Accurate knowledge of gestational age (GA) is a cornerstone of modern maternity care, reducing unnecessary interventions while reducing stillbirths and other obstetric complications. In most LMICs, many women never receive antenatal care or, if they do, present at 20 weeks’ gestation or later. Pregnancy dating using either maternal recall of her last menstrual period or symphysis-fundal height is inaccurate. Although ultrasound is the most accurate approach, lack of expensive equipment and trained staff means accurate GA dating is often unavailable in less-developed countries despite its obvious need. We have developed a mobile health (mHealth) and image analysis ultrasound platform for automated measurement of the fetal transcerebellar diameter (TCD); this measurement correlates with GA and, in the absence of fetal anomalies, is relatively protected from the influence of FGR and remains reliable throughout pregnancy from 14 weeks until term (37+0 weeks).
The TraCer initiative: to develop and validate automated image recognition and measurement software to capture the TCD during an ultrasound examination by minimally-trained health workers. We have (i) derived normalised values of the TCD for every day of gestational age from 16-41 weeks; (ii) developed and validated an automated image capture and measurement tool for TCD; and (iii) created a prototype TraCer mobile health (mHealth) application that can be integrated into the software package of a hand-held ultrasound machine.
We will integrate the TraCer app on a low-cost hand-held device. Currently, the TraCer system consists of a client app that automatically detects and measures the TCD ≈99% of the time and translates this into a GA. PRECISE will introduce the app onto an android tablet with a bluetoothed basic curvilinear ultrasound probe, running the app on a low-cost tablet computer connected to a database server running a web data collection app with automatic transmission of ultrasound data to a centralised database for quality control. This phase will include usability- and beta-testing of the device and iterative feedback to refine the algorithm, if needed.
Qualitative data will be obtained through focus groups and in-depth interviews with health care providers and policy makers to identify current patterns of practice; barriers and facilitators for task-sharing of GA determination using the TraCer app; and what level of heath care worker would be best placed to use it. Analyses will be conducted using nVivo software in local dialect(s), prior to English translation.
To enumerate the prevalence of pregnancy hypertension, accurate blood pressure (BP) measurement is essential. In less-developed countries, pre-eclampsia is frequently undetected as there is low attendance for antenatal care, inadequate training in BP measurement, and insufficient, poorly functioning equipment. With Gates and MRC funding we have developed and clinically validated a semi-automated BP device (Microlife CRADLE Vital Signs Alert (VSA®)) specifically for use in LMICs, including pre-eclampsia. The device comprises a micro-USB port and a sealed rechargeable Li battery pack for charging through generic mobile phone chargers. The manual inflation prolongs battery life. An integrated traffic-light early warning system alerts users to BP abnormalities. Predefined BP thresholds, used as hypertension triggers for the CLIP Trials have been introduced as the amber and red triggers within the traffic light system. As well as being a highly accurate device suitable for LMIC settings, this device costs less than $20 USD, and was named by Innovation Countdown 2030 as one of 30 high impact innovations to save lives (another is the phone oximeter, described below).
Within PRECISE, the CRADLE team will further evaluate the clinical impact of the device in identifying women with pregnancy hypertension in both urban and rural settings. The CRADLE device will be the cornerstone of our identification of hypertensive pregnancies for the epidemiological elements of PRECISE.
In addition to routine clinical data, we will focus on two issues, namely time-of-disease assessment and integrating translational biomarkers into risk assessment.
Time-of-disease risk assessment with miniPIERS, PIERS on the Move and phone oximetry: The LMIC-based demographics-, symptom- and sign-based miniPIERS (Pre-eclampsia Integrated Estimate of RiSk) prediction model provides a simple, evidence-based tool to identify pregnant women at increased risk of death or major hypertensive-related complications. The model includes: parity (nulliparous vs parous); gestational age; headache/visual disturbances; chest pain/dyspnoea; vaginal bleeding with abdominal pain; systolic BP; and dipstick proteinuria (AUC ROC 0.77 [95% CI 0.74–0.80]). A predicted probability ≥25% to define a positive test classifies women with 85.5% accuracy. Thus, miniPIERS identifies women at increased risk of adverse maternal outcomes to guide MgSO4 and antihypertensive therapy, or transfer to a higher level of care. Also, we have developed and externally validated the extended fullPIERS model that includes demographics (gestational age), symptoms (chest pain/dyspnoea), signs (SpO2) and laboratory tests (platelet count, serum creatinine and AST) (AUC ROC 0.88 [95% CI 0·84–0·92]). External validation in the LMIC-based miniPIERS cohort confirms good fullPIERS performance in these settings.
The PIERS On the Move (POM) app integrates miniPIERS with a decision algorithm to provide mHealth support to community health workers screening women for pregnancy hypertension and initiating life-saving therapy within the community before transfer to referral centres for definitive care. Also, we have developed an integrated miniPIERS and fullPIERS app, Kenek PIERS®. SpO2 is a significant independent predictor of risk in women with pregnancy hypertension, and addition of pulse oximetry improves miniPIERS model. The POM app integrates phone oximetry; the Kenek Edge pulse oximeter® plugs into the audio port on a smart device to measure, record and export heart rate and SpO2. A miniPIERS data set-based fetal time-of-disease risk model, for use ≥32+0wks includes maternal age; symptoms (0, 1 or ≥2); and dipstick proteinuria (AUC ROC 0.75 [95% CI 0.71–0.80]). Also, we have developed the WHO Maternal Morbidity Tool to guide care and to track health systems.
PRECISE provides the opportunity to externally validate the miniPIERS and fullPIERS models in new LMIC settings, using standard epidemiological approaches with which we have experience; standard measures of diagnostic accuracy include stratification capacity, calibration ability, and classification. We will develop and validate a fetal outcome prediction model for women with complex pregnancies building on previous miniPIERS experience.
