Objective
We sought to design an emergency department sepsis scoring system to identify risk of intensive care unit (ICU) admission in pregnant and postpartum women.
Study Design
The Sepsis in Obstetrics Score (S.O.S.) was created by modifying validated scoring systems in accordance with recognized physiologic changes of pregnancy. The S.O.S. was applied to a retrospective cohort of pregnant and postpartum patients from February 2009 through May 2011 with clinical suspicion of sepsis. The primary outcome was ICU admission. Secondary outcomes were telemetry unit admission, length of stay, positive blood cultures, positive influenza swabs, perinatal outcome, and maternal mortality. Receiver operating characteristic curves were constructed to estimate the optimal score for identification of risk of ICU admission.
Results
In all, 850 eligible women were included. There were 9 ICU (1.1%) and 32 telemetry (3.8%) admissions, and no maternal deaths. The S.O.S. had an area under the curve of 0.97 for ICU admission. An S.O.S. ≥6 (maximum score 28) had an area under the curve of 0.92 with sensitivity of 88.9%, specificity of 95.2%, positive predictive value of 16.7%, and negative predictive value of 99.9% for ICU admission, with an adjusted odds ratio of 109 (95% confidence interval, 18–661). An S.O.S. ≥6 was independently associated with increased ICU or telemetry unit admissions, positive blood cultures, and fetal tachycardia.
Conclusion
A sepsis scoring system designed specifically for an obstetric population appears to reliably identify patients at high risk for admission to the ICU. Prospective validation is warranted.
The United States currently ranks 47th (of 184 countries) in overall maternal mortality with a rate of 21/100,000, vs an average rate of 16/100,000 in developed countries. Maternal sepsis, especially puerperal sepsis, is a common pregnancy-related condition and in the United States is the fourth leading cause of maternal mortality, accounting for up to 13% of maternal deaths and approximately 5% of maternal admissions to the intensive care unit (ICU). Unfortunately, the frequency of severe sepsis in pregnancy is increasing in the United States, from 1:15,385 in 1998 to 1:7246 in 2008, as well as sepsis-related maternal death, up 10% per year in that same time frame.
The progression from the systemic inflammatory response syndrome (SIRS) to septic shock is clearly defined in the nonpregnant population using specific objective vital signs and laboratory values. Based on these parameters, critical care and infectious disease experts developed management guidelines for severe sepsis and septic shock with the Surviving Sepsis Campaign. This campaign highlighted the need for appropriate assessment of the severity of sepsis to enable early detection of cases at risk for rapid clinical deterioration, leading to the development of many disease severity scoring systems related to sepsis. None included pregnant women in the initial study population and all have been shown to overestimate morbidity and mortality in an obstetric population.
The failure of existing scoring systems to identify risk of morbidity in an obstetric population likely stems from their failure to account for the normal physiologic changes seen in pregnancy. These changes include a decrease in diastolic blood pressure by 5-10 mm Hg in the second trimester with return to baseline by the third trimester, an increase in heart rate by 17% (to 83 ± 10 beats per minute), and an elevation in leukocyte count (up to 16.9/μL by the third trimester and up to 30/μL in labor). Temperature, systolic blood pressure, respiratory rate, blood oxygen saturation (SpO 2 ), and percentage of immature neutrophils are unchanged. Little is known about the effect of pregnancy on lactic acid.
To our knowledge, no study to date has evaluated the use of a pregnancy-specific scoring system for use in an emergency department (ED) to predict clinical deterioration in an obstetric population presenting with signs of sepsis. Therefore, the objective of this study was to evaluate the utility of an ED scoring system designed specifically for an obstetric population to identify risk of ICU admission for pregnant and postpartum women who presented with signs of sepsis.
Materials and Methods
We performed a retrospective cohort study of pregnant and postpartum women evaluated in the ED at Women and Infants Hospital, a large tertiary care obstetric hospital, with suspected SIRS or sepsis from February 2009 through May 2011. To identify women deemed at high risk for sepsis, only those who had blood cultures or an influenza swab sent to the clinical laboratory were included. Blood cultures or an influenza swab were used as surrogate markers for a patient presenting with signs or symptoms concerning for sepsis and were used as the sole enrollment criterion because the aim was to capture patients in whom there was a clinical suspicion of a severe infectious process as determined by the primary ED physician.
