Automated Feature Selection from Medical Literature.

We propose an automated approach to rank the most salient variables related to a certain clinical phenomenon from scientific literature. Our solution is an automated approach to improve the efficiency of the collection of different health-related measures from a population, and to accelerate the discovery of novel associations and dependencies between health-related concepts.

Low Levels of Vitamin D and Silent Myocardial Ischemia in Type 2 Diabetes: Clinical Correlations and Prognostic Significance.

Vitamin D deficiency has a pathogenetic and prognostic role in coronary artery disease and a key role in pain transmission. Diabetic patients have a higher risk of silent myocardial ischemia (SMI) due to diabetic neuropathy. We evaluated the correlation between SMI and Vitamin D serum levels in type 2 diabetic patients and assessed whether SMI patients had a worse survival rate than their symptomatic counterpart. We enrolled 253 patients admitted in our Cardiology Unit and compared them with 50 healthy volunteers. We created three sub-groups: symptomatic MI group (125, 32.4%); SMI group (78, 25.7%), and no-MI group (50, 41.9%). 25(OH)D levels (nmol/L) were lower in the SMI group (34.9 ± 5.8) compared to those in the symptomatic MI (49.6 ± 6.1; p = 0.01), no MI (53.1 ± 6.2; p = 0.001), and control groups (62.1 ± 6.7; p = 0.0001). 25(OH)D levels predicted SMI in diabetic patients, with an inverted odds ratio of 1.11 (p = 0.01). Symptomatic MI group survival was higher than the SMI one (6-year survival rate: 83 vs. 69%; p = 0.01). Diabetic patients with SMI had a higher mortality risk and showed lower 25(OH)D levels than the symptomatic group. This suggests the crucial role that vitamin D has in the pathogenesis of SMI.

A Digital Platform to Support Self-management of Multiple Chronic Conditions (ProACT): Findings in Relation to Engagement During a One-Year Proof-of-Concept Trial.

BACKGROUND: Populations globally are ageing, resulting in higher incidence rates of chronic diseases. Digital health platforms, designed to support those with chronic conditions to self-manage at home, offer a promising solution to help people monitor their conditions and lifestyle, maintain good health, and reduce unscheduled clinical visits. However, despite high prevalence rates of multimorbidity or multiple chronic conditions, most platforms tend to focus on a single disease. A further challenge is that despite the importance of users actively engaging with such systems, little research has explored engagement. OBJECTIVE: The objectives of this study are to design and develop a digital health platform, ProACT, for facilitating older adults self-managing multimorbidity, with support from their care network, and evaluate end user engagement and experiences with this platform through a 12-month trial. METHODS: The ProACT digital health platform is presented in this paper. The platform was evaluated in a year-long proof-of-concept action research trial with 120 older persons with multimorbidity in Ireland and Belgium. Alongside the technology, participants had access to a clinical triage service responding to symptom alerts and a technical helpdesk. Interactions with the platform during the trial were logged to determine engagement. Semistructured interviews were conducted with participants and analyzed using inductive thematic analysis, whereas usability and user burden were examined using validated questionnaires. RESULTS: This paper presents the ProACT platform and its components, along with findings on engagement with the platform and its usability. Of the 120 participants who participated, 24 (20%) withdrew before the end of the study, whereas 3 (2.5%) died. The remaining 93 participants actively used the platform until the end of the trial, on average, taking 2 or 3 health readings daily over the course of the trial in Ireland and Belgium, respectively. The participants reported ProACT to be usable and of low burden. Findings from interviews revealed that participants experienced multiple benefits as a result of using ProACT, including improved self-management, health, and well-being and support from the triage service. For those who withdrew, barriers to engagement were poor health and frustration when technology, in particular sensing devices, did not work as expected. CONCLUSIONS: This is the first study to present findings from a longitudinal study of older adults using digital health technology to self-manage multimorbidity. Our findings show that older adults sustained engagement with the technology and found it usable. Potential reasons for these results include a strong focus on user-centered design and engagement throughout the project lifecycle, resulting in a platform that meets user needs, as well as the integration of behavior change techniques and personal analytics into the platform. The provision of triage and technical support services alongside the platform during the trial were also important facilitators of engagement. INTERNATIONAL REGISTERED REPORT IDENTIFIER (IRRID): RR2-10.2196/22125.

