Sarah Li

Sarah Li

Auto-generated profile, based on publicly available information

Sarah Li

Department of Obstetrics and Gynaecology, National University Hospital, Singapore

Research Focus & Metrics

Academic Career Length

2Years (2025-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
6 August 2026 in Acta Obstetricia et Gynecologica Scandinavica
Verified
Journal Article
Open access
PDF file
KLKelvin Z. X. Lee
Kelvin Z. X. Lee
CPCalista L. S. Por
Calista L. S. Por
TLTak Yein Lim
Tak Yein Lim
PDPradip Dashraath
Pradip Dashraath
WTWei Wei Teo
Wei Wei Teo
SLSarah W. L. Li
Sarah W. L. Li
Introduction: Globally, 1 in 5 maternal deaths are attributable to major postpartum hemorrhage (PPH), despite widespread uterotonic use. Intrauterine tamponade is recommended as second‐line therapy for refractory atonic PPH. Uterine balloon tamponade (UBT) achieves hemostasis through intrauterine pressure. Conversely, the vacuum‐assisted hemorrhage‐control device (VHD) promotes uterine contraction through negative pressure. Limited data compare these modalities. We compared the clinical effectiveness of UBT and VHD in refractory atonic PPH. Material and Methods: This retrospective cohort study was conducted at the National University Hospital in Singapore and included all women treated for primary PPH with either the UBT or VHD between January 2020 and September 2025, during which ROTEM‐guided transfusion and E‐MOTIVE‐aligned practices were implemented. The composite morbidity endpoint, comprising transfusion of ≥4 units of packed red blood cells (PRBC), plasma or cryoprecipitate administration, examination under anesthesia, high dependency or intensive care unit admission, or device failure was compared between groups with logistic regression adjusted for placenta previa major and cesarean delivery. Results: Fifty‐three women met inclusion criteria (33 UBT, 20 VHD). The composite morbidity outcome occurred less frequently with VHD than UBT (45.0% versus 75.8%; unadjusted OR 0.26, 95% CI 0.08–0.86, p = 0.03); after adjusting for placenta previa major and cesarean delivery, this difference was no longer statistically significant (adjusted OR 0.38, 95% CI 0.10–1.41, p = 0.15). Baseline characteristics were similar except for higher rates of placenta previa (45.5% versus 5.0%) and cesarean delivery (78.8% versus 50.0%) in the UBT group. Median estimated blood loss at device insertion was comparable (1500 mL). Among secondary outcomes, VHD was associated with lower median PRBC transfusion (1 vs. 2 units, p = 0.009), fewer women receiving ≥4 PRBC units (5.0% versus 36.4%, p = 0.02), fewer plasma transfusions (5.0% versus 36.4%, p = 0.02), shorter device indwelling duration (10.1 versus 24.3 h, p < 0.001), shorter hospital stay (54 versus 72 h, p = 0.007), and fewer high‐dependency admissions (45.0% versus 72.7%, p = 0.04). Conclusions: The VHD demonstrated outcomes comparable to balloon tamponade in managing refractory atonic PPH, with lower transfusion requirements and shorter hospitalization. VHD may be a useful component of standardized PPH protocols incorporating the E‐MOTIVE bundle and ROTEM‐guided transfusion, warranting further evaluation in broader clinical settings.

28 November 2025 in JMIR Preprints
Verified
Preprint
Open access
SLSarah Li
Sarah Li
DTDavid Y.Y. Tan
David Y.Y. Tan
JZJingxian Zhang
Jingxian Zhang
AMAniza P. Mahyuddin
Aniza P. Mahyuddin
SISebastian E. Illanes
Sebastian E. Illanes
MCMahesh Choolani
Mahesh Choolani
BACKGROUND Maternal outcomes remain inequitable worldwide. Severe morbidity persists, and current risk assessment tools are largely arbitrary, focusing on biomedical factors while overlooking social determinants of health. There is a need for data-driven, artificial intelligence (AI) models to improve early pregnancy risk identification and management. OBJECTIVE To develop and internally validate first-trimester AI-based antenatal risk assessment models across three geographically and socio-ethnically diverse populations (Sweden, Chile, and Singapore), and to compare their performance against existing clinical risk assessment strategies. METHODS Retrospective population-based data from over 500,000 pregnancies from Sweden, Chile, and Singapore were used to develop Machine learning (ML) models predicting a composite of adverse maternal and neonatal outcomes. Models were trained and validated separately for each population using first-trimester variables. Model discrimination, measured by the area under the receiver operating characteristic (AUROC) curve, was compared with corresponding real-world first trimester risk assessment approaches. RESULTS The prevalence of the composite adverse outcome was 10.4% (75,647/727,354) in Sweden, 21.9% (1,302/5,934) in Chile, and 16.3% (6,165/37,813) in Singapore. In Sweden, the guideline-based risk assessment achieved an AUROC of 0.53, compared with 0.65 for the LightGBM model (P<.001). In Chile, the midwifery-led risk assessment achieved an AUROC of 0.52, versus 0.66 from the Traditional ML CatBoost model (P<.001). In Singapore, healthcare professional-based risk assessment reached an AUROC of 0.56, compared with 0.60 for the LightGBM model (P<.05). In the Swedish and Singapore, sociodemographic variables were among the most influential predictive features. CONCLUSIONS AI-based models developed using first-trimester data surpassed that of existing first-trimester clinical risk stratification strategies across all three distinct populations. These findings highlight the potential of integrating social, demographic, and behavioural determinants into AI-driven, clinician-augmented antenatal care frameworks to promote more equitable and personalized pregnancy risk assessment.

