Improving Clinical Prediction of Pulmonary Embolism in Pregnancy and Postpartum by Modifying Wells Score with a Population Specific Risk Factor: A retrospective study

Summary

Objective:

Pulmonary embolism (PE) is a significant cause of maternal mortality, necessitating accurate clinical prediction tools to optimise diagnosis and reduce unnecessary imaging. This study aimed to examine the Wells score's (WS) predictive value for PE, the effect of incorporating sickle cell disease (SCD) on its predictive accuracy and to estimate prevalence of PE in pregnanct and postpartum women.

Methods:

This retrospective study included pregnant and postpartum women who underwent computed tomography-pulmonary angiography (CTPA) for suspected PE at two tertiary hospitals (Sultan Qaboos University Hospital and Royal Hospital in Muscat, Oman) from 2013 to 2019. The relationship between WS and CTPA was evaluated before and after adding 3 points for SCD.

Results:

A total of 193 pregnant and postpartum women with a mean age of 31.45 ± 5.60 years underwent CTPA for suspected PE. PE incidence was 0.31 in 1,000 deliveries and 28 women (14.5%) had PE confirmed in CTPA. WS stratified probability of PE as low in 44 (22.8%) patients, moderate in 142 (73.6%) patients and high in 7 (3.6%) patients. In the women with a low score, 2 (4.5%) had PE, giving 95.2% negative predictive value (NPV) of low WS. High WS had a sensitivity of 33.3% and a specificity of 87.0%. Women with SCD had a higher incidence of PE compared to those without (36.7% versus 10.4%, odds ratio = 4.9, 95% confidence interval: 2.0–12.1; P < 0.001). When modifying WS with additional SCD points, NPV for the low score patients was 94.4%, while for the high score group sensitivity and specificity were 72.3% and 85.7%, respectively.

Conclusion:

The study confirms that low WS effectively excludes PE, reducing the need of imaging in pregnancy and postpartum. Incorporating SCD enhances sensitivity and specificity of a high score aiding to prioritise women for diagnostic imaging or empirical anticoagulation.

Advances in Knowledge

• Low Wells score (WS) can be used as a negative predictor for pulmonary embolism (PE) in pregnant and postpartum women.

• This study proposes a new modification to WS by adding the presence or absence of sickle cells disease (SCD) as a known strong risk factor for PE. The proposed modification enhances the usability of modified WS in pregnant and postpartum women by confirming its negative predictive value and improving its sensitivity.

Application to Patients Care

• This study confirms that when low, WS can be used to exclude PE, hence avoiding performing computed tomography-pulmonary angiography (CTPA) on pregnant or postpartum women suspected to have PE. When high, WS can be used to prioritise women to undergo CTPA to confirm PE.

• When adding SCD to the prediction model, it stratifies positive women, who are more likely to have PE, more accurately. Therefore, these women can undergo a confirmatory testing or empirical anticoagulation.

1. Introduction

Venous thromboembolism (VTE) is characterised by the formation of blood clots within the vascular system, posing significant risks of morbidity and mortality. This term includes both deep vein thrombosis (DVT) and pulmonary embolism (PE). Approximately 75% to 80% of pregnancy-related venous thromboembolism cases result from DVT, whereas 20% to 25% are due to PE.^2,3,4 Overall, the incidence of VTE in this period is estimated to be 1 in 1,000. Pregnant women are 5 times more likely to develop VTE compared to non-pregnant women, with the greater risk occurring after childbirth, especially in the first 6 weeks where the risk is estimated to be 20–35 times higher.^5,6,7 Pregnancy is associated with a hypercoagulable state, arising from elevation of factors VII, X, VIII, fibrinogen and von Willebrand factor, as well as physiological decrease of protein S activity and by acquired activated protein C resistance.^1,8 Additionally, there are anatomic changes, including compression of the inferior vena cava or pelvic veins by the enlarging uterus, which alter venous outflow.⁹ Decreased mobility, obesity, infections, diabetes or multiparity are also significant factors in VTE development.^2,10 In the postpartum period, older age (>35 years old), obesity and smoking, as well as caesarean delivery, preeclampsia, anaemia, antepartum haemorrhage and postpartum infection are known to increase the risk of VTE.¹¹