Outcome Prediction from Behaviour Change Intervention Evaluations using a Combination of Node and Word Embedding.

Findings from randomized controlled trials (RCTs) of behaviour change interventions encode much of our knowledge on intervention efficacy under defined conditions. Predicting outcomes of novel interventions in novel conditions can be challenging, as can predicting differences in outcomes between different interventions or different conditions. To predict outcomes from RCTs, we propose a generic framework of combining the information from two sources - i) the instances (comprised of surrounding text and their numeric values) of relevant attributes, namely the intervention, setting and population characteristics of a study, and ii) abstract representation of the categories of these attributes themselves. We demonstrate that this way of encoding both the information about an attribute and its value when used as an embedding layer within a standard deep sequence modeling setup improves the outcome prediction effectiveness.

Knowledge Extraction and Prediction from Behavior Science Randomized Controlled Trials: A Case Study in Smoking Cessation.

Due to the fast pace at which randomized controlled trials are published in the health domain, researchers, consultants and policymakers would benefit from more automatic ways to process them by both extracting relevant information and automating the meta-analysis processes. In this paper, we present a novel methodology based on natural language processing and reasoning models to 1) extract relevant information from RCTs and 2) predict potential outcome values on novel scenarios, given the extracted knowledge, in the domain of behavior change for smoking cessation.

KITE: An Analytics Platform for Multimorbidity Management.

Managing multimorbidity entails processing distributed, dynamic and heterogeneous data using diverse analytics tools. We present KITE, a Cloud-based infrastructure allowing the aggregation and processing of health data using a dynamic set of analytical components. We showcase KITE in the context of the ProACT project, aiming at advancing home-based integrated care though IoT, analytics and a behavior change framework. We validate the viability of the infrastructure through an application of Bayesian networks to give a probabilistic representation of older individuals based on a variety of factors.

An Analytical Method for Multimorbidity Management Using Bayesian Networks.

This paper describes an application of Bayesian Networks to mo-del persons with multimorbidity using measurements of vital signs and lifestyle assessments. The model was developed as part of a project on the use of wearable and home sensors and tablet applications to help persons with multimorbidity and their carers manage their conditions in daily life. The training data was extracted from TILDA, an open dataset collected from a longitudinal health study of the older Irish population. A categorical BN structure was learnt using a score-based approach, with constraints on the ordering of variables. The prediction accuracy of the model is assessed using the Brier score in a cross-validation experiment. Finally, a user inter-face that allows to set some observed levels and query the resulting margi-nal probabilities from the BN is presented.

Interaction between uric acid and endothelial dysfunction predicts new onset of diabetes in hypertensive patients.

BACKGROUND: Both uric acid and endothelial dysfunction are associated with new occurrence of type-2 diabetes but, at this moment, there is no evidence about a possible interaction between them. We tested, in untreated hypertensive patients, without clinical evidence of vascular damage, the hypothesis that serum uric acid and endothelial dysfunction may interact in predicting new diabetes. METHODS: In 500 uncomplicated hypertensive non diabetic (ADA criteria) patients we evaluated endothelial function, by strain-gauge plethysmography, and uric acid. RESULTS: During the follow-up (median 87.1 months), there were 54 new cases of diabetes (1.8%/year). On univariate analysis, incident diabetes was inversely related with ACh-stimulated FBF (HR=0.65, 95%CI=0.52-0.82; P<0.001) and directly with serum CRP (HR=1.22, 95%CI=1.09-1.37; P<0.001), HOMA-index (HR=1.20, 95%CI=1.05-1.37; P=0.007), fasting insulin (HR=1.05, 95%CI=1.01-1.09; P=0.006) and age (HR=1.03, 95%CI=1.00-1.05; P=0.014). At multiple regression analysis, the interaction between ACh-stimulated FBF and uric acid resulted statistically significant. Similar results were observed for the interaction between FBF and CRP. CONCLUSIONS: Our data clearly demonstrate that the coexistence of both hyperuricemia and reduced endothelium-dependent vasodilation increases the risk to develop new diabetes in hypertensive patients. In addition, mild-inflammation seems to be the mediator of the interaction between endothelial dysfunction and uric acid.

