A clinically actionable and explainable real-time risk assessment framework for stroke-associated pneumonia.

The current medical practice is more responsive rather than proactive, despite the widely recognized value of early disease detection, including improving the quality of care and reducing medical costs. One of the cornerstones of early disease detection is clinically actionable predictions, where predictions are expected to be accurate, stable, real-time and interpretable. As an example, we used stroke-associated pneumonia (SAP), setting up a transformer-encoder-based model that analyzes highly heterogeneous electronic health records in real-time. The model was proven accurate and stable on an independent test set. In addition, it issued at least one warning for 98.6 % of SAP patients, and on average, its alerts were ahead of physician diagnoses by 2.71 days. We applied Integrated Gradient to glean the model's reasoning process. Supplementing the risk scores, the model highlighted critical historical events on patients' trajectories, which were shown to have high clinical relevance.

Effect of the Shanghai Stroke Service System (4S) on the quality of stroke care and outcomes: A prospective quality improvement project.

BACKGROUND: In China, disparities in the quality of stroke care still exist and implementing quality improvement is still a challenge. AIM: The aim of the study was to determine whether the intervention by Shanghai Stroke Service System (4S) has helped improve adherence to stroke care guidelines and patient outcome. METHODS: The 4S is a regional stroke network with real-time data extraction among its 61 stroke centers in Shanghai. A total of 11 key performance indicators (KPIs) were evaluated. The primary outcomes were a composite measure and an all-or-none measure of adherence to 11 KPIs. The secondary outcomes were length of hospital stay and in-hospital mortality. RESULTS: The study enrolled 92,395 patients (mean age 69.0 ± 12.5 years, 65.2% men) with acute ischemic stroke hospitalized within 7 days of onset in Shanghai from January 2015 to December 2020. More patients received guideline recommended care between 2018 and 2020 than those between 2015 and 2017 (composite measure 87.1% vs 83.6%; absolute difference 2.9%, 95% confidence interval (CI) = [2.7%, 3.2%], p < 0.001; all-or-none measure 49.2% vs 44.8% patients; absolute difference 3.5%, 95% CI = [2.7%, 4.2%], p < 0.001). Further analysis of individual KPIs showed an absolute increase in six KPIs ranging from 3.4% to 8.9% (p < 0.001 for all comparisons). Compared with 2015-2017, hospital length of stay was shorter (10.95 vs 11.90 days; absolute difference -1.08, 95% CI = [-1.18, -0.99], p < 0.001), and in-hospital mortality was significantly reduced (risk ratio (RR) = 0.88, 95% CI = [0.79, 0.98], p = 0.01) in 2018-2020. CONCLUSION: The 4S intervention was associated with increased adherence to the stroke care guidelines, which further translated to improved clinical outcomes. TRIAL REGISTRATION: ClinicalTrials.gov identifier: NCT02735226.

Adherence to Stroke Care Performance Measures in Different Regions in China, 2015-2019: Evidence From the Chinese Stroke Center Alliance.

