Discussion

In this population-based cohort of in-hospital AIS, we examined cardiac inpatients who developed stroke, with emphasis on those who did and did not undergo cardiac procedures and cardiac surgeries. Adjusted odds of in-hospital mortality and in-hospital complications did not differ significantly across groups. Among patients who underwent procedures, most strokes were diagnosed on the day of the intervention, but strokes diagnosed later were associated with higher odds of mortality.

Few studies directly compare cardiac patients who develop in-hospital AIS across procedural subgroups. In our population-based analysis of in-hospital AIS restricted to patients with a primary cardiac diagnosis, patients who underwent cardiac procedures or surgery were generally younger and more frequently in the ICU at the time of stroke recognition, whereas those without prior procedures more commonly had heart failure, arrhythmia or endocarditis as the primary cardiac diagnosis. In contrast, ischemic heart disease predominated in the cardiac procedure group and valvular disease in the cardiac surgery group. These differences likely reflect variation in underlying cardiac pathology and care pathways rather than a simple gradient of illness severity. ^{Reference Sultan, Bianco and Kilic21–Reference Karunanantham, Ali, Evans, Webb and Large23} Despite these differences, all groups carried a heavy comorbidity burden, similar to broader in-hospital stroke cohorts. ^{Reference Ignacio, Waraich and Khan4}

Procedure-related strokes were not inherently more lethal than strokes in other cardiac inpatients. This contrasts with surgical cohorts that compare cardiac surgery patients with and without postoperative stroke that report higher perioperative and longer-term mortality in those who suffer stroke. ^{Reference Sultan, Bianco and Kilic21,Reference Karunanantham, Ali, Evans, Webb and Large23} One possible hypothesis for our result is that once a cardiac inpatient develops in-hospital AIS, prognosis is poor irrespective of procedural status. Baseline vulnerability and systems of care may also be important determinants of survival that outweigh cardiac procedures. However, these findings should be interpreted in the context of the study’s observational design. Although we adjusted for measured confounders, residual confounding by indication remains possible, as patients undergoing cardiac procedures or surgery may differ from medically managed patients in underlying cardiac pathology, procedural indications, periprocedural antithrombotic therapy and overall clinical complexity, which were not fully captured in our data.

The timing of stroke diagnosis in relation to cardiac procedures also emerged as an important signal in our cohort. Importantly, there is often a delay in stroke diagnosis among patients who have required anesthesia for intervention, and the timing of the stroke diagnosis will not be precise to the hour or minute despite having performed manual chart review. Nevertheless, a majority of strokes in procedural patients were diagnosed on the day of the intervention, consistent with studies that distinguish early (intraoperative or immediate postoperative) from delayed strokes after cardiac surgery, where early events are often embolic and procedure related, while later events are more often linked to evolving atrial fibrillation, hemodynamic instability and systemic illness, and are associated with poor survival. ^{Reference Gaudino, Di Franco, Rong, Piccini and Mack16,Reference Ivascu, Khan and Rahouma24,Reference Hedberg, Boivie and Engström25}

In our study, we found that strokes diagnosed 2 or more days after a cardiac procedure were associated with higher in-hospital mortality than strokes diagnosed on the same day. This finding may reflect a more vulnerable subset of patients with a complicated postoperative course rather than a direct effect of delayed stroke recognition alone. Delayed strokes may occur in the setting of postoperative complications such as atrial fibrillation, sepsis, low cardiac output, respiratory or renal failure and prolonged immobilization, which can accumulate during the postoperative period and contribute to mortality. ^{Reference Nouh, Amin-Hanjani and Furie3,Reference Shu, Aziz and de Havenon6,Reference Saltman, Silver, Fang, Stamplecoski and Kapral26} Postoperative atrial fibrillation, occurring usually 2–4 days post-procedure, is a plausible contributor and has been reported in up to 30% of cardiac surgical patients, to as high as 40–50% after valve surgery. ^{Reference Gaudino, Di Franco, Rong, Piccini and Mack16,Reference Lin, Kamel, Singer, Wu, Lee and Ovbiagele27} Another hypothesis is that delayed perioperative strokes may reflect a distinct phenotype with more severe strokes, whereas early strokes may include smaller, procedure-related embolic events. However, we cannot confirm this without imaging data, which would be important for identifying more granular patterns in perioperative stroke. Early post-procedural strokes, in contrast to delayed strokes, may also be recognized as procedure-related complications that may trigger more aggressive care pathways, for example, earlier referral to rehabilitation, which could influence outcomes. However, we do not have data on rehabilitation interventions to support this. Ultimately, our findings suggest the need for sustained neurological monitoring during the postoperative period and not only on the day of surgery.

Within the procedural subgroup, we did not find clear evidence that procedure type or receipt of acute reperfusion treatment was independently associated with in-hospital mortality. This may reflect the small sample size, which limits causal inference. Residual confounding and selection bias are likely, as patients chosen for reperfusion or admitted under surgical versus medical services may differ systematically in baseline risk.

