A total of 10.1% (n = 60) healthcare facility patients had a history of former incarceration; those persons were more likely than other TB patients to be infected with lineage 2 M. tuberculosis strains, unemployed, 40–59 years of age, male, of hill tribe ethnicity, and to have a history of previous TB treatment (Table 1).

TB isolates were classified into 4 lineages: lineage 1 (271 [45.8%]), lineage 2 (236 [39.9%]), lineage 3 (7 [1.2%]), and lineage 4 (78 [13.2%]) (Figure 1). When we used the 12-SNP cutoff, the percentage of clustered cases was 6.6% (18 patients) in lineage 1, 46.2% (109 patients) in lineage 2, and 29.5% (23 patients) in lineage 4 (Appendix Table 1). None of the lineage 3 isolates were clustered. We identified 4 large clusters (>10 isolates) of strains with 150 (25.3%) isolates by using the 12-SNP cutoff and 1 large cluster with 33 (5.6%) isolates by using the 5-SNP cutoff (Table 2, Figure 1). The percentage of formerly incarcerated persons within the 4 large clusters was 28.0% (n = 14). In the univariate analysis for all lineages, M. tuberculosis isolates from previously incarcerated TB patients were 4.19 (95% CI 2.11–8.34-fold) more likely to be members of large clusters. After we adjusted for patient age, ethnicity, sex, and previous TB treatment history, prior incarceration remained associated with inclusion in large clusters (adjusted odds ratio [aOR] 4.47, 95% CI 2.05–9.32) (Table 2). Because the 4 large clusters included only lineage 2 isolates, we restricted our analysis to lineage 2 in the multivariate analysis. The odds ratio for prior incarceration decreased modestly (aOR 3.57, 95% CI 1.56–8.15) among lineage 2 isolates after adjustment for age, ethnicity, sex, and previous TB treatment history. Although there was only 1 large cluster with a cutoff of 5 SNPs, the genomic association with incarceration history was stronger than for persons in the large clusters with a cutoff of 12 SNPs (Appendix Table 2).

Figure 2

Figure 2. Distribution of clusters of tuberculosis patients in study of risk for prison-to-community tuberculosis transmission, Thailand, 2017–2020. NNI, nearest-neighbor index.

The proportion of formerly incarcerated persons in a large cluster with the 12-SNP cutoff ranged from 20.0% through 40.0% (Appendix Table 3). A maximum of 4/14 (28.6%) persons in the large clusters had received a TB diagnosis within 2 years after release. Cluster 3 included the highest percentage of formerly incarcerated persons (40.0%); 3 of the 4 formerly incarcerated patients had been incarcerated during 2014–2018, and TB developed within 2 years after they were released. We found little overlap between the periods of incarceration in clusters 1, 2 and 4. The NNIs were close to 1 or >1 (cluster 1, 1.352; cluster 2, 0.980; cluster 3, 1.350; cluster 4, 1.050), which suggested that the patients in the large clusters were not spatially clustered (Figure 2).

Our study found that formerly incarcerated TB patients were 4.7 times more likely than nonincarcerated TB patients to be linked with other patients in large transmission clusters. The association between being in a large cluster and having a previous incarceration history suggests that these genotypes could have circulated in prisons and spread to the community.

The timing of TB diagnosis varied after persons were released from prison. Only 29% of the formerly incarcerated persons in the large clusters received a TB diagnosis within the first 2 years after their release from prison.

Our results suggest that formerly incarcerated persons could be at higher risk for disease progression from latent to active TB for several years after their release from prison. A study from Brazil showed that among 83% of incarcerated persons in whom TB developed, diagnoses were within 2 years of release (5), and another study from Brazil estimated that it took 7 years for TB incidence rates among formerly incarcerated persons to decline to community levels (6). Because of the long-term risk for active TB development in incarcerated persons after they are released, careful follow-up of such incarcerated persons should be a focus of local public health centers in the community (19).

Our study showed that in this setting, formerly incarcerated TB patients were widely distributed throughout a large geographic area (Figure 2). This result was not consistent with that of a previous study in Lima, Peru, which found higher risk for multidrug-resistant TB among persons living near prisons (20). The difference in the spatial distribution of TB observed between the 2 studies is probably associated with the smaller number of more centralized prisons in Chiang Rai than in Lima. Lima has >10 million residents and 7 prisons, and formerly incarcerated persons are likely to reside in areas relatively close to the prisons in which they were incarcerated. In contrast, Chiang Rai Province has only 1.2 million inhabitants dispersed over ≈4,500 square miles and only 2 prisons. In addition, 1 study from Brazil found that prison-related TB genotypes were widely dispersed throughout regions because some inmates were transferred from one prison to another (21). With dissemination of the risk for TB infection throughout large areas, it would be challenging to detect links between formerly incarcerated TB patients and community TB patients and to estimate the effect of incarceration history on the community transmission.

TB control strategies in prisons should focus on not only reducing TB transmission in prison but also on preventing spillover from prison to community. In 2021, the World Health Organization Consolidated Guidelines on Tuberculosis updated the recommendation of systematic screening in prisons from a “conditional recommendation” to a “strong recommendation” but noted that there was “very low certainty of evidence” (22). As a recent modeling analysis showed (6), combined interventions could reduce TB incidence in prisons and in the general population. Options for preventing TB transmission in prisons include screening at the time of entry into the prison, periodic mass screening of incarcerated persons, or both. To control prison-to-community transmission, exit screening before persons return to their respective residential areas and follow-up testing in the community should also be performed (6). Furthermore, although the World Health Organization guidelines conditionally recommend preventive therapy in incarcerated persons depending on resource availability and the local risk for TB (19), few countries with low TB burden and high income have implemented preventive therapy programs for incarcerated persons (23). To reduce illness and death as well as the risk for TB transmission, the benefits of preventive therapy for incarcerated persons in countries with high TB burden should also be considered (6).

Among the limitations of our study, we were not able to establish exact and direct epidemiologic links between incarcerated persons and persons in the general population through contact tracing and are therefore not able to rule out the possibility that formerly incarcerated persons were infected in the community after release rather than vice versa. In addition, our study did not include all TB patients for whom diagnosis was made at the study site because we excluded patients who did not have culture isolates, who died before we contacted them, and who were co-infected with HIV. Therefore, we missed some links that might have affected the number of clusters, the size of clusters, and the proportion of formerly incarcerated persons in large clusters.

In conclusion, our study determined that large clusters included a high percentage of formerly incarcerated TB patients with variable years of incarceration and residential areas. Because prison-related genotypes are circulating in the community, control strategies such as entry and exit screening at release and follow-up screening in the community should be considered to prevent TB-associated illness and death among incarcerated persons and community transmission.