BACKGROUND AND OBJECTIVES: To explore the regional discrepancy of the adherence to guideline-recommended stroke interventions for the stroke belt division (north vs south), the economic development division (east vs middle vs west), and potential interaction. METHODS: We conducted a retrospective observational study using data from the Chinese Stroke Center Alliance from August 2015 to August 2019. The primary outcome was hospital personnel adherence to 11 individual guideline-recommended treatments. The coprimary outcomes included 2 summary measures: a composite score (range, 0 [nonadherence] to 1 [perfect adherence]) and an all-or-none binary outcome for adherence to evidence-based stroke. Regional disparities were assessed according to the stroke belt division and the economic development division and the interaction between these 2 divisions. Multivariate regression models with generalized estimating equations were used to analyze the outcomes. RESULTS: This study included 838,229 patients with acute ischemic stroke from 1,473 hospitals. The overall quality of care in the nonbelt regions (southern China) was higher than in the stroke belt regions (northern China), as reflected by a higher composite score (0.77 vs 0.75; adjusted odds ratio 1.03 [95% CI 1.02-1.04]; p < 0.001) and a higher all-or-none measure (25.5% vs 22.0%; 1.32 [1.17-1.49], p < 0.001). Patients in the East and Central had higher odds of using intravenous tissue-type plasminogen activator (East: 1.81 [95% CI 1.51-2.18], p < 0.001; Central: 1.57 [95% CI 1.26-1.95], p < 0.001), early antithrombotic medications (East: 1.77 [1.49-2.11], p < 0.001; Central: 1.37 [1.12-1.66], p < 0.001), lipid-lowering medications (East: 1.29 [1.08-1.53], p < 0.001), and deep vein thrombosis prophylaxis (East: 1.28 [1.08-1.50], p = 0.003) compared with those in the West. Patients in the nonbelt regions had higher odds of getting dysphagia screening (1.82 [1.55-2.13], p < 0.001) and rehabilitation assessment (which though varied among different economic development levels). Reflected by significant interaction effects, for patients in the East, those in the nonbelt regions had greater odds of receiving anticoagulation (1.62 [1.34-1.96]; p < 0.001) but lower odds of receiving antithrombotic (0.63 [0.52-0.77]; p < 0.001) and antidiabetic medications (0.87 [0.77-0.99]; p = 0.03); for patients in the West, those in the nonbelt regions were less likely to receive antihypertensive (0.64 [0.46-0.88]; p = 0.004) and antidiabetic (0.66 [0.54-0.81]; p < 0.001) medications. DISCUSSION: Stroke care performance measures differed across regions, along the stroke belt division, and the economic development division. The overall quality of care in the non-stroke belt regions was higher than that in the stroke belt regions. The 2 divisions had interaction effects on several individual measures.

Assisting scalable diagnosis automatically via CT images in the combat against COVID-19.

The pandemic of Coronavirus Disease 2019 (COVID-19) is causing enormous loss of life globally. Prompt case identification is critical. The reference method is the real-time reverse transcription PCR (RT-PCR) assay, whose limitations may curb its prompt large-scale application. COVID-19 manifests with chest computed tomography (CT) abnormalities, some even before the onset of symptoms. We tested the hypothesis that the application of deep learning (DL) to 3D CT images could help identify COVID-19 infections. Using data from 920 COVID-19 and 1,073 non-COVID-19 pneumonia patients, we developed a modified DenseNet-264 model, COVIDNet, to classify CT images to either class. When tested on an independent set of 233 COVID-19 and 289 non-COVID-19 pneumonia patients, COVIDNet achieved an accuracy rate of 94.3% and an area under the curve of 0.98. As of March 23, 2020, the COVIDNet system had been used 11,966 times with a sensitivity of 91.12% and a specificity of 88.50% in six hospitals with PCR confirmation. Application of DL to CT images may improve both efficiency and capacity of case detection and long-term surveillance.

Ambulatory blood pressure profile and stroke recurrence.

OBJECTIVES: To establish a new ambulatory blood pressure (ABP) parameter (24-hour ABP profile) and evaluated its performance on stroke outcome in ischaemic stroke (IS) or transient ischaemic attack (TIA) patients. METHODS: The prospective cohort consisted of 1996 IS/TIA patients enrolled for ABP monitoring and a 3-month follow-up for stroke recurrence as outcome. Profile groups of systolic blood pressure (SBP) were identified via an advanced functional clustering method, and the associations of the profile groups and conventional ABP parameters with stroke recurrence were examined in a Cox proportional hazards model. RESULTS: Three discrete profile groups (n=604, 781 and 611 in profiles 1, 2 and 3, respectively) in 24-hour ambulatory SBP were identified. Profile 1 resembled most to the normal diurnal blood pressure pattern; profile 2 also dropped at night, but climbed earlier and with higher morning surge; while profile 3 had sustained higher nocturnal SBP without significant nocturnal SBP decline. The incidence of stroke recurrence was 2.9%, 3.9% and 5.5% in profiles 1, 2 and 3, respectively. After adjustment for covariates, profile 3 was significantly associated with higher risk of stroke recurrence with profile 1 as reference (HR 1.76, 95% CI: 1.00 to 3.09), while no significant difference was observed between profiles 2 and 1 (HR 1.22, 95% CI: 0.66 to 2.25). None of conventional ABP parameters showed significant associations with the outcome. CONCLUSIONS: Ambulatory 24-hour SBP profile is associated with short-term stroke recurrence. Profiles of ABP may help improve identification of stroke recurrence by capturing the additive effects of individual ABP parameters.

