Predicting facility-based delivery in Zanzibar: The vulnerability of machine learning algorithms to adversarial attacks.

Background: Community health worker (CHW)-led maternal health programs have contributed to increased facility-based deliveries and decreased maternal mortality in sub-Saharan Africa. The recent adoption of mobile devices in these programs provides an opportunity for real-time implementation of machine learning predictive models to identify women most at risk for home-based delivery. However, it is possible that falsified data could be entered into the model to get a specific prediction result - known as an "adversarial attack". The goal of this paper is to evaluate the algorithm's vulnerability to adversarial attacks. Methods: The dataset used in this research is from the Uzazi Salama ("Safer Deliveries") program, which operated between 2016 and 2019 in Zanzibar. We used LASSO regularized logistic regression to develop the prediction model. We used "One-At-a-Time (OAT)" adversarial attacks across four different types of input variables: binary - access to electricity at home, categorical - previous delivery location, ordinal - educational level, and continuous - gestational age. We evaluated the percent of predicted classifications that change due to these adversarial attacks. Results: Manipulating input variables affected prediction results. The variable with the greatest vulnerability was previous delivery location, with 55.65% of predicted classifications changing when applying adversarial attacks from previously delivered at a facility to previously delivered at home, and 37.63% of predicted classifications changing when applying adversarial attacks from previously delivered at home to previously delivered at a facility. Conclusion: This paper investigates the vulnerability of an algorithm to predict facility-based delivery when facing adversarial attacks. By understanding the effect of adversarial attacks, programs can implement data monitoring strategies to assess for and deter these manipulations. Ensuring fidelity in algorithm deployment secures that CHWs target those women who are actually at high risk of delivering at home.

Machine learning for maternal health: Predicting delivery location in a community health worker program in Zanzibar.

Background: Maternal and neonatal health outcomes in low- and middle-income countries (LMICs) have improved over the last two decades. However, many pregnant women still deliver at home, which increases the health risks for both the mother and the child. Community health worker programs have been broadly employed in LMICs to connect women to antenatal care and delivery locations. More recently, employment of digital tools in maternal health programs have resulted in better care delivery and served as a routine mode of data collection. Despite the availability of rich, patient-level data within these digital tools, there has been limited utilization of this type of data to inform program delivery in LMICs. Methods: We use program data from 38,787 women enrolled in Safer Deliveries, a community health worker program in Zanzibar, to build a generalizable prediction model that accurately predicts whether a newly enrolled pregnant woman will deliver in a health facility. We use information collected during the enrollment visit, including demographic data, health characteristics and current pregnancy information. We apply four machine learning methods: logistic regression, LASSO regularized logistic regression, random forest and an artificial neural network; and three sampling techniques to address the imbalanced data: undersampling of facility deliveries, oversampling of home deliveries and addition of synthetic home deliveries using SMOTE. Results: Our models correctly predicted the delivery location for 68%-77% of the women in the test set, with slightly higher accuracy when predicting facility delivery versus home delivery. A random forest model with a balanced training set created using undersampling of existing facility deliveries accurately identified 74.4% of women delivering at home. Conclusions: This model can provide a "real-time" prediction of the delivery location for new maternal health program enrollees and may enable early provision of extra support for individuals at risk of not delivering in a health facility, which has potential to improve health outcomes for both mothers and their newborns. The framework presented here is applicable in other contexts and the selection of input features can easily be adapted to match data availability and other outcomes, both within and beyond maternal health.

Exploring the use of mobile phone data for national migration statistics.

Statistics on internal migration are important for keeping estimates of subnational population numbers up-to-date as well as urban planning, infrastructure development and impact assessment, among other applications. However, migration flow statistics typically remain constrained by the logistics of infrequent censuses or surveys. The penetration rate of mobile phones is now high across the globe with rapid recent increases in ownership in low-income countries. Analysing the changing spatiotemporal distribution of mobile phone users through anonymized call detail records (CDRs) offers the possibility to measure migration at multiple temporal and spatial scales. Based on a dataset of 72 billion anonymized CDRs in Namibia from October 2010 to April 2014, we explore how internal migration estimates can be derived and modelled from CDRs at subnational and annual scales, and how precision and accuracy of these estimates compare to census-derived migration statistics. We also demonstrate the use of CDRs to assess how migration patterns change over time, with a finer temporal resolution compared to censuses. Moreover, we show how gravity-type spatial interaction models built using CDRs can accurately capture migration flows. Results highlight that estimates of migration flows made using mobile phone data is a promising avenue for complementing more traditional national migration statistics and obtaining more timely and local data.

Development of a medication review service for patients with enteral tubes in a community teaching hospital.

PURPOSE: The results of a study to develop a hospital-wide medication review service for patients with enteral tubes to improve patient safety are presented. SUMMARY: Inappropriate enteral administration of medications can result in occluded tubes, altered clinical response, and an increase in adverse effects. At Saint Barnabas Medical Center, a 600-bed community teaching hospital located in Livingston, New Jersey, a medication review service for patients with an enteral tube was developed. A phased approach was used. In phase 1, a retrospective chart review revealed that 43% of our patients with enteral tubes received at least one medication that should not be crushed. In phase 2, we identified formulary medications that should not be crushed based on guidance from the Institute for Safe Medication Practices. We added a "do not crush" warning to the identified medications in our electronic medication administration record and automated medication dispensing system. In phase 3, we created an automatic substitution list of medications. Phase 4 involved the development of the program in our health information technology platform. An electronic task list alerted pharmacists about patients with enteral tubes who required medication review and potential medication substitutions, as well as patients with newly removed enteral tubes who can be placed back on their original medications. In phase 5, we provided education to prescribers, nurses, and pharmacists. CONCLUSION: A hospital-wide medication review service for patients with enteral tubes at our community teaching medical center was developed.