Factors in the implementation of ASH SCD guidelines on silent cerebral infarcts and cognition at a single sickle cell center Authors: Jennifer A Afranie-Sakyi, Modupe J Karimi, Valentina Restrepo-Espinosa, Janis Bozzo, Joanna S Cole, Layla N Van Doren, Cecelia L Calhoun Published: 2026 TO THE EDITOR: Neurologic complications of sickle cell disease (SCD) include stroke, silent cerebral infarct (SCI), and cognitive impairment1,2 In 2020, the American Society of Hematology (ASH) released guidelines for the prevention, diagnosis, and treatment of cerebrovascular disease in SCD.1 These guidelines recommend at least 1 brain magnetic resonance imaging (MRI) in adulthood to detect SCI in adults with HbSS and HbSβ-0 thalassemia (HbSβ0) genotypes and using simplified signaling questions to monitor for cognitive impairment. Both recommendations cite low certainty in the evidence. SCI constitutes an infarct measuring ≥3 mm that is visible on 2 planes on T2-weighted brain MRI without an associated neurologic deficit on examination.3 SCI can progress to overt strokes.4,5 Both are associated with cognitive impairment.6 The prevalence of stroke or transient ischemic attack is 11% by the age of <19 years and 24% by the age of 45 years.7 SCI is more prevalent, 39% by the age of 18 years and 53.3% by the age of 30 years.2 Knowledge about the practical implementation of these guidelines is limited. Studies have demonstrated substandard rates of annual transcranial Doppler screening for stroke risk in children with HbSS and HbSβ0.8,9 In this study, we aimed to investigate the implementation of SCI and cognition recommendations before and after publication of ASH guidelines. We conducted a retrospective cohort study of 443 adult patients with a principal diagnosis of SCD (International Classification of Diseases-10 D57∗) who had an inpatient or outpatient encounter within the Yale New Haven Health System from October 2018 through May 2024. Emergency department visits only and sickle cell trait (D57.3∗) were excluded (supplemental Figure 1). Brain MRI was identified using billing codes (70551, 70553, and 70554) and confirmed via manual chart review. Cognitive assessment was defined as the documentation of a validated screening tool (ie, Montreal Cognitive Assessment), neuropsychological testing, or informal surveillance, defined as documentation of a cognitive domain, such as memory recall, and exclusion of vague statements, such as “alert and oriented” or “cognition was normal.” If multiple brain MRI scans or assessments were present, the oldest and most recent dates were captured. The history of stroke and other comorbidities was determined by chart diagnosis. Data was tabulated into REDCap and analyzed using SPSS. Preguideline screening excluded those born after 1 January 2002 (n = 45), and postguideline screening excluded those who died before its publication (n = 8). This study was exempt under the Yale University Institutional Review Board. HbSS was the most common genotype (51.5%), followed by HbSC (30.9%), HbSβ+ (8.6%), and HbSβ0 (4.1%). Most (93.5%) patients had a hemoglobin electrophoresis. The median age was 34 years (range, 18-88). Most had Medicaid (44.7%), commercial insurance (38.7%), or Medicare (9.7%), and most (87.4%) also had at least 1 outpatient hematology visit within the health system. In total, 16.9% had a history of stroke, and 17.6% had a cognitive disorder (Table 1). Table 1.Patient characteristicsCharacteristicN = 443Genotype, n (%) HbSS228 (51.5) HbSC137 (30.9) HbSβ+38 (8.6) HbSβ018 (4.1) Another genotype22 (5.0)Hemoglobin electrophoresis on file, n (%) Yes414 (93.5) No29 (6.5)Age, median (range), y34 (18-88)Sex, n (%) Male205 (46.3) Female253 (53.7)Race, n (%) Asian1 (0.2) Black/African American399 (90.1) Not reported35 (7.9) Native American9 (2.0) White4 (0.4)Ethnicity, n (%) Hispanic or Latino/a34 (7.7) Not Hispanic or Latino/a401 (90.5) Not reported8 (1.8)Insurance type, n (%) Medicaid198 (44.7) Commercial/private169 (38.1) Medicare41 (9.3) Uninsured/self-pay35 (7.9)Income, median (range), US $55 175 (12 066-241 827)Chronic transfusion therapy, n (%) Yes (overall)34 (7.9) RBC exchange28 (82.4) Simple transfusions6 (17.6)Disease-modifying therapy, n (%) Hydroxyurea170 (38.4) Crizanlizumab17 (3.8) l-glutamine8 (1.8) Voxelotor53 (12) No prescription on file229 (51.7)Documented cognitive disorder78 (17.6)History of stroke, n (%) Yes (overall)75 (16.7) Ischemic30 (40) Hemorrhagic12 (16) Unknown etiology33 (44)RBC, red blood cell. Patient characteristics RBC, red blood cell. Overall, 172 of the 443 (38.8%) patients obtained a brain MRI in adulthood, including 106 of the 246 (43.1%) with HbSS or HbSβ0. There was no significant difference before vs after ASH guideline publication (29.02% vs 26.56%; χ2 = 0.351; P = .554). Only 19 adults (4.3%) had at least 1 asymptomatic screening brain MRI, 7 of whom had an allogeneic stem cell transplant. Instead, most (153/172 [89%]) obtained a diagnostic brain MRI for symptoms (eg, headache and focal deficits). Among patients with a brain MRI without a history of stroke, 57 had at least 1 unremarkable brain MRI, and 51 had T2 fluid-attenuated inversion recovery hyperintensities, reported as small vessel disease or microvascular changes. These reports did not include measurements of the signal