20 October 2026 in BMJ Open
Verified
Journal Article
Open access
SLSarah Li
Sarah Li
CPCaitlin Wan Sze Por
Caitlin Wan Sze Por
HRHarshaana Ramlal
Harshaana Ramlal
AMAniza P Mahyuddin
Aniza P Mahyuddin
DTDavid Y Y Tan
David Y Y Tan
SCShiao-Yng Chan
Shiao-Yng Chan

26 August 2026 in Journal of Medical Internet Research
Verified
Journal Article
Open access
SLSarah Li
Sarah Li
DTDavid Y Y Tan
David Y Y Tan
JZJingxian Zhang
Jingxian Zhang
AMAniza P Mahyuddin
Aniza P Mahyuddin
HRHarshaana Ramlal
Harshaana Ramlal
MCMahesh Choolani
Mahesh Choolani
Background: Maternal outcomes remain inequitable worldwide. Severe morbidity persists, and current risk assessment tools are largely arbitrary, focusing on biomedical factors while overlooking social determinants of health. There is a need for data-driven AI models to improve early pregnancy risk identification and management. Objective: The study aimed to develop and internally validate first-trimester AI-based antenatal risk assessment models across three geographically and socioethnically diverse populations (Sweden, Chile, and Singapore) and to compare their performance with existing clinical risk assessment strategies. Methods: We conducted a retrospective population-based study using routinely collected first-trimester data from over 700,000 pregnancies from Sweden, Chile, and Singapore. Separate machine learning models predicting a composite of adverse maternal and neonatal outcomes were trained and internally validated for each population. Input variables were limited to information available at or before 14 weeks’ gestation. Model discrimination, measured by the area under the receiver operating characteristic (AUROC) curve, was compared with corresponding proxies for real-world first-trimester risk assessment approaches in each setting. Model interpretability was assessed using Shapley additive explanations. Results: The prevalence of the composite adverse outcome was 10.40% (75,647/727,354) in Sweden, 21.94% (1302/5934) in Chile, and 16.25% (6145/37,813) in Singapore. In Sweden, the guideline-based risk assessment achieved an AUROC of 0.53, compared with 0.65 for the LightGBM (light gradient boosting machine) model ( P <.001). In Chile, the midwifery-led risk assessment achieved an AUROC of 0.52, versus 0.65 from the traditional machine learning LightGBM model ( P <.001). In Singapore, the health care professional–based risk assessment reached an AUROC of 0.56, compared with 0.60 for the LightGBM model ( P <.05). In the Swedish and Singapore cohorts, sociodemographic variables were among the most influential predictive features. At a false-positive rate of 50.0%, the sensitivities for predicting the primary composite adverse outcome were 69.17%, 69.62%, and 64.93% for the Sweden, Chile, and Singapore models, respectively. At a 30.0% false-positive rate operating point, the sensitivities were lower, at 50.38%, 55.38%, and 44.75%, respectively, but with higher positive predictive values of 16.31%, 34.53%, and 22.32%, respectively. The model calibration plots showed reasonable agreement in Sweden and Singapore, whereas the Chile model showed poorer calibration, with a calibration slope of approximately 1.45, indicating underconfidence. Conclusions: AI-based models developed using first-trimester data generally demonstrated improved performance compared with existing first-trimester clinical risk stratification strategies across three distinct populations. These findings suggest the potential feasibility of population-specific, AI-enabled risk stratification as a clinical decision support tool and highlight the potential value of integrating social, demographic, and behavioral determinants into antenatal risk assessment frameworks to support more equitable and personalized antenatal care.

18 November 2026 in Midwifery
Verified
Journal Article
Closed access
TLTak Yein Lim
Tak Yein Lim
KLKelvin Z.X. Lee
Kelvin Z.X. Lee
HRHarshaana Ramlal
Harshaana Ramlal
CMCaroline J. Hollins Martin
Caroline J. Hollins Martin
CMColin R. Martin
Colin R. Martin
CMCitra N.Z. Mattar
Citra N.Z. Mattar