Clinically, diagnosing VTE in pregnant women presents several challenges due to the overlap of symptoms with normal physiological changes, such as dyspnoea, leg oedema and tachycardia.¹² Delayed diagnosis can lead to severe consequences and a false diagnosis of PE can impact delivery plans, future contraception options and thromboprophylaxis in subsequent pregnancies.¹³ Additionally, D-dimer testing is less useful in pregnant women, as their D-dimer levels are naturally elevated, reducing the test's diagnostic value in this population.¹⁴

Clinical prediction models, such as Wells score (WS) and Geneva Score, estimate the probability of PE in non-pregnant population by assigning values to specific symptoms and risk factors.¹⁵ These scores classify patients into low, intermediate or high risk for PE. WS contains 7 parameters and assigns +3 score for “clinical signs and symptoms of DVT” and “PE is the most or equally likely diagnosis”. Next, +1.5 is added if heart rate is more than 100, immobilisation more than 72 hours or the patient had major surgery in the last 4 weeks and a previous diagnosis of DVT or PE. It assigns +1 to the presence of malignancy or haemoptysis. WS < 2 is believed to be a low probability of PE, while 2–6 is intermediate and >6 is a high probability. However, prediction rules for diagnosing PE in pregnancy are not validated since they were developed from non-pregnant populations.¹⁶ Both the Royal College of Obstetricians and Gynecologists and the Society of Obstetricians and Gynecologists of Canada advise against their use during pregnancy.^17,18 Boureily et al. highlighted the low positive predictive value of these predictors in pregnancy, suggesting the need for pregnancy-specific clinical decision rules.¹³ WS could potentially rule out PE in pregnant or postpartum women, as those with the negative imaging results had WS scores below the high-risk threshold.¹⁹ However, these studies had small sample sizes and lacked precise sensitivity. Overall, evidence supporting the use of clinical predictors for PE in pregnant or postpartum patients remains limited.

Risk stratification, as well as D-dimer testing and deep venous thrombosis ultrasound evaluation, determine whether patients with clinically suspected acute PE need to have computed tomography pulmonary angiogram (CTPA) performed which is a standard in PE diagnosis with a high positive predictive value.²⁰ Studies have shown that CTPA has a sensitivity ranging from 83% to 100%.^{21,22,23,24,25} A retrospective case-control study was performed to compare diagnostic adequacy of CTPA in pregnant/postpartum patients with age-match control from 2008 to 2011. In this study, pregnant and postpartum women had a low yield in CTPA because a higher number of results of CTs were interpreted as limited or non-diagnostic compared to the control population (12% versus 0%). This finding is associated with low opacification of the pulmonary artery in pregnant and postpartum patients, due to physiological change during pregnancy including increased blood and hyperdynamic circulation.²⁶

Sickle cell disease (SCD) is an inherited autosomal recessive haemoglobinopathy, caused by the presence of abnormal haemoglobin, known as haemoglobin S.²⁷ It is believed that pregnancy increases the risk of VTE in SCD patients. Villers et al.'s study on 18,000 SCD deliveries confirmed that DVT and cerebral vein thrombosis are more common in women with SCD compared to those without the disorder with an odds ratio of 2.5 for DVT (95% confidence interval [CI]: 1.5–4.1).^28,29,30 Other studies supported the thesis that women with SCD are at a higher risk of developing VTE during pregnancy and postpartum compared to the general SCD population.^31,32 SCD is a common genetic blood disorder in Arabian Gulf countries, including Oman. In Oman, the neonatal incidence rates of SCD and sickle cell trait are 0.3% and 4.8%, respectively.³³

This study aimed to assess the prevalence of PE in pregnant and postpartum populations in tertiary care centres in Oman, as well as examine the predictive accuracy of WS for PE in pregnant and postpartum women as confirmed by CTPA. Additionally, this study aimed to determine the effect of further modifications of WS by incorporating SCD as a risk factor for thromboembolism.

