Near-Term Prediction of Sudden Cardiac Death in Older Hemodialysis Patients Using Electronic Health Records

Goldstein, Benjamin A.; Chang, Tara I.; Mitani, Aya A.; Assimes, Themistocles L.; Winkelmayer, Wolfgang C.

doi:10.2215/cjn.03050313

Cited by 29 publications

(24 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In a study by Goldstein et al using electronic medical record data, blood pressure, ultrafiltration and serum albumin predicted short-term (within a day of last outpatient hemodialysis) SCD in dialysis patients but cardiac biomarkers were not measured. [ 26 ] The rate of SCD in dialysis patients, 41 per 1000 person-years in our study, was substantially higher than that reported in the general population [ 27 ]. Implantable cardioverter-defibrillators (ICDs) can prevent SCD but are associated with a higher risk of infection [ 28 , 29 ].…”

Section: Discussioncontrasting

confidence: 59%

Association of NTproBNP and cTnI with outpatient sudden cardiac death in hemodialysis patients: the Choices for Healthy Outcomes in Caring for ESRD (CHOICE) study

Kruzan

Herzog

et al. 2016

BMC Nephrol

View full text Add to dashboard Cite

BackgroundSudden cardiac death (SCD) is the most common etiology of death in hemodialysis patients but not much is known about its risk factors. The goal of our study was to determine the association and risk prediction of SCD by serum N-terminal prohormone of brain natriuretic peptide (NTproBNP) troponin I (cTnI) in hemodialysis patients.MethodsWe measured NTproBNP and cTnI in 503 hemodialysis patients of a national prospective cohort study. We determined their association with SCD using Cox regression, adjusting for demographics, co-morbidities, and clinical factors and risk prediction using C-statistic and Net Reclassification Improvement (NRI).ResultsPatients’ mean age was 58 years and 54 % were male. During follow-up (median 3.5 years), there were 75 outpatient SCD events. In unadjusted and fully-adjusted models, NTproBNP had a significant association with the risk of SCD. Analyzed as a continuous variable, the risk of SCD increased 27 % with each 2-fold increase in NTproBNP (HR, 1.27 per doubling; 95 % CI, 1.13–1.43; p < 0.001). In categorical models, the risk of SCD was 3-fold higher in the highest tertile of NTproBNP (>7,350 pg/mL) compared with the lowest tertile (<1,710 pg/mL; HR for the highest tertile, 3.03; 95 % CI, 1.56–5.89; p = 0.001). Higher cTnI showed a trend towards increased risk of SCD in fully adjusted models, but was not statistically significant (HR, 1.17 per doubling; 95 % CI, 0.98–1.40; p = 0.08). Sensitivity analyses using competing risk models showed similar results. Improvement in risk prediction by adding cardiac biomarkers to conventional risk factors was greater with NTproBNP (C-statistic for 3-year risk: 0.810; 95 % CI, 0.757 to 0.864; and continuous NRI: 0.270; 95 % CI, 0.046 to 0.495) than with cTnI.ConclusionsNTproBNP is associated with the risk of SCD in hemodialysis patients. Further research is needed to determine if biomarkers measurement can guide SCD risk prevention strategies in dialysis patients.Electronic supplementary materialThe online version of this article (doi:10.1186/s12882-016-0230-x) contains supplementary material, which is available to authorized users.

show abstract

Section: Discussioncontrasting

confidence: 59%

Association of NTproBNP and cTnI with outpatient sudden cardiac death in hemodialysis patients: the Choices for Healthy Outcomes in Caring for ESRD (CHOICE) study

Kruzan

Herzog

et al. 2016

BMC Nephrol

View full text Add to dashboard Cite

show abstract

“…[14][15][16][17][18][19][20][21] Moreover, when comparing the "important" variables over different time horizons, previous work has similarly suggested that more "dynamic" metrics are important for nearer-term outcomes and more "stable" metrics are important for longer-term events. 17 This finding stresses the importance of machine-learning methods capable of handling large numbers of disparate predictor variables. With a large number of variables, particularly correlated ones, regularized methods such as LASSO are highly effective, since they stabilize regression coefficients via shrinkage.…”

Section: Discussionmentioning

confidence: 99%

Predicting mortality over different time horizons: which data elements are needed?