GRP per capita and hospital characteristics associated with intravenous tissue plasminogen activator adherence rate: evidence from the Chinese Stroke Center Alliance.

BACKGROUND: Timely delivery of intravenous tissue plasminogen activator (IV-rt PA) is pivotal to eligible patients who had a stroke while achieving higher rates of IV-rt PA has been problematic. This paper focuses on investigating influential factors associated with the administration of IV-rt PA, primarily per capita gross regional product (GRP) and healthcare system factors. METHODS: The study included 980 hospitals in the Chinese Stroke Center Alliance where 158 003 patients who had an acute ischaemic stroke received IV-rt PA between August 2015 and August 2019. The adherence rate to IV-rt PA within 4.5 hours time window in each hospital was the primary outcome. Influential factors were grouped into two categories: macroeconomic status and hospital characteristics. The outcome was analysed using multivariable linear regression. RESULTS: GRP per capita (ß=2.37, p<0.001), hospital stroke centre certification (ß=3.77, p<0.001), number of neurologists (ß=0.12, p<0.001), existence of emergency services for neurological treatment (ß=7.43, p=0.014), presence of emergency department (ß=10.03, p=0.019) and cooperating with emergency centre (ß=4.65, p=0.029) were significantly positively associated with the adherence rate to IV-rt PA. CONCLUSIONS: Higher GRP per capita, affluent neurological personnel, well-equipped emergency services for neurological treatment and routine cooperation with the emergency centre were important for enhancing the adherence rate to IV-rt PA among patients who had an acute ischaemic stroke in China.

Sampling Properties of color Independent Component Analysis.

Independent Component Analysis (ICA) offers an effective data-driven approach for blind source extraction encountered in many signal and image processing problems. Although many ICA methods have been developed, they have received relatively little attention in the statistics literature, especially in terms of rigorous theoretical investigation for statistical inference. The current paper aims at narrowing this gap and investigates the statistical sampling properties of the colorICA (cICA) method. The cICA incorporates the correlation structure within sources through parametric time series models in the frequency domain and outperforms several existing ICA alternatives numerically. We establish the consistency and asymptotic normality of the cICA estimates, which then enables statistical inference based on the estimates. These asymptotic properties are further validated using simulation studies.

Who may benefit from lower dosages of intravenous tissue plasminogen activator? Results from a cluster data analysis.

BACKGROUND: The risk of symptomatic intracranial haemorrhage (sICH) after thrombolysis is low but severe. Lower dose of alteplase may reduce the risk of sICH. We aim to identify subsets of patients who could benefit from lower dose of alteplase compared with standard dose. METHODS: Data from two observational registries were pooled together. A total of 3479 patients who had acute ischaemic stroke were entered into the interaction tree model. The response variable was the rate of sICH per the definition of the National Institute of Neurological Disorders and Stroke Study. Clinical improvement was measured by the National Institutes of Health Stroke Scale (NIHSS) and defined as NIHSS 0 or 1 or an improvement of more than 4 points (within 7 days or at discharge). Rare event logistic regression was performed to analyse the OR of safety outcome. RESULTS: To optimise the interaction effect between tissue plasminogen activator (tPA) dosage (standard/lower) and patient subgroups, three subgroups based on the severity of stroke were identified: (1) NIHSS ≤4, (2) NIHSS between 5 and 14, and (3) NIHSS ≥15. The estimated difference of OR of having sICH was 2.71 (95% CI 0.80 to 7.69, p=0.10) for mild, 0.13 (95% CI 0.02 to 0.68, p=0.01) for moderate, and 0.65 (95% CI 0.19 to 2.55, p=0.52) for severe, respectively. In addition, patients who had moderate stroke treated with lower dose had comparable efficacy outcome (OR 1.23, 95% CI 0.71 to 2.13, p=0.45). CONCLUSION: Our analysis demonstrated that in patients who had moderate stroke, lower doses of alteplase are associated with significant sICH reduction and non-inferior performance in efficacy, compared with those in the standard dose group. TRIAL REGISTRATION NUMBER: The TIMS-China was a national prospective stroke registry on thrombolytic therapy using intravenous tPA in patients who had acute ischaemic stroke. The results were initially published in 2012 without a clinical trial registration number. The Shanghai Stroke Service System was registered at www.clinicaltrial.gov (NCT02735226).

