Study investigating Early Warning Score (EWS) -Aiming at predicting which inpatients will get worse-using poor methods, poor reporting and lack of data and details It is biased because it was discovered by a British scientist in a systematic review.
As a result, researchers at the University of Oxford’s Center for Medical Statistics, Steven Gerry, led by MSc say, “clinicians have little knowledge of how such scores work in the clinical setting.”
“Thus, clinicians need to be aware that they rely on these scores to identify clinical deterioration in their patients.”
Research Release Along BMJ May 20th.
EWS is usually aimed at determining if an inpatient may be exacerbated based on vital signs such as heart rate, oxygen levels, and blood pressure, addressing adverse events, and reducing unnecessary deaths. I will.
Scores like this are “very widely used and trusted,” Steven Jerry Medscape News UK..
He also noted the National Institute for National Healthcare Care Excellence Guidelines for Adults with acute illness in hospital It is recommended to use them on admission and every 12 hours thereafter.
As a result, they are “more likely to be used than any other clinical predictive model” and are therefore “increased in pressure” for use in “primary care, ambulances and, in some cases, nursing homes.”
To investigate further, researchers searched the Medline, CINAHL, PsycInfo, and Embase databases to conduct an EWS study developed for inpatients in adult hospitals and published by June 2019. It was
They included 95 studies, 11 of which focused solely on EWS development, 23 of both EWS development and external validation, and 61 of external validation only.
The majority of the 34 unique EWSs were developed for use in the UK (29%) or the US (38%).
The most common predictors were respiratory rate (88%), heart rate (83%), oxygen saturation, body temperature, and systolic blood pressure, each used in 71% of EWS.
In contrast, age is 38% and gender is 9%, which is less common.
The most common predictive outcome was death, which occurred in 44% of development studies and 79% of validation studies. Outcomes had varying forecast periods, with 35% being the most commonly used 24 hours.
Looking at the quality of the study, teams often find that key details of the analytical population are not reported in either developmental or validation studies, there are fewer samples and fewer events in both model development and external validation. I found it too late.
They determined that there were only nine EWSs presented in sufficient detail to allow individual risk prediction, and internal validation was commonly done, but recommendations such as bootstrapping and cross-validation were recommended. The techniques used were rarely used.
The performance of the model was based on assessment discrimination, or which patient produced the outcome of interest, in 82% of the studies, but it used calibration to predict predictive risk and observed event rates. Only 15% decided the agreement between the two.
Finally, the team investigated the risk of bias using the bias assessment risk of the PROBAST predictive model. This includes participant selection, forecasting, results and analysis.
This indicates a high risk of bias in participant selection in 55% of the studies, 5% of the predictors, 66% of the results, and 98% of the analytical methods.
“Scores developed using improper methods will likely result in poor scoring performance,” they said. “Inadequate methods in external validation studies lead to scoring system implementations and predictive abilities. And may give false guarantees about generalizability. “
“It may explain why recent systematic reviews have found little evidence of the clinical efficacy of EWS,” he said.
Researchers are finding it increasingly important that electronic medical records are used to record vital signs and EWS, and that the development of higher scores makes it “important” that future research be “top quality”. Is added.
“The move to electronic implementation of EWS offers the opportunity to introduce better scoring systems, especially with increasing interest in modern model building approaches such as machine learning and artificial intelligence.”
“But this potential may never be achieved if methodologies and reporting criteria are not improved.”
This study was funded by the National Institutes of Health (NIHR) and Cancer Research UK.
Peter J. Watkinson is Sensyne Health’s Chief Medical Officer and holds a stake in the company. Timothy Bonnici receives royalties from Sensyne Health. No other potential conflicts of interest have been declared.
BMJ 2020; 369: m1501 Doi: 10.1136 / bmj.m1501