Association between one-hour post-load plasma glucose levels and vascular stiffness in essential hypertension.

OBJECTIVES: Pulse wave velocity (PWV) is a surrogate end-point for cardiovascular morbidity and mortality. A plasma glucose value ≥155 mg/dl for the 1-hour post-load plasma glucose during an oral glucose tolerance test (OGTT) is able to identify subjects with normal glucose tolerance (NGT) at high-risk for type-2 diabetes (T2D) and for subclinical organ damage. Thus, we addressed the question if 1-hour post-load plasma glucose levels, affects PWV and its central hemodynamic correlates, as augmentation pressure (AP) and augmentation index (AI). METHODS: We enrolled 584 newly diagnosed hypertensives. All patients underwent OGTT and measurements of PWV, AP and AI. Insulin sensitivity was assessed by Matsuda-index. RESULTS: Among participants, 424 were NGT and 160 had impaired glucose tolerance (IGT). Of 424 NGT, 278 had 1-h post-load plasma glucose <155 mg/dl (NGT<155) and 146 had 1-h post-load plasma glucose ≥155 mg/dl (NGT≥155). NGT≥155 had a worse insulin sensitivity and higher hs-CRP than NGT<155, similar to IGT subjects. In addition, NGT ≥155 in comparison with NGT<155 had higher central systolic blood pressure (134±12 vs 131±10 mmHg), as well as PWV (8.4±3.7 vs 6.7±1.7 m/s), AP (12.5±7.1 vs 9.8±5.7 mmHg) and AI (29.4±11.9 vs 25.1±12.4%), and similar to IGT. At multiple regression analysis, 1-h post-load plasma glucose resulted the major determinant of all indices of vascular stiffness. CONCLUSION: Hypertensive NGT≥155 subjects, compared with NGT<155, have higher PWV and its hemodynamic correlates that increase their cardiovascular risk profile.

Association between hemoglobin level and endothelial function in uncomplicated, untreated hypertensive patients.

BACKGROUND AND OBJECTIVES: Hemoglobin (Hb) is an important nitric oxide (NO) buffer and a modulator of NO bioavailability. In addition, endothelial dysfunction is common in hypertensive patients, suggesting a pivotal role of hemoglobin concentration ([Hb]) in vascular function. To investigate the potential role of [Hb] in endothelium-dependent vasodilation, the relationship between Hb and endothelial function was tested in a group of patients with essential hypertension. DESIGN, SETTING, PARTICIPANTS, & MEASUREMENTS: In this retrospective study, 174 nonsmoking, uncomplicated, never-treated hypertensives were enrolled. Endothelium-dependent and -independent vasodilation was assessed by measurement of forearm blood flow response during intra-arterial infusion of increasing doses of acetylcholine (ACh) and sodium nitroprusside (SNP) using strain-gauge plethysmography. Correlation with established risk factors of endothelial dysfunction was performed. RESULTS: The vasodilatory response to ACh was inversely (P < 0.001) related to [Hb], and this relationship was dose dependent (P < 0.001), being minimal at the lowest dose and maximal at the highest dose. No association was found between Hb and the vasodilatory response to SNP. In a multiple linear regression model adjusted for Framingham risk factors (age, sex, BP, cholesterol, body mass index, glucose) and emerging risk factors (homeostasis model assessment index, C-reactive protein, estimated GFR), [Hb] maintained a strong and independent link with the vasodilatory response to ACh (P < 0.001). CONCLUSIONS: In a large group of nonsmoking untreated hypertensives, [Hb] is inversely related to forearm endothelium-dependent vasodilation. [Hb] should be taken into account, especially in conditions associated with low [Hb], when performing vascular function studies.

