Detecting bots in social-networks using node and structural embeddings.

Users on social networks such as Twitter interact with each other without much knowledge of the real-identity behind the accounts they interact with. This anonymity has created a perfect environment for bot accounts to influence the network by mimicking real-user behaviour. Although not all bot accounts have malicious intent, identifying bot accounts in general is an important and difficult task. In the literature there are three distinct types of feature sets one could use for building machine learning models for classifying bot accounts. These feature-sets are: user profile metadata, natural language features (NLP) extracted from user tweets and finally features extracted from the the underlying social network. Profile metadata and NLP features are typically explored in detail in the bot-detection literature. At the same time less attention has been given to the predictive power of features that can be extracted from the underlying network structure. To fill this gap we explore and compare two classes of embedding algorithms that can be used to take advantage of information that network structure provides. The first class are classical embedding techniques, which focus on learning proximity information. The second class are structural embedding algorithms, which capture the local structure of node neighbourhood. We show that features created using structural embeddings have higher predictive power when it comes to bot detection. This supports the hypothesis that the local social network formed around bot accounts on Twitter contains valuable information that can be used to identify bot accounts.

Influence of a long-term proctoring process on the effectiveness of chronic total occlusion percutaneous coronary interventions.

BACKGROUND: Despite the complexity of the chronic total occlusion (CTO) percutaneous coronary intervention (PCI) procedures and unsatisfactory results in centers with low volume experience, the practice of training and certifying operators is not a routine. AIMS: The study aimed to identify factors influencing the effectiveness and complications of PCI CTOs during a proctoring program. METHODS: The study group consisted of 194 consecutive patients (226 PCI CTOs) as part of the proc-toring program. The relationships between clinical and treatment parameters and the experience gained along with the duration of the proctoring program on the effectiveness and safety of the procedure were assessed. RESULTS: The multivariable analysis showed an independent effect of CTO morphology (odds ratio [OR], 0.38; 95% confidence interval [CI], 0.21-0.71; P <0.01) and an independent effect of increasing operator's experience (OR, 2.62; 95% CI, 1.24-5.60; P = 0.01) on the effectiveness of the procedure. The increase in the efficiency of the PCI CTO, related to the treatment experience gained during the program, was observed especially in the first 50 procedures, treatment effectiveness increased from 55% to 72% (P <0.05). The success of procedures was higher in months when ≥3 procedures were performed (75% vs. 52%; P <0.001). Periprocedural complications occurred in 11 patients (4.9%). In the multivariable analysis, no independent factors influencing the risk of complications were identified. CONCLUSIONS: The effectiveness of PCI CTO depended on lesion complexity and broadening oper-ator's experience. No independent factors affecting the risk of complications were identified. The number of >50 procedures under the proctor's supervision should be considered in designing teaching programs.

Gender-Related Differences in Prodromal Multiple Sclerosis Characteristics: A 7-Year Observation Study.

Increasing evidence supports the observation that multiple sclerosis (MS) has a preclinical period, with various prodromal signs and symptoms more frequently represented in patients with confirmed MS many years later. Considering the apparent gender differences in the incidence and clinical course of MS, it remains unclear whether it could be reflected in prodromal symptom features. This study aimed to compare a broad spectrum of prodromal signs and symptoms between males and females in the 7-year period before the definite diagnosis of MS. Data came from the central register of the national payer of services, financed under the public healthcare system in Poland. They covered a 7-year period of patient health record claims, from 2009 to 2016. The following groups of symptoms were significant with women: musculoskeletal (p < 0.001), ophthalmic (p < 0.001), laryngological (p < 0.001), digestive system (p < 0.001), urinary tract (p < 0.001), mental (p < 0.001), cardiovascular (p < 0.001), complaints and headaches (p < 0.001). There was also a weak correlation with head injuries (p = 0.03) while dermatological and reproductive system complaints did not appear to be significant (p < 0.05). For males, the following groups of symptoms were significant: musculoskeletal (p < 0.001), ophthalmic (p < 0.001), laryngological (p = 0.007), cardiovascular system symptoms (p < 0.001), and headaches (p < 0.001). Interestingly, reproductive system problems were overrepresented in the male population (p = 0.008). There was no significant correlation with MS risk for dermatological, digestive, urinary, and mental complaints. Similarly, head injuries were not significant. Our results shed more light on well-known differences in the epidemiological and clinical characteristics between sexes in multiple sclerosis, and show differences in prodromal complaints before MS onset.

