This is a planned secondary analysis of the CRADLE 3 trial; a pragmatic, step-wedge, cluster- randomised control trial that evaluated the CRADLE VSA intervention (CVSA device and training package) in low resource settings.
The CRADLE 3 intervention consisted of implementing the CRADLE VSA device and associated training package in routine community and hospital maternity care in low-resource settings. Prior to implementation of the intervention package, management was based on local guidelines and assessment of patients used varying medical devices, and this was used as a control. The trial was carried out across ten clusters, over eight countries, including Addis Ababa (Ethiopia), Cap Haitien (Haiti), Freetown (Sierra Leone), Harare (Zimbabwe), Gokak (India), Kampala and Mbale (Uganda), Lusaka and Ndola (Zambia), and Zomba and the Southern Region (Malawi). Each cluster included at least one urban or peri-urban secondary or tertiary facility and multiple peripheral hospitals that referred to the region’s central hospital. In total 286 facilities and 536,223 deliveries were included in the CRADLE intervention between April 1 2016 and November 30 2017.
Clusters crossed over from the control to the cradle intervention at a randomly allocated timepoint, at 2 monthly intervals. At this randomly selected timepoint, all existing devices were replaced with the CRADLE VSA and health care providers at the facility were provided with access to the device and training package. Prior to intervention, management was based on local guidelines and assessment of patients used varying medical devices, and this was used as a control. Ethics approval was granted by the King's College London (UK) Research Ethics Subcommittee (LRS-14/15-1484) and in all countries before the start of the trial. Institutional-level consent on behalf of the cluster was obtained. In total 3868 devices were delivered to 286 facilities.
All women identified as pregnant or up to 42 days postnatal presenting to the facility were eligible to be exposed to the intervention. There were no exclusion criteria.
Randomisation: The randomisation was the cluster. A computer-generated randomly allocated sequence run by the CRADLE statistician determined the order in which the clusters received the intervention. All clusters were masked to the order of implementation until 2 months before the intervention. Because of the nature of the intervention the trial was not masked.
At each randomly allocated date, the training package was delivered to health workers in each facility by interactive group sessions. Existing equipment for vital signs observations was replaced with the CRADLE device unless specific function were required (e.g. cyclical BP monitoring in HDU).
The proportion of women referred from periphery facilities to higher-level care was collected from a 4-week period before and another 4-week period repeated 3 months post-implementation. Referrals were either counted from referral registers and compared to number of patients seen in antenatal clinic or admitted to ward, or data was documented prospectively as patients were referred. It was not possible to collect accurate denominator referral data in one large site with multiple referral areas (Kampala) therefore this area was not included.
Maternity staffing levels and access to essential treatment (intensive care beds, capacity for blood transfusion) were also assessed at each facility and recorded throughout the trial period. Major changes to infrastructure, patient payment requirements, or environmental conditions were systematically evaluated each month in each site. The trial ended after 20 months as planned. Reason for referral was documented as infection, bleeding, high blood pressure, labour or other (which included anaemia, malaria, early pregnancy complications).
The primary outcome of the CRADLE-3 trial was a composite of at least one of eclampsia, hysterectomy or maternal death. In this analysis we reviewed the number of patients referred from primary to secondary or tertiary higher-level care for bleeding pre- and post- intervention. We report outcomes from the CRADLE-3 trial related to haemorrhage i.e. death from obstetric haemorrhage or emergency hysterectomy due to obstetric haemorrhage.
We evaluated the effect of implementation on referrals for bleeding through a planned secondary analysis of the CRADLE-3 data. Odds ratios were calculated for each centre, comparing event rates pre- and post- CRADLE intervention. As there was considerable heterogeneity, random effects meta-analysis was used throughout.
For evaluation of outcome the CRADLE-3 trial reports the bent stick analysis. This achieves great stability to the trend-and-step pattern originally proposed in the CRADLE 3 trial, because it allows for separate linear trends in each cluster before and after intervention.
Statistical analyses used Stata, version 14.2 (by PTS). This study is registered with the ISRCTN registry, number ISRCTN41244132.