Developing machine learning models to predict multi-class functional outcomes and death three months after stroke in Sweden.

Globally, stroke is the third-leading cause of mortality and disability combined, and one of the costliest diseases in society. More accurate predictions of stroke outcomes can guide healthcare organizations in allocating appropriate resources to improve care and reduce both the economic and social burden of the disease. We aim to develop and evaluate the performance and explainability of three supervised machine learning models and the traditional multinomial logistic regression (mLR) in predicting functional dependence and death three months after stroke, using routinely-collected data. This prognostic study included adult patients, registered in the Swedish Stroke Registry (Riksstroke) from 2015 to 2020. Riksstroke contains information on stroke care and outcomes among patients treated in hospitals in Sweden. Prognostic factors (features) included demographic characteristics, pre-stroke functional status, cardiovascular risk factors, medications, acute care, stroke type, and severity. The outcome was measured using the modified Rankin Scale at three months after stroke (a scale of 0-2 indicates independent, 3-5 dependent, and 6 dead). Outcome prediction models included support vector machines, artificial neural networks (ANN), eXtreme Gradient Boosting (XGBoost), and mLR. The models were trained and evaluated on 75% and 25% of the dataset, respectively. Model predictions were explained using SHAP values. The study included 102,135 patients (85.8% ischemic stroke, 53.3% male, mean age 75.8 years, and median NIHSS of 3). All models demonstrated similar overall accuracy (69%-70%). The ANN and XGBoost models performed significantly better than the mLR in classifying dependence with F1-scores of 0.603 (95% CI; 0.594-0.611) and 0.577 (95% CI; 0.568-0.586), versus 0.544 (95% CI; 0.545-0.563) for the mLR model. The factors that contributed most to the predictions were expectedly similar in the models, based on clinical knowledge. Our ANN and XGBoost models showed a modest improvement in prediction performance and explainability compared to mLR using routinely-collected data. Their improved ability to predict functional dependence may be of particular importance for the planning and organization of acute stroke care and rehabilitation.

Mediation Analyses of the Mechanisms by Which Socioeconomic Status, Comorbidity, Stroke Severity, and Acute Care Influence Stroke Outcome.

BACKGROUND AND OBJECTIVES: Low socioeconomic status (SES) is associated with increased risk of death and disability after stroke, but interventional targets to minimize disparities remain unclear. We aim to assess the extent to which SES-based disparities in the association between low SES and death and dependency at 3 months after stroke could be eliminated by offsetting differences in comorbidity, stroke severity, and acute care. METHODS: This nationwide register-based cohort study included all 72 hospitals caring for patients with acute stroke in Sweden. All patients registered with an acute ischemic stroke in the Swedish Stroke Register in 2015-2016 who were independent in activities of daily living (ADL) during stroke were included. Data on survival and SES the year before stroke were retrieved by cross-linkage with other national registers. SES was defined by education and income and categorized into low, mid, and high. Causal mediation analysis was used to study the absolute risk of death and ADL dependency at 3 months depending on SES and to what extent hypothetical interventions on comorbidities, stroke severity, and acute care would equalize outcomes. RESULTS: Of the 25,846 patients in the study, 6,798 (26.3%) were dead or ADL dependent 3 months after stroke. Adjusted for sex and age, low SES was associated with an increased absolute risk of 5.4% (95% CI 3.9%-6.9%; p < 0.001) compared with mid SES and 10.1% (95% CI 8.1%-12.2%; p < 0.001) compared with high SES. Intervening to shift the distribution of all mediators among patients with low SES to those of the more privileged groups would result in absolute reductions of these effects by 2.2% (95% CI 1.2%-3.2%; p < 0.001) and 4.0% (95% CI 2.6%-5.5%; p < 0.001), respectively, with the largest reduction accomplished by equalizing stroke severity. DISCUSSION: Low SES patients have substantially increased risks of death and ADL dependency 3 months after stroke compared with more privileged patient groups. This study suggests that if we could intervene to equalize SES-related differences in the distributions of comorbidity, acute care, and stroke severity, up to 40 of every 1,000 patients with low SES could be prevented from dying or becoming ADL dependent.

Death and ADL Dependency After Scoring Zero on the NIHSS: A Swedish Retrospective Registry-Based Study.

Background and Objectives: Of all strokes, mild strokes (defined as 5 points or less on the National Institutes of Health Stroke Scale [NIHSS]) are in the majority. However, up to one-third of patients with mild strokes still exhibit significant deficits 3 months after the stroke. Studies on the presumably mildest strokes, defined by zero points on the NIHSS (0-NIHSS) at admission, are scarce. Hence, we aimed to study patient characteristics and outcomes among patients with 0-NIHSS strokes. Methods: Our retrospective registry-based study included a total of 6,491 adult patients with stroke admitted to 3 different stroke units in Gothenburg, Sweden, from November 2014 to June 2019. Our main outcome was a composite measure including death and activities of daily living (ADL) dependency 3 months after the stroke. Analyses of patient characteristics were followed by adjusted analyses including multiple confounders. Results: In total, 5,945 patients had data on NIHSS at admission, of whom 1,412 (24%) presented with a 0-NIHSS stroke. Among these, the median age was 72 years, 600 (42%) were female, and 86 (6%) had a hemorrhagic stroke. Among previously ADL-independent patients, 65 (6%) were either dead or ADL-dependent 3 months after the stroke. Prestroke physical inactivity (OR 2.48, 95% CI 1.40-4.38) and age (OR 1.05 per gained year, 95% CI 1.02-1.08) significantly increased the risk of death and ADL dependency. Discussion: One of 17 patients has either died or become ADL-dependent 3 months after a 0-NIHSS stroke, stressing that these strokes are not always benign. Older and physically inactive patients are at greater risk of an adverse outcome.