Clustering via hypergraph modularity.

Despite the fact that many important problems (including clustering) can be described using hypergraphs, theoretical foundations as well as practical algorithms using hypergraphs are not well developed yet. In this paper, we propose a hypergraph modularity function that generalizes its well established and widely used graph counterpart measure of how clustered a network is. In order to define it properly, we generalize the Chung-Lu model for graphs to hypergraphs. We then provide the theoretical foundations to search for an optimal solution with respect to our hypergraph modularity function. A simple heuristic algorithm is described and applied to a few illustrative examples. We show that using a strict version of our proposed modularity function often leads to a solution where a smaller number of hyperedges get cut as compared to optimizing modularity of 2-section graph of a hypergraph.

A framework for sensitivity analysis of decision trees.

In the paper, we consider sequential decision problems with uncertainty, represented as decision trees. Sensitivity analysis is always a crucial element of decision making and in decision trees it often focuses on probabilities. In the stochastic model considered, the user often has only limited information about the true values of probabilities. We develop a framework for performing sensitivity analysis of optimal strategies accounting for this distributional uncertainty. We design this robust optimization approach in an intuitive and not overly technical way, to make it simple to apply in daily managerial practice. The proposed framework allows for (1) analysis of the stability of the expected-value-maximizing strategy and (2) identification of strategies which are robust with respect to pessimistic/optimistic/mode-favoring perturbations of probabilities. We verify the properties of our approach in two cases: (a) probabilities in a tree are the primitives of the model and can be modified independently; (b) probabilities in a tree reflect some underlying, structural probabilities, and are interrelated. We provide a free software tool implementing the methods described.

Pharmacotherapy prior to and in acute ischaemic stroke. The use of pharmacotherapy and drug-associated outcomes in real world practice - findings from the Polish Hospital Stroke Registry.

BACKGROUND AND PURPOSE: Stroke is a preventable disease and acute ischaemic stroke can be effectively treated. Specific pharmacotherapy is recommended in either prevention or acute ischemic stroke treatment. We aimed to evaluate the use and the early and late outcomes impact of drugs administered before and in acute ischaemic stroke in a real world practice. MATERIAL AND METHODS: Ischaemic stroke patients hospitalized between 1st March 2007 and 29th February 2008 and reported in Polish Hospital Stroke Registry were analysed. Fully anonymous data were collected with standardized, web-based questionnaire with authorized access. Multivariate regression models were used to adjust for case-mix and evaluate the impact of drugs used prior to or in acute ischaemic stroke on outcomes. The early outcomes were defined as in-hospital mortality or poor outcome (death or dependency - modified Rankin Scale  3) at hospital discharge, while late outcomes covered one-year survival. RESULTS: A total number of 26 153 ischaemic stroke patients (mean age: 71.8 years; females: 51.6%) was reported. The ana-lysis of pharmacotherapy showed that preventive use of hypo-tensive agents, anticoagulants in atrial fibrillation, antiplatelets and statins is inadequate. Regression models confirmed some expected drug benefits and additionally revealed that antihypertensive drugs or aspirin used prior to stroke and oral anticoagulants or statins used in hospital were associated with better stroke outcome. CONCLUSIONS: The prevention of ischaemic stroke needs to be monitored and improved. Evidence-based treatment of acute ischaemic stroke requires further promotion. The benefits of acute ischaemic stroke treatment with statins require to be confirmed in randomized controlled settings.

Can we improve the accuracy of risk assessment in patients with non ST-segment elevation acute coronary syndromes?

