Search

The Big Data for Education Spoke of the NSF Northeast Big Data Innovation Hub and ETS co-sponsored an educational data mining competition in which contestants were asked to predict efficient time use on the NAEP 8th grade mathematics computer-based assessment, based on the log file of a student's actions on a prior portion of the assessment. In this work, a combined approach of process mining and expert feature engineering was used to build a large set of features that were then trained with an Extreme Gradient Boosting machine learning model to classify students based on whether they would use their time efficiently. Predictions were evaluated throughout the competition on half of a hidden data set and then the final results were based on the second half of the hidden data set. The approach used here earned the top score in the competition. The work presented elaborates on the combined technique for analyzing computer-based assessment log-file data with the hope that this approach will offer valuable insights for future predictive model building in educational data mining.

The key message from the 1958 Edelman study states that combinations of external gains or losses of sodium, potassium and water leading to an increase of the fraction (total body sodium plus total body potassium) over total body water will raise the serum sodium concentration ([Na]S), while external gains or losses leading to a decrease in this fraction will lower [Na]S. A variety of studies have supported this concept and current quantitative methods for correcting dysnatremias, including formulas calculating the volume of saline needed for a change in [Na]S are based on it. Not accounting for external losses of sodium, potassium and water during treatment and faulty values for body water inserted in the formulas predicting the change in [Na]S affect the accuracy of these formulas. Newly described factors potentially affecting the change in [Na]S during treatment of dysnatremias include the following: (a) exchanges during development or correction of dysnatremias between osmotically inactive sodium stored in tissues and osmotically active sodium in solution in body fluids; (b) chemical binding of part of body water to macromolecules which would decrease the amount of body water available for osmotic exchanges; and (c) genetic influences on the determination of sodium concentration in body fluids. The effects of these newer developments on the methods of treatment of dysnatremias are not well-established and will need extensive studying. Currently, monitoring of serum sodium concentration remains a critical step during treatment of dysnatremias.

BackgroundAcute kidney injury (AKI) is increasingly encountered in community settings and contributes to morbidity, mortality, and increased resource utilization worldwide. In low-resource settings, lack of awareness of and limited access to diagnostic and therapeutic interventions likely influence patient management. We evaluated the feasibility of the use of point-of-care (POC) serum creatinine and urine dipstick testing with an education and training program to optimize the identification and management of AKI in the community in 3 low-resource countries.Methods and findingsPatients presenting to healthcare centers (HCCs) from 1 October 2016 to 29 September 2017 in the cities Cochabamba, Bolivia; Dharan, Nepal; and Blantyre, Malawi, were assessed utilizing a symptom-based risk score to identify patients at moderate to high AKI risk. POC testing for serum creatinine and urine dipstick at enrollment were utilized to classify these patients as having chronic kidney disease (CKD), acute kidney disease (AKD), or no kidney disease (NKD). Patients were followed for a maximum of 6 months with repeat POC testing. AKI development was assessed at 7 days, kidney recovery at 1 month, and progression to CKD and mortality at 3 and 6 months. Following an observation phase to establish baseline data, care providers and physicians in the HCCs were trained with a standardized protocol utilizing POC tests to evaluate and manage patients, guided by physicians in referral hospitals connected via mobile digital technology. We evaluated 3,577 patients, and 2,101 were enrolled: 978 in the observation phase and 1,123 in the intervention phase. Due to the high number of patients attending the centers daily, it was not feasible to screen all patients to assess the actual incidence of AKI. Of enrolled patients, 1,825/2,101 (87%) were adults, 1,117/2,101 (53%) were females, 399/2,101 (19%) were from Bolivia, 813/2,101 (39%) were from Malawi, and 889/2,101 (42%) were from Nepal. The age of enrolled patients ranged from 1 month to 96 years, with a mean of 43 years (SD 21) and a median of 43 years (IQR 27-62). Hypertension was the most common comorbidity (418/2,101; 20%). At enrollment, 197/2,101 (9.4%) had CKD, and 1,199/2,101 (57%) had AKD. AKI developed in 30% within 7 days. By 1 month, 268/978 (27%) patients in the observation phase and 203/1,123 (18%) in the intervention phase were lost to follow-up. In the intervention phase, more patients received fluids (observation 714/978 [73%] versus intervention 874/1,123 [78%]; 95% CI 0.63, 0.94; p = 0.012), hospitalization was reduced (observation 578/978 [59%] versus intervention 548/1,123 [49%]; 95% CI 0.55, 0.79; p < 0.001), and admitted patients with severe AKI did not show a significantly lower mortality during follow-up (observation 27/135 [20%] versus intervention 21/178 [11.8%]; 95% CI 0.98, 3.52; p = 0.057). Of 504 patients with kidney function assessed during the 6-month follow-up, de novo CKD arose in 79/484 (16.3%), with no difference between the observation and intervention phase (95% CI 0.91, 2.47; p = 0.101). Overall mortality was 273/2,101 (13%) and was highest in those who had CKD (24/106; 23%), followed by those with AKD (128/760; 17%), AKI (85/628; 14%), and NKD (36/607; 6%). The main limitation of our study was the inability to determine the actual incidence of kidney dysfunction in the health centers as it was not feasible to screen all the patients due to the high numbers seen daily.ConclusionsThis multicenter, non-randomized feasibility study in low-resource settings demonstrates that it is feasible to implement a comprehensive program utilizing POC testing and protocol-based management to improve the recognition and management of AKI and AKD in high-risk patients in primary care.

