Organ Dysfunction Score for Obstetric Patients

INTRODUCTION The Sequential Organ Failure Assessment (SOFA) score was initially developed in a consensus meeting organized by the European Society of Intensive Care in 1996. The aim of the score was to quantify and objectively describe the degree of organ dysfunction or failure in critically ill patients (1). The SOFA score describes a continuum process from normality to dysfunction and failure for six organ systems (respiratory, hematologic, hepatic, cardiovascular, neurologic and renal) (1). Although primary objective of SOFA score was descriptive, numerous studies have proven its prognostic value as well. There is a strong association between SOFA, number of dysfunctions upon admission and SOFA trends in the first 48h and ICU mortality (2, 3).

The SOFA score is a key component of the last definition of sepsis. In 2016, the European Society of Critical Care and the American Society of Critical Care developed and validated a new definition of sepsis. Sepsis was then considered as a life-threatening organ dysfunction caused by a dysregulated host response to infection (3). SOFA score was chosen to identify organ dysfunction due to its simplicity and widespread use. Organ dysfunction was defined as a change in total SOFA score of ≥ 2 points (3). This definition was validated by many studies of different countries, including Argentina (4, 5).

The SOFA score was developed and validated in a population of adult critically ill patients. Thus, its calibration for other populations with particular physiological changes, such as children or obstetric patients, could be inadequate. While some scores, like the PELOD (6, 7) or modified SOFA (8) were developed and validated for pediatric patients, there is no one suitable for obstetric patients yet. General SOFA score was associated with mortality among obstetric patients (9-12) but its predictive validity could be improved if adjustments to pregnancy were performed. Although two groups proposed modified SOFA scores adjusted to obstetric patients, these scores presented clear deficiencies and were not validated (13, 14). The development of an adjusted and calibrated organ dysfunction score for obstetric patients, would improve the validity of the score to predict ICU mortality, and would provide a reliable tool to be incorporated into the definition of maternal sepsis (15, 16). Obstetric sepsis account for 11% of global maternal mortality and for 19%, when abortion-related deaths are added, as most of them are due to sepsis (17). Developing a validated instrument to define sepsis in this population, would improve its recognition and would eventually promote its early treatment, which is associated with better outcomes (18).

Barriers for using the non-adjusted SOFA score to diagnose sepsis in obstetric patients are mainly related to renal and cardiovascular systems. SOFA score sets a creatinine of ≥1.2 mg/dl as the lower cut-off point to identify kidney dysfunction (1). However, pregnant patients and those in early puerperium, present an increased in cardiac output, in renal plasma flow and in glomerular filtration rate, which in turn decrease normal creatinine levels to <0.9 mg/dl in any trimester of pregnancy (19, 20) and in the first 2 weeks postpartum (20-22). Thus, a non-adjusted SOFA score underestimates renal dysfunction in obstetric patients. In terms of cardiovascular dysfunction, mean arterial pressure (MAP) normally decreases during the second trimester of pregnancy, up to approximately 65 mmHg. However, for the general SOFA score any value of MAP ≤70 mmHg is considered abnormal, which will overestimate hemodynamic dysfunction during that trimester (1). In addition to the abovementioned required adjustments, respiratory evaluation of SOFA is usually lacking among obstetric patients. This evaluation is performed through the paO2/FiO2 ratio, which requires an arterial blood gas (ABG) extraction (23, 24). ABG measurement could be unavailable in some low-resource settings, but also could be missing due to a deliberate physician decision of avoiding arterial gas extraction in these patients (24). Using the relationship between the peripheral oxygen saturation over the inspired fraction of oxygen (spO2/FiO2) could address this deficiency. The spO2/FiO2 ratio is a validated, widely available and non-invasive surrogate of paO2/FiO2 ratio to evaluate respiratory performance (25, 26). The use of a paO2/FiO2 ratio surrogate, such as the spO2/FiO2, improves SOFA score specificity compared to the decision of considering the missing value as normal (24) and it is a key factor when SOFA score is used to define sepsis (15).

