Identifying Diabetes Related-Complications in a Real-World Free-Text Electronic Medical Records in Hebrew Using Natural Language Processing Techniques.

BACKGROUND: Studies have demonstrated that 50% to 80% of patients do not receive an International Classification of Diseases (ICD) code assigned to their medical encounter or condition. For these patients, their clinical information is mostly recorded as unstructured free-text narrative data in the medical record without standardized coding or extraction of structured data elements. Leumit Health Services (LHS) in collaboration with the Israeli Ministry of Health (MoH) conducted this study using electronic medical records (EMRs) to systematically extract meaningful clinical information about people with diabetes from the unstructured free-text notes. OBJECTIVES: To develop and validate natural language processing (NLP) algorithms to identify diabetes-related complications in the free-text medical records of patients who have LHS membership. METHODS: The study data included 2.3 million records of 41 469 patients with diabetes aged 35 or older between the years 2012 and 2017. The diabetes related complications included cardiovascular disease, diabetic neuropathy, nephropathy, retinopathy, diabetic foot, cognitive impairments, mood disorders and hypoglycemia. A vocabulary list of terms was determined and adjudicated by two physicians who are experienced in diabetes care board certified diabetes specialist in endocrinology or family medicine. Two independent registered nurses with PhDs reviewed the free-text medical records. Both rule-based and machine learning techniques were used for the NLP algorithm development. Precision, recall, and F-score were calculated to compare the performance of (1) the NLP algorithm with the reviewers' comments and (2) the ICD codes with the reviewers' comments for each complication. RESULTS: The NLP algorithm versus the reviewers (gold standard) achieved an overall good performance with a mean F-score of 86%. This was better than the ICD codes which achieved a mean F-score of only 51%. CONCLUSION: NLP algorithms and machine learning processes may enable more accurate identification of diabetes complications in EMR data.

The first three years: The association of early postpartum depressive symptoms with infant and toddler development.

BACKGROUND: The influence of postpartum depression (PPD) on child development has been a source of professional interest and practical relevance. OBJECTIVE: This study investigated the association of early PPD symptoms with developmental domains. DESIGN AND METHOD: This historical cohort study included 574,282 children attending Mother Child Healthcare Centers in Israel from January 1, 2014 to July 31, 2020, who underwent at least one developmental screening examination by public health nurses up to age 36 months, and whose mothers completed the Edinburgh Postnatal Depression Scale (EPDS) postnatally. Developmental milestone tasks included four domains: fine and gross motor, language/communication, and social/behavioral. RESULTS: The rate of failure to complete age-appropriate tasks was higher among children whose mothers had scored ≥ 10 on the EPDS on the majority of tasks in every domain. DISCUSSION: This large population-based study has demonstrated the association between early maternal postnatal depressive symptoms and failure to meet developmental milestones across domains, until three years. Recommendations for practice focus on the mother, the child, and health policy.

Utilizing machine learning to predict unplanned cesarean delivery.

OBJECTIVE: To develop a comprehensive machine learning (ML) model predicting unplanned cesarean delivery (uCD) among singleton pregnancies based on features available at admission to labor. METHODS: A retrospective cohort study from a tertiary medical center. Women with singleton vertex pregnancy of 34 weeks or more admitted for vaginal delivery between March 2011 and May 2019 were included. The cohort was divided into training (80%) and validation (20%) data sets. A separate cohort between June 2019 and April 2021 served as a test data set. Features selection was performed using a Random Forest ML algorithm. RESULTS: The study population included 73 667 women, of which 4125 (6.33%) underwent uCD. The final model consisted of 13 features, based on prediction importance. The XGBoost model performed best with areas under the curve for the training, validation, and test data sets of 0.874, 0.839, and 0.840, respectively. The model showed a 65% positive predictive value for uCD among women in the 100th centile group, and a 99% or more negative predictive value in the less than 50th centile group. Positive and negative predictive values remained high among subgroups with high pretest probability of uCD. CONCLUSION: An ML model for the prediction of uCD provides clinically useful risk stratification that remains accurate across gestational weeks 34-42 and among clinical risk groups. The model may be clinically useful for physicians and women admitted for labor. SYNOPSIS: A machine learning model predicts unplanned cesarean delivery and can inform women's individualized decision making.

Standardization of a Developmental Milestone Scale Using Data From Children in Israel.