Identification of Long Non-Coding RNA Expression Profiles and Co-Expression Genes in Thyroid Carcinoma Based on The Cancer Genome Atlas (TCGA) Database.

BACKGROUND Thyroid carcinoma is a malignancy with high morbidity and mortality. Genetic alterations play pivot roles in the pathogenesis of thyroid carcinoma, where long noncoding RNA (lncRNA) have been identified to be crucial. This study sought to investigate the biological functions of lncRNA expression profiles in thyroid carcinoma. MATERIAL AND METHODS The lncRNAs expression profiles were acquired from The Cancer Genome Atlas (TCGA) database according to 510 thyroid cancer tissues and 58 normal thyroid tissues. By using R package edgeR, differentially expressed RNAs were obtained. Also, an overall survival model was established based on Cox regression and clinical data then testified by Kaplan-Meier plot, receiver operating characteristic (ROC)-curve and C-index analysis. We investigated the co-expressed genes with lncRNAs involved in the prognostic model, as well as Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis was conducted R package clusterProfile. RESULTS A total of 352 lncRNAs were identified as differentially expressed in thyroid carcinoma, and an overall survival model consisting of 8 signature lncRNAs was proposed (ROC=0.862, C-index=0.893, P<0.05), 3 of which (DOCK9-DT, FAM111A-DT, and LINC01736) represent co-expressed mRNAs. However, as an oncogene, only FAM111A-DT increased the prognostic risk in thyroid carcinoma. Furthermore, we found differential genes LINC01016, LHX1-DT, IGF2-AS, ND MIR1-1HG-AS1, significantly related to lymph node metastasis (P<0.05). CONCLUSIONS In this study, we clarified the differential lncRNA expression profiles which were related to the tumorigenesis and prognosis in thyroid carcinoma. Our results provide new rationale and understandings to the pathogenesis and regulatory mechanisms of thyroid carcinoma.

The efficacy and safety of nimodipine in acute ischemic stroke patients with mild cognitive impairment: a double-blind, randomized, placebo-controlled trial.

Nimodipine might be effective in subcortical vascular dementia (VaD). Its benefit in preventing further cognitive decline in patients with acute ischemic stroke (AIS) and vascular mild cognitive impairment (VaMCI) remains to be established. In this multicenter, double-blind trial, we randomly assigned 654 eligible patients to nimodipine 30 mg three times a day or placebo. The primary outcome was any cognitive decline defined by the changes on the Mini-Mental State Examination (ΔMMSE ≤ -3) or vascular AD assessment scale cognitive subscale (ΔADAS-cog ≥ 4) at 6 months. Secondary outcomes included any distribution shift of ΔADAS-cog, ΔMMSE or cognitive improvement defined by ΔADAS-cog ≤ -2, or ΔMMSE ≥ 0. The primary outcome in the nimodipine group and placebo group were similar for ΔMMSE ≤ -3 (4.18% and 7.22%, respectively, P = 0.15) and ΔADAS-cog ≥ 4 (8.36% and 8.93% respectively, P = 0.88). The distribution shift of ΔADAS-cog and ΔMMSE differed significantly between the two groups (P = 0.03 and P = 0.05 respectively). Cognitive improvement occurred in 55.4% in the nimodipine group and 43.6% in the placebo group measured by ΔADAS-cog ≤ -2 (Odds Ratio, 1.54; 95% confidence interval [CI] 1.10-2.14, P < 0.01) or 84.0% and 74.6% respectively by ΔMMSE ≥ 0 (Odds Ratio, 1.79; 95% CI 1.18-2.70, P < 0.01). Nimodipine was associated with better cognitive function in the memory domain. The adverse events rate was similar in two groups. This study is registered with ClinicalTrials.gov, NCT01220622. Nimodipine did not show benefit to prevent cognitive decline in AIS patients with VaMCI, but improved cognition moderately, especially measured in the memory domain.