The main objective of this study is to develop and validate an organ dysfunction score adjusted to obstetric patients (SOFA-OBS) to predict ICU mortality among pregnant and early postpartum patients requiring ICU admission.

1. Background and significance There were two previous attempts to adapt the SOFA score to the obstetric population. In 2016, Blanco Esquivel et al. (13), developed a SOFA score adjusted to obstetric patients; however, the variable chosen to be modified was questionable. The authors proposed lowering normal cut-off point for paO2/FiO2 and SaO2/FiO2, assuming pregnant patients have lower paO2 levels due to a decreased total lung capacity (TLC). However, pregnant patients neither present a reduction in TLC nor a decrease in oxygenation; thus, SOFA score cut-off values for respiratory evaluation should not be modified (27). Moreover, pregnancy renal and hemodynamic changes were not considered. In 2017, the Society of Obstetric Medicine of Australia and New Zealand proposed an obstetric SOFA. This score removed scores of 3 and 4 in each category for simplification and reduced creatinine cut-off point to 1 mg/dl (14). However, this creatinine value is still higher than the expected for obstetric patients. In addition, no adjustments were done for MAP during the second trimester and no validation of the modified obstetric SOFA was performed.

The SOFA-OBS score proposed in this study would be the first adjusting creatinine and MAP to the obstetric population in a prospective and multicenter study, using the best evidence available (28, 29). In addition, it would be the first study evaluating SpO2/FiO2 calibration among these patients, which might improve the specificity and predictive validity of the score. The development and validation of an obstetric SOFA score to be used in the diagnose of maternal sepsis, would contribute to its early recognition and prompt treatment, which instead would hopefully reduce maternal mortality associated with sepsis.

II. Objectives

Primary objective:

• To develop an obstetric SOFA score (SOFA-OBS) adjusted to physiological changes of pregnancy and to validate the SOFA-OBS during the first 24h in ICU (SOFA-OBS24h) to predict ICU mortality among obstetric patients.

Secondary Objectives:

• To validate the SOFA-OBS to predict maternal sepsis among obstetric patients requiring ICU admission
• To validate SOFA-OBS to predict sepsis associated mortality among obstetric patients requiring ICU admission
• To evaluate the mean SOFA-OBS (SOFA-OBS mean) to predict mortality among obstetric patients requiring ICU admission
• To evaluate the maximum SOFA-OBS (SOFA-OBS max) to predict mortality among obstetric patients requiring ICU admission
• To evaluate the impact of SOFA-OBS trend during the first 48 h in ICU (SOFAΔ48) (increase, decrease, no changes) on ICU mortality
• To evaluate the relationship between the presence of each organ dysfunction/failure during the first 24h in ICU and ICU mortality
• To evaluate the relationship between the number of organ dysfunctions/failures during the first 24h in ICU and ICU mortality
• To evaluate the validity of the conventional septic shock definition (15) to predict mortality among obstetric patients admitted to ICU using the SOFA-OBS to define dysfunction
• To evaluate the validity of the second trimester septic shock definition (2T-SS) to predict ICU mortality among obstetric patients using the SOFA-OBS to define dysfunction

Null Hypothesis:

• SOFA-OBS does not have higher predictive validity for ICU mortality among obstetric patients requiring ICU admission than non-adjusted SOFA

Alternative Hypothesis:

• SOFA-OBS has higher predictive validity for ICU mortality among obstetric patients requiring ICU admission than non-adjusted SOFA

III. Material y Methods This study will consist in a development phase and a validation phase of an obstetric SOFA (SOFA-OBS). The investigators will follow TRIPOD (Transparent Reporting of a multivariable prediction model for Individual Prognosis or Diagnosis) guidelines and TRIPOD check-list (30). Following, investigators describes elements not covered elsewhere.

