Evaluation of a novel device for the management of high blood pressure and shock in pregnancy in low-resource settings: Study protocol for a stepped-wedge cluster-randomised controlled trial (CRADLE-3 trial)

Background: Obstetric haemorrhage, sepsis and pregnancy hypertension account for more than 50% of maternal deaths worldwide. Early detection and effective management of these conditions relies on vital signs. The Microlife® CRADLE Vital Sign Alert (VSA) is an easy-to-use, accurate device that measures blood pressure and pulse. It incorporates a traffic-light early warning system that alerts all levels of healthcare provider to the need for escalation of care in women with obstetric haemorrhage, sepsis or pregnancy hypertension, thereby aiding early recognition of haemodynamic instability and preventing maternal mortality and morbidity. The aim of the trial was to determine whether implementation of the CRADLE intervention (the Microlife® CRADLE VSA device and CRADLE training package) into routine maternity care in place of existing equipment will reduce a composite outcome of maternal mortality and morbidity in low- and middle-income country populations. Methods: The CRADLE-3 trial was a stepped-wedge cluster-randomised controlled trial of the CRADLE intervention compared to routine maternity care. Each cluster crossed from routine maternity care to the intervention at 2-monthly intervals over the course of 20 months (April 2016 to November 2017). All women identified as pregnant or within 6 weeks postpartum, presenting for maternity care in cluster catchment areas were eligible to participate. Primary outcome data (composite of maternal death, eclampsia and emergency hysterectomy per 10,000 deliveries) were collected at 10 clusters (Gokak, Belgaum, India; Harare, Zimbabwe; Ndola, Zambia; Lusaka, Zambia; Free Town, Sierra Leone; Mbale, Uganda; Kampala, Uganda; Cap Haitien, Haiti; South West, Malawi; Addis Ababa, Ethiopia). This trial was informed by the Medical Research Council guidance for complex interventions. A process evaluation was undertaken to evaluate implementation in each site and a cost-effectiveness evaluation will be undertaken. Discussion: All aspects of this protocol have been evaluated in a feasibility study, with subsequent optimisation of the intervention. This trial will demonstrate the potential impact of the CRADLE intervention on reducing maternal mortality and morbidity in low-resource settings. It is anticipated that the relatively low cost of the intervention and ease of integration into existing health systems will be of significant interest to local, national and international health policy-makers. Trial registration: ISCRTN41244132. Registered on 2 February 2016. Prospective protocol modifications have been recorded and were communicated to the Ethics Committees and Trials Committees. The adapted Standard Protocol Items: Recommendations for Interventional Trials (SPIRIT) Checklist and the SPIRIT Checklist are attached as Additional file 1. The trial was preceded by a mixed-methods feasibility phase. The aim was to explore the acceptability and feasibility of the CRADLE intervention and its implementation strategies in three non-trial sites representative of the 10 main trial clusters (Ramdurg, Belgaum, India, Bishoftu, Ethiopia, Masvingo, Zimbabwe). Simultaneously, the 10 main trial clusters collected primary outcome data to evaluate the methods of data collection, and to inform the randomisation programme and the sample size calculation. Results were used to optimise the final CRADLE-3 protocol including training materials and implementation strategy for the main trial. The primary aim of the trial was to determine whether implementation of the CRADLE intervention to community and facility maternity care reduces a composite of (all-cause) maternal mortality or major morbidity by ≥ 25%. The trial is complemented by simultaneous process evaluation informed by the Medical Research Council (MRC) guidance. The aim of the process evaluation is: (1) to explore if the CRADLE trial was delivered as intended (fidelity, dose and adaptation); (2) to understand whether, how and why the intervention had an impact, through exploring healthcare provider (HCP) perspectives of their usual care and of the intervention and (3) to explore if the results are likely to be scalable and sustainable. This will include an evaluation of cost-effectiveness. The CRADLE-3 trial was a pragmatic, mixed-methods, multicentre, stepped-wedge cluster-randomised controlled trial of the introduction of the CRADLE intervention (CRADLE VSA device and CRADLE training package) to routine maternity care settings in LMICs. It was informed by the MRC guidance for evaluation of complex interventions [11]. Ten clusters were identified to take part in the trial. Each cluster comprised a secondary or tertiary health facility with multiple satellite primary care centres that referred to the central hospital: Each randomisation cluster crossed over from control to the CRADLE intervention at 2-monthly intervals (Figs. ​(Figs.11 and ​and22 and Additional file 1). At the time each cluster was randomised to receive the CRADLE intervention, all levels of healthcare providers within that cluster, involved in the care of pregnant/postpartum women, had access to the CRADLE intervention. The intervention effect will be determined by comparing data points in the intervention section of the wedge with those in the controlled section. Stepped-wedge cluster-randomised controlled trial design. A schematic representation of the trial design Standard Protocol Items: Recommendations for Interventional Trials (SPIRIT) Figure Women identified as pregnant or within the 6-week postpartum period, presenting for antenatal, intrapartum or postpartum care in cluster catchment areas* within the trial time frame. *Catchment areas were defined in collaboration with local principal investigators, and included the major outreach facilities that result in women being assessed and referred to a defined central facility/ies. These were defined prior to randomisation and remained constant throughout the study period. There were