Background: Understanding the biological underpinnings of stroke recovery is essential for developing predictive models and advancing precision rehabilitation approaches. Immune biomarkers such as neutrophils, lymphocytes, and their ratios (e.g. the NLR) have been posited to reflect systemic inflammatory responses relevant to long-term outcomes including the 90 day mRS. This pilot study examined the predictive value of such immune markers in the earlier post-stroke period relative to impairment (NIHSS) and more granulay functional measures (the Activity Measure for Post-Acute Care AMPAC). Design and Setting: Retrospective observational cohort study conducted at a single tertiary academic medical center using EHR-derived complete blood count (CBC) with differential data. NIHSS and AMPAC scores were scaled to percent of maximum. Participants: Of 402 consecutive stroke admissions, 76 met inclusion criteria (serial NIHSS, AMPAC, and CBC). Mean age was 65.5 ± 13.1; mean initial NIHSS was 6.8 ± 8.9. Methods: Linear mixed-effects models included Age, one immune marker (Neutrophils, Lymphocytes, or NLR), Time-to-Evaluation (TTE), and baseline score, with Patient ID as a random effect. Results: TTE was significantly associated with NIHSS (ß = -0.0119, SE = 0.0050, p = 0.02), consistent with time-dependent neurologic recovery. Lymphocyte count was significantly associated with better NIHSS (ß = -0.0189, SE = 0.00988, p = 0.04). Neutrophil count showed a moderate though statistically insignificant association with AMPAC (ß = -0.0146, SE = 0.0112, p = 0.195). For both NIHSS and AMPAC, Age and NLR effects were small and inconsistent in our models. Discussion/Conclusion: Lymphocyte count showed a consistent negative association with NIHSS, and neutrophil count showed a moderate negative association with AMPAC, both suggesting immune activation states related to poorer recovery. The finding of minimal NLR effect was unexpected, but because immune profiles with concurrent neutrophilia and lymphopenia occur in only a subset of patients, this result may be mainly due to a lack of power. Stratified or interaction-based approaches will likely be needed to fully capture such effects. Our findings provide preliminary support for the biological plausibility of immune predictors in stroke recovery research. Routine clinical labs such as lymphocyte and neutrophil counts can offer a scalable means to integrate immune profiling into predictive models for precision rehabilitation approaches.
Ballenger et al. (2026) studied this question.