Implementation of the Maternal Early Warning Criteria during early postpartum. A prospective observational study

Objective: To evaluate the implementation of an early warning system in obstetric patients (MEWC) during the first two hours after delivery in a single tertiary-care hospital.

The MEWC system implementation was carried out from 15th March to 15th September 2018, over 1166 patients. The parameters collected were systolic and diastolic blood pressure, heart rate, oxygen saturation, diuresis, uterine involution, and bleeding. If a parameter was not within defined limits, an obstetrician first examined the patient, determining the need to call the anesthesiologist. We carried out a sensitivity-specificity study of the trigger and multivariate analysis of the factors involved in developing potentially fatal disorders (PFD), reintervention, critical care admission, and stay.

The protocol was triggered in 75 patients (6.43%). The leading cause of alarm activation was the altered systolic blood pressure (32 [42.7%] patients), and eleven developed PFD. Twenty-eight patients were false-negatives. Sensitivity and specificity of MEWC protocol were 0.28 (0.15, 0.45) and 0.94 (0.93, 0.96), respectively. Multivariate analysis showed a relationship between alarm activation and PFD.

Our MEWC protocol presented low sensitivity and high specificity, having a significant number of false-negative patients.

Pregnancy, Morbidity, Warning system, Obstetric labor complications, Risk management December 12, 2020 1/10 . CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 12, 2020. ; https://doi.org/10.1101/2020.09.15.20194910doi: medRxiv preprint NOTE: This preprint reports new research that has not been certified by peer review and should not be used to guide clinical practice.

1 Lowering morbidity and mortality in the obstetric patient continues to be a 2 quality-of-care criterion for healthcare centers. The development over the few decades 3 shows a remarkable improvement in healthcare indicators in this population group [1]. 4 Although mortality rates are very low for developed countries, the impact is high, both 5 for the social repercussion that it involves and the years of life lost. 6 One of the approaches to reduce maternal morbidity and mortality has been using 7 tools that would enable the rapid identification of patients who would benefit most from 8 an aggressive intervention or a higher level of care. In December 2007, a review on 9 maternal mortality concluded that 40-50% of the UK’s mortality rates are preventable. 10 However, the early warning signs were often not recognized [2]. Since then, most UK 11 hospitals have adopted Maternity Early Obstetric Warning Criteria (MEWS) protocols, 12 with these being an auditable quality criterion for the centres [3,4]. 13 Unfortunately, the implementation of these protocols is not universal, and its use is 14 not common practice outside English-speaking countries. Furthermore, although the 15 adoption of these protocols has shown an improvement in the quality of assistance to 16 pregnant women and the detection of adverse effects, it has not shown any impact on 17 reducing maternal mortality [5]. 18 This study evaluates the usefulness and feasibility of implementing a modified 19 obstetric early warning system in our center and studying its possible relationship with 20 the detection and identification of adverse effects. 21

