Discriminatory power of organ-specific Sequential Organ Failure Assessment (SOFA) scores: Predictors of maternal mortality in obstetric intensive care units (ICU)

ABSTRACT

Background:

Critically ill obstetric patients in intensive care units (ICUs) present unique challenges, and maternal mortality remains a major concern, particularly in low- and middle-income countries. The Sequential Organ Failure Assessment (SOFA) score, though widely used in general ICUs, has not been adequately studied for its organ-specific discriminatory power in predicting maternal mortality. This study aimed to evaluate the prognostic accuracy of individual organ SOFA sub-scores to improve risk stratification in obstetric critical care.

Methods:

This prospective observational study was conducted over one year in the obstetric ICU of a tertiary center in Northern India. A total of 125 obstetric patients (pregnant or within 42 days postpartum) were enrolled. SOFA sub-scores for respiratory, cardiovascular, hepatic, renal, coagulation, and neurological systems were calculated on admission and throughout ICU stay. Patients were classified into survivors and non-survivors. Associations between individual organ dysfunction and maternal mortality were assessed. Discriminatory accuracy of organ-specific SOFA scores was determined using receiver operating characteristic (ROC) curve analysis.

Results:

Mortality increased proportionally with the number of dysfunctional organs (χ² =39.693; P < 0.001). Escalating organ-specific SOFA scores were significantly associated with maternal mortality. ROC analysis revealed that cardiovascular dysfunction had the highest predictive accuracy (area under the curve [AUC] =0.875), followed by hepatic (AUC = 0.808), respiratory (AUC = 0.804), hematological (AUC = 0.763), and renal (AUC = 0.757) dysfunction.

Conclusion:

Cardiovascular, hepatic, and respiratory SOFA sub-scores provide strong discriminatory power for predicting maternal mortality in Obstetric ICU patients. Incorporating organ-specific SOFA assessments may enhance early risk stratification and guide life-saving interventions in resource-limited settings.

Introduction

The management of critically ill obstetric patients in intensive care units (ICU) remains uniquely challenging, as physiological adaptations of pregnancy combined with acute obstetric complications can rapidly progress to life-threatening conditions.[1,2] Early identification of high-risk patients is essential to improve maternal outcomes, particularly in resource-limited settings. The Sequential Organ Failure Assessment (SOFA) score, originally developed for general ICU populations, evaluates dysfunction across six organ systems—respiratory, cardiovascular, hepatic, renal, coagulation, and neurological—assigning scores from 0 to 4, with higher scores indicating greater severity of organ dysfunction and increased risk of mortality.[3] Recent studies have reaffirmed the utility of SOFA and its dynamic changes in predicting outcomes in critically ill patients, including obstetric populations.[4,5]

Despite its widespread use, the application of SOFA in obstetric populations remains an area of ongoing research. In particular, the discriminatory power of individual organ-specific SOFA sub-scores in predicting maternal mortality has not been adequately explored.[6,7,8] Discriminatory power refers to the ability of a prognostic tool to accurately distinguish between survivors and non-survivors. Understanding that organ dysfunctions contribute most significantly to mortality risk could enhance prognostic accuracy and support more targeted clinical interventions.

For clinicians, identifying the most critical organ dysfunctions can facilitate early risk stratification, prioritize monitoring, and guide timely, organ-specific interventions in high-risk patients. This is especially relevant in obstetric ICUs, where rapid clinical deterioration is common and timely decision-making is crucial. Furthermore, improved prognostic precision can assist in optimizing the allocation of limited resources, including ICU beds, ventilatory support, and multidisciplinary care.

From the patient perspective, improved risk prediction enables earlier recognition of deterioration, prompt escalation of care, and potentially better survival outcomes. It also supports more accurate counseling of patients and families regarding prognosis. In low- and middle-income countries, where delays in diagnosis and intervention significantly contribute to maternal mortality, the use of reliable and context-appropriate prognostic tools may help reduce preventable deaths and improve overall quality of care.[9,10]

Traditional ICU severity scoring systems, such as APACHE II and SAPS II, often overestimate mortality in obstetric patients due to the physiological changes associated with pregnancy.[8] In contrast, organ dysfunction–based models like SOFA may provide more accurate prognostic information in this population.[11] Although previous studies have identified common organ dysfunctions—particularly cardiovascular, renal, and neurological involvement—in obstetric ICU admissions[12,13] the relative contribution of each organ-specific SOFA component to maternal mortality remains unclear.