A survey of high dimension low sample size asymptotics.

Peter Hall's work illuminated many aspects of statistical thought, some of which are very well known including the bootstrap and smoothing. However, he also explored many other lesser known aspects of mathematical statistics. This is a survey of one of those areas, initiated by a seminal paper in 2005, on high dimension low sample size asymptotics. An interesting characteristic of that first paper, and of many of the following papers, is that they contain deep and insightful concepts which are frequently surprising and counter-intuitive, yet have mathematical underpinnings which tend to be direct and not difficult to prove.

Effect of a Multifaceted Quality Improvement Intervention on Hospital Personnel Adherence to Performance Measures in Patients With Acute Ischemic Stroke in China: A Randomized Clinical Trial.

Importance: In China and other parts of the world, hospital personnel adherence to evidence-based stroke care is limited. Objective: To determine whether a multifaceted quality improvement intervention can improve hospital personnel adherence to evidence-based performance measures in patients with acute ischemic stroke (AIS) in China. Design, Setting, and Participants: A multicenter, cluster-randomized clinical trial among 40 public hospitals in China that enrolled 4800 patients hospitalized with AIS from August 10, 2014, through June 20, 2015, with 12-month follow-up through July 30, 2016. Interventions: Twenty hospitals received a multifaceted quality improvement intervention (intervention group; 2400 patients), including a clinical pathway, care protocols, quality coordinator oversight, and performance measure monitoring and feedback. Twenty hospitals participated in the stroke registry with usual care (control group; 2400 patients). Main Outcomes and Measures: The primary outcome was hospital personnel adherence to 9 AIS performance measures, with co-primary outcomes of a composite of percentage of performance measures adhered to, and as all-or-none. Secondary outcomes included in-hospital mortality and long-term outcomes (a new vascular event, disability [modified Rankin Scale score, 3-5], and all-cause mortality) at 3, 6, and 12 months. Results: Among 4800 patients with AIS enrolled from 40 hospitals and randomized (mean age, 65 years; women, 1757 [36.6%]), 3980 patients (82.9%) completed the 12-month follow-up of the trial. Patients in intervention group were more likely to receive performance measures than those in the control groups (composite measure, 88.2% vs 84.8%, respectively; absolute difference, 3.54% [95% CI, 0.68% to 6.40%], P = .02). The all-or-none measure did not significantly differ between the intervention and control groups (53.8% vs 47.8%, respectively; absolute difference, 6.69% [95% CI, -0.41% to 13.79%], P = .06). New clinical vascular events were significantly reduced in the intervention group compared with the control group at 3 months (3.9% vs 5.3%, respectively; difference, -2.03% [95% CI, -3.51% to -0.55%]; P = .007), 6 months (6.3% vs 7.8%, respectively; difference, -2.18% [95% CI, -4.0% to -0.35%]; P = .02) and 12 months (9.1% vs 11.8%, respectively; difference, -3.13% [95% CI, -5.28% to -0.97%]; P = .005). Conclusions and Relevance: Among 40 hospitals in China, a multifaceted quality improvement intervention compared with usual care resulted in a statistically significant but small improvement in hospital personnel adherence to evidence-based performance measures in patients with acute ischemic stroke when assessed as a composite measure, but not as an all-or-none measure. Further research is needed to understand the generalizability of these findings. Trial Registration: ClinicalTrials.gov Identifier: NCT02212912.

