Estimating national-level measles case fatality ratios: an updated systematic review and modelling study

Background: To understand current and mitigate future measles mortality burden, it is critical to have robust estimates of measles case fatality. Estimates of measles case fatality ratios (CFR) that are age-, location-, and time-specific are essential to capture variation in underlying population-level factors--such as vaccination coverage and measles incidence--that contribute systematically to increases or decreases in case fatality. In this study, we updated estimates of measles CFR by expanding upon previous systematic reviews and implementing a meta-regression model. Methods: We conducted a literature review of all available data on measles case fatality from 1980 through 2019 from low- and middle-income countries and extracted the most granular information available on measles cases and deaths. Using this data and a suite of covariates related to measles CFR, we implemented a Bayesian meta-regression model to produce estimates of measles CFR by location and age from 1990 to 2019. Findings: We identified 245 sources that contained information on both measles cases and deaths. In 2019, we estimated a mean all-age CFR among community-based settings of 1.32% (95% Uncertainty Interval (UI): 1.28 - 1.36%) and among hospital-based settings of 5.35% (95% UI: 5.08 - 5.64%). In community-based settings, we estimated 2019 CFR to be 3.03% (95% UI: 2.89 - 3.16%) among under-one year olds, 1.63% (95% UI: 1.58 - 1.68%) among 1 to 4 year olds, 0.84% (95% UI: 0.80 - 0.87%) among 5 to 9 year olds, and 0.67% (95% UI: 0.64 - 0.70%) among 10 to 14 year olds. Between 1990 and 2019, we estimated measles vaccination has averted approximately 71 million deaths due to decreased measles mortality. Interpretation: While CFRs have declined, there are still large heterogeneities across locations and ages. Our updated methodologic framework and estimates can be used to evaluate the effect of measles control and vaccination programmes on reducing preventable measles mortality burden. Research in contextO_ST_ABSEvidence before this studyC_ST_ABSTwo previous systematic reviews have synthesized individual studies of measles CFR. The first review, by Wolfson et al., was published in 2009 and used 58 community-based studies in 29 countries to provide global estimates of measles CFR. Wolfson and colleagues published a descriptive analysis suggesting global estimates of CFR with a mean of 3.3%, a median of 3.9%, and range from 0 - 40.1%. For outbreak investigations, results suggested a median CFR of 5.2% (95% CI: 2.6 - 11.6%). These results were the first figures of measles CFR beyond single country-year studies, reports, and investigations; however, this study only included community-based studies, did not produce estimates for other locations or years, and did not stratify by other underlying determinants of mortality, such as development status of each country. The later review by Portnoy and others was published in 2019 and included data from 1980 to 2016 from low- and middle-income countries; studies included reports from both community- based (n=85) and hospital-based (n=39) settings. Following the review, authors used a log- linear prediction model with a select set of covariates, generally understood to be related to measles CFR (previous vaccination history [first dose MCV coverage used as a proxy], estimated measles attack rate) and indirectly associated with measles CFR (under-5 mortality [U5M], total fertility rate, proportion of population living in urban areas, population density). The authors reported predicted CFR stratified by year, country-development status, under-5 mortality rate, care-setting (community versus hospital), age (under- or over-5 years), and calendar year from 1990 to 2030. Results predicted a mean CFR of 2.2% (95% CI: 0.7 - 4.5%) for years 1990-2015, with stratification for community (CFR: 1.5, 95% CI: 0.5 - 3.1%) and hospital-based studies (CFR: 2.9, 95% CI: 0.9 - 6.0%). Added value of this studyOur study produces age-, location- and year-specific estimates of measles CFR from 1990 to 2019 by building on previous estimates in three ways. First, it updates the existing body of evidence to those published through 2020 and non-English studies. Second, it incorporates an explicit conceptual framework based on literature review and expert consultation to identify a suite of covariates demonstrated to be related to measles CFR at the population level. Last, it uses a Bayesian meta-regression model, with a flexible spline component to better capture variation in CFR by age. Implications of all the available evidenceThis model, along with corresponding estimates, can contribute to a deeper understanding of measles CFR and allow for a more robust assessment of vaccination programmes and other interventions to reduce measles mortality burden.