Maternal mortality continues to be a major global health concern, with a disproportionate burden in low- and middle-income countries.[9,10] Addressing this knowledge gap, the present study aimed to evaluate the discriminatory power of individual organ-specific SOFA scores in predicting maternal mortality among critically ill obstetric patients. By identifying key predictors of adverse outcomes, this study seeks to improve risk stratification and support timely, life-saving interventions in obstetric critical care settings.

Material and Methods

This prospective observational cohort study was carried out over a period of one year in the Obstetric ICU at a tertiary care center in Northern India, with the objective of evaluating the discriminatory power of organ-specific SOFA scores in predicting maternal mortality. A total of 125 consecutive obstetric patients admitted to the ICU who fulfilled the eligibility criteria were enrolled. Informed consent was taken from the patient or the legal representative/guardian in case of patients who not were not fully conscious. The study population included pregnant women and postpartum women up to 42 days following termination of pregnancy, regardless of the mode and outcome, whether spontaneous or induced abortion, ectopic pregnancy, molar pregnancy, or delivery. Patients younger than 18 years of age, non-obstetric gynecology cases, those transferred back towards within 24 hours of ICU admission, those with incomplete records, and those who declined consent were excluded. For analysis, women were stratified into two groups based on their outcomes: non-survivors, comprising patients who died during ICU admission, and survivors, representing those who recovered and were successfully discharged. The ethical approval was obtained by the Institutional Ethics Committee of King George's Medical University U.P on 2/06/21 Ethical Number : 539/Ethics /2021.

Data collection process

To study the discriminatory power of organ-specific SOFA scores in predicting maternal mortality, each of the six sub-scores—respiratory, cardiovascular, hepatic, renal, coagulation, and neurological—was analyzed individually as well as in combination. At admission (Day 1) and during subsequent ICU stay, SOFA sub-scores were computed based on clinical and laboratory parameters, and patients were stratified into survivors and non-survivors. The extent of multiorgan dysfunction at admission was assessed by documenting the number of organs involved (SOFA ≥ 1), and survival outcomes were compared across patients with single, two, three, or multiple organ dysfunctions. In addition, the frequency of individual organ dysfunctions was compared between survivors and non-survivors to identify systems most strongly associated with poor outcomes. The severity of each organ-specific SOFA sub-score (ranging from 0 to 4) was further analyzed, and trends in mortality were examined with escalating scores to determine prognostic thresholds. Finally, the discriminatory accuracy of each organ-specific SOFA score was evaluated using receiver operating characteristic (ROC) curve analysis, and the area under the curve (AUC) was calculated to identify which organ systems provided the greatest predictive value for maternal mortality. This combined approach enabled assessment of both the breadth of organ involvement and the depth of dysfunction severity, thereby highlighting the organ systems most critical to outcome prediction in critically ill obstetric patients.

Results

Out of 125 women enrolled in the study, 74.4% (n = 93) of women were survivors, they recovered and were discharged while 25.6% (n = 32) of patients were non-survivors [Table 1].

Table 1.

Survival outcome of study population (n=125)

Outcome of patients	No. of patients (n)	%
Group 1 (Non-survivor)	32	25.6
Group 2 (Survivor)	93	74.4

There was no significant difference in sociodemographic characteristics between non-survivors and survivors. Figure 1 is a histogram showing dispersion of age of study population. Patients enrolled in the study were between 19 and 41 years and mean age was 26.68 ± 4.71 years. Difference in mean age of non-survivors (26.16 ± 4.04 years) and survivors (26.86 ± 4.93 years) was not found to be significant. Survival did not show any significant association with religion, socio-economic status, mode of admission, booking status, parity, and mode of delivery in previous pregnancy.

Figure 1.

Histogram showing dispersion of age of study population

Figure 2 shows the distribution of Study Population based on present obstetric status. Out of 125 women, 88 women were antenatal cases, 29 were post-partum, four were septic abortions and remaining four were ruptured ectopic pregnancies. Out of 88 antenatal cases, 64 (51.2%) women were between 28 and 37 weeks of gestation, 16 (12.8%) were >37 weeks of gestation and remaining 8 (6.4%) were <28 weeks of gestation. Out of 29 post-partum cases, 17 (13.6%) were post (LSCS) Lower Segment Caesarean Section (operated outside) and 12 (9.6%) were post-natal cases.

Figure 2.

