October 21, 2026 interva.net Editorial

One Nurse in Eighty-Four Got an Answer Back

A survey of 143 primary-care nurses in Campo Grande, Brazil, describes the home interview behind maternal death surveillance, and how rarely anything returns to the people who conduct it.

maternal mortalityverbal autopsysurveillancedata quality

In Campo Grande, a city in the middle of Brazil, a maternal death does not end with a piece of paper. Brazilian rules require that the death of any woman of reproductive age be investigated, and in the municipal system the person who usually does the investigating is a nurse from the local basic health unit. She goes to the house and works through the home-interview version of the maternal death investigation form, the one the Ministry of Health labels M3, asking about the pregnancy, the delivery, the care that was sought and the care that could not be found.

That conversation is the field end of what the literature splits into verbal and social autopsy: a family’s account standing in for clinical evidence. Readers of our explainer on the method will recognise the instrument. What is described far less often is who carries it to the door.

What 143 nurses said about the work

Between December 2019 and March 2021 a team from the Federal University of Mato Grosso do Sul put a 34-question instrument to the municipal primary-care nursing staff: 143 respondents of 204 eligible, a response rate of 70.1%, across 71 units in all seven health regions. The results appeared in the Revista Latino-Americana de Enfermagem on 16 October.

Just under three in five, 58.7% or 84 nurses, had ever investigated a maternal death. Word of a death reached them mostly through community health agents (41.5%), then health-surveillance technicians (29.3%), then the family itself (22%). First contact was usually a home visit (52.5%), usually with a second professional present. Two answers describe the conditions: 46.4% said the health-surveillance team gave them no support, and 30.5% said families refused the interview.

Only one variable held up in the regression as a predictor of performance, specific training in maternal death surveillance, with an odds ratio of 5.82 (95% CI 2.46-13.75). The model itself was modest, reaching an AUC of 0.672 and 63.6% accuracy, and it sorted nurses who had not investigated a death (86.2% specificity) far better than those who had (48.2% sensitivity).

The number that is not about training

Of the 84 nurses who had investigated a death, exactly one, 1.2%, reported receiving any recommendation from the municipal committee for the prevention of maternal, child and fetal mortality once a case had been analysed. A third said maternal deaths were discussed in team meetings.

So the sequence ran: the interview happened, the form was filled, the file went up. The committee is not a paper body, having existed in the municipality since 2001, but its analysis, on this account, seldom reached the nurse who had collected the material.

A loop that has been described before

In 2023 the Bulletin of the World Health Organization published a realist synthesis of 59 qualitative studies of people running maternal and perinatal death surveillance and response in low- and middle-income countries.

Where the action cycle worked, learning, vigilance and implemented recommendations fed each other; the enabling condition was a blame-free environment with good leadership. Where it did not, under-reporting and inaccurate data produced thin review and weak recommendations, which produced demotivation and disengagement, and then more under-reporting.

Read against that, the Campo Grande figures are not a story about nurses who cannot be bothered. They describe a loop that lost its return leg. Bangladesh’s national programme, described here in a review of its first decade, shows the other arrangement: every notification paired with a scheduled social autopsy and a committee required to produce an action.

Why the visit is worth the trouble

The record is otherwise wrong. A 2023 meta-analysis of 20 studies found maternal death recording incomplete by 34% (95% CI 28-48) and maternal mortality under-estimated by 39% (30-48). Deaths at the extremes of reproductive age, deaths not classified by a medical examiner or coroner, deaths from indirect causes and deaths at home were the most likely to be missed. The studies it pooled found deaths through record linkage and established cause through medical records and verbal autopsy.

Brazil has its own demonstration. The Campo Grande authors note that when officials investigated 88% of the roughly 64,000 deaths of women of reproductive age recorded in 2019, the count of officially recognised maternal deaths rose by 32.7%, citing a Ministry of Health epidemiological bulletin.

Worldwide, WHO’s analysis of deaths from 2009-20 puts haemorrhage at 27%, indirect obstetric causes at 23% and hypertensive disorders at 16%, with hypertensive deaths proportionally highest in Latin America and the Caribbean. That is the picture behind an estimated 260,000 maternal deaths in 2023 and a global ratio of 197 per 100,000 live births.

What this evidence is, and is not

A descriptive-analytical cross-sectional survey in one state capital, with a convenience sample, a self-administered questionnaire and roughly 30% non-response. Collection ran almost entirely through the COVID-19 period, when home visits and collective activities in primary care were curtailed, and the authors warn that the most overloaded nurses were probably the least likely to answer, so the picture may be rosier than the reality.

Nothing here tests whether an investigation reached the right cause of death. No reference standard, no comparison against medical records or a panel, and therefore no accuracy estimate for the interview as a cause-of-death instrument. What it measures is process: who does the work, with what training, with whose backing, and what comes back.

The authors conclude with a plea for permanent education. The finding that travels further is smaller and harder to fix. A system that asks a nurse to sit with a grieving family, and then tells her nothing about what was learned, is asking her to keep doing it on trust.

Sources

← Back to the news digest