2. Methods

This retrospective study was performed at two tertiary hospitals in Muscat, Oman (Sultan Qaboos University Hospital and Royal Hospital). Women with the clinical suspicion of PE who were pregnant or in the postpartum period and needed to have CTPA performed between January 2013 to December 2019 were included. At the hospitals, the decision to perform CTPA in the studied population is guided by individual medical history, clinical assessment, ultrasonography of the lower extremity veins, D-dimer level and the treating team judgement.

Data were collected from the electronic medical record system of the hospitals. Firstly, a list with all women who had CTPA ordered in that period was retrieved. Any patients who were not pregnant or in the postpartum period were excluded as well as if the indication for the CTPA was not the suspicion of PE. Specifically, the inclusion criteria were women age 18 to 45 years with confirmed intra-uterine pregnancy, at any gestational age with any mode of delivery up to 42 days post-partum. Women were excluded if they were outside the targeted age group and if they had a molar pregnancy. Further information was gathered from the electronic medical systems of both institutions, including demographic information, medical history, pregnancies details, WS parameters and other features that might have contributed to the clinician decision to suspect PE (e.g., body mass index, temperature, cough, palpitation, oxygen saturation, autoimmune disorders or family history of VTE). This information was essential to answer one of the WS parameters: “PE is the most or equally likely diagnosis” as many factors contribute to the clinician's impression in suspecting and prioritising PE as a diagnosis. To decrease the bias in reporting clinical parameters, WS was calculated before recording the CTPA results. The report of the CTPA was then determined to be positive for diagnosis of PE, negative or inconclusive.

To calculate the sensitivity of high-risk WS the number of positive diagnoses in the high-risk group was divided by the sum of that number to the positive CTPA results in the low-risk group. The specificity was estimated by dividing number of patients classified as low-risk group with negative results by the sum of them and the number of negative CTPA diagnoses in the high-risk group.

To estimate the predictive value of the proposed changed WS, 3 points were added to the count in patients with diagnosed SCD. WS assigns +1, +1.5 or + 3 points. It assigns +3 points to criteria that have higher weight clinically such as “alternative diagnosis is less likely than PE” and “symptoms and signs of DVT present”. As the literature supports the fact that SCD is a high risk for VTE, the current authors chose to treat SCD as a high risk factor for thromboembolism and would therefore add a value of +3 to the score. The literature states that SCD increases the risk of VTE significantly; pregnancy-related VTE in women with SCD appears to be 1.5–5 times higher than in the general pregnant population.^19,30,31

The obtained data were entered and analysed using the Statistical Package for Social Science (SPSS), Version 23 (IBM Corp., Armonk, New York, USA). Descriptive statistics were shown as frequencies, percentage, mean and standard deviation as well as the prevalence of PE. For inferential analysis, the Chi-square test (or Fisher's Exact tests for 2 × 2 table) was calculated to determine the relationship between variables.

3. Results

A total of 193 pregnant or postpartum women who underwent CTPA for suspected PE were included in this study. The mean age of the participants was 31.45 ± 5.60 years (range: 19–45 years). The mean gravity was 3.23 ± 2.33 (range: 1–7) and parity of 2.01 ± 1.80 (range:0–8). The majority of women were in their third trimester (n = 108, 56%), followed by those in the postpartum period (n = 66, 34.2%). There were fewer participants in the second trimester (n = 14, 7.3%) and the first trimester (n = 4, 2.1%). Additionally, 15 women (7.8%) had multiple pregnancies [Table 1]. Overall, 28 women (14.5%) had PE confirmed by CTPA. The number of deliveries at Royal Hospital and at Sultan Qaboos University Hospital was 93,383 during the study period. The PE incidence rate has been calculated to be 0.31 in 1,000 deliveries [Fig. 1].

Table 1.

Characteristics of pregnant and postpartum women with pulmonary embolism (Wells score criteria are bolded).