Distribution of study population based on present obstetric status

Figure 3 shows the relationship between the number of organ dysfunctions on day-1 and survival outcomes. Patients with single-organ dysfunction were almost exclusively survivors, while the risk of mortality rose sharply with increasing number of dysfunctional organs. Notably, three or more organ failures were predominantly seen in non-survivors, with 32 cases of three-organ dysfunction and 27 cases of four-organ dysfunction resulting in poor outcomes. In contrast, survivors were more common among patients with up to two-organ dysfunctions. The association between the extent of multiorgan dysfunction and mortality was highly significant (χ² =39.693; P < 0.001), emphasizing that survival decreases proportionally with the number of organs involved.

Figure 3.

Association between number of dysfunctional organs on D1 and survival outcomes

The analysis of individual organ system and multiorgan dysfunction on day-1 revealed a significant association with survival outcomes as depicted in Figure 4. Cardiovascular (CVS), renal, and hepatic dysfunction were more frequently involved among non-survivors compared to survivors, with highly significant P values (P = 0.001, 0.006, and <0.001, respectively), indicating their strong prognostic impact. Although respiratory and central nervous system (CNS) dysfunction were nearly universal across both groups and did not show discriminative value. Hepatic dysfunction emerged as the most striking predictor, being present in 65.6% of non-survivors versus 23.7% of survivors. Importantly, the coexistence of multiorgan involvement, particularly combinations such as CVS + renal and CVS + renal + respiratory dysfunction, showed a very strong association with mortality (P < 0.001), with non-survivors disproportionately affected. These findings highlight that while isolated organ dysfunctions contribute to worse outcomes, the presence of multiple organ dysfunctions, especially involving the cardiovascular system, is a critical determinant of poor survival.

Figure 4.

Association of Single and multiorgan dysfunction on D-1 with Survival

Figure 5 shows that majority of cases having D-1 SOFA scores 6 to 13 were survivors, while majority of cases with D-1 SOFA scores ≥14 were non-survivors. This difference was found to be significant statistically (χ² =55.514; P < 0.001)

Figure 5.

Comparison of D-1 SOFA scores with maternal mortality. SOFA = Sequential Organ Failure Assessment

Higher individual organ SOFA scores were strongly associated with increased mortality. For the respiratory system, mortality rose steeply with maximum SOFA scores of 4, where 51.8% were non-survivors compared to only 6.4% at a score of 3 (P < 0.001). Hematological dysfunction also showed progressive risk, with non-survivor rates increasing from 16.6% at SOFA 1 to 76.4% at SOFA 3 (P < 0.001). Hepatic involvement emerged as a critical predictor, with mortality reaching 100% at SOFA 3–4 and 53.3% at SOFA 2 (P < 0.001). Cardiovascular dysfunction was highly significant, with 60.8% mortality at SOFA 4 (P < 0.001). Renal impairment similarly demonstrated worsening outcomes as scores increased, with more than half of patients at SOFA 3–4 not surviving (P < 0.001). By contrast, CNS scores were universally high (SOFA 4 in all patients) and thus non-discriminatory. Overall, escalating SOFA scores in respiratory, hematological, hepatic, cardiovascular, and renal systems were strongly predictive of mortality, underscoring the prognostic importance of organ-specific dysfunction severity [Table 2].

Table 2.

Relationship between maternal mortality and maximum individual organ SOFA scores

Organ	SOFA Score	Total no. of patients (n)	Non- Survivor (n=32)		Survivor (n=93)		χ 2	“P”
			No.	%	No.	%
Respiratory	0	0	0	0	0	0	34.571	<0.001
	1	0	0	0	0	0
	2	9	0	0	9	100
	3	62	4	6.4	58	93.5
	4	54	28	51.8	26	48.1
Hematological	0	36	3	8.3	33	91.6	30.623	<0.001
	1	30	5	16.6	25	83.3
	2	40	10	25.0	30	75.0
	3	17	13	76.4	4	23.5
	4	2	1	50.0	1	50.0
Hepatic	0	76	6	7.9	70	92.1	37.881	<0.001
	1	15	6	40.0	9	60.0
	2	30	16	53.3	14	46.7
	3	3	3	100	0	0.0
	4	1	1	100	0	0.0
CVS	0	64	1	1.6	62	98.4	56.004	<0.001
	1	1	0	0.0	1	100.0
	2	0	0	0.0	0	0.0
	3	9	0	0.0	9	100.0
	4	51	31	60.8	20	39.2
CNS	0	0	0	0.0	0	0.0	–	–
	1	0	0	0.0	0	0
	2	0	0	0.0	0	0.0
	3	0	0	0.0	0	0.0
	4	125	32	25.6	93	74.2
Renal	0	37	1	2.7	36	97.3	20.473	<0.001
	1	44	11	25.0	33	75.0
	2	21	8	38.1	13	61.9
	3	8	4	50.0	4	50.0
	4	15	8	53.3	7	46.7