The Women and Infants ED serves as both an obstetric triage unit and a free-standing emergency room, and thus is the point of entry to the hospital for every patient, regardless of gestational age or pregnancy status. It is staffed by obstetricians who are employed as full-time ED providers. This ED is equipped with an electronic medical record that is programmed to alert the provider when a patient meets ≥2 of the following criteria, which were taken from the standard SIRS criteria: mean arterial pressure ≤65, systolic blood pressure ≤90, heart rate ≥110, respiratory rate ≥22, temperature ≥38°C or ≤36°C, and leukocyte count ≥14/μL, <4/μL, or >10% immature neutrophils. The physician then determines if there is a high clinical suspicion for sepsis and makes the decision to draw blood cultures or perform an influenza swab.
Chart abstraction was conducted for women who presented to the ED at any gestational age through the first 2 postpartum weeks. All women had been discharged from the hospital prior to data collection and analysis. Exclusion criteria included a known or suspected ectopic pregnancy, multiple gestation (because of concern regarding additional hemodynamic changes from a multiple compared to a singleton gestation), transfer from an outside hospital (because of the possibility of treatment prior to presentation), or subsequent delivery at an outside hospital (because of inability to ascertain neonatal outcome). The population of this study was restricted to the ED because the scoring system that was developed was modeled after validated ED scoring systems.
Vital sign and laboratory data were individually abstracted from the electronic medical record by 2 investigators (C.M.A. and T.N.A.). Maximum temperature, heart rate, respiratory rate, leukocyte count, percentage of immature neutrophils, lactic acid, and minimum temperature, SpO 2 , and systolic blood pressure for the duration of the ED course were recorded. Other maternal data collected were age, race, body mass index (BMI), gestational age, and medical comorbidities. Fetal data collected included presence or absence of fetal tachycardia, defined as a heart rate of >160 beats per minute.
The Sepsis in Obstetrics Score (S.O.S.), a scoring system modified in accordance with recognized physiologic changes of pregnancy to assess specifically for sepsis, was created ( Figure 1 ). The S.O.S. combines and modifies the elements of the Rapid Emergency Medicine Score (REMS), a validated ED scoring system created by taking those elements of Acute Physiology and Chronic Health Evaluation (APACHE) II easily obtained in the prehospital setting, and the SIRS/sepsis criteria as described by the Surviving Sepsis Campaign. The APACHE II score provides an estimate of ICU mortality based on 12 vital sign and laboratory parameters and additionally takes into account acute and chronic disease states, including chronic organ insufficiency or history of immunocompromise. REMS is a strong predictor of in-hospital mortality and is based on only age and vital sign data recorded in the ED. Both scoring systems use the most abnormal recorded value for each parameter to determine the overall score. To create a sepsis-specific scoring system, the S.O.S. combines parameters taken from APACHE II and REMS (temperature, heart rate, respiratory rate, oxygen saturation, and leukocyte count) and the SIRS criteria (systolic blood pressure, leukocyte count, percentage of immature neutrophils, and lactic acid). To make the score pregnancy-specific, parameters with known changes in pregnancy, as described above, were modified accordingly (systolic blood pressure, heart rate, and leukocyte count). The S.O.S. scoring procedure is similar to that of APACHE II and REMS with scores ranging from 0–4 for each variable, with a maximum score of 28. A score of 0 indicates a normal value. Higher points were given for both high and low abnormal values. If a parameter was not recorded in the medical record, a score of 0 was assigned, and that parameter was therefore considered normal in the analysis.