Characteristic	Total number of patients	n (%)
Mean age in years ± SD (range)	193	31.45 ± 5.60 (19–45)
Mean BMI in kg/m2 ± SD	167	31.10 ± 7.62
BMI >30 in kg/m2	167	91 (54.5)
Mean O2 saturation in % ± SD	192	93.5 ± 7.03
Heart rate >100 in beats/min	193	117 (60.6)
Mean systolic BP in % ± SD	192	125.53 ± 19.96
Mean diastolic BP in % ± SD	192	74.7 ± 13.97
Mean temperature °C in % ± SD	181	36.9 ± 3.9
Clinical signs or symptoms of DVT	193	4 (2.1)
Previously diagnosed DVT or PE	193	10 (5.2)
Alternative diagnosis is less likely than PE	193	120 (62.2)
Immobilisation or major surgery in last 4 weeks	193	128 (66.3)
Haemoptysis in %	193	8 (4.1)
Malignancy within last 6 months	193	2 (1.0)
Family history of VTE	193	1 (0.5)
Coagulation disorder	193	11 (5.7%)
History of varicose veins	193	1 (0.5)
Sickle cell disease	193	30 (15.5)
Autoimmune disorder	193	11 (5.7)
Chest pain	193	90 (46.6)
Dyspnoea	193	110 (57)
Palpitation	193	29 (15)
Cough	193	38 (19.7)

SD = standard deviation; BMI = body mass index; BP = blood pressure; DVT = deep vein thrombosis; PE = pulmonary embolism; VTE = venous thromboembolism.

Fig. 1.

Chart showing the study's population. CTPA = computed tomography pulmonary angiogram; PE = pulmonary embolism.

The most common site of PE was the combination of lobar, segmental and subsegmental PE, occurring in 8 women (28.6%). This was followed by subsegmental PE in 7 women (25%), segmental and subsegmental PE in 6 women (21.4%) and segmental PE in 4 women (14.3%). Additionally, there were 3 cases of bilateral PE (10.7%), 1 case of amniotic fluid PE (3.6%), and 5 inconclusive cases (17.9%).

The majority of women had a moderate WS (n = 142, 73.6%), while 44 women (22.8%) had a low WS and 7 women (3.6%) had a high WS. Among those with a low WS, 40 (90.9%) had a negative CTPA, 2 (4.5%) had a positive result and another 2 (4.5%) had an inconclusive CTPA. For those with a moderate WS, 108 (76.1%) had a negative CTPA, 25 (17.6%) had a positive CTPA and 9 (6.3%) had inconclusive CTPA results. A total of 7 women (3.6%) had high WS and of those only 1 (14.3%) had PE on CTPA [Table 2].

Table 2.

The distribution of the Wells score categories and their relation to computed tomography-pulmonary angiography results.

	CTPA results
Category of MWS	Negative	Positive	Inconclusive	Total	P value
Low	40	2	2	44	0.139
Moderate	108	25	9	142
High	6	1	0	7
Total	154	28	11	193

MWS = modified Wells score; CTPA = computed tomography-pulmonary angiography.

The Chi-square tests show no significant relationship between the WS category and the CTPA results (likelihood ratio = 6.944; P = 0.139). This suggests that the WE may not be a strong predictor of CTPA results in this sample.

Among the elements of the WE, only immobilisation ≥3 days showed a significant predictive value (P = 0.003). This suggests that immobilisation is a significant risk factor for PE in this population.

Other parameters as clinical signs and symptoms of DVT, PE is the most or equally likely diagnosis, heart rate >100, surgery in the previous 4 weeks, prior PE or DVT, haemoptysis and active malignancy were not considered as predictive (P < 0.05). Furthermore, family history of thrombosis, as well as palpitation, chest pain, dyspnoea and cough were not conclusive either.

When comparing women with SCD and those without, women with SCD had a higher incidence of PE, with 11 out of 28 (39.3%) women had positive PE on CTPA, compared to 17 out of 163 (10.4%) women without SCD (odds ratio = 4.9, 95% confidence interval: 2.0–12.1; P < 0.001). Incorporating SCD into the WS by adding 3 points significantly enhances its predictive value for PE in pregnant women with SCD (P < 0.001) [Table 4]. A total of 18 more women were classified as having a high probability of PE which is an increase by 257.1% compared to the WS. Sensitivity of the high probability of PE for the WS was 85.7%, while specificity was estimated to be 72.3% [Table 5].