Characteristics of in-hospital stroke patients in Sweden: A nationwide register-based study.

INTRODUCTION: Few studies have reported the characteristics of patients with in-hospital stroke (IHS) including the reason for hospitalization and invasive procedures before the stroke. We aimed to extend current knowledge. PATIENTS AND METHODS: All adult patients with IHS in Sweden during 2010-2019 registered in the Swedish Stroke Register (Riksstroke) were included. The cohort was cross-linked to the National Patient Register and data extracted on background diagnoses, main discharge diagnoses, and procedure codes for the hospitalization when IHS occurred and any hospital-based healthcare contacts within 30 days before IHS. RESULTS: 231,402 stroke cases were identified of which 12,551 (5.4%) were in-hospital and had corresponding entries in the National Patient Register. Of the IHS patients, 11,420 (91.0%) had ischemic stroke and 1131 (9.0%) hemorrhagic stroke; 5860 (46.7%) of the IHS patients had at least one invasive procedure prior to ictus. 1696 (13.5%) had a cardiovascular procedure and 560 (4.5%) a neurosurgical procedure. 1319 (10.5%) patients only had minimally invasive procedures such as blood product transfusion, hemodialysis, or central line insertion. Common discharge diagnosis in patients with no invasive procedures were cardiovascular disorders, injuries, and respiratory disorders. DISCUSSION AND CONCLUSION: One in every 17 strokes in Sweden occur in a hospital. In this unselected large cohort the previously reported major causes for in-hospital stroke, cardiovascular and neurosurgical procedures, preceded IHS in only 18.0% of cases suggesting that other etiologies are more common than previously reported. Future studies should aim at determining absolute risks of stroke after surgical procedures and ways of risk reduction.

Socioeconomic status and stroke severity: Understanding indirect effects via risk factors and stroke prevention using innovative statistical methods for mediation analysis.

BACKGROUND: Those with low socioeconomic status have an increased risk of stroke, more severe strokes, reduced access to treatment, and more adverse outcomes after stroke. The question is why these differences are present. In this study we investigate to which extent the association between low socioeconomic status and stroke severity can be explained by differences in risk factors and stroke prevention drugs. METHODS: The study included 86 316 patients registered with an ischemic stroke in the Swedish Stroke Register (Riksstroke) 2012-2016. Data on socioeconomic status was retrieved from the Longitudinal integrated database for health insurance and labour market studies (LISA) by individual linkage. We used education level as proxy for socioeconomic status, with primary school education classified as low education. Stroke severity was measured using the Reaction Level Scale, with values above 1 classified as severe strokes. To investigate the pathways via risk factors and stroke prevention drugs we performed a mediation analysis estimating indirect and direct effects. RESULTS: Low education was associated with an excess risk of a severe stroke compared to mid/high education (absolute risk difference 1.4%, 95% CI: 1.0%-1.8%), adjusting for confounders. Of this association 28.5% was an indirect effect via risk factors (absolute risk difference 0.4%, 95% CI: 0.3%-0.5%), while the indirect effect via stroke prevention drugs was negligible. CONCLUSION: Almost one third of the association between low education and severe stroke was explained by risk factors, and clinical effort should be taken to reduce these risk factors to decrease stroke severity among those with low socioeconomic status.

In-Hospital Delays in Stroke Thrombolysis: Every Minute Counts.

BACKGROUND AND PURPOSE: Intravenous thrombolysis is a well-established treatment for acute ischemic stroke. Our aim was to quantify the effect of each minute delay in door-to-needle time (DNT) on 90-day survival, intracerebral hemorrhagic complication <36 hours, and functional outcomes at 3 months, in routine clinical practice. METHODS: Our nationwide registry-based study included 14 132 adult patient admissions with ischemic stroke receiving intravenous thrombolysis from 2010 to 2017. Outcomes were analyzed using multivariable logistic regression, adjusting for potential confounders. RESULTS: Median DNT was 47 minutes, with an improvement from 65 to 38 minutes during the study. Median age was 74 years, and median National Institutes of Health Stroke Scale 8 points. We found a significant impact of each minute delay in DNT with reduced odds of survival by 0.6%, increased odds of intracerebral hemorrhagic and worse activities of daily living by 0.3%, and worse living conditions and mobility by 0.4%. CONCLUSIONS: Improving DNT is a key factor in achieving good outcomes after stroke. We estimate that in Sweden alone in 2017, compared with 2010, the shorter DNT achieved have saved 38 lives, avoided 8 intracerebral hemorrhagic transformations, and spared, respectively, 36, 51, and 52 patients from a worsening in activities of daily living, living conditions, and mobility. DNT is sensitive for interventions and should be targeted in quality improvement efforts.

