NeoEPOCH: Neonatal Electronic health record data for Point Of Care Health research and continuous incremental improvement in neonatal care

在英国，每11个婴儿中就有一个早产；其中许多需要新生儿护理，涉及婴儿护理的每个部分的医疗决策（比如如何最好地支持呼吸）。许多需要新生儿护理的婴儿都有影响其一生的医疗和神经问题；这些可能受到新生儿住院期间所作决定的影响。确定哪种临床决策（即使用哪种类型的呼吸支持）是最好的理想方法是随机试验。在这里，每个婴儿都有平等的机会得到每一个决定（呼吸支持的类型），这是随机选择的，就像扔硬币一样。通过包括很多婴儿，我们可以找出哪种决定（呼吸支持类型）最有效。不幸的是，随机试验是非常昂贵和繁重的，因为它是耗时的收集所有的信息。因此，只有一小部分新生儿临床决定在随机试验中得到了检验，因此大多数决定只是有根据的猜测。我想让随机临床试验更便宜、更容易，这样所有日常的新生儿决策都可以基于最好的研究——随机临床试验。我的计划是去掉最昂贵的部分，数据收集。目前，这是通过付钱给一名研究人员来检查每个婴儿的医疗记录来完成的。我想直接从婴儿的电子健康记录（EHR）中获取所有信息，这是一种计算机化的医疗记录。自2007年以来，所有英国新生儿单位都使用了一个电子病历系统。该电子病历中关于所有接受新生儿护理的婴儿的信息（除去姓名和其他识别信息）保存在国家新生儿研究数据库（NNRD）中。NNRD已经用于其他目的，我想把它用于随机试验，使它们更便宜、更容易。这些试验被称为“护理点试验”（PoCT）。有许多问题需要解决。首先，电子病历中的信息可能不完全准确。我们用1来处理这个。查看其他电子病历系统，找出衡量和报告新生儿电子病历准确性的最佳方法2。看看其他人是如何使电子病历数据更准确的。使用这些技术，并通过集群随机对照试验来衡量它们是否有效；这就像一个更大的随机临床试验，但不是每个婴儿都有平等的机会得到一个或另一个决定，每个新生儿单位都有平等的机会采用一种或另一种技术。我们认为，如果父母更多地参与婴儿的电子病历数据，它将更准确。我们将通过要求父母帮助我们帮助他们使用孩子的电子病历数据来测试这一点。然后，我们将在同一群随机试验中进行试验，看看它是否使电子病历更准确。我们认为采取这些步骤将使电子病历对PoCT足够准确，但我们也将在数学上研究如何使用电子病历数据，即使它不是完全准确的。其次，我们要确保PoCT将以正确的方式提出正确的研究问题。我们用1来做。询问感兴趣的群体（医生、护士、家长、病人等），他们认为PoCT最重要的“结果”是什么。“结果”是一项试验的结果，比如婴儿在家是否需要氧气。询问同一组最重要的临床决策（如如何支持婴儿的呼吸或开始用什么牛奶喂养）在PoCT中进行测试。第三，我们想表明我们可以使用新生儿电子病历进行PoCT。为此，我们将在20个新生儿单位进行为期一年的小型PoCT。我们将衡量它的准确性，以及它是否更容易、更便宜。最后，我们希望确保PoCT的知识能够快速有效地在NHS中传播，以改善婴儿的照顾方式，因此我们将从已经这样做的另一个卫生系统的成功和错误中学习。

1 in 11 UK babies are born prematurely; many of these need neonatal care that involves medical decisions about every part of a baby's care (like how best to support breathing). Many babies who require neonatal care have medical and neurological problems that affect them throughout their lives; these may be influenced by decisions made during their neonatal stay. The ideal way to work out which clinical decision (i.e. which type of breathing support to use) is the best is a randomised trial. Here each baby has an equal chance of being given each decision (type of breathing support) and this is chosen by chance, like tossing a coin. By including lots of babies we can work out which decision (type of breathing support) works best. Unfortunately randomised trials are very expensive and burdensome because it is time-consuming to collect all the information. As a result, only a small number of neonatal clinical decisions have been tested in randomised trials and so most decisions are only educated guesses. I want to make randomised clinical trials cheaper and easier so all day-to-day neonatal decisions can be based on the best research - randomised clinical trials. I plan to do this is by getting rid of the most expensive part, data collection. At the moment this is done by paying a researcher to check each baby's medical notes. I want to get all the information straight from a baby's electronic health record (EHR), a computerised version of the medical notes. Since 2007 all UK neonatal units have used one EHR system. Information from this EHR, on all babies admitted for neonatal care (with names and other identifying information removed), is held in the National Neonatal Research Database (NNRD). The NNRD is already used for other purposes, I want to use it for randomised trials to make them much cheaper and easier. These can be called Point-of-Care Trials (PoCT).There are a number of problems to solve. Firstly, information in the EHR may not be completely accurate. We will deal with this by1. Looking at other EHR systems to work out the best way to measure and report the accuracy of the neonatal EHR 2. Looking to see how other people have made EHR data more accurate3. Using these techniques and measuring to see if they work using a cluster randomised controlled trial; this is like a larger randomised clinical trial but instead of each baby having an equal chance of getting one decision or another, each neonatal unit will have an equal chance of having one technique or anotherWe think that if parents are more involved in their baby's EHR data it will be more accurate. We will test this by asking parents to help us to help them use their baby's EHR data. Then we will try this out in the same cluster randomised trial to see if it makes EHR more accurate. We think that taking these steps will make the EHR accurate enough for PoCT, but we will also work out mathematically how to use EHR data even if it is not completely accurate.Secondly, we want to make sure that PoCT will ask the right research questions in the right ways. We will do this by1. Asking interested groups (doctors, nurses, parents, patients, etc.) what they think the most important "outcomes" for PoCT are. An "outcome" is a result of a trial, like whether a baby needs oxygen at home2. Asking the same groups what the most important clinical decisions (like how to support a baby's breathing or what milk to start feeding with) are to test in PoCTThirdly, we want to show that we can carry out a PoCT using the neonatal EHR. To do this we will run a small PoCT over 20 neonatal units and one year. We will measure how accurate it is and whether it is easier and cheaper.Finally we want to make sure that knowledge from PoCT can be quickly and effectively communicated across the NHS to improve the way babies are looked after so we will learn from the successes and mistakes of another health system that already does this.