Table 4.

The distribution of the changed Wells score categories and their relation to computed tomography-pulmonary angiography results after modification with sickle cell disease.

Changed Wells score	Negative	Positive	Inconclusive	Total	P value
Low	34	2	2	38	<0.001
Moderate	107	14	9	130
High	13	12	0	25
Total	154	28	11	193

Table 5.

Positive and negative predictive values for low and high clinical probabilities according to the modified Wells' score.

Wells score	Positive CTPA	Negative CTPA	Total
Low	2	34	36
High	12	13	25
Total	14	47	61
Positive predictive value for low Wells score, n (%)		2/36 (5.6)
Negative predictive value for low Wells score, n (%)		34/36 (94.4)
Positive predictive value for high Wells score, n (%)		12/25 (48)
Negative predictive value for high Wells score, n (%)		13/25 (52)
Sensitivity of high probability of PE, %		72.3
Specificity of high probability of PE, %		85.7

CTPA = computed tomography-pulmonary angiography.

Table 3.

Positive and negative predictive values for low and high clinical probabilities according to the Wells score.

Wells score	Positive CTPA	Negative CTPA	Total
Low	2	40	42
High	1	6	7
Total	3	46	49
Positive predictive value for low Wells score, n (%)		2/42 (4.8)
Negative predictive value for low Wells score, n (%)		40/42 (95.2)
Positive predictive value for high Wells score, n (%)		1/7 (14.3)
Negative predictive value for high Wells score, n (%)		6/7 (85.7)
Sensitivity of high probability of PE, %		33.3
Specificity of high probability of PE, %		87.0

CTPA = computed tomography-pulmonary angiography.

4. Discussion

PE remains a significant cause of maternal mortality worldwide, highlighting the importance of studying this condition in pregnant and postpartum women. Despite advancements in diagnostic and therapeutic strategies, the incidence of PE in pregnancy continues to pose a serious threat due to physiological changes that increase thromboembolic risk. Therefore, understanding the predictive value of diagnostic tools such as the WS is crucial for improving maternal outcomes.

In the current study, 28 women (14.5%) had PE confirmed by CTPA. The majority of women were in their third trimester and postpartum period. This distribution aligns with the increased risk of thromboembolic events during the later stages of pregnancy and the postpartum period as shown in other studies.³ The current study reported an incidence of PE of 0.31 per 1,000 deliveries which is lower than what is estimated in the literature (0.2–1.5 per 1,000 deliveries).² This could be due to the fact that the current study included only PE confirmed positive by CTPA but some of the inconclusive CTPA could also have been positive. In addition, the incidence could vary depending on what source was chosen for calculations.⁴³ For example, when the discharge diagnosis is used, the incidence of PE would be higher because the majority of inconclusive CTPA were treated for PE and that was included in the discharge summary as a diagnosis. Moreover, this calculation is based only on hospitalised patients. It is possible that some of the maternal mortality cases, who had no CTPA or weren't at the hospital, could have been due to PE.²

This study showed no significant relationship between the original WS and CTPA results (P = 0.139). This finding suggests that the WS, in its current form, is not sufficiently predictive of PE in the current study population. The sensitivity and specificity for the high-risk WS were 33.3% and 87.0%, respectively, indicating a high rate of false negatives. However, a low WS can be used to exclude PE as it has a high negative predictive value of 95.2%. In a study of 103 patients, of whom 31 with confirmed PE by CTPA, Touhami et al. showed that sensitivity, specificity, positive predictive value and negative predictive value of the high-risk WS was 40.7%, 81.5%, 44% and 79.4%, respectively.²⁹ Although the current study's sample size was bigger, the results are comparable to Touhami et al.

