Evaluating the Manitoba Infant Feeding Database: a Canadian infant feeding surveillance system

Classifying early infant feeding status from clinical notes using natural language processing and machine learning

The objective of this study is to develop and evaluate natural language processing (NLP) and machine learning models to predict infant feeding status from clinical notes in the Epic electronic health records system. The primary outcome was the classification of infant feeding status from clinical notes using Medical Subject Headings (MeSH) terms. Annotation of notes was completed using TeamTat to uniquely classify clinical notes according to infant feeding status. We trained 6 machine learning models to classify infant feeding status: logistic regression, random forest, XGBoost gradient descent, k-nearest neighbors, and support-vector classifier. Model comparison was evaluated based on overall accuracy, precision, recall, and F1 score. Our modeling corpus included an even number of clinical notes that was a balanced sample across each class. We manually reviewed 999 notes that represented 746 mother-infant dyads with a mean gestational age of 38.9 weeks and a mean maternal age of 26.6 years. The most frequent feeding status classification present for this study was exclusive breastfeeding [n = 183 (18.3%)], followed by exclusive formula bottle feeding [n = 146 (14.6%)], and exclusive feeding of expressed mother's milk [n = 102 (10.2%)], with mixed feeding being the least frequent [n = 23 (2.3%)]. Our final analysis evaluated the classification of clinical notes as breast, formula/bottle, and missing. The machine learning models were trained on these three classes after performing balancing and down sampling. The XGBoost model outperformed all others by achieving an accuracy of 90.1%, a macro-averaged precision of 90.3%, a macro-averaged recall of 90.1%, and a macro-averaged F1 score of 90.1%. Our results demonstrate that natural language processing can be applied to clinical notes stored in the electronic health records to classify infant feeding status. Early identification of breastfeeding status using NLP on unstructured electronic health records data can be used to inform precision public health interventions focused on improving lactation support for postpartum patients.

Robust estimation of infant feeding indicators by data quality assessment of longitudinal electronic health records from birth up to 18 months of life

BACKGROUND AND OBJECTIVE

The Baby-Friendly Hospital Initiative (BFHI) is an international strategy aimed at improving breastfeeding practices in health care services. Regular monitoring of indicators is key for BFHI implementation and maintenance. Currently, routine data collected from electronic health records (EHR) is an excellent source for infant feeding monitoring, however data quality (DQ) assessment should be undertaken. The aim of this research is to enable robust estimations of infant feeding indicators through DQ assessment of routine EHR data.

MATERIALS AND METHODS

We use the longitudinal series of healthcare contacts belonging to 6427 children born from 2009 to 2018 in the Health Area V of Murcia (Spain). Longitudinal data came from EHR at hospital discharge and community infant health reviews up to 18 months. The data of each healthcare contact contained a 24-h recall of infant feeding. We perform a DQ process in three phases: (1) an assessment of each-single-contact and the definition of their infant feeding status; (2) a longitudinal DQ assessment of completeness and consistency of the series of contacts to obtain meta-information that guides the duration calculus, for each case, of the different types of breastfeeding: exclusive breastfeeding (EBF), full breastfeeding (FBF) and any breastfeeding (ABF); and finally (3) a robust estimation of indicators and description of DQ of each indicator.

RESULTS

We found deficiencies of DQ in 30.42% of single contacts for EBF, 19.02% for FBF and 22.50% for ABF that were used to establish the infant feeding status. However, after longitudinal DQ assessment, we obtained valid and reliable data rates for most indicators such as "median duration of breastfeeding" nearly 90%, both for FBF and ABF, not so for EBF.

CONCLUSIONS

Despite the DQ deficiencies found in raw data, the DQ assurance approach by indicators proposed in this work, allowed us to obtain a robust estimation of indicators with a significant percentage of subjects with valid information for ABF and FBF monitoring. The estimations were consistent with results previously published. The methodology provided with this study allows a continuous and reliable population monitoring of infant feeding indicators of BFHI from routine EHR data.