Bruno Masquelier

Background: Excess mortality estimates depend on an expected baseline, an unobservable quantity typically derived from models with prescribed functional forms and short reference periods that exclude exceptional events. Singular Spectrum Analysis (SSA) offers a nonparametric alternative but requires subjective component selection and is sensitive to outliers. We combined SSA with automated component selection and outlier interpolation to estimate long-term baselines and compile a catalog of excess mortality events (EMEs). Methods: We analyzed weekly all-cause mortality (2015–2024) for six metropolitan areas with contrasting climatic, demographic, and epidemiological contexts: Brussels, Antwerp, and Liège (Belgium), and Athens, Thessaloniki, and Larisa (Greece). SSA was applied to standardized log-mortality rates; components were selected using Monte Carlo SSA, and COVID-19 periods were interpolated iteratively. EMEs were defined as weeks exceeding a quasi-Poisson two-z-score threshold, then screened for coincidence with heat waves, cold waves, COVID-19 periods, and influenza epidemics. Results: Baselines were consistent within countries: annual cycles in Belgium, with an additional significant semiannual cycle in Greece reflecting summer mortality. We cataloged 170 EMEs spanning 456 weeks and 14,210 excess deaths. Pandemic weeks accounted for 77% of excess mortality, and 96 weeks coincided with temperature extremes; 56 EMEs, representing 3.7% of excess mortality, remained unmatched. Conclusions: SSA yields reproducible, low-assumption baselines that capture country-specific seasonality. Because flexible baselines can absorb recurrent seasonal mortality, component selection is a substantive choice that should be reported explicitly, together with its sensitivity.

Objectives: We evaluate the utility of mobile phone surveys (MPS) for estimating child mortality in Burkina Faso, where armed conflict and forced displacements hamper traditional face‐to‐face data collection on an important share of the territory. Methods: We estimated under‐five mortality from truncated birth histories collected in an MPS from September 2021 to October 2022. The sample consisted of telephone numbers collected in a prior face‐to‐face survey and an additional set of numbers generated through random digit dialling (RDD). We compared regionally disaggregated mortality estimates from the MPS with those from the 2021 Demographic and Health Survey (DHS), which could not cover parts of the national territory due to insecurity. Results: Of 52,650 attempted phone numbers, 11,387 women of reproductive age were interviewed. Of these, 8077 were reached through RDD (36% lived in conflict‐affected areas at the time of data collection) and 3310 via previously collected numbers (42% lived in conflict‐affected areas). In regions not directly affected by conflict, under‐five mortality rates have remained relatively stable over time: from 45 deaths per 1000 live births in 2018–2019 (95% CI: 36–54) to 58 in 2020–2022 (95% CI: 46–69). This finding mirrors the trend observed in the DHS data. In regions affected by armed conflict, however, the MPS data indicate a substantial increase in under‐five mortality: From 47 deaths per 1000 live births in 2018–2019 (95% CI: 38–57) to 82 in 2020–2022 (95% CI: 67–98). This mortality reversal in conflict‐affected areas is not apparent in the DHS, most likely due to the exclusion of the most insecure survey clusters from its sampling frame. Conclusion: MPS offers a viable alternative or complement to traditional face‐to‐face surveys for estimating child mortality in populations where insecurity limits face‐to‐face survey data collection.

Under-five mortality estimates for low- and middle-income countries are primarily derived from detailed birth or pregnancy histories collected through in-person household surveys. Such surveys are, however, resource intensive and vulnerable to interruption during epidemic outbreaks and other crises. Remotely deployed mobile phone surveys can circumvent these disadvantages, but their suitability for measuring population-level mortality has not been demonstrated. In this contribution, we examine Malawian mobile phone survey data from the Summary Birth Histories, Truncated Pregnancy Histories, and Full Pregnancy Histories instruments for estimating under-five mortality. Considering the limited penetration and the unequal distribution of mobile phones in Malawi, quota sampling was used to ensure representation of population subgroups where mobile phone ownership is low, and poststratification methods were applied to further attenuate selection bias. Resulting probabilities of dying, or q(x)—before 28 days, 12 months, and 60 months of life—are compared against external estimates from a recent Demographic and Health Survey, a Multiple Indicator Cluster Survey, and model-based estimates from the UN Inter-agency Group of Child Mortality Estimation. Mobile phone survey estimates using the Summary Birth Histories capture the historical trends of q(12m) and q(60m) up to 2018, but they are less reliable for the most recent years. Compared with external sources, estimates from the Truncated Pregnancy Histories appear to be biased downward. Estimates of q(28d), q(12m), and q(60m) from the Full Pregnancy Histories are in line with those published by the UN Inter-agency Group, but they are also suggestive of a mortality excess during the COVID-19 outbreak in 2020‒2022. We conclude that mobile phone surveys are a promising method for collecting under-five mortality data, and particularly so via the Full Pregnancy Histories instrument.