Goldstein

Pencina

Montez‐Rath

et al. 2016

Journal of the American Medical Informatics Association

View full text Add to dashboard Cite

Objective: Electronic health records (EHRs) are a resource for "big data" analytics, containing a variety of data elements. We investigate how different categories of information contribute to prediction of mortality over different time horizons among patients undergoing hemodialysis treatment. Material and Methods: We derived prediction models for mortality over 7 time horizons using EHR data on older patients from a national chain of dialysis clinics linked with administrative data using LASSO (least absolute shrinkage and selection operator) regression. We assessed how different categories of information relate to risk assessment and compared discrete models to time-to-event models. Results: The best predictors used all the available data (c-statistic ranged from 0.72-0.76), with stronger models in the near term. While different variable groups showed different utility, exclusion of any particular group did not lead to a meaningfully different risk assessment. Discrete time models performed better than time-to-event models. Conclusions: Different variable groups were predictive over different time horizons, with vital signs most predictive for near-term mortality and demographic and comorbidities more important in long-term mortality.Key words: Electronic Health Records, hemodialysis, ESRD, predictive modeling OBJECTIVEThe increasing availability of electronic health records (EHRs) offers vast and unique opportunities for biomedical research, at the core of which are predictive modeling and analytics. The presence of diverse data elements allows for the construction of prediction models using a wealth of predictors that may not all be available in more standard settings. Depending on the EHR and other linkable sources, it is generally possible to ascertain information on patients' demographics, health service utilization, diagnosed comorbidities, prescribed medications, results from laboratory tests, and vital signs. In total, these represent a source of "big data" analytics in clinical research. Owing to these variable data elements, EHRs present the opportunity to develop risk models over a range of time horizons. Studies have predicted events from within the next 12 hours 1 to up to 8 years. 2Examples include predicting the risk of adverse outcomes, including administrative events such as hospital readmission, 3 and discrete clinical events such acute kidney injury. 4 As such, each study has abstracted different pieces of information from its respective EHR.-With this in mind, we sought to explore the role EHR data elements play in predicting mortality over different time horizons. This could inform future model development efforts by prioritizing limited resources toward collecting the predictors that provide the most useful

show abstract

“…The AUROC for traditional statistical mortality and hospitalization prediction models usually fall in the range between 0.65 and 0.75 31 . In nephrology, prediction of sudden cardiac death in older HD patients was an early example of employing an advanced ML method where a random forest model yielded a AUROC of 0.79 32 . In another example, Mezzatesta et al used SVM, to predict the risk of ischemic heart disease in dialysis patients with an accuracy of approximately 92% 33 …”

Section: Applications In Kidney Diseasementioning

confidence: 99%

Artificial intelligence enabled applications in kidney disease

Chaudhuri

Long

Zhang

et al. 2020

Seminars in Dialysis

View full text Add to dashboard Cite

Artificial intelligence (AI) is considered as the next natural progression of traditional statistical techniques. Advances in analytical methods and infrastructure enable AI to be applied in health care. While AI applications are relatively common in fields like ophthalmology and cardiology, its use is scarcely reported in nephrology. We present the current status of AI in research toward kidney disease and discuss future pathways for AI. The clinical applications of AI in progression to end‐stage kidney disease and dialysis can be broadly subdivided into three main topics: (a) predicting events in the future such as mortality and hospitalization; (b) providing treatment and decision aids such as automating drug prescription; and (c) identifying patterns such as phenotypical clusters and arteriovenous fistula aneurysm. At present, the use of prediction models in treating patients with kidney disease is still in its infancy and further evidence is needed to identify its relative value. Policies and regulations need to be addressed before implementing AI solutions at the point of care in clinics. AI is not anticipated to replace the nephrologists’ medical decision‐making, but instead assist them in providing optimal personalized care for their patients.

show abstract

Near-Term Prediction of Sudden Cardiac Death in Older Hemodialysis Patients Using Electronic Health Records

Cited by 29 publications

References 26 publications

Association of NTproBNP and cTnI with outpatient sudden cardiac death in hemodialysis patients: the Choices for Healthy Outcomes in Caring for ESRD (CHOICE) study

Association of NTproBNP and cTnI with outpatient sudden cardiac death in hemodialysis patients: the Choices for Healthy Outcomes in Caring for ESRD (CHOICE) study

Predicting mortality over different time horizons: which data elements are needed?

Artificial intelligence enabled applications in kidney disease

Contact Info

Product

Resources

About