BACKGROUND: In patients with non-ST segment elevation acute coronary syndromes (NSTE-ACS), the long-term risk of deathand myocardial infarction (MI) is estimated by scores based on noninvasively derived variables. Much less is known about the relation between the degree of atherosclerotic burden in the coronary tree and the long-term risk of patients with NSTE-ACS. AIM: To evaluate the accuracy of a wide spectrum of coronary angiographic and clinical data in predicting outcomes ina long-term follow-up of patients successfully treated invasively for NSTE-ACS. METHODS: The study group consisted of 112 consecutive patients (age 62 ± 10 years; 76 men) treated invasively for NSTE-ACS.27 (24%) patients had a history of diabetes mellitus (DM) and 37 (33%) patients a history of MI. The coronary angiograms priorto intervention were evaluated blindly for the four angiographic scores: (1) Stenosis score derived from the assessment of thedegree of stenosis in 15 segments of the coronary tree; (2) Vessel score showing the number of main vessels stenosed > 70%; (3) Extensity score assessing the proportion of lumen length irregularity in 15 segments; and (4) Complexity score describingthe number of complex plaques. The angiographic analysis also focused on the flow, presence of thrombus and collateralsupply prior to intervention (according to TIMI) and the size of the culprit lesion vessel. The intervention was successful in 95% of cases. All patients were followed-up for 6-24 months for the occurrence of death or MI. RESULTS: In the follow-up period, the composite end point of death or MI occurred in 20 (17%) patients. In order to indicate therisk predictors from the group of clinical and angiographic variables (age, sex, history of DM, history of MI, four angiographicscores and culprit lesion vessel characterisation), logistic regression analysis was performed. The independent angiographic predictors of composite end point (selected by forward conditional selection) were stenosis score (OR 1.13; 95% CI 1.05-1.2;p < 0.001) and size of the vessel (OR 0.08; 95% CI 0.01-0.6; p = 0.02). CONCLUSIONS: Our preliminary data shows that attempting to add angiographic variables into the risk assessment scoring systems in order to strengthen their predictive accuracy is justified.

Impact of diabetes on survival in patients with ST-segment elevation myocardial infarction treated by primary angioplasty: insights from the POLISH STEMI registry.

BACKGROUND: It has been shown that, among patients with ST-segment elevation myocardial infarction (STEMI), diabetes is associated with a significantly higher mortality. The aim of this study was to investigate in a large cohort of patients the impact of diabetes on mortality in a large cohort of patients with STEMI treated with primary angioplasty. METHODS: Our population is represented by consecutive patients with STEMI treated by primary angioplasty and enrolled in the POLISH registry in 2003. All clinical, angiographic, and follow-up data were prospectively collected. Diagnosis of diabetes was based on history of diabetes at admission. RESULTS: Among 7193 patients, 877 (12.2%) had diabetes at admission. Diabetes was associated with more advanced age (p<0.0001), higher prevalence of female gender (p<0.0001), hyperlipidemia (p<0.0001), shock at presentation (p<0.0001), renal failure (p<0.0001), previous myocardial infarction (p<0.0001), more often treated after 6h from symptom onset (p<0.0001). Diabetes was associated with more extensive coronary artery disease (p<0.0001), less often treated with stenting (p<0.0001). Diabetes was significantly associated with impaired epicardial reperfusion (TIMI 0-2: OR [95% CI]=1.81 [1.5-2.18], p<0.0001), that persisted after correction for baseline confounding factors (OR [95% CI]=1.33 [1.075-1.64], p=0.009). At a mean follow-up of 524+/-194 days, diabetes was associated with higher mortality (unadjusted cumulative mortality: 23.5% vs. 12.6%, unadjusted HR=1.95 [1.66-2.3], p<0.0001), that persisted after correction for confounding factors (adjusted cumulative mortality: 13.3% vs. 10.7%, adjusted HR=1.23 [1.04-1.46], p=0.013). CONCLUSIONS: This study shows that among STEMI treated by primary angioplasty diabetes is independently associated with impaired epicardial reperfusion and higher mortality.

