Causes and circumstances of maternal death: a secondary analysis of the Community-Level Interventions for Pre-eclampsia (CLIP) trials cohort
Bibliographic record
Abstract
BACKGROUND: Incomplete vital registration systems mean that causes of death during pregnancy and childbirth are poorly understood in low-income and middle-income countries. To inform global efforts to reduce maternal mortality, we compared physician review and computerised analysis of verbal autopsies (interpreting verbal autopsies [InterVA] software), to understand their agreement on maternal cause of death and circumstances of mortality categories (COMCATs) in the Community-Level Interventions for Pre-eclampsia (CLIP) cluster randomised trials. METHODS: The CLIP trials took place in India, Pakistan, and Mozambique, enrolling pregnant women aged 12-49 years between Nov 1, 2014, and Feb 28, 2017. 69 330 pregnant women were enrolled in 44 clusters (36 008 in the 22 intervention clusters and 33 322 in the 22 control clusters). In this secondary analysis of maternal deaths in CLIP, we included women who died in any of the 22 intervention clusters or 22 control clusters. Trained staff administered the WHO 2012 verbal autopsy after maternal deaths. Two physicians (and a third for consensus, if needed) reviewed trial surveillance data and verbal autopsies, and, in intervention clusters, community health worker-led visit data. They determined cause of death according to the WHO International Classification of Diseases-Maternal Mortality (ICD-MM). Verbal autopsies were also analysed by InterVA computer models (versions 4 and 5) to generate cause of death. COMCAT analysis was provided by InterVA-5 and, in India, by physician review of Maternal Newborn Health Registry data. Causes of death and COMCATs assigned by physician review, Inter-VA-4, and InterVA-5 were compared, with agreement assessed with Cohen's κ coefficient. FINDINGS: Of 61 988 pregnancies with successful follow-up in the CLIP trials, 143 maternal deaths were reported (16 deaths in India, 105 in Pakistan, and 22 in Mozambique). The maternal death rate was 231 (95% CI 193-268) per 100 000 identified pregnancies. Most deaths were attributed to direct maternal causes (rather than indirect or undetermined causes as per ICD-MM classification), with fair to good agreement between physician review and InterVA-4 (κ=0·56 [95% CI 0·43-0·66]) or InterVA-5 (κ=0·44 [0·30-0·57]), and InterVA-4 and InterVA-5 (κ=0·72 [0·60-0·84]). The top three causes of death were the same by physician review, InterVA-4, and InterVA-5 (ICD-MM categories obstetric haemorrhage, non-obstetric complications, and hypertensive disorders); however, attribution of individual patient deaths to obstetric haemorrhage varied more between methods (physician review, 38 [27%] deaths; InterVA-4, 69 [48%] deaths; and InterVA-5, 82 [57%] deaths), than did attribution to non-obstetric causes (physician review, 39 [27%] deaths; InterVA-4, 37 [26%] deaths; and InterVA-5, 28 [20%] deaths) or hypertensive disorders (physician review, 23 [16%] deaths; InterVA-4, 25 [17%] deaths; and InterVA-5, 24 [17%] deaths). Agreement for all nine ICD-MM categories was fair for physician review versus InterVA-4 (κ=0·48 [0·38-0·58]), poor for physician review versus InterVA-5 (κ=0·36 [0·27-0·46]), and good for InterVA-4 versus InterVA-5 (κ=0·69 [0·59-0·79]). The most commonly assigned COMCATs by InterVA-5 were emergencies (68 [48%] of 143 deaths) and health systems (62 [43%] deaths), and by physician review (India only) were health systems (seven [44%] of 16 deaths) and inevitability (five [31%] deaths); agreement between InterVA-5 and physician review (India data only) was poor (κ=0·04 [0·00-0·15]). INTERPRETATION: Our findings indicate that InterVA-5 is less accurate than InterVA-4 at ascertaining causes and circumstances of maternal death, when compared with physician review. Our results suggest a need to improve the next iteration of InterVA, and for researchers and clinicians to preferentially use InterVA-4 when recording maternal deaths. FUNDING: University of British Columbia (grantee of the Bill & Melinda Gates Foundation).
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".