Kidney disease is prevalent in low-resource settings worldwide, but tests for its diagnosis are often unavailable. The saliva urea nitrogen (SUN) dipstick is a laboratory and electricity independent tool, which may be used for the detection of kidney disease. We investigated the feasibility and performance of its use in diagnosing kidney disease in community settings in Africa. Adult patients at increased risk of kidney disease presenting to three community health centres, a rural district hospital and a central hospital in Malawi were recruited between October 2016 and September 2017. Patients underwent concurrent SUN and creatinine testing at enrolment, and at 1 week, 1 month, 3 months and 6 months thereafter. Of 710 patients who presented at increased risk of kidney disease, 655 (92.3%) underwent SUN testing at enrolment, and were included (aged 38 (29-52) years, 367 (56%) female and 333 (50.8%) with HIV). Kidney disease was present in 482 (73.6%) patients and 1479 SUN measurements were made overall. Estimated glomerular filtration rate (eGFR) correlated with SUN (r=-0.39; ptest pad #1) compared with patients with non-elevated SUN (p

Substantial heterogeneity in practice patterns around the world has resulted in wide variations in the quality and type of dialysis care delivered. This is particularly so in countries without universal standards of care and governmental (or other organizational) oversight. Most high-income countries have developed such oversight based on documentation of adherence to standardized, evidence-based guidelines. Many low- and lower-middle-income countries have no or only limited organized oversight systems to ensure that care is safe and effective. The implementation and oversight of basic standards of care requires sufficient infrastructure and appropriate workforce and financial resources to support the basic levels of care and safety practices. It is important to understand how these standards then can be reasonably adapted and applied in low- and lower-middle-income countries.