A low WS demonstrated a high negative predictive value (>95%), suggesting that PE can be reasonably excluded in this group, potentially reducing unnecessary CTPA exposure. In clinical practice, clinicians may opt for close monitoring, D-dimer testing or alternative imaging (e.g., lower limb ultrasonography) instead of immediate CTPA, balancing the risks of radiation exposure and contrast-related complications against the low likelihood of PE. This proposed use is also supported by O'Connor et al.¹⁹

This inadequacy of WS necessitates a re-evaluation of the scoring criteria to enhance its predictive power in pregnant and postpartum women. Interestingly, among the various elements of the WS, only immobilisation for ≥3 days was found to have a significant predictive value (P = 0.003). This underscores the importance of considering prolonged immobilisation as a critical risk factor in this population. The lack of significance for other parameters, such as heart rate >100, recent surgery and clinical signs of DVT, indicates that these factors may not be highly relevant in the context of pregnancy and postpartum risk for PE.

The current study's findings strongly support the inclusion of SCD as a significant risk factor for PE. The substantial increase in PE incidence among women with SCD (36.7%) compared to those without (10.4%) highlights the profound impact of SCD on thromboembolic risk. The significant association (P = 0.003) underscores the necessity of incorporating SCD into the risk assessment models for PE, particularly in populations with a higher prevalence of SCD, such as the Omani population. Adding 3 points for SCD to the WS significantly enhanced its predictive value for PE, as evidenced by an increase in sensitivity of 72.3% and specificity of 85.7%. This substantial improvement demonstrates the effectiveness of this adjustment in better identifying women at risk of PE. The high sensitivity ensures that most cases of PE are correctly identified, reducing the likelihood of missed diagnoses. However, there is still a notable rate of false positives, which must be managed clinically to avoid unnecessary interventions.

Compared to other studies investigating the clinical accuracy of WS in pregnancy and postpartum, this study has a larger sample size than many other publications. It also investigated individual components of WS and their predictability of PE individually in the obstetric population. Furthermore, incorporating SCD in WS as a population specific risk factor is a novel concept and is important to enhance the clinical prediction of PE in populations where SCD is prevalent.

However, this study has several limitations that should be considered. The sample size, while adequate, may not be large enough to detect subtle differences in PE risk across different subgroups. Additionally, the retrospective design may introduce selection bias, as well as the data collected from the medical records systems by different investigators. Excluding patients with inconclusive CTPA results may introduce selection bias, as these cases could include undiagnosed PE. However, since the CTPA is the gold standard for diagnosing PE, the exclusion of inconclusive cases was necessary as there was no way to confirm which of them is true PE or not. In some CTPA reports, the terminology used varies, and the description of the pulmonary tree levels (main, segmental, and subsegmental branches) is inconsistent. It is important to highlight that this study was conducted in the Omani population, where SCD is a common disorder. This risk factor might not be as relevant in other populations and that introduces a limitation on the utility of the SCD-modified WS. Findings may not be generalised to low SCD prevalence populations and further testing in larger, high-prevalence populations is needed.

5. Conclusions

The study highlights that a low WS effectively excludes PE with a high negative predictive value but the high score's sensitivity limits its standalone diagnostic utility in pregnant and postpartum women. Further modification with a population specific risk factors for PE such as SCD may result in enhancing its predictive accuracy by improving its sensitivity.

Authors' Contribution

Maryam Al Shukri: Conceptualization, Formal analysis, Writing-original draft, writing-review and editing. AlBayan Al Musalami: Data curation, Formal Analysis. Karolina Stachyra: Data curation, Formal analysis. Elias Houari: Data curation, Formal analysis. Vaidyanathan Gowri: Writing- review & editing. Sumaya Al Amri: Resources. Waoud Al Saidi: Data curation. Khadija D.B. Ishaq: Data curation. Silja A. Pillai: Conceptualization, Resources.

Ethics Statement

The study was approved by the Medical Research and Ethics Committee of the College of Medicine and Health Sciences at Sultan Qaboos University (MREC #2220) and the ethical committee of Royal Hospital (SRC#3/2021).

Conflict of Interest

The authors declare no conflicts of interest.

Funding

No funding was received for this study.

Data Availability

Data is available upon reasonable request from the corresponding author.