The purpose of this study is to conduct Verbal Autopsies of deaths ( stillbirths and neonatal deaths together) identified in the BetterBirth trial to identify their potential causes, timing, and social determinants.

In order to be included in this study, one must be the mother or family member of an individual who was enrolled in the BetterBirth Trial and died. Women who did not consent to follow-up in the BetterBirth study will be excluded.

We identified these families by using data from the BetterBirth Trial, in which there was a perinatal death where one or more pause points was observed by BetterBirth staff (FADA) at 2 months or 6 months post-intervention launch. Our analyses focus on data surrounding the perinatal deaths.

Respondents for the neonatal deaths and stillbirths include mother or the family members of the baby. Mothers had consented to follow-up as part of the BetterBirth Trial.

These individuals will be initially contacted by telephone. Individuals will be asked to participate in verbal autopsy, and, if they agree, a date and time for the interview will be determined. A data collector from CEL or PSI will visit the participant's home on an agreed-upon date and written consent will be obtained. In the event that participants cannot be reached by phone, study staff members will attempt to make contact via Accredited Social Health Activists (ASHAs), who are appointed by the National Rural Health Mission and act as conduits between patients and the health system. The data collector will attempt to call the relevant ASHA worker and request her assistance in establishing contact with the mother. The ASHA will be asked to provide a telephone number at which the mother can be reached, if possible.

Final attempts at contact will be made via home visit if participants cannot be reached either via telephone or with ASHA engagement. Individuals who are reached at home will be introduced to the study and asked to participate using the appropriate recruitment script. For those individuals who agree to participate, data collectors will schedule a time to return to participants' homes for interviews to be conducted. (Note: Home visits are a culturally acceptable practice in this context; for example, home visits were used in the BetterBirth Trial for collection of 7-day outcomes of mothers and newborns.)

Data will be collected using the WHO standard tools for conducting stillbirth or neonatal verbal autopsies, modified for the local context. A section on social autopsy will also be incorporated in the existing study instrument. These tools are standardized and adapted to use in these settings and have been implemented before in other studies in similar communities in U.P. (example: "AMANHI" and "Impact of topical application of cold-pressed sunflower seed oil with improved massage practices on neonatal mortality: a cluster randomized controlled trial in rural North India"). A team of specially trained data collectors (given the sensitive nature of the verbal autopsy questionnaire) will collect the data on existing Android based data collection platform.

Data will be collected by specially trained data collectors. They will collect the data at participants' homes. Participants will include the mother (in case of neonatal death) and any additional person who was present during the events that led to the death. Additional interviewees may include the delivered mother's husband, mother, mother-in-law, other family member, neighbor, etc. The delivered mother will be asked which individuals, if any, should be included in the interview. The mother and the individuals they indicate will be consented individually but interviewed together.

Each interview will take approximately 60 minutes.

A team of specially trained data collectors will be hired. Data will be collected on tablets using a data collection platform that has in-built checks of missing values, range checks, skip patterns etc. to minimize the error in data collection. A GPS monitoring of location of interview will be conducted. The data will be checked at an aggregate level for heaping, interviewer-specific patterns, etc. Forms identified with errors will be sent back for verification and re-entry, but audit-trail will be captured to ensure that the full trail from original data and changes will be maintained with timestamp, GPS and user information. On-site quality assurance would involve both real-time data checks as well as GPS verification of the location of data collection. Data access will be restricted by users, with different privileges for data collectors, supervisors and study investigators. All personal identification information including names, etc. will be encrypted, and only a unique identification number to identify individuals will be made available for analysis. Data will be hosted over a secured network.

Every case will be coded for assigning underlying cause of death, antecedent causes and timing of death by trained and certified physicians. A WHO standard software and coding protocol will be used by these physicians to assign causes of death. The data collected through tablets will be sent to the central server. The narratives of every case will be collected on paper and a scan of the narrative will be sent along with the data. This entire data will be uploaded on the physician coding software from which the verbal autopsy cases will be allotted to physicians for coding. Each case will be independently coded by two physicians and in case of mismatch in the underlying cause of death between the two physicians, a third physician will independently code for the cause of death. Finally, if none of the three underlying causes of death match, a higher level physician arbitrator will resolve the case - this will be done to minimize unclassified deaths. A physician coordinator, who has extensive prior experience of coding verbal autopsy data will oversee the entire process and resolve issues with the help of supervisors of the data collection team. All the cases will be de-identified before uploading on the software for coding.

The verbal autopsy findings will be helpful to understand the result of the BetterBirth Trial. In addition, the findings will inform future investigations of maternal and newborn harm, especially those further upstream in health system and policy design.

Inclusion Criteria:

• Mother, family members, and community members of deceased babies who enrolled in the BetterBirth Trial

Exclusion Criteria:

• Mothers and family members members who did not consent to follow-up in the BetterBirth Trial

Respondents for the neonatal deaths and stillbirths include mother, family members, or community members of the baby. Mothers had consented to follow-up as part of the BetterBirth Trial.

A de-identified data repository will hold the verbal autopsy data collected, including all data fields collected. At the end of the study, de-identified data will be shared in a public domain.

Participant who consented to be interviewed (verbal autopsy) about a perinatal death where one or more pause points was observed by BetterBirth staff (FADA) at 2 months or 6 months post-intervention launch

Adverse event data were collected over the study period (July 3, 2017- August 27, 2017).

There were no adverse events among the individuals who were interviewed for verbal autopsy.

Please note our outcomes represent cases of perinatal mortality. These cases were expected as our sample was based on perinatal mortality cases from the BetterBirth trial; we performed verbal autopsies on all cases of perinatal mortality in the trial.