Objective
We sought to compare the predictive power of published modified obstetric early warning scoring systems (MOEWS) for the development of severe sepsis in women with chorioamnionitis.
Study Design
This was a retrospective cohort study using prospectively collected clinical observations at a single tertiary unit (Chicago, IL). Hospital databases and patient records were searched to identify and verify cases with clinically diagnosed chorioamnionitis during the study period (June 2006 through November 2007). Vital sign data (heart rate, respiratory rate, blood pressure, temperature, mental state) for these cases were extracted from an electronic database and the single worst composite recording was identified for analysis. Global literature databases were searched (2014) to identify examples of MOEWS. Scores for each identified MOEWS were derived from each set of vital sign recordings during the presentation with chorioamnionitis. The performance of these MOEWS (the primary outcome) was then analyzed and compared using their sensitivity, specificity, positive and negative predictive values, and receiver-operating characteristic curve for severe sepsis.
Results
Six MOEWS were identified. There was wide variation in design and pathophysiological thresholds used for clinical alerts. In all, 913 women with chorioamnionitis were identified from the clinical database. In all, 364 cases with complete data for all physiological indicators were included in analysis. Five women developed severe sepsis, including 1 woman who died. The sensitivities of the MOEWS in predicting the severe deterioration ranged from 40–100% and the specificities varied even more ranging from 4–97%. The positive predictive values were low for all MOEWS ranging from <2–15%. The MOEWS with simpler designs tended to be more sensitive, whereas the more complex MOEWS were more specific, but failed to identify some of the women who developed severe sepsis.
Conclusion
Currently used MOEWS vary widely in terms of alert thresholds, format, and accuracy. Most MOEWS have not been validated. The MOEWS generally performed poorly in predicting severe sepsis in obstetric patients; in general severe sepsis was overdetected. Simple MOEWS with high sensitivity followed with more specific secondary testing is likely to be the best way forward. Further research is required to develop early warning systems for use in this setting.
Sepsis describes a systemic inflammatory reaction to infection in usually sterile tissues. It progresses through a spectrum of severity, each stage carrying increasing risk of mortality. Maternal sepsis is a leading direct cause of preventable maternal morbidity and mortality. The physiological changes of pregnancy can mimic the early stages of sepsis and can also accelerate its progress. These factors can delay the recognition of sepsis until severe deterioration has occurred.
The early detection of sepsis is key to optimizing outcomes. Early warning systems that track observed vital signs and trigger medical response at threshold limits have been developed to improve identification of all patients at risk of imminent deterioration, irrespective of their diagnosis. In the nonobstetric population, the modified early warning score (MEWS) has been shown to predict intensive care unit (ICU) transfer and mortality. However, when the same system was applied to an obstetric population with infection, the predictive power was found to be poor. In recognition of the altered physiology of pregnancy, there have been efforts to adapt these systems for use in maternity care, and modified obstetric early warning scoring systems (MOEWS) have been recommended for clinical use. These are mostly unvalidated and no single system has been accepted for maternity care. Therefore, a variety of different MOEWS remain in use. The aim of this study was to compare the predictive power of MOEWS in women with chorioamnionitis.
Materials and Methods
Global literature databases (MEDLINE, PubMed) and clinical guidance publications were searched to identify working examples of early warning systems specifically used in maternity care. Several different titles were identified for these early warning scores/charts; for consistency we will refer to all of these obstetric systems as MOEWS. MOEWS were included if they had clear instructions, in English, such that the scoring system could be easily applied to a dataset of clinical vital sign observations. Six published MOEWS (A-F) were identified representing the 2 most common methods of track-and-trigger early warning systems: color-coded trigger bands (single/multiple indicator trigger systems) and numerical scoring triggers (or aggregate-weighted scoring systems). The scoring thresholds for each system investigated in this study are outlined in Table 1 . For each early warning system, the threshold score indicating the highest risk level that should prompt immediate senior medical review was identified as the trigger and was then used for further analysis.