The primary outcome was ICU admission within 48 hours of presentation to the ED. Secondary outcomes included telemetry unit admission, length of hospital stay, mortality, positive blood cultures, positive influenza swabs, antibiotic use, and adverse perinatal outcome, defined as a composite outcome of fetal or neonatal death, respiratory distress syndrome, grade III or IV intraventricular hemorrhage, necrotizing enterocolitis, or sepsis within 72 hours of birth. The criteria for ICU admission for sepsis are not standardized in our hospital; however in general, patients who meet criteria for septic shock and have a vasopressor requirement or require mechanical ventilation are transferred to the ICU. The criteria by which patients should be sent to the telemetry unit include a diagnosis of SIRS or sepsis requiring early goal-directed therapy.
The hypothesis of this study was that the S.O.S., a pregnancy-specific sepsis scoring system, would accurately identify women at risk of ICU transfer for sepsis. In 2 prior studies, approximately 2% of patients who presented to the ED with a clinically suspected infection concerning for sepsis were admitted to the ICU within 48 hours of admission. Therefore, to detect an area under the receiver operating characteristic curve (AUC-ROC) of 0.7 vs 0.5 (no association) with an alpha of 0.05, 80% power, and an estimated 2% ICU admission rate, required 850 patients. Descriptive statistics were used to compare variables by χ 2 or t test. The AUC-ROC and 95% confidence interval were calculated by simple logistic regression. Sensitivity and specificity at different S.O.S. values were determined from the ROC curve. The sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of the S.O.S. were then compared to other validated ED scoring systems. All statistical analyses were completed using SAS 9.2. (SAS Institute, Cary, NC). This study was approved by the Women and Infants Hospital Institutional Review Board (no. 11-0059).
Results
From February 2009 through May 2011, 850 eligible women were evaluated in the ED and included in this study. Nine were admitted to the ICU (1.1%), 32 to the telemetry unit (3.8%), and none died. The most common diagnosis at presentation was influenza-like illness (ILI) (60.4%), followed by nonrespiratory viral syndrome (11.1%), pyelonephritis (5.3%), endometritis (4.5%), pneumonia (2.4%), mastitis (1.2%), chorioamnionitis (0.7%), and septic abortion (0.6%).
The S.O.S. ranged from 0–13, with a mean of 1.3 and a median of 0. The overall AUC for the S.O.S. was 0.97. The ROC curve for the composite score is shown in Figure 2 . This curve was used to estimate the optimal S.O.S. cutoff for identification of risk of ICU admission. Using a cutoff of ≥6, the S.O.S. had a sensitivity of 88.9%, a specificity of 95.2%, a PPV of 16.7%, and a NPV of 99.9%. The AUC for a cutoff of 6 was 0.92.
Those with an S.O.S. ≥6 and <6 were similar demographically. There was a statistical difference in age but no difference in BMI, race/ethnicity, insurance coverage, gestational age at presentation, or medical comorbidities ( Table 1 ). Of those patients with an S.O.S. of ≥6, the most common diagnoses at presentation were pyelonephritis (25%), ILI (25%), and endometritis (10.4%). In contrast, those with an S.O.S. of <6 most commonly presented with ILI (62.6%) and nonrespiratory viral syndrome (11.4%) ( Table 2 ).
