Search

To develop and evaluate a high-dimensional, data-driven model to identify patients at high risk of clinical deterioration from routinely collected electronic health record (EHR) data.In this single-center, retrospective cohort study, 488 patients with single-ventricle and shunt-dependent congenital heart disease6 months old were admitted to the cardiac intensive care unit before stage 2 palliation between 2014 and 2019. Using machine-learning techniques, we developed the Intensive care Warning Index (I-WIN), which systematically assessed 1028 regularly collected EHR variables (vital signs, medications, laboratory tests, and diagnoses) to identify patients in the cardiac intensive care unit at elevated risk of clinical deterioration. An ensemble of 5 extreme gradient boosting models was developed and validated on 203 cases (130 emergent endotracheal intubations, 34 cardiac arrests requiring cardiopulmonary resuscitation, 10 extracorporeal membrane oxygenation cannulations, and 29 cardiac arrests requiring cardiopulmonary resuscitation onto extracorporeal membrane oxygenation) and 378 control periods from 446 patients.At 4 hours before deterioration, the model achieved an area under the receiver operating characteristic curve of 0.92 (95% confidence interval, 0.84-0.98), 0.881 sensitivity, 0.776 positive predictive value, 0.862 specificity, and 0.571 Brier skill score. Performance remained high at 8 hours before deterioration with 0.815 (0.688-0.921) area under the receiver operating characteristic curve.I-WIN accurately predicted deterioration events in critically-ill infants with high-risk congenital heart disease up to 8 hours before deterioration, potentially allowing clinicians to target interventions. We propose a paradigm shift from conventional expert consensus-based selection of risk factors to a data-driven, machine-learning methodology for risk prediction. With the increased availability of data capture in EHRs, I-WIN can be extended to broader applications in data-rich environments in critical care.

Accurately predicting drought a few months in advance is important for drought mitigation and agricultural and water resources management, especially for a river basin like that of the Yellow River in North China. However, summer drought predictability over the Yellow River basin is limited because of the low influence from ENSO and the large interannual variations of the East Asian summer monsoon. To explore the drought predictability from an ensemble prediction perspective, 29-year seasonal hindcasts of soil moisture drought, taken directly from several North American multimodel ensemble (NMME) models with different ensemble sizes, were compared with those produced by combining bias-corrected NMME model predictions and variable infiltration capacity (VIC) land surface hydrological model simulations. It was found that the NMME/VIC approach reduced the root-mean-square error from the best NMME raw products by 48% for summer soil moisture drought prediction at the lead-1 season, and increased the correlation significantly. Within the NMME/VIC framework, the multimodel ensemble mean further reduced the error from the best single model by 6%. Compared with the NMME raw forecasts, NMME/VIC had a higher probabilistic drought forecasting skill in terms of a higher Brier skill score and better reliability and resolution of the ensemble. However, the performance of the multimodel grand ensemble was not necessarily better than any single model ensemble, suggesting the need to optimize the ensemble for a more skillful probabilistic drought forecast.

The main objective of this study is to analyse the relationship between atmospheric circulation and high precipitation in Poland, using selected classifications of circulation types. Daily precipitation totals from 48 synoptic stations and three classifications of circulation types ( Litynski, 1969; Niedzwiedz, 1981, 2008; Ustrnul, 1997 ) were used. These classifications were described and compared. The relationships between high precipitation and atmospheric circulation were analysed by (i) determining those circulation types that favour the occurrence of high precipitation, (ii) analysing the spatial differentiation of the probability of high precipitation events for chosen cyclonic circulation types and (iii) the usefulness of the classifications to analyse relations between high precipitation in Poland and atmospheric circulation was assessed. Any study of relations between atmospheric circulation and climate elements demands a preceding detailed assessment of the usefulness of a given classification for a defined problem. The results for relations between high precipitation occurrence and circulation depend on the rules applied to classify circulation types. Ustrnul’s classification seems to be best suited to study the relationship between heavy precipitation and circulation in the southern part of Poland. This result is based on two methods, the range of probabilities for high precipitation in each circulation type of a given classification (Pr diff ) and the Brier skill score (BSS). The Niedzwiedz classification mainly produces very good results in central southern Poland where its probability range is clearly wider than elsewhere. Both methods seem to be less useful to study the relationships between high precipitation and circulation in northern Poland. Litynski’s classification delivers a greater role of the cyclonic types in forming high precipitation, but no definite prevailing type can be identified at many of the stations. When analysing relations, regional classification of circulation should be applied rather than macro-scales ones.

A novel approach is presented for the evaluation of circulation type classifications (CTCs) in terms of their capability to predict surface climate variations. The approach is analogous to that for probabilistic meteorological forecasts and is based on the Brier skill score. This score is shown to take a particularly simple form in the context of CTCs and to quantify the resolution of a climate variable by the classifications. The sampling uncertainty of the skill can be estimated by means of nonparametric bootstrap resampling. The evaluation approach is applied for a systematic intercomparison of 71 CTCs (objective and manual, from COST Action 733) with respect to their ability to resolve daily precipitation in the Alpine region. For essentially all CTCs, the Brier skill score is found to be higher for weak and moderate compared to intense precipitation, for winter compared to summer, and over the north and west of the Alps compared to the south and east. Moreover, CTCs with a higher number of types exhibit better skill than CTCs with few types. Among CTCs with comparable type number, the best automatic classifications are found to outperform the best manual classifications. It is not possible to single out one ‘best’ classification for Alpine precipitation, but there is a small group showing particularly high skill.

We derive autocorrelation-robust asymptotic variances of the Brier score and Brier skill score, which are generally applicable in circumstances with weak serial correlation. A simulation experiment and an empirical application from macroeconomics underscore the importance of taking care of serial correlation. We find that the conventional variances are too conservative to account for the sampling variability in estimating the Brier (skill) score.