Can Natural Language Processing Improve Adnexal Torsion Predictions?

STUDY OBJECTIVE: To create a decision support tool based on machine learning algorithms and natural language processing (NLP) technology, to augment clinicians' ability to predict cases of suspected adnexal torsion. DESIGN: Retrospective cohort study SETTING: Gynecology department, university-affiliated teaching medical center, 2014-2022. PATIENTS: This study assessed risk-factors for adnexal torsion among women managed surgically for suspected adnexal torsion based on clinical and sonographic data. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: The dataset included demographic, clinical, sonographic, and surgical information obtained from electronic medical records. NLP was used to extract insights from unstructured free text and unlock them for automated reasoning. The machine learning model was a CatBoost classifier that utilizes gradient boosting on decision trees. The study cohort included 433 women who met inclusion criteria and underwent laparoscopy. Among them, 320 (74%) had adnexal torsion diagnosed during laparoscopy, and 113 (26%) did not. The model developed improved prediction of adnexal torsion to 84%, with a recall of 95%. The model ranked several parameters as important for prediction. Age, difference in size between ovaries, and the size of each ovary were the most significant. The precision for the "no torsion" class was 77%, with a recall of 45%. CONCLUSIONS: Using machine learning algorithms and NLP technology as a decision-support tool for the diagnosis of adnexal torsion is feasible. It improved true prediction of adnexal torsion to 84% and decreased cases of unnecessary laparoscopy.

Are There Similarities in Pregnancy Complications and Delivery Outcomes among Sisters?

This retrospective cohort study evaluated pregnancy outcomes and similarities between pairs of nulliparous sisters with a singleton fetus who delivered between 2013 and 2020. The "Sister-1 group" was defined as the sibling who delivered first, while the "Sister-2 group" included the siblings who gave birth after Sister-1. Obstetrical complications and delivery outcomes were compared. The relative risk for recurrence of a complication in Sister-2 was calculated. The study included 743 sister pairs. There were no between-group differences in maternal BMI, gestational age at delivery, gravidity, smoking, or epidural rates. The Sister-2 group was older than the Sister-1 group (26.4 ± 5 vs. 25.8 ± 4.7 years, respectively, p = 0.05). Higher birthweights and more large-for-gestational-age infants characterized the Sister-2 group compared with the Sister-1 group (3241 ± 485 g vs. 3148 ± 536 g, p < 0.001 and 7.7% vs. 4.8%, p = 0.025, respectively). There were no between-group differences in the rate of small-for-gestational-age, gestational diabetes, hypertensive disorders, pre-term births, vacuum extraction, or cesarean deliveries. Logistic regression analysis found that if Sister-1 underwent vacuum extraction, her sibling had an increased risk for vacuum delivery (adjusted RR 3.03, 95% CI 1.4−6.7; p = 0.003) compared with those whose sibling (Sister-1) did not. There was a three-fold risk of vacuum extraction delivery between sisters. This finding could be related to biological inheritance, environmental factors, and/or psychological issues that may affect similarities between siblings' delivery outcomes.

Factors Affecting Clinical over and Underestimation of Fetal Weight-A Retrospective Cohort.

Clinical estimation of fetal weight is an integral component of obstetric care that might dictate the timing and mode of delivery. Inaccurate fetal weight estimation might result in unnecessary interventions or in underestimating potential risks, resulting in inappropriate intrapartum care. This retrospective study assessed factors associated with under- or overestimation of birthweight and evaluated the obstetric implications. It included singleton births ≥24 w with clinically estimated fetal weight (EFW) up to 1 week before delivery, during 2014−2020. Estimates >±10% of the actual birthweight were considered inaccurate and categorized as overestimation (>10% heavier than the actual birthweight) or underestimation (>10% smaller than the birthweight). Multivariable logistic regression was performed to reveal factors associated with inaccurate EFW. Maternal characteristics and obstetric outcomes were compared. The primary outcomes for the overestimation group were the neonatal composite adverse outcome, induction of labor and cesarean delivery rates. The primary outcomes for the underestimation group were rates of shoulder dystocia, 3rd- or 4th-degree perineal lacerations, and failed vacuum extraction. Among 38,615 EFW, 5172 (13.4%) were underestimated, 6695 (17.3%) were overestimated and 27,648 (69.3%) accurate. Multivariable logistic regression found increasing gestational age as an independent risk-factor for underestimation (odds ratio (OR) 1.15 for every additional week, 95% confidence interval (CI) 1.12−1.2). Major factors independently associated with overestimation were nulliparity (OR 1.95, CI 1.76−2.16), maternal obesity (OR 1.52, CI 1.33−1.74), smoking (OR 1.6, CI 1.33−1.93), and oligohydramnios (OR 1.92, CI 1.47−2.5). Underestimation was an independent risk-factor for shoulder dystocia (OR 1.61, CI 1.05−2.46) and 3rd- or 4th-degree perineal lacerations (OR 1.59, CI 1.05−2.43). Overestimation was an independent risk-factor for neonatal composite adverse outcome (OR 1.15, CI 1.02−1.3), induced labor (OR 1.30, CI 1.21−1.40) and cesarean delivery (OR 1.59, CI 1.41−1.79). Clinicians should be aware of factors and adverse obstetric implications associated with over- or underestimation of birthweight.

Can the Need for Invasive Intervention in Tubo-ovarian Abscess Be Predicted? The Implication of C-reactive Protein Measurements.

STUDY OBJECTIVE: To evaluate the clinical parameters of hospitalized patients with pelvic inflammatory disease (PID) for the presence of tubo-ovarian abscess (TOA) and predict the need for intervention. DESIGN: A prospective cohort study. SETTING: A tertiary care university medical center. PATIENTS: Ninety-four patients were diagnosed with complicated PID and hospitalized between 2015 and 2017. INTERVENTIONS: Patients with PID were treated with parenteral antibiotics according to Centers for Disease Control guidelines. Demographic, clinical, sonographic, and laboratory data for patients with PID were analyzed. Inflammatory markers including C-reactive protein (CRP), white blood cells (WBCs), erythrocyte sedimentation rate (ESR), and clinical parameters were collected at admission and during hospitalization. MEASUREMENTS AND MAIN RESULTS: Forty-eight of 94 patients (51.1%) hospitalized with complicated PID were diagnosed with TOA sonographically. CRP levels were the strongest predictor of TOA, followed by WBC count, ESR, and fever on admission. The areas under the receiver operating characteristic (ROC) curve for CRP, WBC, ESR, and fever were .92, .75, .73 and .62, respectively. CRP specificity was 93.4% and sensitivity was 85.4% for predicting TOA, with cutoff value of 49.3 mg/L. Twelve patients (25%) failed conservative management and underwent surgical intervention including laparoscopy (n = 7), computed tomography (CT)-guided drainage (n = 4), and laparotomy (n = 1). In this group, CRP levels significantly increased from admission to day 1 and day 2 during hospitalization (128.26, 173.75, and 214.66 mg/L, respectively; p < .05 for both). In the conservative management group, CRP levels showed a plateau from admission to day 1 and then a decrease until day 3 (110, 120.49, 97.52, and 78.45 mg/L, respectively). CONCLUSION: CRP is a sensitive, specific inflammatory marker for predicting TOA in patients with complicated PID, and levels >49.3 mg/L suggest the presence of TOA. In the TOA group, CRP level trends correlated well with success or failure of conservative management. Increasing CRP levels during treatment may be used as an indicator of the need for invasive intervention, and daily CRP measurements can help predict the need for invasive intervention.