Acute Febrile Illness Surveillance for Estimating Population Immunity, Dominican Republic, 2021

Disclaimer: Early release articles are not considered as final versions. Any changes will be reflected in the online version in the month the article is officially released.

Author affiliation: Harvard Humanitarian Initiative, Cambridge, Massachusetts, USA (E.J. Nilles); Brigham and Women’s Hospital, Boston, Massachusetts, USA (E.J. Nilles, P. Jarolim); Ministry of Public Health and Social Assistance, Santo Domingo, Dominican Republic (C.T. Paulino, R.S. Ramm); Yale School of Medicine, New Haven, Connecticut, USA (M. Vazquez); National University Pedro Henríquez Ureña, Santo Domingo, Dominican Republic (W. Duke); Harvard Medical School, Boston (P. Jarolim); London School of Hygiene & Tropical Medicine, London, UK (A. Kucharski); The University of Queensland, Brisbane, Queensland, Australia (C.L. Lau)

Cross-sectional population-representative serosurveys currently serve as the standard for estimating population immunity. However, the discrete timeframe, high cost, logistical complexity, and long timelines of that survey method often limit its utility during rapidly evolving outbreaks, when timely public health decision-making is essential. The COVID-19 pandemic evidenced this limitation, when the processing of serologic samples for a national household serosurvey in the Dominican Republic (1) in 2021—although methodologically rigorous—failed to keep up with a rapidly shifting postvaccine and postvariant immune landscape.

Researchers have referenced models based on reported case and death data to estimate cumulative infections (2), but those approaches rely on strong assumptions about testing access and outcome ascertainment and may be prone to error for pathogens with high proportions of asymptomatic infection. Moreover, such models estimate incidence rather than the level of immune protection in the population, which is the quantity most directly relevant for public health decision-making for pathogens that generate only partial or nonsterilizing immunity. There is a need, therefore, for alternative surveillance approaches, particularly those that can monitor changes over time and are rapid, scalable, low-cost, and feasible during periods of widespread social disruption (3).

We evaluated whether routinely collected serologic data from an acute febrile illness (AFI) clinical surveillance platform could serve as a proxy for estimating population immunity, using COVID-19 as proxy. We compared SARS-CoV-2 spike antibody data (Roche Diagnostics, https://www.roche.com) from 2 sources collected during July–October 2021 in the same Dominican Republic provinces: a longitudinal AFI surveillance system embedded in routine healthcare settings ( “surveillance”) (4), which included routine blood collection for serologic testing; and a multistage, population-representative household serologic survey (“survey”) (1). We matched surveillance participants to survey participants by propensity score 1:5 using age and number of COVID-19 vaccine doses, reflecting a pragmatic approach based on covariates routinely available in surveillance systems. We looked at baseline characteristics before and after matching (Table). To evaluate the potential for type II error (failing to detect a meaningful difference when one exists), we estimated the detectable difference in the proportion above variant-specific protection thresholds given the matched sample sizes (surveillance, n = 115; survey, n = 575), which provided ≥80% power (2-sided α = 0.05) to detect absolute differences of ≈6–8 percentage points. For each variant-specific protection threshold, we estimated the proportion of participants exceeding the threshold within each group using exact binomial methods. We evaluated differences between groups using 2-sample tests for equality of proportions with continuity correction. Those tests evaluate a 2-sided null hypothesis of equal proportions. We used the prespecified +10-percentage-point margin as a benchmark of public health relevance and not as a formal statistical equivalence test (Appendix).