Sample size calculation:

The sample size was estimated considering the main objective, which includes validity assessment of an organ dysfunction score modified to obstetric patients (SOFA-OBS), through the AUROC. The investigators considered information published by Oliveira Neto et al. (31), which reports an obstetric mortality of 2.6% in a 673 patients cohort. In order to detect an AUROC of 0.9, an alpha error of 5% and a power of 80%, 130 patients are required (4 presenting the outcome and 126 not presenting it). Statistical package pROC of the statistical software R was used to calculate sample size.

Theoretical and operational variables definitions:

Outcome variables:

1. Obstetric patients ICU mortality: defined elsewhere. This outcome cannot be blinded to in site researchers; however, it is a definite outcome, not requiring any interpretation. Moreover, the final data analysis will be performed by a separate team from the one collecting them.
2. Sepsis-related obstetric ICU mortality: defined elsewhere Although ICU mortality is an objective outcome, as previously mentioned, the attribution of sepsis as cause of death requires some interpretation. This outcome variable cannot be blinded, but the investigators will use an accepted and concise definition supported by objective variables, such as a change in SOFA score of at least 2 points and a suspected infection triggering cultures or antibiotics administration. The final data analysis will be performed also by a separate team from the one collecting data.

Predictor variables:

1. Sequential Organ Failure Assessment (SOFA) and modified SOFA scores (obstetric SOFA: SOFA-OBS): The investigators decided to adapt the Sequential Organ Failure Assessment (SOFA) score to the obstetric population. The SOFA score is used worldwide and is simple. It also has high predictive validity for mortality among critically ill patients in general and critically ill obstetric patients, in particular. However, it was not calibrated for obstetric patients. The SOFA score is a key component of last sepsis definition (15). Although sepsis is an important cause of maternal mortality, a SOFA score adjusted to obstetric patients, which could be incorporated in sepsis definition, is still unavailable. SOFA score will be measured according to its original version. Six organ systems (respiratory, hematologic, hepatic, cardiovascular, neurologic and renal) are evaluated (1). The worst value for each organ on a day is recorded and a score of 0 to 4 is assigned to each value according to predefined ranges (1, 2). Normal values are assigned 0 points and abnormal values are assigned 1 to 4 points (from least to most abnormal); where 1 and 2 represents organ dysfunction and 3 and 4 represents organ failure (2). Total SOFA score is the results from the sum of the score of each organ (1, 2). The obstetric SOFA (SOFA-OBS) adjusts renal and cardiovascular variables to physiological changes of pregnancy. Moreover, it simplifies respiratory evaluation, in terms of replacing arterial blood gases (ABG) for peripheral oxygen saturation/fraction of inspired oxygen (spO2/FiO2), when ABGs were not available. Variables unchanged during pregnancy (neurologic, hepatic and platelets), won't be modified. Organ evaluation according to SOFA and SOFA-OBS score:

• Neurologic: will be performed through the Glasgow Coma Scale (GCS) (32). If the patient is under sedation, the GCS before sedation (assumed) will be used (3). Should this data be unavailable, GCS recorded will be recorded at 15 (normal) (ECG= 15).
• Hepatic: will be assessed by measuring total bilirubin levels, as for the original SOFA, because they do not change with pregnancy (19).
• Hematologic: will be evaluated according to platelet counts, using the same ranges as those from the original SOFA, because platelets count during most pregnancy period are normally ≥150000/mm3, which is the lower cut-off point for the SOFA score (19).
• Renal: will be assessed with creatinine adjusted to normal changes of pregnancy and early puerperium. Creatinine will be considered abnormal at a cut-off point of ≥0.9 mg/dl (79.58 umol/l), according to data from a systematic review (28), population studies of different regions (20, 21, 33-35) and consensus from general practice (19). The investigators consider this value unifies values described in the literature, decreases false positives and increments specificity.
• Cardiovascular: for the first and third trimesters of pregnancy, normal mean arterial pressure (MAP) is the same as that of the general population. So, original SOFA cut-off point of abnormality (MAP <70 mmHg) will be considered. During the second trimester, MAP usually descends, thus, cut-off point of abnormality and consequently a SOFA score= 1, will be taken at a MAP of < 65 mmHg, according to Meah and col. metanalysis (29).
• Respiratory: for patients with arterial blood gases available (ABG) the worst value of arterial partial pressure of oxygen/ fraction of inspired oxygen (paO2/FiO2) ratio will be taken. When ABGs were unavailable, the worst value of pulse oximetry saturation/fraction of inspired oxygen (spO2/FiO2) ratio will be considered, following previously proposed ranges (24, 36).
• For spO2/FiO2 measurements normal oximetry readings should be ensured. This requires: no position changes neither aspiration 10 minutes before measurement and no changes in ventilatory setting 30 minutes before measurement (26). SOFA will be recorded daily, taking the worst value of each parameter (from 0 to 4). Daily total SOFA will represent the sum of the 6 parameters (0-24)(2). Measurements will be taken daily, as far as the patient is pregnant or ≤3 days postpartum, during the whole ICU-LOS and up to discharge or dead. The investigators will specifically measure:
• SOFA24h: SOFA during the first 24h in the ICU
• SOFAmax: maximum SOFA during ICU-LOS
• SOFAprom: mean SOFA during ICU-LOS
• SOFAΔ48: difference between SOFA at 48 h from admission and SOFA on admission

1. Sepsis: the following parameters will be recorded on admission and at any point during ICU-LOS .2.1 Sepsis: life-threatening organ dysfunction caused by a dysregulated host response to infection (15).

2.2 Organ dysfunction: acute change in total SOFA score of ≥ 2 puntos as a result of an infection. A previous SOFA is assumed to be 0 in a patient with no preexistent infection (15).

2.3 Suspected infection: condition requiring cultures order and/or antibiotic onset (15).

2.4 Septic shock: sepsis with persistent hypotension requiring vasopressors to maintain MAP > 65 mmHg and having a serum lactate level > 18 mg/dl (> 2 mmol/L) despite adequate volume resuscitation (15). This definition applies without modifications in pregnant patients during their first or third trimesters, because no changes in MAP are observed at these two points, compared to non-pregnant population.

2.5 Septic shock in the second trimester (SS-2T): sepsis with persistent hypotension requiring vasopressors to maintain MAP > 60 mmHg and having a serum lactate level > 18 mg/dl (> 2 mmol/L), despite adequate volume resuscitation.

Septic shock definition for the general ICU population considers hypotension when MAP persists < 65 mmHg requiring vasopressors, while SOFA score considers hypotension when MAP <70 mmHg. In order to preserve consistency, for septic shock definition during the second trimester, hypotension will be considered when MAP <60 mmHg, while SOFA-OBS score will define hypotension (SOFA= 1) when MAP <60 mmHg.

To blind assessment of predictors for the outcome will be impossible. Investigators collecting SOFA and SOFA-OBS scores will be aware of patients' outcomes (mortality). In order to avoid classification bias, predictive variables will be defined in the same way by all centers, using objective and clear criteria. Moreover, investigators will be asked to record raw data for SOFA components, instead of the scoring itself. The REDcap system will calculate the score for each parameter and the total SOFA score, according to predefined criteria.

