June 2017
Rebecca Pullon has recently completed her DPhil with the Department of Engineering. Her thesis focused on the development of an early warning scoring system during pregnancy, working on data from the 4P Project, funded by the Oxford NIHR Biomedical Research Centre. Here she writes about the challenges of designing an obstetric early warning score system.
While the maternal mortality rate in the UK and many other developed countries is now low, the rate of pregnancy-induced complications remains high. Complications such as pre-eclampsia, postpartum haemorrhage, or sepsis, can have long term effects for the mother and child if not treated in a timely manner. Since 2007, the UK Centre for Enquiry into Maternal Deaths have emphasised the importance of routine monitoring of pregnant women and recommended the implementation of an early warning score specific to obstetrics.
Early Warning Scores (EWS) have become a common place bedside tool in the non-pregnant adult population, used to track vital signs and initiate a clinical response when necessary (a type of “track-and-trigger” chart). In 2012, the National Early Warning Score (NEWS) was implemented in the UK, a national standardised evidence-based scoring system for adults. However, standard adult alerting systems are not appropriate for the obstetric population since physiology adapts to the pregnant state, and no UK national EWS for the obstetric population has yet been developed. Some EWS systems have been adapted for use in pregnancy, and “obstetric EWS” (ObsEWS) are now implemented in 99% of midwifery centres in NHS secondary care organisations; a significant rise from 2007 when only 19% of UK maternity units used any form of EWS for obstetric patients. An Irish national ObsEWS has been published, and in 2014 Mhyre et al. proposed an ObsEWS for the United States. However, most of the obstetric modifications to EWS are based on limited knowledge of physiology in pregnancy. A recent review of UK ObsEWS systems by Smith et al. found huge heterogeneity in the vital sign cut-off values and when a clinical response should be alerted. Furthermore, none of these systems accounted for the physiological changes associated with gestation.
The 4P study (Predicting Physiological Patterns in Pregnancy) is collecting vital sign information from 1000 women in Oxford, Newcastle, and London, who are experiencing a normal pregnancy. The vital sign data collected from this population will form a database of the expected physiology during pregnancy. It will allow us to update obstetric vital sign values from studies that are now several decades old (such as the AVON study), and establish centiles of expected physiology during pregnancy. Preliminary analysis on population demographics and obstetric history indicate similarities to the English population.
The development of an ObsEWS faces many of the same challenges as standard adult EWS: the decision as to where to set the cut-off values for each vital sign, the range of scores assigned per vital sign, at what aggregate score should a clinical response be triggered, and how the system should be validated. Yet there are additional challenges for the ObsEWS, such as:
(1) How can a scoring system that changes with gestation be presented on a paper bedside chart?
(2) How should the system be validated in a population with a low alert rate?
The first of these challenges is practical. A bedside tool must be easy to use and available. These requirements may not be met if all 40 weeks of gestation are fitted onto a single chart, or if many different charts for each gestational age are constructed. One solution to this problem could be to construct an ObsEWS per trimester of gestation, so that there are three charts that cover the antenatal period, appropriate for that stage of gestation. This would require the approximation of centiles for each trimester but would still capture significant physiological changes during pregnancy. A similar set of charts could be constructed for the postnatal period. However, fortunately, this challenge may not remain a challenge for long. There are now several examples of electronic vital sign recording systems successfully deployed within the NHS. The SEND system in Oxford allows patients’ vital signs to be entered and reviewed via a bedside tablet, and is now used for routine observations on adult wards in all acute NHS hospitals in Oxfordshire. Such a system could store an algorithm and thus automatically compute an EWS score for the appropriate stage of gestation once a set of vital signs has been entered. Furthermore, electronic systems would allow more complex EWS scoring systems to be developed, such as a scoring system that incorporates the correlation between vital signs.
The second of these challenges – method for validation of an ObsEWS system – is a much tougher challenge. The first step for validation of an EWS should be the establishment of an outcome metric, i.e. what physiological condition the chart should identify. Morbidity (death), the typical endpoint for non-obstetric EWS systems, is rare in obstetrics. Admission to ICU is difficult to consistently define between hospitals. Pregnancy-induced complications typically have also been difficult to consistently define, however in 2016 a WHO working group developed a definition for maternal morbidity which includes severe and less-severe cases. This could be a good place to start. The next step in validation is to carry out a prospective trial. Clinical intervention trials are not practical. They require cluster-randomisation, because two different alerting systems cannot be deployed within the same institution, and therefore take a considerable amount of time and money to complete. An alternative approach could be a stepped-wedge implementation, as recently demonstrated in the deployment of the SEND electronic EWS recording system in Oxfordshire. This approach proposes that evaluating roll-outs by using the effects of the steps (no randomisation) helps to manage the effects of other changes over time, which is a major drawback of the “before-and-after” study approach. As obstetric complications are fortunately rare, a prospective study of any design would have to span multiple centres over a considerable time-period.
In conclusion, an evidence-based ObsEWS system is required to appropriately monitor obstetric health. The 4P study is collecting vital sign information from a large number of women to provide an initial database with which to inform clinical thresholds. Additional challenges must be overcome before an ObsEWS can be implemented or validated. However, all these steps aim to improve the existing system, when perhaps a new tool is required to reduce the burden of pregnancy-induced complications.