Scaling Up an Integrated Approach to Improve Delivery Care in North Guatemala With Stepped Wedge Design

Overview

"¡Que Vivan las Madres!: Venga a tener su parto al CAP" (QVLM) is a guatemalan quasi-experimental study that has been performed from January 2014 to January 2017 by the Epidemiological Research Center in Sexual and Reproductive Health (CIESAR) in Guatemala in coordination with PRONTO International and University of San Francisco, California. This project has been financed by Grands Challenges Canada' "Save Lives at Birth, A Grand Challenge for Development" partnership that includes USAID, Norwegian ministry of foreign affairs, Bill\&Melinda Gates foundation, UKaid. This project has applied a stepped wedge design (SWD) over 6 zones or clusters. Each one of the zones contains from 4 to 6 communities, each one with the presence of one second level health facility (known in Spanish as CAP, Centro de Atención Permanente). These health centers are the next level in attention after home, traditional and empirical attention. Communities around the selected health centers are mostly rural and have the worst maternal health indicators in the country. These health centers are expected to have enough equipment and personnel to attend the deliveries that occur in their communities. This study was performed in Huehuetenango and Alta Verapaz districts in north Guatemala. Each one with 3 zones for a total of 6 zones. The study follows a Stepped Wedge Design, in which all 6 zones are eventually intervened, but at different regular periods of time (each period is 4 months long). This project applies a package of 3 simultaneous interventions in each zone with the purpose of increasing institutional deliveries and improving deliveries attention in public health centers. This intervention plan has been implemented in a pilot study reported in (Kestler et. al, 2013).

Individual delivery data has been collected in site from health centers records. Thus, the study does not depend on official data from the ministry of health. Not only is MOH data hard to obtain, it also may not contain the details we are interested in that are put by hand by doctors and nurses in clinic histories. In order to get uniform data that fits the study needs, a parallel monitoring process has been set up. Data for mortality and morbidity events that occured in communities during the study is limited. We don't have the community counterpart of our indicators. Government data of perinatal deaths in community may be obtained with 1 or 2 years of delay, however there may be an underreporting of such events. Likewise, there is no data source for maternal morbidity in community deliveries since community practice is usually not documented nor reported. The concept of perinatal mortality used in this study is not the standard concept that includes death from 28th gestational week to 7 days after birth. The perinatal mortality outcome in this study takes into account only deaths that occur during care in health centers (approximately 48 hours since birth). This is intentional since the aim of this study is to improve clinics' skilled delivery care, thereferore it is concerned on reducing the newborn deaths that occur during care given in site. Along this study, the focus is in these subset of perinatal deaths. The APGAR measure 1 minute and 5 minutes after birth, and the reported procedures for the newborn are used to identify cases of perinatal morbidity. This study is targetted to the low level health centers in Huehuetenango and Alta Verapaz. This includes 33 health centers. This is a fixed number of health centers and there is no control of the patients that got treated at the health centers. This is the reason why there is no sample size assessment. The selection of the health centers was limited by the definition of CAP (second level health facilities according to local definition) and the amount of patients that received the intervention by attending their deliveries in health centers can not be controlled. Having data on all vaginal and uncomplicated deliveries that occur in health centers, the analysis can be done by individual cases and also by aggregating data in time periods. Also, analysis methodology of trials with stepped wedge design has been developed in the last years. A generalized mixed model for a multilevel and longitudinal analysis is going to be used in order to account for different sources of variability within and between health centers and times. So far (April 2017) we have done preliminary analysis this way and are confortable with this methodology selection.