22 Study design and setting 23 This study analyses the implementation of a MEWC protocol in a tertiary-care hospital 24 (Hospital Fundaci´ on Jim´ enez D´ ıaz, Madrid, Spain). The hospital has all the medical and25 surgical specialties but does not have a specific building for gynecology and obstetrics. 26 We designed a single-arm prospective cohort study developed during the first six months 27 after implementing the protocol (first day of implementation: 15th March 2018). Our 28 center’s ethics committee approved our research before the start of patient recruitment 29 (Date of Approval: 12th March 2018, Code: FJD-MEOWS-17-01. S1 File, original in 30 Spanish). This study followed the Strengthening the Reporting of Observational Studies 31 in Epidemiology (STROBE) reporting guideline for cohort studies [6]. 32 Sample size calculation 33 For calculating the sample size [7], we considered that our model should have a 34 minimum sensitivity and specificity of 90%. We knew that the incidence of maternal 35 morbidity in our environment was 5% [8]. Thus, the required sample size with a 36 precision of 0.1, a type I error of 0.05, and a dropout rate of 10% were 692 patients. 37 Since the number of births in the hospital was around one thousand between 1st 38 September 2017 and 28th February 2018, we decided to include all patients who gave 39 birth between 15th March and 15th September 2018, except those who refused to 40 provide their data to the study after informed consent. 41 Procedures 42 We designed our center’s MEWC protocol based on the one described by Mhyre et 43 al. [9]. We monitored systolic and diastolic blood pressure, heart rate, oxygen 44 saturation, and diuresis at 10, 30, 60, 90, and 120 minutes. By consensus between 45 December 12, 2020 2/10 . CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 12, 2020. ; https://doi.org/10.1101/2020.09.15.20194910doi: medRxiv preprint anesthesia and obstetrics departments, we included uterine involution measurement (by 46 manual exploration), diuresis, and any bleeding greater than 500 ml in the protocol. We 47 defined bleeding or lack of uterine involution as obstetric causes. S2 File shows the data 48 collection form. 49 Two months before the start of implementation, we provided training sessions on the 50 usefulness of the MEWC protocols and the process of implementing them in our 51 hospital to professionals from both services, including midwives. After implementing the 52 protocol, nurses and midwives monitored patients for 2 hours after giving birth or 53 undergoing a cesarean delivery by filling in the data form, either in the delivery room or 54 in the PACU. They also collected data on previous parity, preeclampsia, and multiple 55 births as possible risk factors. For each variable in the protocol, specific values obliged 56 the midwives to call the obstetricians, who had to assess the patient within 15 minutes. 57 If the obstetricians did not resolve the alarm, they would make a second call to the 58 anesthesiologists. Independent of the form’s data collection, the anesthesiologists 59 followed all patients who had given birth until discharge. 60 Outcomes 61 Following the recommendations made by the WHO [10], the outcome of our study was 62 the rate of potential fatal disorder (PFD) during the stay, defined as one of the 63 following criteria: access to a critical care unit (CCU), surgery within two hours of 64 delivery, or length of stay of more than seven days depending on the activation of the 65 alert [11]. As secondary outcomes, we measured the alert activation relation on each of 66 the PFD criteria separately. 67 Statistical analysis 68 We used R v4.0.2 (The R Foundation for Statistical Computing, Vienna, Austria) and 69 RStudio 1.2.5033 (RStudio PBC, Boston, USA) to perform statistical analysis. We 70 analyzed outcomes depending on the alarm activation. We described discrete and 71 continuous variables as number and percentage and median (interquartile range [IQR]), 72 and their differences analyzed using the Pearson test or the Wilcoxon rank-sum tests. 73 We calculated the sensitivity, specificity, positive and negative predictive value (PPV 74 and NPV) of alert activation for PFD and each criterion separately, with a 95% 75 confidence interval. We performed a multivariate logistic analysis to study the 76 association of the outcomes with alarm activation, preeclampsia, parity, type of delivery, 77 and multiple births. We presented the results in forest plots as an odds ratio with a 78 95% confidence interval. We used Cox regression for multivariate analysis of length of 79 stay, showing the results in forest plot as a hazard ratio with a 95% confidence interval. 80 To avoid errors by multiple comparisons, we calculated the respective q-value for each 81 p-value to maintain a false discovery rate below 5% [12]. We considered comparisons in 82 which the p-value and q-value were below .05 as being statistically significant. We 83 provide the original study databases, the step-by-step statistical analysis, and the 84 document in R-Markdown format in S3 File, S4 File, and S5 File. 85

