A retrospective study design was used to retrieve information on pregnant women who were admitted to the maternal ICU of a maternal and new-born tertiary care centre in south India, during the period from January 2011 to December 2012. The study included all pregnant women who were admitted to the ICU during the study period and excluded women admitted to the ICU who were not pregnant. Details including age, body mass index, prior obstetric history, associated medical and surgical co-morbidities, cause of ICU admissions, details of management, duration of ICU stay and organ dysfunction or failure were abstracted from the medical records database.

A maximum SOFA score was determined (range 0-4, with 4 being the worst score) for each of the six organ systems- respiratory, coagulation, hepatic, cardiovascular, neurologic and renal- which was used to derive SOFA score [11].The maximum SOFA score reflected the worst score for the woman during the entire duration of stay in the ICU and it did not reflect a snap shot of a single particular day [11]. A maximum SOFA score of ≥1 and ≤ 2 was used to define organ dysfunction and a score of ≥3 was considered to define organ failure [11]. An aggregate total maximum SOFA score (range 0-24) was derived from the maximum SOFA score for each individual organ [11].

The extracted data were first entered into a de-identified Microsoft Office Excel spreadsheet that conformed to the Declaration of Helsinki protocol and was then exported into a statistical software for analysis (STATA Version 9, College Station, Texas, USA). The distribution of SOFA scores was assessed for normality by using the Shapiro Wilk Test (p=0.0004). The SOFA scores for each organ were compared between mothers who survived and non-survivors by using the non-parametric Wilcoxon ranksum test, as the SOFA scores were not normally distributed. The predictive and discriminatory abilities of the SOFA scores for identifying survivors among women with SMM were ascertained by using tests of diagnostic effectiveness, including sensitivity, specificity and area under the Receiver Operator Characteristic (ROC) curve. Statistical significance was pre-defined as a value <0.05.

The study included 85 women with a mean (SD) age of 27.02 (4.25) years (range 20 to 38 years). Seventy three (85.88%) of these 85 women were admitted for obstetrics related conditions. The maternal characteristics of the 85 women have been presented in [Table/Fig-1]. The mean duration of stay in the maternal ICU was 2.45 (2.12) days and it ranged from 1 to 13 days. Sixty nine (81.18%, 95% CI: 71.82, 81.44) women had SMM and 5 (5.88%, 95% CI: 2.19, 12.55) of the 85 women did not survive. All the 5 women who did not survive had SMM. Two of the five women who did not survive had single organ involvement, while one woman each had 2, 4 and 6 organ involvements respectively.

The maximum and total SOFA scores were compared between women with SMM and those without it (see [Table/Fig-2]) and between women who survived with SMM and women who did not survive with SMM (see [Table/Fig-3]). The ability of the maximum SOFA score which was evaluated for each of the individual organs to predict or discriminate maternal mortality in women with SMM has been presented in [Table/Fig-4]. The ability of the total SOFA score to predict or discriminate maternal mortality in women with SMM has been presented in [Table/Fig-5]. The positive likelihood ratios for total SOFA scores of ≥10 and ≥12 were 19.20 and 38.40 respectively and the accuracies for total SOFA scores of ≥10 and ≥12 were 94.20% and 95.65% respectively. Considering that the pre-test probability for maternal deaths in women with SMM was 7.25%, the post-test probabilities for maternal deaths in women with SMM were estimated for the cut off values with the highest accuracies (SOFA score ≥10 and ≥12) were 60.01% and 75.01% respectively.

Previous studies done on populations which were admitted to ICUs have reported maternal mortality rates that ranged from 2.3% to 27% [12–15]. Several studies have reported that the major causes of admissions to ICUs, amongst pregnant women, were obstetrics related [12,13,16]. Several studies have also reported a 40-65% rate of organ failure among pregnant women with SMM, who were admitted to ICU [17,18]. Consistent with these studies, we found a maternal mortality rate of 5.88 % (7.25% in women with SMM), that nearly 86% of the admissions to the ICU were done for obstetrics related causes and that nearly 51% of pregnant women who were admitted to ICUs had at least one organ involvement. Assessment of organ dysfunction and organ failure is essential in the management of SMM. Several scales or scores are used to assess severity and predict outcomes in an ICU, including the APACHE II or the SAPS II scores [12–15,17,18]. The SOFA score offers several advantages, including the use of variables that can be easily measured without the need of very complex resources and the ease of standardizing these variables. Consistent with other studies, respiratory failure was significantly associated with prognosis in this study [12,13,18] However, a closer look at the 95% confidence intervals around the point estimates for the area under ROC curves showed that upper limits were >0.75 for all organs, except coagulation. This suggested that the SOFA scores for all organs except coagulation could help in discriminating prognosis in pregnant women with SMM. A lack of association with coagulation has been reported earlier and it maybe related to the transitory nature of thrombocytopaenia and its reversibility [11,19]. The use of the total SOFA scores was found to be a good indicator of prognosis in this population, with a value of ≥ 10 or ≥ 12, which significantly increased the probability of maternal deaths in pregnant women with SMM.

The use of the SOFA scores, and the total SOFA score in particular, offers several opportunities for the obstetric community to attempt to reduce maternal mortality by targeting SMM. The total SOFA score may be used as a cut off for referring pregnant women with SMM to higher centres. The referral may pragmatically be made from a secondary care centre to a tertiary care centre or from a tertiary care centre to a dedicated ICU. The use of the SOFA scores has to be considered pragmatically at a primary care centre, given the resources which are available. However, a cut off of >0 may be considered to be pragmatic for referral from a primary care centre. At the secondary levels, an individual organ score of ≥1 should alert the obstetrics team to the possibility of organ dysfunction. It might be prudent for secondary care centres to refer pregnant women with SMM if the individual organ SOFA scores are ≥1 or total SOFA score is ≥ 3, especially when one considers that organ dysfunction and failure may involve multiple organs, that the SOFA score may not be able to identify early organ dysfunction, and the availability and accessibility to maternal ICU centres in India. Most obstetric units are already using modified early obstetric warning scores (MEOWS) [20]. An evaluation of organ dysfunction and calculation of SOFA score should be triggered when there are four yellow or two red of any parameters in the MEOWS.

The retrospective nature of data collection and the representativeness of the study population maybe considered as a limitation.

Further studies, including those done in more diverse ICU settings, will help in determining the replication and generalizability of results.The SOFA scores can be used to ascertain organ dysfunction and failure and they can form the basis for referrals to higher centres.

The SOFA scores are useful for determining prognoses of pregnant women with SMM and they will help in identifying pregnant women with SMM, who may develop worse outcomes.