no exclusion criteria. The CRADLE intervention consisted of two components and is described according to the TIDieR Checklist [12]. The Microlife® CRADLE VSA is a hand-held, upper-arm, semi-automated device that measures blood pressure and pulse. The device has undergone extensive testing for accuracy and is one of few blood pressure devices to have been validated as accurate in pregnancy, including in pre-eclampsia and hypotension [10], as well as non-pregnant adults [13]. The device incorporates a traffic-light early warning system that alerts all levels of HCP to abnormalities in blood pressure and pulse secondary to obstetric haemorrhage, sepsis and pregnancy hypertension. The thresholds that trigger the traffic lights were determined through prediction studies [14, 15]. The traffic light early warning system triggers are shown in Fig. ​Fig.3.3. A ‘red light’ and ‘up arrow’ displayed following vital signs measurement of a pregnant/postpartum woman indicate severe hypertension and should prompt intervention and/or referral. Likewise, a ‘red light’ and ‘down arrow’ with haemorrhage or sepsis should prompt immediate action. A ‘yellow light’ and either ‘down arrow’ or ‘up arrow’ indicate less urgent need for assessment, intervention and/or referral. It is proposed that the traffic lights will alert users to abnormal vital signs, including those without formal healthcare training. This should enable earlier management to improve outcomes that would directly benefit the woman and her unborn/newborn child. Vital Signs Alert (VSA) device traffic-light early warning system display options. Legend: SBP systolic blood pressure; DBP diastolic blood pressure; SI Shock Index The production costs are less than US$20 per unit and has been designed to be simple to use, even by unskilled personnel after minimal training. The device fulfils the World Health Organisation requirements for automated devices used in low-resource settings [13]. Other developments suited to LMICs include a micro-USB charging ability. The CRADLE VSA is manufactured in Taiwan. As part of the CRADLE-3 trial intervention, the CRADLE VSA was incorporated into routine maternity care. Primary, secondary and tertiary facilities were allocated devices according to their delivery rate, staffing numbers and number of beds per ward. Pre-existing blood pressure measurement devices were removed from clinical areas unless existing equipment had functionality designed for that area, e.g. repeated automated measures in an operating theatre or a high dependency area, and this was left to the discretion of the lead clinician. The intervention users included every cadre of healthcare professional involved in maternity care within the cluster. This included community healthcare providers (cHCPs) where they were supported at district level and involved in provision of routine maternity care. The CRADLE research group created a simple CRADLE training package for prospective CRADLE VSA users. The training package consisted of short animated films, an interactive session, action prompt cards attached to the CRADLE VSA and posters. There were two sets of training materials available, one for facility HCP (fHCPs) and one for cHCP with very limited resources or no formal training. All materials were translated into local language where required. The CRADLE package content covered: At the randomised time point the local implementation team (clinical research officers responsible for ongoing CRADLE outcome data collection and site principal investigators) attended face-to-face one-off training with the research team lasting approximately 5 h. The implementation team and research team subsequently delivered one-off group training sessions lasting 2–4 h to local stakeholders and representative HCP from each of the clinical areas in the cluster. Attendees were given training materials and CRADLE VSA to disseminate to their clinical areas. The implementation team continued to visit clinical areas regularly to collect outcome data therefore providing ongoing support to HCP. The core components of the intervention (the CRADLE VSA, animated films, posters and content of the training presentation) were standardised across all clusters. Delivery of the core components could be adapted to meet the needs of the cluster. This intervention was compared to routine maternity care. This involved blood pressure monitoring with a variety of blood pressure devices that were available locally and management according to local guidelines. The rate of a composite of maternal mortality or major morbidity (one of maternal death, eclampsia or emergency hysterectomy with no double counting per 10,000 deliveries in each cluster each month). We will report the effect of the intervention on the primary endpoint, on each of the three components, and on each of the secondary endpoints. Maternal death was defined as death during pregnancy or within 42 days of delivery (or last contact day if contact not maintained to 42 days). Eclampsia was defined as occurrence of generalised convulsions with increased blood pressure during pregnancy, labour or within 42 days of delivery in the absence of epilepsy or another condition predisposing to convulsions. Emergency hysterectomy was defined as surgical removal of all or part of the uterus. Maternal deaths from all causes were collected with additional information regarding the cause of maternal death collected to determine the potential for impact of the CRADLE VSA. The percentage of deaths that occurred as a result of obstetric haemorrhage, pregnancy-related sepsis and hypertensive disorders of pregnancy will be presented across all clusters pre and postintervention with adjusted risk ratios (RRs). If there are other large groups of other clinically important causes of death, e.g. early pregnancy complications, these will also be defined. Maternal secondary outcome measures: Number of stillbirths per 10,000 deliveries per month. Number of neonatal deaths per 10,000 deliveries per month. The implementation and impact of the intervention in each site was evaluated by both quantitative and qualitative measures. The fidelity, dose, reach