86 During the study period, there were 1169 deliveries at our hospital. Only three patients 87 declined to participate in the study. Since there were no losses to follow-up, we included 88 1166 patients. Fig 1 shows the STROBE flow chart. 89 The patients had a median age of 34 (31-37) years, without previous deliveries, and 90 had a rate of cesarean delivery and instrumental delivery of 23.2% and 15.1%, 91 December 12, 2020 3/10 . CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 12, 2020. ; https://doi.org/10.1101/2020.09.15.20194910doi: medRxiv preprint Fig 1. Patients Flowchart according to the STROBE statement respectively. The protocol alarm was activated in 75 patients (6.43%). The cohort of 92 subjects with an activated alarm had a higher rate of cesarean delivery (37.3% vs 22.3%, 93 p = 0.005), preeclampsia (17.3% vs 4.8%, p ¡ 0.001), and multiple birth (6.7% vs 1.5%). 94 Also, the PFD rate was higher (14.7% vs 2.6%, p ¡ 0.001), as was the CCU admission 95 rate (12% vs 0.8%, p ¡ 0.001) and length of stay (median: 2.9 [2.3-3.6 vs 2.5 [2.1-3.1] 96 days, p = 0.005). Table 1 shows the demographic data and the outcome results 97 according to MEWC protocol activation. 98 T able 1. Demographic data and outcome results according to MEWC protocol alarm activation. V ariable Overall No Alarm Activated Alarm p-value (n = 1166) (n = 1091) (n = 75) Age, median (IQR) 34 (31-37) 34 (31-37) 36 (29.5-39) 0.23 P revious deliveries, n (%) 0.63 0 782 (67.1) 730 (66.9) 52 (69.3) 1 269 (23.1) 251 (23.0) 18 (24.0) ≥ 2 115 (9.9) 110 (10.1) 5 (6.7) Diabetes, n (%) 99 (8.5) 90 (8.2) 9 (12.0) 0.26 P reeclampsia, n (%) 65 (5.6) 52 (4.8) 13 (17.3) <0.001* T ype of delivery, n (%) 0.005* Eutocic delivery 719 (61.7) 685 (62.8) 34 (45.3) Instrumental delivery 176 (15.1) 163 (14.9) 13 (17.3) Cesarean delivery 271 (23.2) 243 (22.3) 28 (37.3) M ultiple deliveries, n (%) 21 (1.8) 16 (1.5) 5 (6.7) 0.009* P otential F atal Disorder, n (%) 39 (3.3) 28 (2.6) 11 (14.7) <0.001* Intervention, n (%) 15 (1.3) 12 (1.1) 3 (4.0) 0.07 Critical Care Admission, n (%) 18 (1.5) 9 (0.8) 9 (12.0) <0.001* Length of Stay (h), median (IQR) 62 (52-75) 61 (52-75) 71 (54.5-86.5) 0.005* Length of Stay < 7 days, n (%) 13 (1.1) 11 (1.0) 2 (2.7) 0.2 *: statistical significance. The leading cause of alarm activation was altered systolic blood pressure (32 [42.7%] 99 patients), followed by obstetric causes (24 [32%] patients). The median time of alarm 100 activation, obstetric assessment, call to the anesthesiologist, and anesthetic assessment 101 was 10, 11, 15, and 20 minutes from the start of postpartum monitoring (Table 2). 102 Our MEWC had a sensitivity of 0.28 (0.15, 0.45), a specificity of 0.94 (0.93, 0.96), a 103 December 12, 2020 4/10 . CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 12, 2020. ; https://doi.org/10.1101/2020.09.15.20194910doi: medRxiv preprint T able 2. Causes for activation and limits of the MEWC protocol’s trigger thresholds and assessment time from patient monitoring. V ariable Activated Alarm (n = 75) Lower boundary Upper boundary Alarm trigger, n (%) Oxygen saturation 2 (2.7) 95% Systolic arterial pressure 32 (42.7) 90 mmHg 160 mmHg Diastolic arterial pressure 3 (4.0) 40 mmHg 100 mmHg Heart rate 14 (18.7) 50 bpm 120 bpm Obstetrics (uterine involution and bleeding) 24 (32.0) 500 ml Urine output 0 (0) 17.5 ml/h T ime to activate the alarm (min), median (IQR) 10 (0-60) T ime to obstetrics assessment (min), median (IQR) 11 (2.5-60) T ime to call the anesthesiologist (min) 15 (4.5, 61.5) T ime to anesthetic assessment (min) 20 (7-65.5) positive predictive value of 0.15 (0.08-0.25) and a negative predictive value of 0.97 (0.96, 104 0.98) for PFD detection (Table 3). Furthermore, the protocol showed a specificity of 105 0.94 and a negative predictive value of 0.99 for each of the PFD criteria separately 106 (intervention, CCU admission, and length of stay longer than seven days). 107 T able 3. Sensitivity , specificity , positive and negative predictive values (with 95% confidence interval) presented by the activation of our MEWC protocol alarm for the detection of PFD and its respective criteria Activated Alarm Sensitivity Specificity Positive predictive value Negative predictive value P otencial f atal disorder 0.28 (0.15, 0.45) 0.94 (0.93, 0.96) 0.15 (0.08, 0.25) 0.97 (0.96, 0.98) Intervention 0.20 (0.04, 0.48) 0.94 (0.92, 0.95) 0.04 (0.01, 0.11) 0.99 (0.98, 0.99) Critical Care Admission 0.50 (0.26, 0.74) 0.94 (0.93, 0.96) 0.12 (0.06, 0.22) 0.99 (0.98, 1.00) Length of stay < 7 days 0.15 (0.02, 0.45) 0.94 (0.92, 0.95) 0.03 (0.00, 0.09) 0.99 (0.98, 0.99) With regard to multivariate regression analysis, we only found a relationship with 108 PFD with 2 factors, preeclampsia [Odds Ratio = 7.81 (3.50, 16.7), p < 0.001] and 109 MEWC alarm activation [Odds Ratio = 3.97 (1.63, 8.95), p = 0.001] (Fig 2). The only 110 factor related to the intervention was the MEWC alarm activation (Odds Ratio = 4.73 111 [1.04, 15.7], p = 0.02, Fig 3). The main factor related to critical care admission was 112 preeclampsia (Odds Ratio = 23.2 [7.75-75.3], p < 0.001), followed by MEWC alarm 113 activation (Odds Ratio = 9.73 [2.98-31.5], p < 0.001) and cesarean delivery (Odds Ratio 114 = 4.40 [1.22-18], p = 0.03) (Fig 4). We found no relationship between length of stay 115 and MEWC alarm activation in the Cox regression (Fig 5). 116 No significant p-value was rejected after calculating its q-value within the multiple 117 comparability study (S4 File). 118

119 The first thing that draws attention to implementing the MEWC in our hospital is the 120 data on sensitivity, specificity, positive, and negative predictive value of alarm activation. 121 Our protocol’s sensitivity is very low (0.28), while specificity and negative predictive 122 value are very high (0.94 and 0.99, respectively). For practical purposes, these results 123 mean that we can be confident that patients without activation of the MEWC protocol 124 alarm are doubtful to suffer a PFD. However, the cost was not to detect many patients 125 who developed PFD (28 out of 39, 71.1% of total cases). Although we do not know the 126 December 12, 2020 5/10 . CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 12, 2020. ; https://doi.org/10.1101/2020.09.15.20194910doi: medRxiv preprint Fig 2. F orest plot of multivariate logistic analysis of the influence of MEWC activation, patient comorbidity and delivery on PFD. We present the results as an odds ratio with a 95% confidence interval. Results less than 1, left of the y-axis, imply risk reduction. We accepted p < 0.05 as significant. Fig 3. F orest plot of multivariate logistic analysis of the influence of MEWC activation, patient comorbidity and delivery on intervention. We present the results as an odds ratio with a 95% confidence interval. Results less than 1, left of the y-axis, imply risk reduction. We accepted p < 0.05 as significant. December 12, 2020 6/10 . CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 12, 2020. ; https://doi.org/10.1101/2020.09.15.20194910doi: medRxiv preprint Fig 4. F orest plot of multivariate logistic analysis of the influence of MEWC activation, patient comorbidity and delivery on critical care unit admission. We present the results as an odds ratio with a 95% confidence interval.

