Fairness is a critical trait in decision making. As machine-learning models are increasingly being used in sensitive application domains (e.g. education and employment) for decision making, it is crucial that the decisions computed by such models are free of unintended bias. But how can we automatically validate the fairness of arbitrary machine-learning models? For a given machine-learning model and a set of sensitive input parameters, our Aeqitas approach automatically discovers discriminatory inputs that highlight fairness violation. At the core of Aeqitas are three novel strategies to employ probabilistic search over the input space with the objective of uncovering fairness violation. Our Aeqitas approach leverages inherent robustness property in common machine-learning models to design and implement scalable test generation methodologies. An appealing feature of our generated test inputs is that they can be systematically added to the training set of the underlying model and improve its fairness. To this end, we design a fully automated module that guarantees to improve the fairness of the model.We implemented Aeqitas and we have evaluated it on six stateof-the-art classifiers. Our subjects also include a classifier that was designed with fairness in mind. We show that Aeqitas effectively generates inputs to uncover fairness violation in all the subject classifiers and systematically improves the fairness of respective models using the generated test inputs. In our evaluation, Aeqitas generates up to 70% discriminatory inputs (w.r.t. the total number of inputs generated) and leverages these inputs to improve the fairness up to 94%. CCS CONCEPTS• Software and its engineering → Software testing and debugging;

AimsRevascularization is frequently advocated to improve ventricular function and prognosis for patients with heart failure due to coronary artery disease, especially when there is evidence of extensive myocardial viability. Methods and resultsPatients with heart failure, coronary artery disease, and a left ventricular (LV) ejection fraction ,35%, who had a substantial volume of viable myocardium with contractile dysfunction assessed by any standard imaging technique, were randomly assigned to a strategy of conservative management vs. angiography with the intent of percutaneous or surgical revascularization. Patients requiring revascularization for angina or too frail for surgery were excluded. Only 138 of the planned 800 patients were enrolled because of withdrawal of funding due to slow recruitment. Also, a larger trial (The Surgical Treatment for Ischemic Heart Failure Trial) addressing a similar question became available, which investigators were encouraged to join. Of 69 patients assigned to the invasive strategy, 6 refused angiography, 2 died as a result of the diagnostic procedure, 14 were considered unsuitable for revascularization, 2 refused surgery, and 45 had revascularization. After a median follow-up of 59 (inter-quartile range: 33 -63) months, there were 51 (37%) deaths; 25 (37%) in those assigned to the conservative strategy, and 26 (38%) in those assigned to the invasive strategy, 13 (29%) of whom had been revascularized. ConclusionA conservative management strategy may not be inferior to one of coronary arteriography with the intent to revascularize in patients with heart failure, LV systolic dysfunction, and extensive myocardial viability. However, this study was underpowered and, further, larger trials are required to settle this issue.

No abstract

ObjectiveTo investigate the prognostic effect of newly diagnosed diabetes mellitus (NDM) and impaired glucose tolerance (IGT) post myocardial infarction (MI).Research Design and MethodsRetrospective cohort study of 768 patients without preexisting diabetes mellitus post-MI at one centre in Yorkshire between November 2005 and October 2008. Patients were categorised as normal glucose tolerance (NGT n = 337), IGT (n = 279) and NDM (n = 152) on pre- discharge oral glucose tolerance test (OGTT). Primary end-point was the first occurrence of major adverse cardiovascular events (MACE) including cardiovascular death, non-fatal MI, severe heart failure (HF) or non-haemorrhagic stroke. Secondary end-points were all cause mortality and individual components of MACE.ResultsPrevalence of NGT, impaired fasting glucose (IFG), IGT and NDM changed from 90%, 6%, 0% and 4% on fasting plasma glucose (FPG) to 43%, 1%, 36% and 20% respectively after OGTT. 102 deaths from all causes (79 as first events of which 46 were cardiovascular), 95 non fatal MI, 18 HF and 9 non haemorrhagic strokes occurred during 47.2 ± 9.4 months follow up. Event free survival was lower in IGT and NDM groups. IGT (HR 1.54, 95% CI: 1.06–2.24, p = 0.024) and NDM (HR 2.15, 95% CI: 1.42–3.24, p = 0.003) independently predicted MACE free survival. IGT and NDM also independently predicted incidence of MACE. NDM but not IGT increased the risk of secondary end-points.ConclusionPresence of IGT and NDM in patients presenting post-MI, identified using OGTT, is associated with increased incidence of MACE and is associated with adverse outcomes despite adequate secondary prevention.

No abstract

Background: Most patients with heart failure due to left ventricular systolic dysfunction (LVSD) secondary to coronary artery disease (CAD) have evidence of myocardium in jeopardy (reversible ischaemia andyor stunning hibernation). It is not known whether revascularisation in such cases is safe or beneficial. Aims: To determine whether revascularisation will improve the survival of patients with LVSD and heart failure secondary to CAD and myocardium in jeopardy. Methods: This is a randomised controlled trial comparing revascularisation or not, in addition to optimal medical therapy with ACE inhibitors, beta-blockers, aldosterone antagonists and an anti-thrombotic agent. Patients must have heart failure requiring treatment with diuretics, a left ventricular ejection fraction -35% and evidence of coronary disease. Myocardial viability and ischaemia are assessed by a broad range of techniques including stress echocardiography and nuclear imaging. All imaging tests are reviewed in core laboratories to ensure uniform reporting. Any conventional revascularisation technique is permitted. The primary outcome measure is all cause mortality. Symptoms, quality of life and health economic issues will also be explored. Assuming an annual mortality of 10% in the control group and allowing for substantial cross-over rates, a study of 800 patients followed for 5 years has 80% power with an alpha of 0.05 (two-sided) to show a 25% reduction in mortality with revascularisation. Results: At the time of writing 180 patients have been screened for inclusion, 111 have consented to participate and 70 have been randomised. The results of viability testing are awaited in 22 patients. Twenty-six patients had been investigated for myocardial viability andyor by angiography prior to consent, as part of the routine practice in that cardiology department. Of 68 patients who have completed assessment only after consent, 47 (69%) were included. The principal reason for drop-out between consent and randomisation was lack of evidence of myocardial ischaemia or hibernation. Conclusion: The HEART trial will help to determine whether investigation of myocardial ischaemia andyor viability with a view to revascularisation should become part of the routine care of patients with heart failure due to LVSD and CAD.

Background-The genesis of symptoms in patients with heart failure (HF) and normal ejection fraction (HFNEF) is unclear. Most investigations of HFNEF have focused on cardiac function at rest although most of these patients are breathless only on exercise. Stress-induced impairment in systolic or diastolic function could result in these symptoms. Method and Result-Forty-one patients with HFNEF and 29 controls underwent dobutamine stress echocardiography with color tissue Doppler imaging. Wall motion score index and regional myocardial systolic velocity (Sm) were measured at and peak stress. Systolic (Sa), early diastolic (Ea), and late diastolic (Aa) mitral annular velocities were averaged over the 6 periannular sites. Mitral annular long-axis velocity was lower in the HFNEF than controls at rest. Global, regional, and long-axis systolic function did not worsen with stress in the HFNEF group. The Ea decreased and the E/Ea increased with stress in the HFNEF but not in controls. The 6-minute walk distance was shorter and negatively correlated to the E/EA ratio at rest and stress in the HFNEF group. Conclusion-Impaired diastolic reserve results in stress-induced increase in the left ventricular end-diastolic pressure in patients with HFNEF giving rise to exercise intolerance. (Circ Heart Fail. 2010;3:35-43.)

No abstract