| MOEWS | ||||||
|---|---|---|---|---|---|---|
| Color-coded MEOWS | ||||||
| MOEWS A | ||||||
| Variable | Red | Amber | 0 | Amber | Red | Trigger |
| Pulse rate, beats/min | ≤59 | 60–69 | 70–109 | 110–139 | ≥140 | Score ≥2 Ambers or ≥1 Red |
| Systolic BP, mmHg | ≤89 | 90–99 | 100–149 | 150–159 | ≥160 | |
| Diastolic BP, mmHg | ≤79 | 80–89 | ≥90 | |||
| Respiratory rate, breaths/min | ≤10 | 11–20 | 21–29 | ≥30 | ||
| Temperature, °C | ≤34.9 | 35.0–35.9 | 36.0–37.9 | 38–38.9 | ≥39.0 | |
| O2 Sats, % | ≤89 | 90–100 | ||||
| Mental state | Alert | Voice | Pain/unresponsive | |||
| MOEWS B | ||||||
| Variable | Red | Amber | 0 | Amber | Red | Trigger |
| Pulse rate, beats/min | ≤39 | 40–49 | 50–99 | 100–119 | ≥120 | Score ≥2 Ambers or ≥1 Red |
| Systolic BP, mmHg | ≤89 | 90–99 | 100–159 | 160–169 | ≥170 | |
| Diastolic BP, mmHg | ≤99 | 100–109 | ≥110 | |||
| Respiratory rate, breaths/min | 0–10 | 11–20 | 21–29 | ≥30 | ||
| Temperature, °C | ≤34.9 | 35.0–35.9 | 36.0–37.9 | ≥38.0 | ||
| O2 Sats, % | ≤94 | 95–100 | ||||
| Mental state | Alert | Voice | Pain/unresponsive | |||
| MOEWS C | ||||||
| Variable | Red | Amber | 0 | Amber | Red | Trigger |
| Pulse rate, beats/min | ≤39 | 40–49 | 50–99 | 100–119 | ≥120 | Score ≥2 Ambers or ≥1 Red |
| Systolic BP, mmHg | ≤89 | 90–99 | 100–149 | 150–159 | ≥160 | |
| Diastolic BP, mmHg | ≤89 | 90–100 | ≥100 | |||
| Respiratory rate, breaths/min | 0–10 | 11–20 | 21–29 | ≥30 | ||
| Temperature, °C | ≤34.9 | 35.0–35.9 | 36.0–37.9 | ≥38.0 | ||
| O2 Sats, % | ≤94 | 95–100 | ||||
| Mental state | Alert | Voice | Pain/unresponsive | |||
| Aggregate score MOEWS | ||||||||
|---|---|---|---|---|---|---|---|---|
| MOEWS D | ||||||||
| Variable | 3 | 2 | 1 | 0 | 1 | 2 | 3 | Trigger |
| Pulse rate, beats/min | ≤39 | 40–59 | 60–74 | 75–104 | 105–109 | 110–129 | ≥130 | Medium risk |
| Systolic BP, mmHg | ≤79 | 80–89 | 90–139 | 140–149 | 150–199 | ≥200 | Score 4–5 | |
| Respiratory rate, breaths/min | ≤5 | 5–9 | 10–14 | 15–19 | 20–24 | 25–29 | ≥30 | High risk |
| Temperature, °C | ≤34.9 | 35.0–35.9 | 36.0–37.9 | 38.0–38.4 | ≥38.5 | Score ≥6 | ||
| O2 Sats, % | ≤87 | 88–89 | 90–94 | 95–100 | ||||
| Mental state | Alert | Voice | Pain | Unresponsive | ||||
| MOEWS E | ||||||||
| Variable | 3 | 2 | 1 | 0 | 1 | 2 | 3 | Trigger |
| Pulse rate, beats/min | ≤39 | 40–49 | 50–109 | 110–119 | 120–129 | ≥130 | Medium risk | |
| Systolic BP, mmHg | ≤69 | 70–79 | 80–99 | 100–149 | 150–159 | 160–199 | ≥200 | Score 4–5 |
| Diastolic BP, mmHg | ≤39 | 40–89 | 90–99 | 100–109 | ≥110 | High risk | ||
| Respiratory rate, breaths/min | ≤5 | 5–9 | 10–19 | 20–24 | 25–29 | ≥30 | Score ≥6 | |
| Temperature, °C | ≤34.9 | 35.0–35.9 | 36.0–37.9 | 38.0–38.9 | 39.0–39.9 | ≥40 | ||
| O2 Sats, % | ≤95 | 96–100 | ||||||
| Mental state | Alert | Voice | Pain | Unresponsive | ||||
| MOEWS F | ||||||||
| Variable | 3 | 2 | 1 | 0 | 1 | 2 | 3 | Trigger |
| Pulse rate, beats/min | ≤59 | 60–110 | 111–149 | ≥150 | Medium risk | |||
| Systolic BP, mmHg | ≤79 | 80–89 | 90–139 | 140–149 | 150–159 | ≥160 | Score ≥4 or single parameter = 3 | |
| Diastolic BP, mmHg | ≤89 | 90–99 | 100–109 | ≥110 | High risk trigger | |||
| Respiratory rate, breaths/min | ≤9 | 10–17 | 18–24 | 25–29 | ≥30 | Score ≥6 | ||
| Temperature, °C | ≤33.9 | 34.0–35.0 | 35.1–37.9 | 38.0–38.9 | ≥39.0 | |||
| O2 Sats, % | ≤94 | 95–100 | ||||||