Rest of variables that will be collected for the study at each center:

1. Hospital general data:
2. Localization b. Third level center c. Type of health care: public-private- social security d. Total bed number e. 24h obstetrician available f. Total number of births per year g. High-risk pregnancy care h. Neonatology unit i. 24h anesthesiologist available j. Type of ICU: general, obstetric k. 24 h board certified or trained in critical care physician: l. ICU bed number m. General ICU admissions per year (Mean ± SD) and total number of ICU admissions during the study period n. Obstetric ICU admissions per year (Mean ± SD) and total number of obstetric ICU admissions during the study period
3. Patients' data:
4. Age
5. Source of admission to ICU: emergency department, ward, operating room, other hospital) <!-- -->
6. Comorbidity, measured with the Charlson Comorbidity Index (CCI) (37). Causes of admission to ICU: obstetric (occurring only in pregnant/postpartum patients) vs. non obstetric (also occurring in nonpregnant patients) (10).
7. Specific diagnoses of admission
8. Obstetric history:
9. Gestational age on admission to ICU (weeks) ii. Gestational age on admission classified in trimesters (29): i. First trimester: 6-13 weeks ii. Second trimester: 14-27 weeks iii. Third trimester: ≥28 weeks iii. Parity. iv. Prenatal care v. Should pregnancy terminate during ICU-LOS: gestational age at termination will be recorded as well as fetal/neonatal outcome.
10. ICU data: i. Severity of illness during the first 24 h in ICU through the Simplified Acute Physiologic Score (SAPS II) (38) ii. ICU interventions: advanced respiratory support (high Flow nasal cannula [HFNC], Non-invasive ventilation [NIV], invasive mechanical ventilation [IMV]); renal replacement therapy (RRT) and hemodynamic monitoring. iii. Documented infection in ICU: Yes, No; type of infection (pneumonia, urinary tract infection, catheter related bacteremia, bacteriemia of unknown source, endometritis, septic abortion, chorioamnionitis, surgical site infection.
11. Additional outcome variables: i. Total number of livebirths during the study period. ii. ICU-length of stay (ICU-LOS). iii. Hospital-LOS. iv. Additional mortality data: i. Maternal mortality rate (MMR): is defined as female deaths from any cause related to or aggravated by pregnancy or its management, occurring during pregnancy or within 6 weeks of termination of pregnancy, irrespective of the duration and site of the pregnancy, for a specific period of time and expressed over 100000 live births (17). The investigators will adapt this definition, as only pregnant patients or those in their early puerperium (up to 3 days after birth) will be included in this study Maternal mortality ratio will be defined for this study as: Non-surviving obstetric patients admitted to ICU during the study period/ number of live births among the same population during the same study period. ii. MMR associated with direct causes= non-surviving obstetric patients admitted to ICU for a direct cause during the study period/ number of live births among the same population during the same study period. iii. MMR associated with indirect causes= non-surviving obstetric patients admitted to ICU for an indirect cause during the study period/ number of live births among the same population during the same study period. ii. Hospital mortality: i. Obstetric patient hospital mortality: Number of obstetric patients who required ICU admission during the study period and did not survive at hospital discharge/ all obstetric patients who required ICU admission during the study period. iii. Pregnancy outcomes i. Termination of pregnancy during ICU-LOS: Yes/No 1. Gestational age at termination of pregnancy (39)=
12. Term: ≥ 37 weeks
13. Preterm: < 37 weeks
14. Extrem preterm: 22-27+6 weeks ii. Early preterm: 28-31+6 weeks iii. Late preterm: 32-36 +6 weeks ii. Discharged pregnant iv. Fetal/neonatal outcomes: i. Survivors: both livebirths and discharged pregnancies will be included in this category/total number of obstetric patients admitted to the ICU during the study period. ii. Non survivors: both spontaneous abortions and perinatal deaths will be included in this category/ total number of obstetric patients admitted to the ICU during the study period: 1. Spontaneous abortions: those occurring < 22 weeks 2. Perinatal deaths: fetal (≥ 22 weeks) and early neonatal deaths (0-6 days after birth) (39). IV. Handling missing data The investigators will set Redcap with required fields, which will not allow researchers to continue filling forms if data is lacking. In addition, a research assistant will check the database weekly in order to look for missing or incongruent data, contacting local researchers to complete or review data. To avoid bias related to potential missing data, while assuming it is missing at random, primary objective analysis will be replicated after multiple imputations.