Endothelial dysfunction, ADMA and insulin resistance in essential hypertension.

BACKGROUND: Endothelial dysfunction and insulin resistance (IR) are associated with essential hypertension and other cardiovascular risk factors. Asymmetric dimethylarginine (ADMA), an endogenous inhibitor of nitric oxide synthase, induces endothelial dysfunction in different setting of patients. However, at this moment no data are available about the role of ADMA and IR to induce endothelial dysfunction in an independent way or combined between them. In this study, we investigated, in 63 hypertensives and 21 normotensive healthy subjects, the relationship between ADMA and IR and their possible interaction on endothelial function. METHODS: ADMA plasma levels were measured by high-performance liquid chromatography, and IR by homeostasis model assessment (HOMA). Endothelial function was estimated by intra-arterial infusion of acetylcholine (ACh) and sodium nitroprusside at increasing doses. RESULTS: Hypertensive patients had significantly higher ADMA, insulin, HOMA and C-reactive protein (CRP) values than normotensive controls (P<0.0001). There were no significant differences in mean l-arginine/ADMA ratio between groups. ACh-stimulated forearm blood flow (FBF) was significantly reduced in hypertensive patients (P<0.0001). In hypertensive group, HOMA was the strongest determinant of FBF, accounting for the 45.5% of its variation. ADMA and gender were the independent determinants of HOMA, accounting for 12.3% and 8.3% of its variation, respectively. CONCLUSIONS: The association between ADMA and IR contributes to identify a possible novel mechanism by which ADMA promotes vascular damage, increasing individual cardiovascular risk in hypertensive patients. However, this hypothesis should be tested in a larger study group.

Insulin-like growth factor-1 and glomerular filtration rate in hypertensive patients.

OBJECTIVE: Essential hypertension is a clinical condition associated with insulin resistance and progressive impairment of renal function that increases cardiovascular events. Insulin-like growth factor (IGF)-1, which is inversely related to insulin levels, increases renal blood flow and glomerular filtration rate (GFR). The aim of the present study was to investigate the relationship between circulating IGF-1 levels and GFR in a group of never treated hypertensive patients. METHODS: The study population consisted of 537 outpatients presenting at Catanzaro University Hospital. To participate in this study, patients have to had a systolic clinic blood pressure (BP) of more than 140 and less than 180 mmHg or a diastolic BP of more than 90 and less than 100 mmHg or both on at least two separate visits. Blood samples were at least obtained after 8-10 h in fasting conditions. The GFR was estimated by the modification of diet in renal disease equation. Serum creatinine was measured in the laboratory by an automated technique. Insulin sensitivity was estimated by using the homeostasis model assessment index calculated from the fasting glucose and insulin concentrations. RESULTS: Both fasting insulin and homeostasis model assessment significantly (P < 0.0001) show an inverse relationship with GFR decline, whereas IGF-1 presents a significant and direct relationship with it. As expected, IGF-1 and fasting insulin resulted in an inverse relationship between them (r = -0.318; P < 0.0001). The strongest predictor of GFR resulted IGF-1, accounting for a 9.8% of its variation; the addition of fasting insulin and systolic BP accounts for another 3.7% of the variation. CONCLUSION: We demonstrate a significant relationship between IGF-1 and GFR in a large sample of never treated hypertensive patients, probably as consequence of insulin resistance/hyperinsulinemia, which is a very frequent condition in high BP.