LBP? Subsequently, prognostic factors influencing outcomes for CLBP have been described, however results remain conflicting due to methodological weaknesses. This retrospective observational cohort study examined prognostic risk factors for PT outcomes in CLBP treatment using a sub-type of AI. Bootstrap random forest supervised machine learning analysis was employed to identify the outcomes-associated variables. The top variables identified as predictive were: FOTO™ predicted functional status (FS) change score; FOTO™ predicted number of visits; initial FS score, age; history of jogging/walking, obesity, and previous treatments; provider education level; medication use; gender. This article presents how AI can be used to predict risk prognostic factors in healthcare research. Improving predictive accuracy helps clinicians predict outcomes and determine most appropriate plans of care and may impact research attrition rates.

Rationale Low back pain (LBP) is a leading cause of disability in the United States creating substantial economic hardship through multiple negative social, financial, and health effects. Chronic low back pain (CLBP) accounted for just above half of patients treated in physical therapy (PT) clinics for LBP. However, a systematic review and meta-analysis showed only small benefit from PT in the treatment of CLBP. Preliminary evidence suggests that clinician-level variables such as post-professional education may affect treatment outcomes, but requires further investigation to determine whether patients with CLBP would reap greater benefit from treatment by providers with post-professional training. This study examined the relationship between levels of clinician training and patient-reported outcomes in the treatment of CLBP. Methods PTs were surveyed using a large commercial patient outcome assessment system to determine their level of post-professional education then categorized by education level. Outcomes for 2497 patients with CLBP were extracted from the database. To account for the possibility that clinicians with higher levels of training are referred more clinically complex patients, a machine learning approach was used to identify predictive variables for clinician group, which were then used to construct propensity scores as a means to account for differences between groups. Differences in functional status score change among the pooled data were analyzed using linear models adjusted for propensity scores. Results There were no clinically meaningful differences in patient outcomes when comparing clinicians with post-professional training to those without. The propensity score method proved to be a valuable way to account for differences at baseline between the groups. Conclusion Post-professional training does not appear to contribute to improved patient outcomes in the treatment of CLBP. This study demonstrates that propensity score analysis can be used to ensure that differences observed are true and not due to differences at baseline.

The cortical motor system can be reorganized following a stroke, with increased recruitment of the contralesional hemisphere. However, it is unknown whether a similar hemispheric shift occurs in the somatosensory system to adapt to this motor change, and whether this is related to movement impairments. This proof-of-concept study assessed somatosensory evoked potentials (SEPs), P50 and N100, in hemiparetic stroke participants and age-matched controls using high-density electroencephalograph (EEG) recordings during tactile finger stimulation. The laterality index was calculated to determine the hemispheric dominance of the SEP and re-confirmed with source localization. The study found that latencies of P50 and N100 were significantly delayed in stroke brains when stimulating the paretic hand. The amplitude of P50 in the contralateral (to stimulated hand) hemisphere was negatively correlated with the Fügl–Meyer upper extremity motor score in stroke. Bilateral cortical responses were detected in stroke, while only contralateral cortical responses were shown in controls, resulting in a significant difference in the laterality index. These results suggested that somatosensory reorganization after stroke involves increased recruitment of ipsilateral cortical regions, especially for the N100 SEP component. This reorganization delays the latency of somatosensory processing after a stroke. This research provided new insights related to the somatosensory reorganization after stroke, which could enrich future hypothesis-driven therapeutic rehabilitation strategies from a sensory or sensory-motor perspective.

The cortical motor system can be reorganized following a stroke, with an increasing recruitment of the contralesional hemisphere. However, it is unknown whether a similar hemispheric shift occurs in the somatosensory system to adapt to this motor change, and whether this is related to movement impairments. This proof-of-concept study assessed somatosensory evoked potentials (SEPs), P50 and N100, in hemiparetic stroke participants and age-matched controls using high-density electroencephalograph (EEG) recordings during tactile finger stimulation. The laterality index was calculated to determine hemispheric dominance of the SEP and re-confirmed with source localization. The study found that latencies of P50 and N100 were significantly delayed in stroke brains when stimulating the paretic hand. The amplitude of P50 was negatively correlated with Fügl-Meyer Upper Extremity Motor Score in stroke. Bilateral cortical responses were detected in stroke, while only contralateral cortical responses were shown in controls, resulting in a significant difference in the laterality index. These results suggested that somatosensory reorganization after stroke involves increased recruitment of ipsilateral cortical regions. This reorganization delays the latency of somatosensory processing after a stroke. This research provided new insights related to the somatosensory reorganization after stroke, which could enrich future hypothesis-driven therapeutic rehabilitation strategies from a sensory or sensory-motor perspective.

Rationale: Low back pain (LBP) is a leading cause of disability in the United States creating substantial hardships through negative social, financial, and health effects. Chronic low back pain (CLBP) accounted for above half of patients treated in physical therapy (PT) clinics for LBP. However, research shows small benefit from PT in CLBP treatment. Preliminary evidence suggests clinician‐level training variables may affect outcomes, but requires further investigation to determine whether patients with CLBP benefit from treatment by providers with post‐professional training. This study examined the relationship between clinician training levels and patient‐reported outcomes in CLBP treatment.Methods: Physical therapies were surveyed using a large patient outcome assessment system to determine and categorise them by level of post‐professional education. To account for the possibility that clinicians with higher levels of training are referred more‐complex patients, a machine learning approach was used to identify predictive variables for clinician group, then to construct propensity scores to account for differences between groups. Differences in functional status score change among pooled data were analysed using linear models adjusted for propensity scores.Results: There were no clinically meaningful differences in patient outcomes when comparing clinician post‐professional training level. The propensity score method proved to be a valuable way to account for differences at baseline between groups.Conclusion: Post‐professional training does not appear to contribute to improved patient outcomes in the treatment of CLBP. This study demonstrates that propensity score analysis can be used to ensure that differences observed are true and not due to differences at baseline.