SOFA=Sequential Organ Failure Assessment; CVS=Cardiovascular; CNS=Central nervous system

Figure 6 illustrates the relationship between maximum SOFA sub-scores for individual organ systems and maternal mortality, showing that higher dysfunction severity was consistently associated with worse outcomes. Respiratory failure showed sharply increased mortality at SOFA 4 (51.8%), while hematological dysfunction peaked at 76.4% mortality at SOFA 3. Hepatic dysfunction was a particularly strong predictor, with 100% mortality at SOFA 3–4 and 53.3% even at SOFA 2. Cardiovascular dysfunction emerged as the strongest determinant overall, with 60.8% mortality at SOFA 4, whereas nearly all patients with lower scores survived. In contrast, CNS scoring lacked discriminatory power since all patients had a SOFA score of 4, though 25.6% were non-survivors. Renal dysfunction also demonstrated a progressive trend, with mortality reaching 53.3% at SOFA 4. Collectively, these findings emphasize the prognostic weight of cardiovascular, hepatic, and respiratory dysfunction in predicting mortality.

Figure 6.

Relationship of maximum SOFA sub scores with Maternal Mortality rate in a- Respiratory system, b-Hematological system, c - Hepatic system, d-CVS system, e - CNS system, f - Renal System. SOFA = Sequential Organ Failure Assessment

The ROC analysis revealed that among individual organ SOFA scores, cardiovascular dysfunction had the highest discriminatory power for predicting mortality (AUC = 0.875, 95% CI: 0.811–0.940, P < 0.001), followed by hepatic (AUC = 0.808) and respiratory (AUC = 0.804) systems. Hematological (AUC = 0.763) and renal (AUC = 0.757) scores also showed significant predictive value (P < 0.001). In contrast, CNS dysfunction (AUC = 0.500, P = 1.000) had no discriminatory ability. These findings highlight that cardiovascular, hepatic, and respiratory SOFA sub-scores are the most reliable predictors of mortality among organ systems [Table 3].

Table 3.

Discriminatory power of maximum SOFA score of individual organ systems

Variable	Area under curve	SE	“P”	95% CI
				Lower bound	Upper bound
Respiratory	0.804	0.042	<0.001	0.721	0.887
Hematological	0.763	0.051	<0.001	0.662	0.863
Hepatic	0.808	0.047	<0.001	0.716	0.901
CVS	0.875	0.033	<0.001	0.811	0.940
CNS	0.500	0.059	1.000	0.384	0.616
Renal	0.757	0.046	<0.001	0.667	0.846

SOFA=Sequential Organ Failure Assessment; CVS=Cardiovascular; CNS=Central nervous system

Discussion

Critical illness during pregnancy and the postpartum period remains a significant contributor to maternal morbidity and mortality, particularly in developing countries where the burden is disproportionately high. In about 2% of women, critical care is required and when required, maternal mortality is higher ranging from 3.4 to 14%.[14,15] The present study reaffirms the crucial role of prognostic scoring systems, especially the SOFA score, in predicting outcomes among obstetric ICU patients.[16] Overall, our findings are consistent with existing literature demonstrating that increasing SOFA scores correlate strongly with higher mortality,[17] while also offering important obstetric-specific insights that distinguish this study from prior research.

Our study demonstrated that both total SOFA scores and maximum organ-specific SOFA scores were strongly associated with ICU mortality. Patients with Day-1 SOFA scores between 6 and 13 were predominantly survivors, whereas those with scores ≥ 14 had significantly higher mortality (P < 0.0001). These findings agree with earlier study by Fuchs et al.,[18] which established the SOFA score as a reliable predictor of mortality in critically ill patients. Similarly, Baradari et al. (2016)[19] reported that admission, mean, and maximum SOFA scores were significantly higher in non-survivors, with excellent predictive accuracy (AUC 0.988). In obstetric populations, Kaye et al.[20] (2017) also observed significantly higher SOFA scores among non-survivors, reinforcing the applicability of SOFA in this subgroup.

However, unlike many previous studies that primarily focused on total SOFA scores, the present study provides a more granular analysis of individual organ dysfunction, highlighting their differential contributions to mortality prediction.

