Effectiveness of key - population COVID - 19 prevention depends on overall community - level control: a correctional facility outbreak study in Guangdong, China

Abstract

Introduction

In December 2022, the zero - COVID strategy was amended, eliminating mandatory testing and isolation. The purpose of this study was to explore the effectiveness of targeted prevention measures in a key population (inmates and prison guards) and to provide data for similar situations in the future.

Methods

After the 2022 policy changes, a comprehensive 40 - day study was executed within the prison’s enclosed environment. 79,075 nucleic acid samples from facility entrants were collected, complemented by near-full-length viral genome sequencing using a targeted next-generation sequencing (tNGS) approach, to discern the impact of external non-pharmaceutical interventions (NPIs) on viral transmission within the institution.

Results

After two years without COVID − 19 cases, the prison’s streak ended abruptly due to changes in external NPIs. After two years without COVID-19 cases, the prison’s streak ended abruptly following the December 2022 policy shift. Continuous internal NPIs delayed the epidemic peak by ~ 5 days compared to the community. Critically, tNGS revealed nine distinct SARS-CoV-2 strains co-circulating on the first day of widespread testing among ordinary inmates (Dec 17, 2022)—providing direct genomic evidence of multiple external introductions. New variants continued to emerge over time. The 20× coverage decreased significantly in the later stage of the epidemic (p < 0.0001). The 20×coverage decreased significantly in the later stage of the epidemic (p < 0.0001).

Conclusions

This study highlights that facility-specific NPIs—such as closed-loop management, quarantine for new admissions, and cohorting—can only delay, but not prevent, outbreaks driven by imported infections. It underscores the necessity for enhanced surveillance and expeditious testing in densely populated or confined quarters to identify and mitigate infection sources quickly. tNGS is an indispensable tool for validating superspreading events, deciphering transmission intricacies, and distinguishing strains among cases.

Clinical trial number

Not applicable.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12879-026-12725-9.

Introduction

The coronavirus disease 2019 (COVID − 19) pandemic has had a profound impact on global public health, economies, and daily life, making continuous surveillance of the virus and evaluation of control measures an urgent necessity. Among the various strategies, non - pharmaceutical interventions (NPIs) have been crucial in curbing the spread of SARS - CoV − 2 [1]. Correctional facilities, with their dense populations and restricted spaces, are critical areas for outbreak surveillance and intervention [2]. Although the general public often assumes that prisoners are at a lower risk of contracting COVID − 19 due to the closed nature of prisons, in reality, they remain a high - risk group [3]. Despite the World Health Organization’s recognition of the need for intensified SARS - CoV − 2 surveillance in such high - risk environments (https://www.who.int/publications/i/item/WHO − 2019 - nCoV - clinical − 2022.2), the investigation of the impact of NPIs in these settings remains inadequate. For example, in prisons, stringent measures such as enhanced isolation and quarantine protocols have been implemented [4], but their actual effectiveness has been rarely reported.

Prisons achieved the remarkable feat of zero infections by purchasing sufficient personal protective equipment (PPE), arranging shift work for staff to prevent virus introduction, and setting up separate areas for confirmed and suspected cases [5, 6], and the effectiveness of these measures was particularly notable during the implementation of the dynamic zero - COVID policy across society. However, the effectiveness of prison - specific NPIs is inherently intertwined with broader community - level control measures. While prior research exists on NPIs, few have explored the effect of broad policy shifts on closed institutions. Our study aims to fill this gap by providing the evaluation within a correctional facility.

Study Setting. The correctional facility under investigation maintained a dynamic population of approximately 3600 inmates, with daily admissions and releases. The staff comprised 519 correctional officers and approximately 200 administrative and support personnel. Inmates were housed in three separate cell units (A, B, and C), each accommodating roughly 1200 individuals; inter-unit movement was strictly prohibited. Within each unit, prisoners (8–14 per room) shared common areas for dining, washing, and engaged in eight hours of communal labor daily, creating frequent close-contact opportunities. Correctional officers patrolled in groups of five and had occasional verbal interactions with inmates. To mitigate external virus introduction, staff operated on a 7-days-on/7-days-off rotating schedule, with off-duty movements monitored; re-entry was barred if their movements overlapped with confirmed cases, and returning staff were required to present a negative nucleic acid test and normal temperature. Family visits were conducted remotely; in-person visits required a negative test within the prior seven days.

By conducting a rigorous 40 - day investigation in a penal institution after the adjustment of the zero - COVID policy in 2022 [7], we provide the first comprehensive analysis of how societal policy shifts influence prison - specific NPIs. The policy shift provided an excellent opportunity for observation, and by comparing the prison’s epidemic situation before and after the policy change, we clearly demonstrate the effectiveness of local key - population protection. The following sections will elaborate on the details of this study, which not only fills a critical research gap but also offers valuable insights for future public health strategies in similar settings.