15. Statistical analysis Categorical variables will be presented as numbers (%) and continuous variables as mean ± SD or median [IQR], according to their distribution. The Shapiro-Wilk test will determine whether data is normally distributed or not. Categorical variables will be presented as number and percentage (%). Normally distributed continuous variables will be compared with Student t-test and non-normally distributed continuous data with Wilcoxon test. Categorical variables will be compared with chi-square or Fisher test. Multiple comparisons of parametric, non-parametric and categorical variables will be analyzed with one factor ANOVA, Kruskal Wallis o multiple chi-square tests, respectively. SOFA-OBS predictive validity for ICU mortality will be evaluated through its capacity to distinguish survivors from non-survivors (discrimination) and through the agreement between observed and predictive probabilities of death (calibration). Discrimination will be assessed by the area under the receiver operating characteristic curve and its 95% confidence interval (AUROC IC 95%) and calibration by the Hosmer-Lemeshow statistics. The same analysis will be repeated for the general SOFA score. Finally, both scores (SOFA-OBS and SOFA) performances will be compared by AUROC and c-index. Predictive validity for mortality of each organ dysfunction, measured by maximum SOFA of each organ, will be assessed by AUROC. Bootstrapping will be used for internal validation. The relation between the presence of organ dysfunction/failure during the first 24h in ICU and ICU mortality will be assessed through Chi2 test, measuring the percentage of non-survivors among all patients presenting each category of organ dysfunction/failure. Creatinine cut-off points for each SOFA score category to predict ICU mortality will be determine using Fisher algorithm to obtain ordinal variables with k +1 categories. The relation between trends in SOFA score during the first 48h (increase, decrease, no change) and ICU mortality, will be assessed through Chi2 multiple test. Predictive validity of mortality of a change in SOFA-OBS ≥2 among obstetric patients with suspected infection in the ICU will be assessed through AUROC and its 95% CI. This analysis will be repeated for general SOFA score and both results will be compared. The relationship between predictive and outcome variables, will be evaluated through multivariate analysis, adjusting for potential confounders. A p value <0.05 will be considered significant. SPSS20 will be used for data analysis. VI. Procedures and tools for collecting data and data quality control Data for this study will be collected as part of standard care in the ICU. In the case the patient does not require daily laboratory, data will be record as missing.
16. Tools for collecting data: the same case report form (CRF) will be used across all centers. Data will be collected through REDCap.
17. Data quality control: variables will be clearly defined in the manual of procedures, to ensure data validity. Data will be initially validated and audited by REDCap in terms of mistakes associated with typing or normal ranges. A research assistant will perform a weekly database monitoring, looking for data consistency and quality. If any inconsistency, error or missingness are detected, local researchers will be contacted to correct or fill them. Predetermined virtual meetings will be arranged for training and addressing queries. All sites will use the same definitions for variables. Local investigators will be asked to insert raw data into REDCap and the system will calculate different scores (e.g. SAPS II, SOFA, etc.). VII. Ethical considerations in human research and informed consent This is an observational study using data normally collected for clinical practice. The study was approved by the institutional review board of the Argentinian Society of Critical Care Medicine (Identifier PRIISABA-15646). As the study is also prospective, informed consent will be obtained before enrolling participants. The study will be performed in accordance with ethical standards laid down in the 1964 Declaration of Helsinki and later amendments and with national and regional guidelines for research in human beings (Argentinian National Ministry of Health-Res. 1480/2011 and Law for protecting rights of participants in Health Research 3301/2009 of the city of Buenos Aires, respectively). Each participant site, will obtain approval from their IRB and will follow their requirements. The organizer site will get a copy of the IRB approval from each center. Confidentiality and privacy will be protected according to the Argentinian National Law 25326 for Protecting personal data. Each site and patient will get a code number to keep anonymity. No personal data will be inserted into REDcap.