Exponential Family Functional data analysis via a low-rank model.

In many applications, non-Gaussian data such as binary or count are observed over a continuous domain and there exists a smooth underlying structure for describing such data. We develop a new functional data method to deal with this kind of data when the data are regularly spaced on the continuous domain. Our method, referred to as Exponential Family Functional Principal Component Analysis (EFPCA), assumes the data are generated from an exponential family distribution, and the matrix of the canonical parameters has a low-rank structure. The proposed method flexibly accommodates not only the standard one-way functional data, but also two-way (or bivariate) functional data. In addition, we introduce a new cross validation method for estimating the latent rank of a generalized data matrix. We demonstrate the efficacy of the proposed methods using a comprehensive simulation study. The proposed method is also applied to a real application of the UK mortality study, where data are binomially distributed and two-way functional across age groups and calendar years. The results offer novel insights into the underlying mortality pattern.

The network of Shanghai Stroke Service System (4S): A public health-care web-based database using automatic extraction of electronic medical records.

Background Several stroke outcome and quality control projects have demonstrated the success in stroke care quality improvement through structured process. However, Chinese health-care systems are challenged with its overwhelming numbers of patients, limited resources, and large regional disparities. Aim To improve quality of stroke care to address regional disparities through process improvement. Method and design The Shanghai Stroke Service System (4S) is established as a regional network for stroke care quality improvement in the Shanghai metropolitan area. The 4S registry uses a web-based database that automatically extracts data from structured electronic medical records. Site-specific education and training program will be designed and administrated according to their baseline characteristics. Both acute reperfusion therapies including thrombectomy and thrombolysis in the acute phase and subsequent care were measured and monitored with feedback. Primary outcome is to evaluate the differences in quality metrics between baseline characteristics (including rate of thrombolysis in acute stroke and key performance indicators in secondary prevention) and post-intervention. Conclusions The 4S system is a regional stroke network that monitors the ongoing stroke care quality in Shanghai. This project will provide the opportunity to evaluate the spectrum of acute stroke care and design quality improvement processes for better stroke care. A regional stroke network model for quality improvement will be explored and might be expanded to other large cities in China. Clinical Trial Registration-URL http://www.clinicaltrials.gov . Unique identifier: NCT02735226.

Length of Stay, Mortality, Cost, and Perceptions of Care Associated With Transition From an Open to Closed Staffing Model in the Cardiac Intensive Care Unit.

BACKGROUND: Organizational models in the intensive care unit (ICU) have classically been described as either closed or open, depending on the presence or absence of a dedicated ICU team. Although a closed model has been shown to improve patient outcomes in medical and surgical ICUs, the merits of various care models have not been previously explored in the cardiac ICU (CICU) setting. METHODS: From November 2012 to March 2014, data were prospectively collected on all admissions before and after transition from an open to closed CICU at our institution. Baseline clinical variables, illness severity, admission and discharge diagnoses, resource use, and outcomes were recorded. Anonymous surveys were also collected from nursing and resident trainee participants to evaluate the influence of unit structure on perceptions of care. Descriptive statistics were used, and logistic regression modeling was performed to examine the impact of unit structure on mortality. RESULTS: The study consisted of 670 patients, 332 (49.6%) of whom were admitted to the open CICU model and 338 (50.4%) of whom were admitted to the closed model. Neither CICU nor hospital mortality differed between the open and closed units, though length of stay was shorter in the closed CICU. Additionally, nurses and resident trainees reported that the closed CICU allowed for better communication, collaboration, and education. CONCLUSIONS: Although there was no significant impact of unit structure on patient outcomes in this single-center study, the closed CICU model was associated with better perceptions of care.

Artificial intelligence in healthcare: past, present and future.