Time Trends and Monthly Variation in Swedish Acute Stroke Care.

Background and Purpose: Studies of monthly variation in acute stroke care have led to conflicting results. Our objective was to study monthly variation and longitudinal trends in quality of care and patient survival following acute stroke. Methods: Our nationwide study included all adult patients (≥18 years) with acute stroke (ischemic or hemorrhagic), admitted to Swedish hospitals from 2011 to 2016, and that were registered in The Swedish Stroke Register (Riksstroke). We studied how month of admission and longitudinal trends affected acute stroke care and survival. We also studied resilience to this variation among hospitals with different levels of specialization. Results: We included 132,744 stroke admissions. The 90-day survival was highest in May and lowest in January (84.1 vs. 81.5%). Thrombolysis rates and door-to-needle time within 30 min increased from 2011 to 2016 (respectively, 7.3 vs. 12.8% and 7.7 vs. 28.7%). Admission to a stroke unit as first destination of hospital care was lowest in January and highest in June (78.3 vs. 80.5%). Stroke unit admission rates decreased in university hospitals from 2011 to 2016 (83.4 vs. 73.9%), while no such trend were observed in less specialized hospitals. All the differences above remained significant (p < 0.05) after adjustment for possible confounding factors. Conclusion: We found that month of admission and longitudinal trends both affect quality of care and survival of stroke patients in Sweden, and that the effects differ between hospital types. The observed variation suggests an opportunity to improve stroke care in Sweden. Future studies ought to focus on identifying the specific factors driving this variation, for subsequent targeting by quality improvement efforts.

Diurnal variations in the quality of stroke care in Sweden.

OBJECTIVES: A recent study of acute stroke patients in England and Wales revealed several patterns of temporal variation in quality of care. We hypothesized that similar patterns would be present in Sweden and aimed to describe these patterns. Additionally, we aimed to investigate whether hospital type conferred resilience against temporal variation. MATERIALS AND METHODS: We conducted this nationwide registry-based study using data from the Swedish Stroke Register (Riksstroke) including all adult patients registered with acute stroke between 2011 and 2015. Outcomes included process measures and survival. We modeled time of presentation as on/off-hours, shifts, day of week, 4-hour, and 12-hour time blocks. We studied hospital resilience by comparing outcomes across hospital types. RESULTS: A total of 113 862 stroke events in 72 hospitals were included. The process indicators and survival all showed significant temporal variation. Door-to-needle (DTN) time within 30 minutes was less likely during nighttime than daytime (OR 0.50; 95% CI 0.41-0.60). Patients admitted during off-hours had lower odds of direct stroke unit (SU) admission (OR 0.72; 95% CI 0.70-0.75). 30-day survival was lower in nighttime vs daytime presentations (OR 0.90, 95% CI 0.84-0.96). The effects of temporal variation differed significantly between hospital types for DTN time within 30 minutes and direct SU admission where university hospitals were more resilient than specialized non-university hospitals. CONCLUSIONS: Our study shows that variation in quality of care and survival is present throughout the whole week. We also found that university hospitals were more resilient to temporal variation than specialized non-university hospitals.

Patients with acute stroke are less likely to be admitted directly to a stroke unit when hospital beds are scarce: A Swedish multicenter register study.

INTRODUCTION: It is well established that managing patients with acute stroke in dedicated stroke units is associated with improved functioning and survival. The objectives of this study are to investigate whether patients with acute stroke are less likely to be directly admitted to a stroke unit from the Emergency Department when hospital beds are scarce and to measure variation across hospitals in terms of this outcome. PATIENTS AND METHODS: This register study comprised data on patients with acute stroke admitted to 14 out of 72 Swedish hospitals in 2011-2014. Data from the Swedish stroke register were linked to administrative daily data on hospital bed occupancy (measured at 6 a.m.). Logistic regression analysis was used to analyse the association between bed occupancy and direct stroke unit admission. RESULTS: A total of 13,955 hospital admissions were included; 79.6% were directly admitted to a stroke unit from the Emergency Department. Each percentage increase in hospital bed occupancy was associated with a 1.5% decrease in odds of direct admission to a stroke unit (odds ratio = 0.985, 95% confidence interval = 0.978-0.992). The best-performing hospital exhibited an odds ratio of 3.8 (95% confidence interval = 2.6-5.5) for direct admission to a stroke unit versus the reference hospital. DISCUSSION AND CONCLUSION: We found an association between hospital crowding and reduced quality of care in acute stroke, portrayed by a lower likelihood of patients being directly admitted to a stroke unit from the Emergency Department. The magnitude of the effect varied considerably across hospitals.