The global nephrology community recognizes the need for a cohesive strategy to address the growing problem of end-stage kidney disease (ESKD). In March 2018, the International Society of Nephrology hosted a summit on integrated ESKD care, including 92 individuals from around the globe with diverse expertise and professional backgrounds. The attendees were from 41 countries, including 16 participants from 11 low- and lower-middle-income countries. The purpose was to develop a strategic plan to improve worldwide access to integrated ESKD care, by identifying and prioritizing key activities across 8 themes: (i) estimates of ESKD burden and treatment coverage, (ii) advocacy, (iii) education and training/workforce, (iv) financing/funding models, (v) ethics, (vi) dialysis, (vii) transplantation, and (viii) conservative care. Action plans with prioritized lists of goals, activities, and key deliverables, and an overarching performance framework were developed for each theme. Examples of these key deliverables include improved data availability, integration of core registry measures and analysis to inform development of health care policy; a framework for advocacy; improved and continued stakeholder engagement; improved workforce training; equitable, efficient, and cost-effective funding models; greater understanding and greater application of ethical principles in practice and policy; definition and application of standards for safe and sustainable dialysis treatment and a set of measurable quality parameters; and integration of dialysis, transplantation, and comprehensive conservative care as ESKD treatment options within the context of overall health priorities. Intended users of the action plans include clinicians, patients and their families, scientists, industry partners, government decision makers, and advocacy organizations. Implementation of this integrated and comprehensive plan is intended to improve quality and access to care and thereby reduce serious health-related suffering of adults and children affected by ESKD worldwide.

BackgroundBioimpedance analysis (BIA) is well established to assess body composition. Agreements between single- and multi-frequency bioimpedance (SF-BIA, MF-BIS) measurements in subjects undergoing 6 or 3 times/week hemodialysis (HD) were analyzed.MethodsTotal body water (TBW) and intra- and extracellular fluid (ICF, ECF) of subjects enrolled in the Frequent Hemodialysis Network (FHN) Daily Trial (www.clinicaltrials.gov No. NCT00264758) were measured with a Hydra 4200 at baseline (BL) and at 5 months (M5). Volumes were computed using SF (at 50 kHz) and MF approaches. Agreement was assessed by means of linear regression and Bland-Altman analysis and treatment effects by t test.Results35 subjects (17 on the more frequent regimen, 26 males, 20 African-American, 48 ± 9 years, pre-HD weight 84 ± 19 kg) were studied. Assessments with SF-BIA and MF-BIS correlated significantly at BL and M5 in both arms. No proportional errors, but systematic biases over the entire range of values were found at BL and M5. Agreement did not differ between subjects randomized to either HD treatment arm at both time points. MF-BIS appears to have better precision than SF-BIA allowing the observation of a significant treatment effect by the intervention [-1.5 (95% CI -2.5 to -0.5) l] on ECF, not found for ECF SF-BIA. Precision also affected the statistical power of the SF-BIA data in the current analysis.ConclusionBoth methods showed agreement without significant proportional errors regardless of HD frequency and can be used for longitudinal analyses. SF-BIA has lower precision which needs thorough consideration in the design of future trials with similar outcomes.

BackgroundCurrent evidence does not clearly identify the contribution of kidney function decline and mortality to racial disparities in end-stage renal disease (ESRD) incidence. We used observed estimated glomerular filtration rate (eGFR) to project the time of onset of kidney failure and examined mortality to better understand these racial disparities.Study designRetrospective cohort.Setting & participantsAdult members of Kaiser Permanente Southern California in 2003-2009 with more than 2 serum creatinine tests and more than 180 days between tests: 526,498 whites, 350,919 Hispanics, 136,923 blacks, and 105,476 Asians.PredictorRace/ethnicity.OutcomesESRD (dialysis or transplantation); mortality.MeasurementseGFR decline was modeled using linear regression. Kidney failure was projected based on predicted eGFR