Methods

Study population, setting, and data collection

Throughout the duration of the dynamic zero-COVID policy, daily nucleic acid testing for SARS-CoV-2 was mandatory, with immediate isolation of those testing positive. Given the high-risk nature of the prison environment, the policy of daily testing remained rigorously in place following the relaxation measures introduced on December 7, 2022. From December 21 to 25, a negative nucleic acid test result within 48 hours was obligatory. After December 26, nucleic acid testing became mandatory only for those entering, exiting, or transferring between sections. Inmates who test positive are transferred to a medical facility for treatment to mitigate the spread of disease.

During the brief study period from December 2022 to January 2023, NPIs were enforced. Proof of vaccination was required for entry into the prison. Newly admitted prisoners were temporarily accommodated in an auxiliary facility, undergoing mandatory testing and awaiting negative results before being integrated into the main prison population.

Laboratory procedures

Real-time Quantitative Polymerase Chain Reaction (RT-qPCR) testing was employed as delineated in prior studies [8]. RT-qPCR testing was applied universally to all individuals in the facility during the study period (n = 79,075). For near-full-length SARS-CoV-2 sequencing by tNGS, we selected all RT-qPCR-positive samples with high viral loads (cycle threshold [CT] values ≤ 32) from two critical timepoints: 69 samples collected on December 17, 2022 (initial outbreak phase) and 39 samples collected on January 6, 2023 (late-phase outbreak), using the Genolab M platform.

The variant calling process was conducted using Varscan2 [9], with established criteria requiring a minimum read depth of 20X (each base in the genome covered by an average of 20 reads) and alignment against the reference SARS-CoV-2 sequence, Wuhan-Hu-1 (NC_045512.2). Only variants meeting quality control standards with a median allele frequency of at least 0.8 were retained. Maximum likelihood phylogenetic trees were reconstructed using FastTree v2.1.11 with default parameters, employing the nearest-neighbor interchange heuristic for topology optimization. Node support was assessed through 1000 bootstrap replicates [10]. Clade categorization was facilitated by NextCladeCLI 2.3.0 (https://clades.nextstrain.org/] [11]. Sequences achieving ≥80% coverage of the reference genome were included in phylogenetic analysis. Sequences were considered part of a distinct transmission cluster if they differed by no more than two nucleotides [12]. These sequences were then displayed via iTOL [13].

Results

Delayed but inevitable outbreak following policy relaxation

Our study evaluated a cohort within a correctional facility in China, analyzing 79,075 nucleic acid samples (one sample per individual) collected via RT-qPCR from December 5, 2022, to January 15, 2023. During the dynamic zero - COVID policy period, the prison achieved long - term zero infections (for three years) through strict measures such as full - closure management. However, following the policy adjustment on December 7, 2022, the first positive case among newly admitted prisoners was recorded within five days (December 12). Infections among prison guards emerged on December 15, followed by cases in the ordinary prison population (those who had completed quarantine and tested negative upon entry) on December 17. An exponential surge in infection rates began on December 18, peaking on December 26, followed by a gradual decline. Analysis by demographic segments revealed staggered peak incidence rates: newly admitted prisoners peaked on December 25, prison guards on December 26, and the ordinary prison population on December 28 (Fig. 1a, Supplemental Fig. 1a). This 3-day lag between subgroup peaks corresponded with differential external exposure risks and internal movement patterns within the facility. These findings underscore the limitations of localized containment efforts in the context of broader policy changes. While strict measures within a single facility can delay the spread of the epidemic, they cannot prevent the outbreak of imported infections when the overall epidemic prevention strategy shifts.

Fig. 1.

Epidemiological and phylogenetic overview of SARS-CoV-2 in a correctional facility setting. (A) the epidemiological trajectory of SARS-CoV-2 infections across different population segments within the prison, charting epidemic curves. Incidence of daily confirmed cases is represented by colored bars—the prison guards (green), newly admitted (blue), and the ordinary prisoners (red)—with the overarching positivity rate delineated by a purple line. Red box indicates the date of national policy adjustment (December 7, 2022). (B) the shifts in variant dominance observed through successive testing periods. (C) the varied distribution of SARS-CoV-2 variants among initial subpopulations. The participant counts for units A, B, C, newly admitted prisoners, and correctional staff are 21, 20, 9, and 22, respectively. (D) the genomic elucidation of SARS-CoV-2 transmission pathways within the correctional ecosystem. Transmission clusters defined as sequences differing by ≤ 2 nucleotides