Only 9% of the total cohort received acute treatment. This low rate could reflect multiple challenges to this population, including delayed stroke recognition, diagnostic uncertainty, recent procedures and treatment contraindications. In our cohort, among patients with documented reasons for withholding IVT, recent surgery and presentation beyond the treatment window were the most common factors. However, determining treatment eligibility retrospectively remains challenging, and the observed treatment rate may reflect both appropriate patient selection and potential treatment gaps. ^{Reference Ignacio, Abdalrahman and Kaveeta1}

Our results have important implications for the care of cardiac inpatients. First, they reinforce that in-hospital AIS in cardiac patients carries a high risk of death, highlighting the importance of prevention. These include optimizing atrial fibrillation detection and prompt resumption of anticoagulation after cardiac procedures. ^{Reference Nouh, Amin-Hanjani and Furie3,Reference Shu, Aziz and de Havenon6} For patients in whom atrial fibrillation is not identified during hospitalization, post-discharge rhythm monitoring such as repeat Holter or extended ambulatory ECG monitoring should be considered.Reference Spooner, Messé and Chaturvedi ²⁸ Second, strokes recognized later after the procedure also carried substantial mortality, emphasizing the need for sustained monitoring and vigilance across all cardiac wards and ICUs. In practice, neurological monitoring may include protocolized bedside neurological examinations during the immediate postoperative period for high-risk patients, minimization or structured interruption of sedation when feasible to enable meaningful assessment, and standardized screening tools such as BE-FAST (Balance, Eyes, Face drooping, Arm weakness, Speech difficulty and Time), Rapid Arterial oCclusion Evaluation (RACE) scale or a modified NIHSS to ensure consistent detection by nursing and physician staff. Emerging non-invasive technologies, such as wearable monitors and pressure-sensitive mattresses, may further aid early recognition in sedated or obtunded patients. ^{Reference Ignacio, Abdalrahman and Kaveeta1,Reference Nouh, Amin-Hanjani and Furie3,Reference Shu, Aziz and de Havenon6,Reference Ganesh29} Third, ICU admission after stroke was frequent, particularly following cardiac surgery and procedures. This likely reflects routine postoperative management and baseline critical care needs rather than stroke severity and should not be interpreted as a stroke-related outcome. Nevertheless, it highlights that a substantial proportion of in-hospital AIS in cardiac patients occurs in higher-acuity settings, which may inform monitoring practices and resource planning. ^{Reference Ignacio, Abdalrahman and Kaveeta1,Reference Shu, Aziz and de Havenon6}

This study has several strengths. This population-based cohort of in-hospital AIS across an entire provincial health system reflects real-world practice in stroke centers and non-stroke centers. Manual chart review further supplemented administrative data with stroke etiologies and verification of cardiac procedures. However, it also has several limitations. Selection bias and confounding by indication may have affected comparisons between patients with and without cardiac procedures, as these groups may represent distinct underlying pathologies and levels of clinical complexity that were not fully accounted for in our data. Its retrospective observational design precludes causal inference, and residual confounding is likely despite adjustment for key covariates. Additionally, anticoagulation status at the time of stroke onset was not available and could not be included for adjustment in the regression models. Administrative data may also misclassify diagnoses, procedures and complications. We only considered the most proximal procedure and did not assess the impact of multiple procedures on outcomes. Data on ICU admission were limited since we could not distinguish between routine post-procedure ICU care and ICU admission driven by clinical need or medical complexity. Finally, local organization of cardiac and stroke services may limit generalizability to other settings. Our study was conducted within a province-wide integrated stroke system with centralized stroke expertise at two comprehensive stroke centers and standardized care pathways. Differences in hospital resources, availability of stroke specialists, access to acute reperfusion therapies, postoperative monitoring practices and patient case mix across healthcare systems may influence both stroke recognition and outcomes. Therefore, replication in other settings with different models of cardiac and stroke care will be important to determine whether these findings are consistent across healthcare systems.

Future work should further characterize stroke related to cardiac conditions. NeuroVISION-2, a large international cohort study designed to elucidate mechanisms and risk factors for noncardiac perioperative stroke, provides a model that could be adapted to cardiac periprocedural stroke. ³⁰ Similar efforts are needed to clarify mechanisms, identify high-risk periods and determine modifiable factors in this population. Studies should also evaluate optimal care pathways for in-hospital AIS in cardiac patients (e.g., stroke unit or ICU vs ward care) and assess system performance, including time to recognition, treatment use and outcomes. ^{Reference Ignacio, Abdalrahman and Kaveeta1,Reference Nouh, Amin-Hanjani and Furie3,Reference Shu, Aziz and de Havenon6} Promising parallel research is exploring approaches to earlier neurological detection, such as structured bedside assessments paired with simple continuous monitoring tools (e.g., limb actigraphy and bed-pressure monitoring to flag asymmetric movement and trigger rapid imaging). ^{Reference Ignacio and Pensato31} A multifaceted research approach will be essential to advance our understanding of this special population.