Dataset Information

Improving preterm newborn identification in low-resource settings with machine learning.

ABSTRACT: BACKGROUND:Globally, preterm birth is the leading cause of neonatal death with estimated prevalence and associated mortality highest in low- and middle-income countries (LMICs). Accurate identification of preterm infants is important at the individual level for appropriate clinical intervention as well as at the population level for informed policy decisions and resource allocation. As early prenatal ultrasound is commonly not available in these settings, gestational age (GA) is often estimated using newborn assessment at birth. This approach assumes last menstrual period to be unreliable and birthweight to be unable to distinguish preterm infants from those that are small for gestational age (SGA). We sought to leverage machine learning algorithms incorporating maternal factors associated with SGA to improve accuracy of preterm newborn identification in LMIC settings. METHODS AND FINDINGS:This study uses data from an ongoing obstetrical cohort in Lusaka, Zambia that uses early pregnancy ultrasound to estimate GA. Our intent was to identify the best set of parameters commonly available at delivery to correctly categorize births as either preterm (<37 weeks) or term, compared to GA assigned by early ultrasound as the gold standard. Trained midwives conducted a newborn assessment (<72 hours) and collected maternal and neonatal data at the time of delivery or shortly thereafter. New Ballard Score (NBS), last menstrual period (LMP), and birth weight were used individually to assign GA at delivery and categorize each birth as either preterm or term. Additionally, machine learning techniques incorporated combinations of these measures with several maternal and newborn characteristics associated with prematurity and SGA to develop GA at delivery and preterm birth prediction models. The distribution and accuracy of all models were compared to early ultrasound dating. Within our live-born cohort to date (n = 862), the median GA at delivery by early ultrasound was 39.4 weeks (IQR: 38.3-40.3). Among assessed newborns with complete data included in this analysis (n = 468), the median GA by ultrasound was 39.6 weeks (IQR: 38.4-40.3). Using machine learning, we identified a combination of six accessible parameters (LMP, birth weight, twin delivery, maternal height, hypertension in labor, and HIV serostatus) that can be used by machine learning to outperform current GA prediction methods. For preterm birth prediction, this combination of covariates correctly classified >94% of newborns and achieved an area under the curve (AUC) of 0.9796. CONCLUSIONS:We identified a parsimonious list of variables that can be used by machine learning approaches to improve accuracy of preterm newborn identification. Our best-performing model included LMP, birth weight, twin delivery, HIV serostatus, and maternal factors associated with SGA. These variables are all easily collected at delivery, reducing the skill and time required by the frontline health worker to assess GA. TRIAL REGISTRATION:ClinicalTrials.gov Identifier: NCT02738892.

SUBMITTER: Rittenhouse KJ

PROVIDER: S-EPMC6392324 | biostudies-literature | 2019

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Improving preterm newborn identification in low-resource settings with machine learning.

Rittenhouse Katelyn J KJ Vwalika Bellington B Keil Alexander A Winston Jennifer J Stoner Marie M Price Joan T JT Kapasa Monica M Mubambe Mulaya M Banda Vanilla V Muunga Whyson W Stringer Jeffrey S A JSA

PloS one 20190227 2

<h4>Background</h4>Globally, preterm birth is the leading cause of neonatal death with estimated prevalence and associated mortality highest in low- and middle-income countries (LMICs). Accurate identification of preterm infants is important at the individual level for appropriate clinical intervention as well as at the population level for informed policy decisions and resource allocation. As early prenatal ultrasound is commonly not available in these settings, gestational age (GA) is often es ...[more]