Cost-effectiveness acceptability curves--caveats quantified.

Cost-effectiveness acceptability curves (CEACs) have become widely used in applied health technology assessment and at the same time are criticized as unreliable decision-making tool. In this paper we show how using CEACs differs from maximizing expected net benefit (NB) and when it can lead to inconsistent decisions. In the case of comparing two alternatives we show the limits of the discrepancy between CEAC and expected NB approach and link it with expected value of perfect information. We also show how the shape of CEAC is influenced by the skewness of estimate of expected NB distribution, the correlation between cost and effect estimates and their variance. In the case of more than two options we show when using CEACs can lead to non-transitive choices in pair-wise comparisons and when it lacks independence of irrelevant alternatives property in joint comparisons.

Influence of different antiplatelet treatment regimens for primary percutaneous coronary intervention on all-cause mortality.

AIMS: The aim of this analysis was to examine the influence of different in-cath-lab antiplatelet regimens for the primary percutaneous coronary intervention (PCI) on all-cause mortality. METHODS AND RESULTS: The study group consisted of 7193 patients (pts) undergoing primary PCI in 38 centres in 2003 in Poland. All patients received pretreatment with 300 mg of aspirin, 992 pts (14%) received glycoprotein (GP) IIb/IIIa inhibitors, 2690 pts (37%) were treated with 300 mg loading dose of clopidogrel, and 1566 (22%) received combined antiplatelet treatment with both GP IIb/IIIa inhibitors and clopidogrel. Remaining 1945 patients (27%) did not receive GP IIb/IIIa inhibitors or clopidogrel. Primary endpoint of the study was all-cause mortality up to 1 year from ST-segment elevation myocardial infarction (STEMI). One year mortality rates in the four groups were: 10.4%, 9.0%, 9.7%, and 15.3%, respectively. Propensity-adjusted survival analysis showed significant reduction of mortality for combination therapy with GP IIb/IIIa inhibitors and clopidogrel, clopidogrel alone, and GP IIb/IIIa inhibitors alone over aspirin alone. No additive effect on survival was seen for a combination therapy with GP IIb/IIIa inhibitors and clopidogrel in comparison to treatment with clopidogrel alone. CONCLUSION: In this large cohort, multicentre STEMI registry in-cath-lab use of GP IIb/IIIa inhibitors and clopidogrel alone or in combination was associated with the reduction of 1 year all-cause mortality in the setting of primary PCI in comparison with aspirin only. However, the use of GP IIb/IIIa inhibitors on top of 300 mg loading dose of clopidogrel did not further reduce mortality.

Pulse pressure--independent predictor of poor early outcome and mortality following ischemic stroke.

BACKGROUND: Pulse pressure (PP) in acute stroke may be related to the outcome. The link between PP in the first week following ischemic stroke and early outcome was assessed. METHODS: We calculated mean PPs during the first 7 days after stroke onset in 1,677 patients. Poor outcome at hospital discharge was defined as a modified Rankin scale score of 3 or more points or death. Logistic regression was developed to evaluate PP as an independent predictor of early outcome. RESULTS: For patients with poor outcomes the mean PP during the first week was higher than that for patients with non-poor outcomes. A logistic regression model confirmed that elevated mean PP was independently associated with poor outcome at discharge and 30-day mortality. CONCLUSION: Elevated PP during the acute phase of ischemic stroke is an independent predictor of poor early outcome at hospital discharge and 30-day mortality.

The cost comparison of rhythm and rate control strategies in persistent atrial fibrillation.

UNLABELLED: Our sub study was designed to analyze the cost effectiveness of two alternative treatment strategies with a view to improved allocation of the limited therapeutic resources. To that effect we conducted detailed analysis of the related costs and other relevant data collected in the course of the HOT CAFE study. METHODS: The prospective costs related to 205 patients randomly assigned to rhythm or rate control were traced over a 12 month period. Since, both strategies produced similar clinical outcomes a cost minimization analysis was undertaken. The cost of diagnostic and treatment procedures, including hospitalization, outpatient visits, drugs and physicians consultations were estimated for both groups. RESULTS: The study population comprised 205 patients (mean age 60.8 year; 35% females). A hundred and one patients were randomly assigned to the rate control group with the pharmacological heart rate frequency optimization treatment combined with Holter monitoring. A hundred and four patients were randomized to sinus rhythm (SR) restoration with its subsequent maintenance with sequential antiarrhythmic drug treatment. There was no significant difference in the composite primary end-point (all-cause mortality, number of thromboembolic and major bleeding events). The hospital admissions rate was significantly higher in the rhythm control than the rate control arm (202 vs. 5, respectively). The conservative strategy involving pharmacological ventricular rate control proved to be less costly than rhythm control (1225 euros vs. 2526 euros; p<0.001). The main cost driver behind the established difference was the cardioversion related hospitalization. CONCLUSIONS: The cost effectiveness appraisal seems to have supported the rate control strategy as less costly due to the lower hospitalization rate as a major cost carrier.

Acute ischemic stroke care and outcome in centers participating in the Polish National Stroke Prevention and Treatment Registry.

BACKGROUND AND PURPOSE: Significant intercenter variability in quality of care and stroke outcomes was found in many countries. The aim of the study was to compare the acute ischemic stroke care and outcomes in centers participating in the Polish National Stroke Prevention and Treatment Registry. METHODS: The World Health Organization Stepwise Approach to Stroke Surveillance-based questionnaire was used to collect data on patients admitted to participating centers between December 1, 2001, and July 31, 2002. To ensure data quality, only centers reporting representative sample of patients were analyzed. Ischemic stroke patient characteristics, in-hospital care, and early outcomes (adjusted for case mix) were compared for participating centers. RESULTS: There were 26 of 48 centers that met inclusion criteria, with a total of 8736 patients (52% women; mean age 71 years, with a range among institutions from 68 to 75 years). Significant differences between centers were observed for distribution of risk factors and in-hospital care. The rates for death and poor outcome (defined as a Rankin score > or =3 or death) ranged from 8.0% to 31.8% and from 44.2% to 74.7%, respectively. After adjusting for case mix, the death or poor outcome prognoses remained significantly different between centers. CONCLUSIONS: The observed significant differences between Polish stroke centers indicate the need for improvement of patient education, effective stroke risk factor control, and standardized in-hospital care.

Influence of gender on baseline features and clinical outcomes among 17,370 patients with confirmed ischaemic stroke in the international stroke trial.

AIM: We sought to determine whether there were differences between men and women with acute stroke in their baseline characteristics and outcome in a large cohort of patients randomized in the International Stroke Trial (IST). METHODS: Of the 19,435 patients randomized in the IST, 17,370 had an ischemic stroke confirmed by CT scan or autopsy (8,003 female and 9,367 male). In males and females, we compared baseline characteristics (age, frequency of atrial fibrillation, pre-stroke administration of aspirin and systolic blood pressure, conscious level, stroke syndrome) and outcome at 14 days and 6 months (death, complications, dependency, recovery, place of residence). We developed a specific logistic regression model to adjust for case-mix in order to evaluate the separate influence of gender on outcome. RESULTS: Female patients were older, suffered more frequently from atrial fibrillation, had higher systolic blood pressure at randomization and generally had more severe strokes (a higher proportion were unconscious or drowsy or had a total anterior circulation syndrome). Females had higher 14-day and 6-month case fatality and were more likely to be dead or dependent at six months (and consequently more likely to require institutional or residential care). Gender was an independent predictor of death or dependency at 6 months. CONCLUSIONS: The adverse effect of female gender on outcome indicates that further research to explore the underlying biological mechanism is justified, and that more intensive acute and long-term treatment may be needed to improve outcome among female patients with stroke.