Figure

75% protection against symptomatic infection for Mu (101.23), Delta (101.88), BA.1 (102.80), and BA.4/5 (103.06). Threshold for XBB.1 was inferred based on ≈10-fold lower neutralizing response relative to BA.4/5. C) Dot-whisker plots showing estimated proportion of participants with antibody levels corresponding to >75% protection by variant or subvariant (underlying data in Appendix Table 1). Dots indicate point estimates; whiskers indicate 95% CIs. Protection thresholds taken from previously published variant-specific correlates of protection (5). Estimates show percentages of persons above those thresholds (uncertainty in thresholds [reported 95% CIs] not propagated into percentages).” />

Figure. Spike antibody responses by surveillance and population survey sampling methods for study of acute febrile illness surveillance similar to household serosurvey for estimating population immunity, Dominican Republic. The study compared…

We included in the matched analysis all 115 surveillance participants and 575 matched survey counterparts (Table; Figure, panel A). We assessed seroprevalence, mean spike antibody levels, and the proportion of persons above thresholds corresponding to 75% protection against symptomatic SARS-CoV-2 infection (5). Unmatched mean antibody titers were similar: 2.4 log10 BAU/mL (95% CI 2.1–2.6) in the surveillance group versus 2.6 log10 BAU/mL (95% CI 2.5–2.7) in the survey group (p = 0.08 by 2-sample t-test). Results in the matched dataset were consistent (2.4 log10 BAU/mL in the surveillance group, 2.5 log10 BAU/mL in the survey group; p = 0.36). Matched seroprevalence was also comparable (surveillance: 88%; survey: 92%; p = 0.21 by 2-sample test for equality of proportions), and the percentage of persons with ≥75% protection differed by ≤6% (range 1.6%–5.4%) across all evaluated variants, with no statistically significant differences between groups (p = 0.18–0.97) (Figure, panels B, C; Appendix Table 1). Although the 95% CIs for the between-group differences in inferred protection slightly exceeded the prespecified +10-percentage-point margin for some variants, the observed differences were small (≤5 percentage points) and centered near zero, and the study had ≥80% power to detect differences of ≈6–8 percentage points, making large discrepancies between sampling frames unlikely.

Given their low cost, integration within routine care, and ability to sample broad community segments, AFI surveillance platforms may complement or substitute for traditional serosurveys when rapid situational awareness or repeated assessments are needed. Unlike cross-sectional serosurveys, this approach could also support ongoing monitoring over time (4,5), enabling near real-time tracking of antibody levels and inferred immunity and informing decisions about school reopening, travel restrictions, and vaccine targeting. Pooling data from multiple sites could further improve representativeness and reduce spatial or demographic bias.

The primary limitation of this analysis is its cross-sectional design, which limits generalizability across timepoints. In addition, the study population had high exposure to SARS-CoV-2 vaccination and infection, which might not reflect settings with lower transmission and vaccination coverage (6). Nonetheless, our findings underscore the potential value of AFI and other clinical surveillance platforms as scalable, cost-effective tools for monitoring population immunity, particularly during pandemics, when speed, affordability, and feasibility are critical.

Dr. Nilles is a clinician and infectious disease epidemiologist at the Brigham and Women’s Hospital and Harvard Medical School in Boston, Massachusetts, USA, and the director of the Infectious Diseases and Epidemics Program at the Harvard Humanitarian Initiative, Cambridge, Massachusetts, USA. His primary research interests include epidemic transmission and control, surveillance of emerging diseases, and seroepidemiology.

Top

We extend thanks to the many persons who volunteered to participate in this study. We also thank the study staff that collected the field data, the US Centers for Disease Control and Prevention Central America Regional Office, the Dominican Republic Ministry of Health and Social Assistance, and the Pedro Henriquez Ureña National University for their commitment and support for the study.

E.J.N. is the principal investigator on a US Centers for Disease Control and Prevention–funded award that funded the study (U01GH002238), and W.D., C.L.L., A.K., and M.V. have received salary support, consultancy fees, or travel paid through this award. C.T. and R.S.R. are employees of the Ministry of Health and Social Assistance, Dominican Republic, an office subcontracted with funds from the US Centers for Disease Control and Prevention award. A.K. is supported by the Welcome Trust, United Kingdom. We declare no other competing interests.

Source