PMID: 30811399

Similar Datasets

Project description:BackgroundMaternal and neonatal mortality rates remain high in many low-income and middle-income countries. Different approaches for the improvement of birth outcomes have been used in community-based interventions, with heterogeneous effects on survival. We assessed the effects of women's groups practising participatory learning and action, compared with usual care, on birth outcomes in low-resource settings.MethodsWe did a systematic review and meta-analysis of randomised controlled trials undertaken in Bangladesh, India, Malawi, and Nepal in which the effects of women's groups practising participatory learning and action were assessed to identify population-level predictors of effect on maternal mortality, neonatal mortality, and stillbirths. We also reviewed the cost-effectiveness of the women's group intervention and estimated its potential effect at scale in Countdown countries.FindingsSeven trials (119,428 births) met the inclusion criteria. Meta-analyses of all trials showed that exposure to women's groups was associated with a 37% reduction in maternal mortality (odds ratio 0.63, 95% CI 0.32-0.94), a 23% reduction in neonatal mortality (0.77, 0.65-0.90), and a 9% non-significant reduction in stillbirths (0.91, 0.79-1.03), with high heterogeneity for maternal (I(2)=58.8%, p=0.024) and neonatal results (I(2)=64.7%, p=0.009). In the meta-regression analyses, the proportion of pregnant women in groups was linearly associated with reduction in both maternal and neonatal mortality (p=0.026 and p=0.011, respectively). A subgroup analysis of the four studies in which at least 30% of pregnant women participated in groups showed a 55% reduction in maternal mortality (0.45, 0.17-0.73) and a 33% reduction in neonatal mortality (0.67, 0.59-0.74). The intervention was cost effective by WHO standards and could save an estimated 283,000 newborn infants and 41,100 mothers per year if implemented in rural areas of 74 Countdown countries.InterpretationWith the participation of at least a third of pregnant women and adequate population coverage, women's groups practising participatory learning and action are a cost-effective strategy to improve maternal and neonatal survival in low-resource settings.FundingWellcome Trust, Ammalife, and National Institute for Health Research Collaboration for Leadership in Applied Health Research and Care for Birmingham and the Black Country programme.

Project description:BackgroundMillions of newborns die annually from preventable causes, with the highest rates occurring in Africa. Reducing neonatal mortality requires investment to scale hospital care, which includes providing hospitals with appropriate technology to care for small and sick newborns. Expensive medical devices designed for high-resource settings often fail to withstand conditions in low-resource hospitals, including humidity, dust, frequent user turnover, complex maintenance, lack of stable power, or difficulty sourcing expensive consumables. Rigorous evaluation protocols are needed to identify effective, affordable, rugged, and easy-to-use medical devices appropriate for quality hospital-based newborn care in low-resource hospitals.MethodsWe developed an evidence-based technology review process to identify medical devices suitable for small and sick newborn care in low-resource hospitals. The eight-step process consists of: identifying devices needed for effective newborn care; defining Target Product Profiles (TPPs); identifying commercially-available products that may meet TPPs; conducting desk research to evaluate technologies against TPPs; performing technical performance verification testing under laboratory conditions; verifying technical performance after exposure to heat, humidity, dust, and power loss; performing usability evaluations with nurses, and qualifying devices that pass all steps. Devices were purchased, installed, and monitored in newborn wards across Kenya, Malawi, Nigeria, and Tanzania.ResultsOf 271 devices considered, only 45 (16.6%) met corresponding TPPs based on desk research. Thirty-nine were purchased and evaluated in the laboratory; five (12.8%) failed to meet TPPs. Thirty-four products passing laboratory evaluation underwent short-term environmental testing; only one (2.9%) device failed. Thirty-seven products underwent usability testing with 127 clinicians; surprisingly, 14 (37.8%) failed to meet TPPs. Twenty-three products passed all evaluations, and 2457 devices were installed across 65 newborn wards in Kenya, Malawi, Nigeria, and Tanzania. Continuous device monitoring reported minimal device failures, with failed devices typically returned to service within two days, resulting in an average uptime (service days divided by days installed) of 99%.ConclusionAn evidence-based device selection process can improve procurement of effective, affordable, rugged, usable newborn care devices for low-resource hospitals, and feedback to manufacturers can improve device quality. Similar processes could be adapted beyond newborn care to identify medical devices suitable for implementation in any low-resource setting.

Project description:Background: Task sharing of TTE may improve capacity for heart failure diagnosis and management in patients in remote, low-resource settings but the impact on diagnostic accuracy and patient outcomes has not been studied.Objectives: Determine feasibility and impact of non-expert training in transthoracic echocardiography (TTE) to improve the diagnosis and outcomes of patients with suspected heart failure in Uganda.Methods: This two-part study examined an innovative training program to develop TTE competency among non-experts and used a pre-post design to determine the impact of decentralized TTE. Four of 8 non-experts (50%) passed a three-part training course. The training comprised of distance learning through a web-based curriculum, a 2-day hands-on workshop with cardiologists, and independent practice with remote mentorship. Continuous measures were compared (pre- vs. post-TTE) using t-tests or Wilcoxon rank-sum tests as distributionally appropriate and categorical variables assessed through chi-square testing. Sensitivity and specificity were calculated according to standard methodology comparing diagnosis pre- and post-TTE during phase 2.Results: Performance in the post-training phase showed good agreement with expert categorization (? = 0.80) with diagnostic concordance in 421 of 454 studies (92.7%). TTE changed the preliminary diagnosis in 81% of patients, showing low specificity of clinical decision-making alone (14.2%; 95% CI 10.1-19.2%). Dilated cardiomyopathy, hypertensive heart disease with preserved systolic function, and right heart failure were the most underdiagnosed conditions prior to TTE while hypertensive heart disease with decreased systolic function was the most over-diagnosed condition.Conclusions: In conclusion, non-expert providers can achieve a high level of proficiency for the categorization of heart failure using handheld TTE in low-resource settings and use of telemedicine and remote mentorship may improve performance and feasibility. The addition of TTE resulted in substantial improvement in etiological specificity. Further study is needed to understand implications of this strategy on healthcare utilization, long-term patient outcomes, and cost.