and adaptation (whether the intervention is implemented as planned, to the intended population and what amount) was determined by measuring: The potential impact of the intervention, the acceptability and potential mechanisms of action were explored through the following methods: The following outcomes were evaluated to explore the potential for the results to be scalable and sustainable: Quantitative process data will be integrated into the outcome datasets to examine whether the effects of the CRADLE intervention differ by implementation. Qualitative methods will capture emerging changes in implementation, experiences of the intervention and unanticipated or complex causal pathways in addition to generating new theory about potential mechanisms of action. Quantitative and qualitative analyses will build upon one another, with qualitative data used to explain quantitative findings, and quantitative data used to test hypotheses generated by qualitative data. The cost consequence analysis will be modelled on the basis of the equipment required and the health outcomes influenced as recorded in the trial. The incremental financial and economic cost of implementing the CRADLE intervention (device and training package) over the trial will be quantified, thereby evaluating the financial sustainability of the interventions. Research costs will be excluded from the costs of the intervention. Each cluster was described according to the number of primary, secondary and tertiary facilities and their referral distances. Details on staffing levels and availability of key resources, such as intensive care beds, blood transfusion and magnesium sulphate use, was collated. In addition, major changes to the trial catchment, such as changes to infrastructure, policy, patient payment requirements or environmental conditions, were systematically reviewed each month. Methods of data collection were discussed and optimised based on the existing resources available in each site. Outcomes were triangulated across multiple sources (including referral registers, ward registers, patient records, local mortality and morbidity records and active case finding) to ensure data completeness and all outcomes checked to avoid double counting. Outcome data were recorded over the 20-month period. Consistency and quality of source data was monitored by the research midwife/assistant at each cluster (monthly). Data were entered onto an online database (MedSciNet). This allows for extensive monitoring and query processing features, as well as a comprehensive alerting system to identify missing data. The trial coordinator will monitor data entry continuously on MedSciNet. Ten percent of the source data were validated by the primary investigator and a proportion of this was reviewed by the trial coordinator. Patient records are identified with a unique identification number generated sequentially and no identifying data are stored. The unit of randomisation is the cluster (or clusters), rather than individual women. Large variation in the primary event rate was anticipated between clusters; therefore, a restricted method of randomisation was used with zero rank correlation between events per month and order of randomisation. The sequence of timing for receiving the CRADLE intervention was determined by computer-generated random numbers by the CRADLE statistician. The clusters were masked to the order of implementation until they are informed of their allocation 2 months prior to their implementation date (to enable training to be arranged). Sample size estimation has been carried out by the CRADLE statistician, using Stata version 13.1 and the methods of Hemming and Girling [16]. This was informed by data from the feasibility phase. For the purpose of the power calculation an assumption that there are at least 4000 deliveries per centre per month was made (or 8000 per cluster period of 2 months) and at least nine clusters, each observed for 20 months (10 time periods of 2 months each). We anticipated a baseline event rate of 1% and anticipate a 25% reduction in this to 0.75% post intervention. We would require a total of 6300 outcome events, with a coefficient of variation of 0.4 and an Intra Cluster Correlation of 0.0085, to have power of 99%. We would require a total of 2450 outcome events with a coefficient of variation of 0.4, to have power of 95%. The effect of the intervention on the primary outcome, on each of the three components, and on the secondary outcomes specified above, will be reported. Results will be reported firstly as odds ratios (ORs), with risk ratios (RRs) as a secondary comparison if the appropriate models converge. Within the stepped-wedge cluster design it remains appropriate to analyse outcome data from clusters individually, despite randomising cluster time points with some clusters paired [17]. The main statistical analysis method will be by logistic regression, using generalised estimating equations with a population-averaged model [18] (Stata command xtgee) with fixed centre effects and separate fixed linear trends in each centre for changes in outcome over time. Autoregressive correlation will account for decreasing correlations between observations over time. This model outperformed the multilevel model structure of Hussey and Hughes [19], being apparently less susceptible to bias. It gives equal weight to each woman, rather than each centre, and (being population-averaged) reports effects averaged over the length of the trial. Simulations studies suggest that using time categories, rather than linear trends do not correct well for separate time trends by centre, and can cause convergence problems. Results will be expressed as ORs; at the low event rates expected (≤ 5%), ORs and RRs are reasonably close (in the simplest case, for 5% and 3.75%, RR = 0.75, OR = 0.74). A secondary analysis to obtain RRs using a log link will be attempted, but simulation studies suggest that the convergence may be poor. We will adjust for centre effects and linear time trends and the interaction between them. This will effectively adjust for differences in baseline availability of resources between clusters.