| Mental state | Alert | Voice/pain/unresponsive | ||||||
The identified MEOWS (A-F) were applied to a vital sign dataset generated from a cohort of women with chorioamnionitis from a single tertiary care maternity hospital (Chicago, IL). Detailed information on the methods of case ascertainment and data retrieval has been previously published. Briefly, multiple search strategies were used to identify cases of clinically diagnosed chorioamnionitis during an 18-month period (June 2006 through November 2007). The hospital perinatal database was searched for the terms “febrile in labor,” “chorioamnionitis,” “therapeutic antibiotics,” or “maternal ICU transfer.” The hospital billing database was searched for obstetric cases with International Classification of Diseases, Ninth Revision codes for endomyometritis, chorioamnionitis, or sepsis. A care-quality database of patients transferred to intensive care was searched for maternity cases with febrile morbidity. Chorioamnionitis was defined clinically as maternal pyrexia in labor (≥38°C) associated with uterine tenderness, maternal or fetal tachycardia, or purulent/foul-smelling amniotic fluid. Electronic patient records for all identified cases were manually reviewed by 1 member of the research team to identify cases meeting these inclusion criteria. There were no specific exclusion criteria.
Maternal demographics and obstetric data (maternal history and peripartum outcomes) were abstracted from the patient record. In addition, each set of vital signs (heart rate, respiratory rate, blood pressure, temperature, mental state) recorded during labor was retrieved from the electronic medical record. Measurements taken within 40 minutes of epidural placement were excluded as transient changes in vital signs could be attributed to this procedure. Otherwise, for each woman the single worst composite set of recorded vital signs (generating the highest MEWS score) during the admission episode was selected. Only cases with complete vital sign data were included in the analysis. These vital signs were then used to generate early warning scores according to the instructions for each MOEW system, as outlined in Table 1 . Oxygen saturation data were not available and therefore assumed to be normal in each case for analysis. The proportion of cases reaching the warning trigger score was calculated for each system with 95% confidence intervals (CIs). Test characteristics (sensitivity, specificity, positive predictive value, negative predictive value) were calculated for each MOEWS trigger for severe sepsis, ICU transfer, or death. The standard American College Chest Physicians/Society of Critical Care Medicine 2001 consensus definition for severe sepsis was used. As a summary measure of the diagnostic performance of each test the receiver-operating characteristic (ROC) curve was calculated along with the area under the ROC (AUROC).
All patient information was abstracted from the electronic patient database into an anonymized spreadsheet (Microsoft Excel 2010, Microsoft Corp, Redmond, WA), and Stata (STATA/IC v13.1; StataCorp LP, College Station, TX) was used to generate early warning system scores from raw vital signs, as well as to perform all statistical analyses. The Northwestern University Institutional Review Board approved analysis.
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