Project description:BackgroundThe impending scale up of noncommunicable disease screening programs in low- and middle-income countries coupled with limited health resources require that such programs be as accurate as possible at identifying patients at high risk.ObjectiveThe aim of this study was to develop machine learning-based risk stratification algorithms for diabetes and hypertension that are tailored for the at-risk population served by community-based screening programs in low-resource settings.MethodsWe trained and tested our models by using data from 2278 patients collected by community health workers through door-to-door and camp-based screenings in the urban slums of Hyderabad, India between July 14, 2015 and April 21, 2018. We determined the best models for predicting short-term (2-month) risk of diabetes and hypertension (a model for diabetes and a model for hypertension) and compared these models to previously developed risk scores from the United States and the United Kingdom by using prediction accuracy as characterized by the area under the receiver operating characteristic curve (AUC) and the number of false negatives.ResultsWe found that models based on random forest had the highest prediction accuracy for both diseases and were able to outperform the US and UK risk scores in terms of AUC by 35.5% for diabetes (improvement of 0.239 from 0.671 to 0.910) and 13.5% for hypertension (improvement of 0.094 from 0.698 to 0.792). For a fixed screening specificity of 0.9, the random forest model was able to reduce the expected number of false negatives by 620 patients per 1000 screenings for diabetes and 220 patients per 1000 screenings for hypertension. This improvement reduces the cost of incorrect risk stratification by US $1.99 (or 35%) per screening for diabetes and US $1.60 (or 21%) per screening for hypertension.ConclusionsIn the next decade, health systems in many countries are planning to spend significant resources on noncommunicable disease screening programs and our study demonstrates that machine learning models can be leveraged by these programs to effectively utilize limited resources by improving risk stratification.

Project description:IntroductionNeonatal mortality associated with preterm birth can be reduced with antenatal corticosteroids (ACS), yet <10% of eligible pregnant women in low-middle income countries. The inability to accurately determine gestational age (GA) leads to under-identification of high-risk women who could receive ACS or other interventions. To facilitate better identification in low-resource settings, we developed a color-coded tape for uterine height (UH) measurement and estimated its accuracy identifying preterm pregnancies.MethodsWe designed a series of colored-coded tapes with segments corresponding to UH measurements for 20-23.6 weeks, 24.0-35.6 weeks, and >36.0 weeks GA. In phase 1, UH measurements were collected prospectively in the Democratic Republic of Congo, India and Pakistan, using distinct tapes to address variation across regions and ethnicities. In phase 2, we tested accuracy in 250 pregnant women with known GA from early ultrasound enrolled at prenatal clinics in Argentina, India, Pakistan and Zambia. Providers masked to the ultrasound GA measured UH. Receiver operating characteristics (ROC) analysis was conducted.Results1,029 pregnant women were enrolled. In all countries the tapes were most effective identifying pregnancies between 20.0-35.6 weeks, compared to the other GAs. The ROC areas under the curves and 95% confidence intervals were: Argentina 0.69 (0.63, 0.74); Zambia 0.72 (0.66, 0.78), India 0.84 (0.80, 0.89), and Pakistan 0.83 (0.78, 0.87). The sensitivity and specificity (and 95% confidence intervals) for identifying pregnancies between 20.0-35.6 weeks, respectively, were: Argentina 87% (82%-92%) and 51% (42%-61%); Zambia 91% (86%-95%) and 50% (40%-60%); India 78% (71%-85%) and 89% (83%-94%); Pakistan 63% (55%-70%) and 94% (89%-99%).ConclusionsWe observed moderate-good accuracy identifying pregnancies ? 35.6 weeks gestation, with potential usefulness at the community level in low-middle income countries to facilitate the preterm identification and interventions to reduce preterm neonatal mortality. Further research is needed to validate these findings on a population basis.

Dataset Information

Improving preterm newborn identification in low-resource settings with machine learning.

Publications

Improving preterm newborn identification in low-resource settings with machine learning.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets