Mandatory, voluntary, repetitive, or one-off post-migration follow-up for tuberculosis prevention and control: A systematic review

Abstract

Background: Post-migration follow-up of migrants identified to be at-risk of developing tuberculosis during the initial screening is effective, but programmes vary across countries. We aimed to review main strategies applied to design follow-up programmes and analyse the effect of key programme characteristics on reported coverage (i.e., proportion of migrants screened among those eligible for screening) or yields (i.e., proportion of active tuberculosis among those identified as eligible for follow-up screening).

Methods and findings: We performed a systematic review and meta-analysis of studies reporting yields of follow-up screening programmes. Studies were included if they reported the rate of tuberculosis disease detected in international migrants through active case finding strategies and applied a post-migration follow-up (defined as one or more additional rounds of screening after finalising the initial round). For this, we retrieved all studies identified by Chan and colleagues for their systematic review (in their search until January 12, 2017) and included those reporting from active follow-up programmes. We then updated the search (from January 12, 2017 to September 30, 2022) using Medline and Embase via Ovid. Data were extracted on reported coverage, yields, and key programme characteristics, including eligible population, mode of screening, time intervals for screening, programme providers, and legal frameworks. Differences in follow-up programmes were tabulated and synthesised narratively. Meta-analyses in random effect models and exploratory analysis of subgroups showed high heterogeneity (I2 statistic > 95.0%). We hence refrained from pooling, and estimated yields and coverage with corresponding 95% confidence intervals (CIs), stratified by country, legal character (mandatory versus voluntary screening), and follow-up scheme (one-off versus repetitive screening) using forest plots for comparison and synthesis. Of 1,170 articles, 24 reports on screening programmes from 7 countries were included, with considerable variation in eligible populations, time intervals of screening, and diagnostic protocols. Coverage varied, but was higher than 60% in 15 studies, and tended to be lower in voluntary compared to compulsory programmes, and higher in studies from the United States of America, Israel, and Australia. Yield varied within and between countries and ranged between 53.05 (31.94 to 82.84) in a Dutch study and 5,927.05 (4,248.29 to 8,013.71) in a study from the United States. Of 15 estimates with narrow 95% CIs for yields, 12 were below 1,500 cases per 100,000 eligible migrants. Estimates of yields in one-off follow-up programmes tended to be higher and were surrounded by less uncertainty, compared to those in repetitive follow-up programmes. Yields in voluntary and mandatory programmes were comparable in magnitude and uncertainty. The study is limited by the heterogeneity in the design of the identified screening programmes as effectiveness, coverage and yields also depend on factors often underreported or not known, such as baseline incidence in the respective population, reactivation rate, educative and administrative processes, and consequences of not complying with obligatory measures.

Conclusion: Programme characteristics of post-migration follow-up screening for prevention and control of tuberculosis as well as coverage and yield vary considerably. Voluntary programmes appear to have similar yields compared with mandatory programmes and repetitive screening apparently did not lead to higher yields compared with one-off screening. Screening strategies should consider marginal costs for each additional round of screening.

Fig 1. Prisma flow chart.

Fig 2

Forest plot: Coverage by follow-up (A) and legal scheme (B). Coverage: defined as the proportion of migrants who presented for screening among all migrants identified as eligible for follow-up screening. ES: estimate. Diamonds: point estimates. Horizontal bands: 95% binomial exact confidence intervals. Circles: Indicate plot weights by sample size according to the inverse of the variance. Larger circles indicate larger sample size and lower variance. Data on individuals presented for follow-up deviate in Mor and colleagues [14] and Pang and colleagues [13] by n = 1 from those reported in the paper to ensure that studies can be included in the plot, as studies with 100% coverage (i.e., numerator equals denominator) would otherwise be excluded.

Fig 3

Forest plot: Eligible-population-yield by follow-up (A) and legal scheme (B). Eligible-population-yield: defined as the fraction of active TB cases with TB disease among those identified as eligible for follow-up screening, per 100,000. ES: estimate. Diamonds: point estimates. Horizontal bands: 95% binomial exact confidence intervals. Circles: Indicate plot weights by sample size according to the inverse of the variance. Larger circles indicate larger sample size and lower variance. Estimates for Berrocal-Almanza and colleagues [35] are shown as “excluded” as the numerator equals zero.

Fig 4

Forest plot: Eligible-population-yield (A) and Coverage (B), by country. Eligible-population-yield: defined as the fraction of active TB cases with TB disease among those identified as eligible for follow-up screening, per 100,000. Coverage: defined as the proportion of migrants who presented for screening among all migrants identified as eligible for follow-up screening. ES: estimate. Diamonds: point estimates. Horizontal bands: 95% binomial exact confidence intervals. Circles: Indicate plot weights by sample size according to the inverse of the variance. Larger circles indicate larger sample size and lower variance. Data on individuals presented for follow-up deviate in Mor and colleagues [14] and Pang and colleagues [13] by n = 1 from those reported in the paper to ensure that studies can be included in the plot, as studies with 100% coverage (i.e., numerator equals denominator) would otherwise be excluded. Estimates for Berrocal-Almanza and colleagues [35] are shown as “excluded” as the numerator equals zero.