abnormalities required for the definition of SCI. Only 1 radiology report included the SCI measurement (supplemental Figure 2). Binary logistic regression analysis revealed a history of stroke (odds ratio [OR], 10.026; P < .001), mood disorder (OR, 3.348; P < .001), and cognitive disorder (OR, 2.143; P = .013) to be strong predictors of receiving a brain MRI. Genotype was also a predictor (P = .013), with HbSβ0 (OR, 5.851; P = .002), compared to HbSS, although limited because of the sample size difference. Cognitive assessment occurred at least once in 200 of the 443 patients (45.1%; Figure 1A). Most (67.6%) underwent informal surveillance, obtained without using a validated tool. Only 8.6% were screened using the Montreal Cognitive Assessment, whereas 20.4% had undergone formal neuropsychological testing, with a mean age at testing of 12 years (Figure 1C). Overview of SCI screening, brain MRI, and cognitive impairment assessments. (A) Patients who underwent a brain MRI, SCI screening, and cognitive impairment assessment (N = 443). (B) Patients who underwent a brain MRI and cognitive impairment assessment before and after publication of the ASH 2020 guidelines on cerebrovascular disease in SCD. (C) Tools used for cognitive impairment assessment. MiniCog, mini cognitive assessment instrument; MMSE, mini mental status exam; MoCA, Montreal Cognitive Assessment. Cognitive assessments decreased from before to after ASH guideline publication (36.2% vs 27.4%; χ2 = 7.492; P = .006). Predictors of cognitive screening were history of stroke (OR, 3.455; P < .001), history of cognitive disorder (OR, 9.476; P < .001), and mood disorder (OR, 4.839; P < .001). Age, sex, disease-modifying therapy, income, and insurance were not significant predictors of brain MRI or cognitive assessments. However, further exploratory analysis of age and insurance coverage suggests that there is a complex interaction at some levels. In this single-center study, there was low implementation of ASH 2020 cerebrovascular disease recommendations for SCI and cognition impairment, consistent with low implementation of other SCD-related guidelines8, 9, 10; <50% of eligible patients underwent a cognitive assessment or had a brain MRI in adulthood, and only 4.3% had an MRI for SCI screening. There was no improvement in rates after guideline publication. Most brain MRI scans were diagnostic, obtained because of neurologic symptoms, and documentation of MRI signaling abnormalities lacked standard terminology and measurement criteria necessary to identify SCI. Abnormalities were instead described as “nonspecific” or as “microvascular changes and/or small vessel disease.”11 Under current practice, we miss opportunities to identify and intervene on SCI and other lesions, leading to a decrease in quality of life and an increase in morbidity and mortality. Monitoring for cognitive impairment was inconsistent. Most assessments were completed without validated screening tools (informal surveillance), which is user dependent, lacks reproducibility, and is prone to bias. There is less access to specialists for cognitive evaluation in adult patients. The absence of a preferred simplified validated tool in the guidelines may be impeding its implementation. Outside of transient COVID-19 pandemic–related disruptions, when most ambulatory visits were virtual, and resources were limited (noted by a decline in cognitive assessments and brain MRI scans from 2019 to 2020), the reasons for poor implementation of the guidelines are multifactorial and nuanced. Implementation of evidence-based guidelines into routine clinical care is often delayed, even more so in SCD.10,12, 13, 14, 15 Potential barriers at the clinician and/or system level16 include competing clinical priorities, potential insurance restrictions for screening MRI, limited access to specialists, and fragmentation of care during transition from pediatric to adult care. An incomplete SCI documentation system wide hinders implementation and detection. There are several limitations to our study. It was a retrospective study at a single comprehensive SCD center, which is not generalizable for most adult patients with SCD. The retrospective design relies on documentation that may be erroneous or incomplete and lacks standardized adjudication of the imaging. Cognitive assessments may be underestimated if not explicitly documented. To address these, we plan to expand this study to additional SCD centers to investigate barriers to guideline implementation.17 A prospective study evaluating the association between cognitive impairment, pain, and prevalence of SCIs is underway, which will provide a more accurate estimate of the prevalence of SCIs and cognitive impairment. We must ensure that recommendations are applied to all adults with SCD, not only symptomatic adults, so that all can benefit. Future research is critical for strong implementation strategies,18, 19, 20 such as education of stakeholders, including radiologists, hematologists, and neurologists, to develop standardized radiology reports and integration of cognitive tools into the workflow. With strong strategies, we will improve the quality of life of this vulnerable, historically marginalized population. Conflict-of-interest disclosure: The authors declare no competing financial interests.