Objectives: In many low‐ and middle‐income countries, perinatal mortality estimates are derived retrospectively from periodically conducted household surveys. Mobile phone surveys offer advantages in terms of cost and ease of implementation. However, their suitability for monitoring perinatal mortality has not been established.Methods: We use data from the Malawi Rapid Mortality Mobile Phone Survey (RaMMPS) to estimate perinatal mortality rates from two versions of the survey instrument: a full pregnancy history and a shorter truncated pregnancy history. Female respondents of reproductive age were randomly allocated to either of these instruments. The sample was generated through random digit dialling with active strata monitoring. Post‐stratification weighting was used to correct for sample selection bias, and estimates are reported with bootstrap confidence intervals. We estimated the stillbirth rate as the synthetic cohort probability of a foetal death with 28+ weeks of gestation over all pregnancies reaching the same gestational age. The perinatal and extended perinatal mortality rates were defined as the probabilities of dying between 28 weeks and 7 or 28 days of life, respectively. RaMMPS estimates are compared to the 2015–2016 Malawi Demographic and Health Survey and estimates published by the United Nations Inter‐agency Group for Child Mortality Estimation.Results: Truncated and full pregnancy histories were administered for 2093 and 2067 women, respectively. Weighted point estimates of the stillbirth (19.81 deaths per 1000 pregnancies, 95%‐confidence interval (CI): 14.11–25.62), perinatal (42.41, 95%‐CI: 33.91–50.92), and extended perinatal mortality rates (50.11, 95%‐CI: 41.56–58.84) from the full pregnancy history instrument are in line with Demographic and Health Survey and United Nations Inter‐agency Group for Child Mortality Estimation estimates. In comparison, the mortality estimates from the truncated pregnancy history instrument are higher, but this difference only approaches statistical significance in the case of the stillbirth rate. Post‐stratification weighting produces a small upwards adjustment in the estimates.Conclusion: Mobile phone surveys are a promising method for collecting perinatal mortality data. The full pregnancy history instrument produces more plausible results than the shorter truncated pregnancy history questionnaire where the window of retrospection is restricted.

Volume 52 - Article 16 | Pages 479–518

In many low and middle-income countries, adult mortality estimates are derived from surveys and censuses conducted through face-to-face interviews. These interviews can be time-intensive and are often impractical during health crises or humanitarian emergencies. The expansion in cellphone ownership and network coverage has created new opportunities for collecting demographic data through mobile phone surveys, but our understanding of selection biases and reporting errors of such data remains incomplete. This study reports on adult mortality estimates obtained through mobile phone surveys conducted in Burkina Faso, Malawi, and the Democratic Republic of the Congo in 2021 and 2022. To mitigate respondent fatigue and network interruptions, we used a shortened version of the set of questions generally used in surveys to ask about the survival of respondents’ siblings. We found substantial differences between mortality estimates obtained from mobile phone interviews and those from face-to-face demographic surveys. Mortality rates from the mobile phone surveys were also approximately half those expected from United Nations estimates. We attribute this underestimation to inaccuracies in reporting of ages and the timing of sibling deaths in the shortened sibling instrument. After imputing ages and dates based on full sibling histories collected in previous face-to-face surveys, mortality rates were more consistent with other data sources. Mobile phone surveys are promising for the measurement of adult mortality in settings where face-to-face surveys are hindered, but they are susceptible to reporting errors. More research is needed on the best set of questions to use for capturing recent adult deaths.

Volume 51 - Article 45 | Pages 1429–1470

In countries without adequate death registration systems, adult mortality is often estimated using orphanhood-based methods. The HIV pandemic breaches several assumptions of these methods, for example, by increasing the correlation between maternal and child survival. Using microsimulations we generated 1,152 populations facing HIV epidemics and evaluated different orphanhood-based estimates against the underlying mortality rates. We regressed survivorship probabilities on proportions of respondents with surviving mothers, adjusting for trends in seroprevalence and coverage of antiretroviral therapy, to obtain new coefficients. We tested the different methods on survey and census data from 16 African countries with high HIV prevalence. We found that the original orphanhood method underestimates mortality during an AIDS epidemic, but better estimates can be obtained using new coefficients applied to synthetic measures of maternal survival. The resulting estimates agree well with those of the United Nations Population Division. Orphanhood-based estimates can fill data gaps in adult mortality, including in countries with high HIV prevalence.

Background: In low- and middle-income countries with limited death registration statistics, adult mortality rates are commonly estimated through sibling survival histories (SSH). In full SSH, respondents are asked about either the age, or the age and time of death, of each of their siblings in turn. Full SSH allow direct mortality estimation but can be time-consuming to collect. In this study, we introduce a new indirect estimation method using summary SSH, requiring only a limited set of questions to produce recent mortality estimates. Methods: We developed a set of 192 microsimulations representing a wide range of fertility and mortality patterns, and reconstructed summary SSH within these simulations as if they had been collected from adults aged 15–49. For each age group of respondents, we calculated coefficients that convert the proportion of adult siblings who died in the previous 5 years into age-specific mortality rates. We then evaluated the performance of this new method with real data, using 154 Demographic and Health Surveys. Results: The new indirect method provides mortality rates that are consistent with direct estimates from full SSH. Across all DHS, the mean absolute percentage error in the risk of dying in adulthood (ages 15–49) is 6% for both men and women. In all but one survey, 95% confidence intervals around the direct and indirect estimates overlap. As with direct estimates of adult mortality from SSH, the indirect estimates remain, however, lower than those of the Population Division of the United Nations. Conclusions: Summary questions on sibling survival can be included in censuses and rapid turn-around surveys for the measurement of recent adult mortality.