less than 1, left of the y-axis, imply risk reduction. We accepted p < 0.05 as significant. Fig 5. F orest plot of multivariate logistic analysis of the influence of MEWC activation, patient comorbidity and delivery on length of stay . We present the results as an odds ratio with a 95% confidence interval. Results less than 1, left of the y-axis, imply risk reduction. We accepted p < 0.05 as significant.December 12, 2020 7/10 . CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 12, 2020. ; https://doi.org/10.1101/2020.09.15.20194910doi: medRxiv preprint causes of such low sensitivity levels, we suspect that it is probably related to the 127 patient’s monitoring period or the different parameters that form the protocol. 128 Our study monitored the patients for only 2 hours after delivery. However, other 129 studies with better sensitivity results had a much more excellent follow-up. Singh et 130 al. [13] published the first study to validate MEWC, carrying out a prospective 131 observational study in 676 obstetric patients to detect morbidity between the 20th week 132 of gestation and six months postpartum. 30% of the patients had an altered trigger. 133 They defined a MEWC sensitivity of 0.98 and a specificity of 79%. The introduction of 134 such an extended period of obstetric monitoring is probably impossible to implement in 135 our center. Given our hospital’s structure, our patients go from the delivery room to 136 conventional hospitalization rooms, where the monitoring is not comparable to that of a 137 post-anesthesia recovery unit. Still, it makes us suspect that if we extend the 138 monitoring period from admission to 24 hours after birth, we should reduce the number 139 of false negatives. 140 On the other hand, the protocol modification by adding obstetric factors has also 141 been able to modify the sensitivity and specificity of the protocol. Initially, we can 142 think that the more complex a protocol is, the more difficult it is to generate false 143 negatives, and the more secure its implementation is. However, this is not as true as it 144 seems. Mhyre et al. [9] described that the simpler the warning criteria, the shorter the 145 time for recognition, diagnosis, and treatment of women who are more at risk of 146 developing obstetric complications. The more uncomplicated measures are more reliable, 147 less vulnerable to human calculation errors, and have higher reproducibility. Mhyre et 148 al. [9] insist on creating multidisciplinary teams that define their warning criteria for 149 each center and review the evidence to date using these criteria9. In 2015 Edwards et 150 al. [14] compared the predictive value of six different MEWS to identify severe sepsis in 151 women with chorioamnionitis and informed of a sensitivity range of 40%-100% with a 152 degree of 4%-97% specificity. The authors concluded that the MEWC tools with simpler 153 designs tended to be more sensitive, while more complex ones tended to be more 154 specific. By adding the obstetric items, we probably complicated the protocol and 155 caused an increase in specificity and a decrease in sensitivity. 156 Likewise, a complex MEWC protocol that causes a high number of false positives is 157 a severe limitation in its application since it can lead to professional fatigue due to the 158 high number of false alarms15. Of course, no MEWC protocol is ideal, and we fully 159 agree with Friedman et al. [15]. They argued that since each hospital has a unique 160 environment, each hospital should develop its protocol and improve it, depending on its 161 particular outcomes. We are currently in that phase, the improvement phase. Our study 162

encourage us to persevere in perfecting the protocol, either by increasing the 163 time of monitoring the patient or by altering the variables to be monitored. However, 164 further studies are necessary to validate and generalize MEWC principles. 165

166 Our MEWC protocol presented low sensitivity and high specificity with a high negative 167 predictive value, having many false-negative patients. Our study showed an association 168 between alarm activation and PFD, reintervention rate, and critical care admission rate, 169 with preeclampsia being the most related factor in all of them. 170 Supporting information 171 S1 File. Ethics committe approval document. Original in Spanish. Date of 172 Approval: 12th March 2018, Code: FJD-MEOWS-17-01. 173 December 12, 2020 8/10 . CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 12, 2020. ; https://doi.org/10.1101/2020.09.15.20194910doi: medRxiv preprint S2 File. Data collection form 174 S3 File. Original study databases. 175 S4 File. Statistical analysis. Html format. 176 S5 File. Statistical analysis raw R code. R-Markdown format. 177