Artificial intelligence (AI) aims to mimic human cognitive functions. It is bringing a paradigm shift to healthcare, powered by increasing availability of healthcare data and rapid progress of analytics techniques. We survey the current status of AI applications in healthcare and discuss its future. AI can be applied to various types of healthcare data (structured and unstructured). Popular AI techniques include machine learning methods for structured data, such as the classical support vector machine and neural network, and the modern deep learning, as well as natural language processing for unstructured data. Major disease areas that use AI tools include cancer, neurology and cardiology. We then review in more details the AI applications in stroke, in the three major areas of early detection and diagnosis, treatment, as well as outcome prediction and prognosis evaluation. We conclude with discussion about pioneer AI systems, such as IBM Watson, and hurdles for real-life deployment of AI.

The Statistics and Mathematics of High Dimension Low Sample Size Asymptotics.

The aim of this paper is to establish several deep theoretical properties of principal component analysis for multiple-component spike covariance models. Our new results reveal an asymptotic conical structure in critical sample eigendirections under the spike models with distinguishable (or indistinguishable) eigenvalues, when the sample size and/or the number of variables (or dimension) tend to infinity. The consistency of the sample eigenvectors relative to their population counterparts is determined by the ratio between the dimension and the product of the sample size with the spike size. When this ratio converges to a nonzero constant, the sample eigenvector converges to a cone, with a certain angle to its corresponding population eigenvector. In the High Dimension, Low Sample Size case, the angle between the sample eigenvector and its population counterpart converges to a limiting distribution. Several generalizations of the multi-spike covariance models are also explored, and additional theoretical results are presented.

Substantial Progress Yet Significant Opportunity for Improvement in Stroke Care in China.

BACKGROUND AND PURPOSE: Stroke is a leading cause of death in China. Yet the adherence to guideline-recommended ischemic stroke performance metrics in the past decade has been previously shown to be suboptimal. Since then, several nationwide stroke quality management initiatives have been conducted in China. We sought to determine whether adherence had improved since then. METHODS: Data were obtained from the 2 phases of China National Stroke Registries, which included 131 hospitals (12 173 patients with acute ischemic stroke) in China National Stroke Registries phase 1 from 2007 to 2008 versus 219 hospitals (19 604 patients) in China National Stroke Registries phase 2 from 2012 to 2013. Multiple regression models were developed to evaluate the difference in adherence to performance measure between the 2 study periods. RESULTS: The overall quality of care has improved over time, as reflected by the higher composite score of 0.76 in 2012 to 2013 versus 0.63 in 2007 to 2008. Nine of 13 individual performance metrics improved. However, there were no significant improvements in the rates of intravenous thrombolytic therapy and anticoagulation for atrial fibrillation. After multivariate analysis, there remained a significant 1.17-fold (95% confidence interval, 1.14-1.21) increase in the odds of delivering evidence-based performance metrics in the more recent time periods versus older data. The performance metrics with the most significantly increased odds included stroke education, dysphagia screening, smoking cessation, and antithrombotics at discharge. CONCLUSIONS: Adherence to stroke performance metrics has increased over time, but significant opportunities remain for further improvement. Continuous stroke quality improvement program should be developed as a national priority in China.

A SPARSE REDUCED RANK FRAMEWORK FOR GROUP ANALYSIS OF FUNCTIONAL NEUROIMAGING DATA.

In spatial-temporal neuroimaging studies, there is an evolving literature on the analysis of functional imaging data in order to learn the intrinsic functional connectivity patterns among different brain regions. However, there are only few efficient approaches for integrating functional connectivity pattern across subjects, while accounting for spatial-temporal functional variation across multiple groups of subjects. The objective of this paper is to develop a new sparse reduced rank (SRR) modeling framework for carrying out functional connectivity analysis across multiple groups of subjects in the frequency domain. Our new framework not only can extract both frequency and spatial factors across subjects, but also imposes sparse constraints on the frequency factors. It thus leads to the identification of important frequencies with high power spectra. In addition, we propose two novel adaptive criteria for automatic selection of sparsity level and model rank. Using simulated data, we demonstrate that SRR outperforms several existing methods. Finally, we apply SRR to detect group differences between controls and two subtypes of attention deficit hyperactivity disorder (ADHD) patients, through analyzing the ADHD-200 data.