In our cohort, cardiovascular, hepatic, and renal dysfunctions were significantly more common among non-survivors. Among these, cardiovascular dysfunction demonstrated the highest discriminatory power (AUC = 0.875), followed by hepatic and respiratory systems. This finding aligns with Kaye et al.,[20] who also identified cardiovascular dysfunction as a strong predictor of mortality. The prominent role of cardiovascular dysfunction in our study can be attributed to the high prevalence of shock in obstetric ICU admissions, which is a well-known adverse prognostic factor. The SOFA scoring system defines cardiovascular dysfunction based on vasopressor requirement, making it particularly sensitive in detecting hemodynamic instability. This differs from other scoring systems such as (MODS) Multiple Organ Dysfunction Syndrome and the Brussels score that rely on more complex or indirect parameters (e.g. pressure-adjusted heart rate or hypotension with acidemia), thereby underscoring the clinical practicality and relevance of SOFA in obstetric critical care.

A notable finding of our study is the strong association between hepatic dysfunction and mortality, with mortality reaching 100% at higher SOFA sub-scores (3–4). While hepatic dysfunction has been recognized in previous studies, its pronounced predictive value in our cohort appears more significant than reported in earlier obstetric ICU studies, suggesting possible population-specific differences or variations in disease severity and case mix.

Consistent with established literature, we also observed that mortality increased sharply with the number of organ dysfunctions, with patients having three or more failing organs showing significantly higher mortality (P < 0.001). This is in line, with the findings from Kaye et al.,[16] where non-survivors almost universally had multiorgan dysfunction. These findings reinforce the concept that cumulative organ failure has a synergistic negative impact on survival.

Despite these consistencies, an important distinction of our study lies in the limited discriminatory power of neurological and renal SOFA sub-scores, which contrasts with some previous reports. The poor predictive value of the neurological component can be explained by several obstetric-specific factors. Firstly, Glasgow Coma Scale (GCS) scoring is often unreliable in sedated ICU patients, as highlighted by Lambden et al. (2019),[21] who proposed a modified SOFA (mSOFA) excluding the CNS component. Secondly, in obstetric patients, reduced GCS is frequently due to eclampsia, a condition that is often rapidly reversible following delivery and appropriate management. As a result, high neurological SOFA scores in such cases do not necessarily translate into increased mortality, thereby reducing its predictive accuracy.

Similarly, although acute kidney injury (AKI) is generally considered a poor prognostic marker in critically ill patients, its predictive value was limited in our study. This can be attributed to the reversible nature of obstetric AKI, which is commonly due to acute tubular necrosis resulting from hemorrhagic hypovolemia or sepsis.[21] Prompt correction of hypovolemia and timely management often lead to complete renal recovery, thereby weakening its association with mortality.[22] This highlights a key difference between obstetric and non-obstetric ICU populations, where renal dysfunction may carry different prognostic implications.[23]

Another important observation is that central nervous system dysfunction was non-discriminatory in ROC analysis, likely due to uniformly high scores across patients, further supporting the argument for reconsidering its inclusion in obstetric-specific prognostic models.

The overall mortality rate observed in our study was comparable to that reported by Baradari et al.[19] but higher than that reported by Kaye et al.[20] This discrepancy may be explained by differences in case mix, severity of illness, and healthcare settings, emphasizing the importance of contextual interpretation of prognostic scores.

Limitations

Single center design, limited sample size, observational nature, and resource limitations within the Obstetric ICU, such as staffing levels, availability of equipment, and variations in clinical practices, may impact patient outcomes and the interpretation of study results.

Conclusion and Future Recommendations

The present study underscores the prognostic value of organ-specific SOFA sub-scores in predicting maternal outcomes in obstetric ICU patients. Cardiovascular, hepatic, and respiratory dysfunctions emerged as the most significant predictors of mortality, highlighting their pivotal role in early clinical deterioration. The findings suggest that focused monitoring of these systems, along with timely interventions, can substantially improve maternal survival in obstetric ICUs. Future research should aim to validate these results in larger multicentric cohorts and across varying healthcare settings to establish population-specific thresholds. Integrating organ-based SOFA monitoring into standardized obstetric critical care protocols and electronic health systems could facilitate real-time risk stratification and optimize resource allocation, particularly in low- and middle-income countries. Additionally, developing modified obstetric-specific SOFA tools incorporating pregnancy-related physiological adaptations may further enhance predictive accuracy and clinical applicability.

Conflicts of interest

There are no conflicts of interest.

Funding Statement

Nil.