1. de Miguel Sesmero JR, Cacho PM, Solano AM, Feu JMO, G´ omez MG, Prieto AP, et al. Mortalidad materna en Espa˜ na en el periodo 2010-2012: resultados de la encuesta de la Sociedad Espa˜ nola de Ginecolog´ ıa (SEGO). Progresos de Obstetricia y Ginecolog´ ıa. 2015;. 2. McClure JH, Cooper GM, Clutton-Brock TH, Centre for Maternal and Child Enquiries. Saving mothers’ lives: reviewing maternal deaths to make motherhood safer: 2006-8: a review. British Journal of Anaesthesia. 2011;107(2):127–132. doi:10.1093/bja/aer192. 3. Bogod D, Mushambi M, Craig SK. Guidelines for the provision of anaesthesia services for an obstetric population. London, Royal College of Anaesthetists. 2019;. 4. Armitage M, Eddleston J, Stokes T, Guideline Development Group at the NICE. Recognising and responding to acute illness in adults in hospital: summary of NICE guidance. BMJ (Clinical research ed). 2007;335(7613):258–259. doi:10.1136/bmj.39272.679688.47. 5. Umar A, Ameh CA, Muriithi F, Mathai M. Early warning systems in obstetrics: A systematic literature review. PloS One. 2019;14(5):e0217864. doi:10.1371/journal.pone.0217864. 6. von Elm E, Altman DG, Egger M, Pocock SJ, Gøtzsche PC, Vandenbroucke JP. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) Statement: Guidelines for reporting observational studies. International Journal of Surgery. 2014;12(12):1495–1499. doi:10.1016/j.ijsu.2014.07.013. 7. Buderer NM. Statistical methodology: I. Incorporating the prevalence of disease into the sample size calculation for sensitivity and specificity. Academic Emergency Medicine: Official Journal of the Society for Academic Emergency Medicine. 1996;3(9):895–900. doi:10.1111/j.1553-2712.1996.tb03538.x. 8. Cueto Hern´ andez I. An´ alisis de la mortalidad y morbilidad materna seg´ un criterios de la organizaci´ on mundial de la salud y del Euro-Peristat en el per´ ıodo 2011-2015 en el Hospital General Universitario Gregorio Mara˜ n´ on [info:eu-repo/semantics/doctoralThesis]. Universidad Complutense de Madrid. Madrid, Espa˜ na; 2018. Available from:https://eprints.ucm.es/47109/. 9. Mhyre JM, D’Oria R, Hameed AB, Lappen JR, Holley SL, Hunter SK, et al. The maternal early warning criteria: a proposal from the national partnership for maternal safety. Obstetrics and Gynecology. 2014;124(4):782–786. doi:10.1097/AOG.0000000000000480. December 12, 2020 9/10 . CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 12, 2020. ; https://doi.org/10.1101/2020.09.15.20194910doi: medRxiv preprint 10. Say L, Souza JP, Pattinson RC, WHO working group on Maternal Mortality and Morbidity classifications. Maternal near miss–towards a standard tool for monitoring quality of maternal health care. Best Practice & Research Clinical Obstetrics & Gynaecology. 2009;23(3):287–296. doi:10.1016/j.bpobgyn.2009.01.007. 11. Behling DJ, Renaud M. Development of an obstetric vital sign alert to improve outcomes in acute care obstetrics. Nursing for Women’s Health. 2015;19(2):128–141. doi:10.1111/1751-486X.12185. 12. Storey JD. A direct approach to false discovery rates. Journal of the Royal Statistical Society: Series B (Statistical Methodology). 2002;64(3):479–498. doi:10.1111/1467-9868.00346. 13. Singh S, McGlennan A, England A, Simons R. A validation study of the CEMACH recommended modified early obstetric warning system (MEOWS). Anaesthesia. 2012;67(1):12–18. doi:10.1111/j.1365-2044.2011.06896.x. 14. Edwards SE, Grobman WA, Lappen JR, Winter C, Fox R, Lenguerrand E, et al. Modified obstetric early warning scoring systems (MOEWS): validating the diagnostic performance for severe sepsis in women with chorioamnionitis. American Journal of Obstetrics and Gynecology. 2015;212(4):536.e1–8. doi:10.1016/j.ajog.2014.11.007. 15. Friedman AM. Maternal early warning systems. Obstetrics and Gynecology Clinics of North America. 2015;42(2):289–298. doi:10.1016/j.ogc.2015.01.006. December 12, 2020 10/10 . CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 12, 2020. ; https://doi.org/10.1101/2020.09.15.20194910doi: medRxiv preprint . CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 12, 2020. ; https://doi.org/10.1101/2020.09.15.20194910doi: medRxiv preprint . CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 12, 2020. ; https://doi.org/10.1101/2020.09.15.20194910doi: medRxiv preprint . CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 12, 2020. ; https://doi.org/10.1101/2020.09.15.20194910doi: medRxiv preprint . CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 12, 2020. ; https://doi.org/10.1101/2020.09.15.20194910doi: medRxiv preprint . CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 12, 2020. ; https://doi.org/10.1101/2020.09.15.20194910doi: medRxiv preprint