| Variable | S.O.S. ≥6 | S.O.S. <6 | P value |
|---|---|---|---|
| Age, y | (n = 48) | (n = 802) | |
| Mean (SD) | 24.0 (6.5) | 26.3 (6.1) | .01 |
| Median (min-max) | 22.5 (15-42) | 26.0 (15-43) | |
| Pregnancy status, n (%) | (n = 48) | (n = 802) | |
| Pregnant | 35 (72.9) | 732 (91.3) | < .0001 |
| Postpartum | 13 (27.1) | 70 (8.7) | |
| Gestational age, wk | (n = 35) | (n = 732) | |
| Mean (SD) | 26.6 (7.0) | 23.6 (10.0) | .12 |
| Median (min-max) | 26.9 (12.8-40) | 24.1 (2.7-42.3) | |
| No. of days postpartum, d | (n = 13) | (n = 70) | |
| Mean (SD) | 6.9 (3.7) | 6.9 (3.4) | .90 |
| Median (min-max) | 6 (1-13) | 7 (0-14) | |
| BMI | (n = 43) | (n = 743) | |
| Mean (SD) | 26.4 (7.8) | 28.5 (6.9) | .07 |
| Median (min-max) | 25.7 (16.9-64.5) | 27.5 (13.5-69.3) | |
| Race/ethnicity, n (%) | (n = 45) | (n = 773) | |
| White | 16 (35.6) | 362 (46.8) | .08 |
| Black | 3 (6.7) | 100 (12.9) | |
| Hispanic | 23 (51.1) | 287 (37.1) | |
| Asian | 3 (6.7) | 13 (1.7) | |
| Multiracial | 0 (–) | 4 (0.5) | |
| Other | 0 (–) | 7 (0.9) | |
| Insurance, n (%) | (n = 41) | (n = 662) | |
| None | 1 (2.4) | 13 (1.9) | .88 |
| Government | 25 (61.0) | 428 (64.6) | |
| Private | 15 (36.6) | 221 (33.4) | |
| Substance abuse, n (%) | (n = 39) 8 (20.5) | (n = 626) 105 (16.8) | .54 |
| Tobacco, n (%) | (n = 39) 7 (17.9) | (n = 626) 88 (14.1) | .50 |
| Alcohol, n (%) | (n = 39) 2 (5.1) | (n = 626) 10 (1.6) | .15 |
| Illegal drugs, n (%) | (n = 39) 2 (5.1) | (n = 626) 19 (3.0) | .35 |
| Hypertension, n (%) | (n = 48) 1 (2.1) | (n = 798) 41 (5.1) | .51 |
| Diabetes, n (%) | (n = 48) 1 (2.1) | (n = 798) 20 (2.5) | 1.0 |
| Asthma, n (%) | (n = 48) 8 (16.7) | (n = 798) 190 (23.8) | .25 |
| Thyroid disease, n (%) | (n = 48) 1 (2.1) | (n = 798) 32 (4.0) | 1.0 |
| HIV, n (%) | (n = 48) 0 (–) | (n = 798) 3 (0.4) | 1.0 |
| Current antibiotic use, n (%) | (n = 48) 1 (2.1) | (n = 793) 50 (6.3) | .35 |
| Variable | S.O.S. ≥6 | S.O.S. <6 | P value |
|---|---|---|---|
| Admission to ICU, n (%) | (n = 48) 8 (16.7) | (n = 802) 1 (0.1) | < .0001 |
| Admission to telemetry unit, n (%) | (n = 40) 16 (33.3) | (n = 801) 16 (2.0) | < .0001 |
| Length of hospital stay, d | (n = 42) | (n = 192) | |
| Mean (SD) | 4.4 (2.9) | 2.8 (1.6) | .0004 |
| Median (min-max) | 3.5 (0-14) | 2 (0-9) | |
| Working diagnosis, n (%) | (n = 48) | (n = 796) | |
| Pyelonephritis | 12 (25.0) | 33 (4.2) | |
| ILI | 12 (25.0) | 498 (62.6) | < .0001 |
| Endometritis | 5 (10.4) | 33 (4.2) | |
| Nonrespiratory viral syndrome | 3 (6.3) | 91 (11.4) | |
| Septic abortion | 2 (4.2) | 3 (0.4) | |
| Chorioamnionitis | 2 (4.2) | 4 (0.5) | |
| Pneumonia | 1 (2.1) | 19 (2.4) | |
| Mastitis | 1 (2.1) | 9 (1.1) | |
| Other | 10 (20.8) | 106 (13.3) | |
| Positive blood cultures, n (%) | (n = 39) 12 (30.8) | (n = 141) 12 (8.5) | .0003 |
| Positive influenza swabs, n (%) | (n = 27) 4 (14.8) | (n = 720) 100 (13.9) | .78 |
| Fetal tachycardia, n (%) | (n = 30) 18 (60.0) | (n = 598) 77 (12.9) | < .0001 |
| Composite perinatal outcome, n (%) (Fetal or infant death, RDS, grade III/IV IVH, stage 2 or 3 NEC, sepsis) | (n = 35) 2 (5.7) | (n = 716) 47 (6.6) | 1.0 |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