Frequent hemodialysis can alter volume status, blood pressure, and the concentration of osmotically active solutes, each of which might affect residual kidney function (RKF). In the Frequent Hemodialysis Network Daily and Nocturnal Trials, we examined the effects of assignment to six compared with three-times-per-week hemodialysis on follow-up RKF. In both trials, baseline RKF was inversely correlated with number of years since onset of ESRD. In the Nocturnal Trial, 63 participants had non-zero RKF at baseline (mean urine volume 0.76 liter/day, urea clearance 2.3 ml/min, and creatinine clearance 4.7 ml/min). In those assigned to frequent nocturnal dialysis, these indices were all significantly lower at month 4 and were mostly so at month 12 compared with controls. In the frequent dialysis group, urine volume had declined to zero in 52% and 67% of patients at months 4 and 12, respectively, compared with 18% and 36% in controls. In the Daily Trial, 83 patients had non-zero RKF at baseline (mean urine volume 0.43 liter/day, urea clearance 1.2 ml/min, and creatinine clearance 2.7 ml/min). Here, treatment assignment did not significantly influence follow-up levels of the measured indices, although the range in baseline RKF was narrower, potentially limiting power to detect differences. Thus, frequent nocturnal hemodialysis appears to promote a more rapid loss of RKF, the mechanism of which remains to be determined. Whether RKF also declines with frequent daily treatment could not be determined.

Predialysis volume overload is the sum of interdialytic weight gain (IDWG) and residual postdialysis volume overload. It results mostly from failure to achieve an adequate volume status at the end of the dialysis session. Recent developments in bioimpedance spectroscopy and possibly relative plasma volume monitoring permit noninvasive volume status assessment in hemodialysis patients. A large proportion of patients have previously been shown to be chronically volume overloaded predialysis (defined as >15% above 'normal' extracellular fluid volume, equivalent to >2.5 liters on average), and to exhibit a more than twofold increased mortality risk. By contrast, the magnitude of the mortality risk associated with IDWG is much smaller and only evident with very large weight gains. Here we review the available evidence on volume overload and IDWG, and question the use of IDWG as an indicator of 'nonadherence' by describing its association with postdialysis volume depletion. We also demonstrate the relationship between IDWG, volume overload and predialysis serum sodium concentration, and comment on salt intake. Discriminating between volume overload and IDWG will likely lead to a more appropriate management of fluid withdrawal during dialysis. Consensually, the present authors agree that this discrimination should be among the primary goals for dialysis caretakers today. In consequence, we recommend objective measures of volume status beyond mere evaluations of IDWG.

ObjectiveIncreased body mass index (BMI) is associated with reduced all-cause and cardiovascular (CV) mortality in hemodialysis (HD) patients, whereas CV risk increases with BMI in the general population. In the general population, obesity is associated with inflammation, decreased high-density lipoprotein (HDL) cholesterol, increased low-density lipoprotein (LDL) cholesterol, and triglycerides (TGs), all risk factors for CV disease. Low-density lipoprotein cholesterol does not predict CV risk in HD, whereas increased C-reactive protein and interleukin-6 (IL-6), low HDL and apolipoprotein (apo) AI, and increased fasting TGs do predict risk. Renal failure is associated with dyslipidemia and inflammation in normal-weight patients. We hypothesized that the effects of obesity may be obscured by renal failure in HD.MethodsWe explored the relationship between adipose tissue pools and distribution, i.e., subcutaneous adipose tissue (SAT) and visceral adipose tissue (VAT) (measured by magnetic resonance imaging) and measures of inflammation (C-reactive protein, IL-6, ceruloplasmin, and alpha1 acid glycoprotein), HDL and LDL cholesterol, total TGs, apo AI, apo B, apo CII (an activator of lipoprotein lipase), apo CIII (an inhibitor of lipoprotein lipase), and the adipokines, leptin and adiponectin, in 48 patients with prevalent HD.Results and conclusionsTotal TG concentrations were positively correlated with VAT controlled for age, sex, and weight. Both apo CII and apo CIII were correlated only with VAT. Adiponectin was inversely correlated with VAT, and leptin was positively associated with SAT. C-reactive protein and alpha1 acid glycoprotein were weakly associated with SAT, whereas ceruloplasmin was strongly associated with VAT according to multiple regression analysis. In contrast, apo B, LDL, apo AI, HDL, and IL-6 were not correlated with any measure of body composition, potentially mitigating the effects of obesity in HD.