Importance: Routine developmental screening tests for children are used worldwide for early detection of developmental delays. However, assessment of developmental milestone norms lacks strong normative data, and there are inconsistencies among different screening tools. Objective: To establish milestone norms and build an updated developmental scale. Design, Setting, and Participants: This is a cross-sectional, population-based study conducted between 2014 and 2020. Developmental assessments were conducted by trained public health nurses, documented in national maternal child health clinics, known as Tipat Halav, which serve all children in Israel. Participants included all children born between January 2014 and September 2020, who were followed at the maternal child health clinics from birth to age 6 years. Exclusion criteria were preterm birth, missing gestational age, low birth weight (<2.5 kg), abnormal weight measurement (<3% according to standardized child growth charts), abnormal head circumference measurement (<3% or >97% according to standardized child growth charts), and visits without developmental data or without the child's age. Data analysis was performed from September 2020 to June 2021. Exposures: In total, 59 milestones in 4 developmental domains were evaluated, and the achievement rate per child's age was calculated for each milestone. Main Outcomes and Measures: A contemporary developmental scale, the Tipat Halav Israel Screening (THIS) Developmental Scale, was built, presenting the 75%, 90%, and 95% achievement rates for each milestone. The THIS scale was compared with other commonly used screening tests, including the Denver Developmental Screening Test II (Denver II), the Alberta Infant Motor Scale (AIMS), and the Centers for Disease Control and Prevention (CDC) Developmental Assessment. Results: A total of 839 574 children were followed in the maternal child health clinics between January 2014 and September 2020 in Israel, and 195 616 children were excluded. A total of 3 774 517 developmental assessments were performed for the remaining 643 958 children aged 0 to 6 years (319 562 female children [49.6%]), resulting in the establishment of new developmental norms. In terms of the comparable milestones, THIS milestones had a match of 18 of 27 (67%) with the Denver II, 7 of 7 (100%) with AIMS, and 10 of 19 (53%) with the CDC Developmental Assessment. The remaining unmatched milestones were achieved earlier in the THIS scale compared with other screening tools. Conclusions and Relevance: The THIS developmental scale is based on the largest population evaluated to date for developmental performance, representing the heterogeneous, multicultural population comprising this cohort. It is recommended for further evaluation worldwide.

Machine learning based study of longitudinal HbA1c trends and their association with all-cause mortality: Analyses from a National Diabetes Registry.

OBJECTIVE: The association of long-term HbA1c variability with mortality has been previously suggested. However, the significance of HbA1c variability and trends in different age and HbA1c categories is unclear. RESEARCH DESIGN AND METHODS: Data on patients with diabetes listed in the Israeli National Diabetes Registry during years 2012-2016 (observation period) were collected. Patients with >4 HbA1c measurements, type 1 diabetes, eGFR < 30mg/ml/min, persistent HbA1c < 6% or malignancy were excluded. Utilizing machine learning methods, patients were classified into clusters according to their HbA1c trend (increasing, stable, decreasing). Mortality risk during 2017-2019 was calculated in subgroups defined by age (35-54, 55-69, 70-89 years) and last HbA1c (≤7% and >7%) at end of observation period. Models were adjusted for demographic, clinical and laboratory measurements including HbA1c, standard deviation (SD) of HbA1c and HbA1c trend. RESULTS: This historical cohort study included 293,314 patients. Increased HbA1c variability (high SD) during the observation period was an independent predictor of mortality in patients aged more than 55 years (p < 0.01). The HbA1c trend was another independent predictor of mortality. Patients with a decreasing versus stable HbA1c trend had a greater mortality risk; this association persisted in all age groups in patients with HbA1c > 7% at the end of the observation period (p = 0.02 in age 35-54; p < 0.01 in aged >55). Patients with an increasing versus stable HbA1c trend had a greater mortality risk only in the elderly group (>70), yet in both HbA1c categories (p < 0.01). CONCLUSIONS: HbA1c variability and trend are important determinants of mortality risk and should be considered when adjusting glycaemic targets.

Noncontact tremor characterization using low-power wideband radar technology.

Continuous monitoring and analysis of tremor is important for the diagnosis and establishment of treatments in many neurological disorders. This paper describes noncontact assessment of tremor characteristics obtained by an experimental new ultrawideband (UWB) system. The system is based on transmission of a wideband electromagnetic signal with extremely low power, and analysis of the received signal, which is composed of many propagation paths reflected from the patient and its surroundings. A description of the physical principles behind the technology, a criterion, and efficient algorithms to assess tremor characteristics from the bulk UWB measurements are given. A feasibility test for the technology was conducted using a UWB system prototype, an arm model that mimics tremor, and a reference video system. The set of UWB system frequencies and amplitudes estimations were highly correlated with the video system estimations with an average error in the scale of 0.1 Hz and 1 mm for the frequency and amplitude estimations, respectively. The new UWB-based system does not require attaching active markers or inertial sensors to the body, can give displacement information and kinematic features from multiple body parts, is not limited by the range captured by the optical lens, has high indoor volume coverage as it can penetrate through walls, and does not require calibration to obtain amplitude estimations.