Multi-source viral introductions drive initial strain diversity

Molecular characterization of the SARS - CoV − 2 isolates, via tNGS, identified the BA0.5.2 Omicron variant as predominant, detected in 101 of 108 successfully sequenced isolates. The proportion of BA0.5.2 decreased from 97% (67/69) on December 17 to 87% (34/39) on January 6 (Fisher’s exact test, p = 0.042) (Fig. 1B, Supplemental Table 2). Analysis within specific units revealed significant diversity in circulating strains. For example, in Unit A, 12 isolates were successfully sequenced, with 11 belonging to the 22B clade (including 1 BA0.5.2.48, 2 BF0.7.14, and 8 DY0.3) and 1 belonging to B0.1.1.1. Similarly, in Unit B, 20 isolates were sequenced, with 19 belonging to the 22B clade (including 9 BA0.5.2.49, 1 BF0.7.14, 1 DY0.1, and 7 DY0.3) and 1 belonging to B0.1.1.1. Unit C showed 9 isolates, all within the 22B clade (including 2 BA0.5.2.49, 6 DY0.3, and 1 DY0.3), with 1 belonging to B0.1.1.1.

Among newly admitted prisoners, 22 isolates were sequenced, all within the 22B clade (including 1 BA0.5.2, 1 BA0.5.2.48, 2 BA0.5.2.49, 1 BF0.7.14.4, 1 BF0.7.14.5, 1 DY0.1, and 15 DY0.3). Prison guards yielded 6 isolates, all within the 22B clade (including 2 BA0.5.2.48, 1 BA0.5.2.49, 2 DY0.3, and 1 DY0.4). Notably, in the second batch of samples (January 6, 2023), 37 of 39 sequenced isolates belonged to the 22B clade, while 1 belonged to B0.1.1.529 and 1 to BA0.2.

The detection of BA0.5.2 in prison guards and newly admitted prisoners on the same day that long - term inmates tested positive, alongside the presence of B0.1.1 in long - term inmates, demonstrates the diversity of transmission routes and multiple sources of introduction, leading to initial strain diversity. Additionally, the emergence of new strains such as B0.1.1.529 and BA0.2 after 15 days indicates continuous external introduction or undetected internal transmission, highlighting the challenges in containment (Fig. 1C). These findings underscore the importance of comprehensive genomic surveillance to identify and manage multiple transmission sources effectively.

Genomic surveillance reveals superspreading and transmission complexity

Phylogenetic analysis, constrained by the exclusion of 34 samples due to inadequate genomic coverage ( < 80%), constructed a lineage tree from 74 samples. This tree elucidated the transmission pathways across different inmate demographics and cell units, revealing single - nucleotide polymorphisms characteristic of clonal spread. For the first day of testing, 9 distinct viral strains (B0.1.1, BA0.5.2, BA0.5.2.48, BA0.5.2.49, BF0.7.14, BF0.7.14.1, DY0.1, DY0.3, DY0.4) were identified among 41 positive samples from long - term inmates. Phylogenetic analysis based on mutation sites (Fig. 1D) revealed that the B0.1.1 strain in Unit A and Unit B was phylogenetically distant from other strains, particularly those from newly admitted prisoners and prison guards, indicating multiple sources of introduction and diverse transmission routes. This finding underscores that outbreaks in confined spaces, such as prisons, can be driven by multiple co - circulating strains, making these settings particularly susceptible to superspreading events.

The diversity of circulating strains, even within confined units, points to the occurrence of distinct clonal expansions among both prison staff and newly admitted prisoners, indicative of complex intra - facility transmission dynamics. Notably, in Unit B, 22 confirmed positive cases were associated with 9 distinct variants, and three superspreading events, each involving four or more individuals, were identified. For example, in Unit B, individuals SD82, SD84, SD89, SD93, SD95, and SD104 were linked to one event, while another involved SD80, SD81, SD90, and SD100. Among newly admitted prisoners, cases SD33, SD37, SD41, and SD42 were identified on the first day. Crucially, identical genetic variations were observed in viral strains shared between recent entrants and long - term inmates over a 20 - day period, particularly in pairs CD57 and CD143, and CD86 and CD141. This highlights the critical role of clonal strains in driving superspreading phenomena and underscores the necessity of precise genomic surveillance to effectively manage COVID − 19 transmission in such settings.

Additionally, tNGS indirectly reflects the integrity of viral genomes. In this study, a significant reduction in 20X (as defined in the Methods) coverage was observed in the later stages of the outbreak (p < 0.0001), further emphasizing the importance of genomic sequencing. Despite initial viral detection using RT-qPCR, only 8 of 39 samples from the second collection (January 6, 2023) could be fully sequenced (genomic coverage > 80%), compared to 66 of 69 in the first collection (December 17, 2022). This discrepancy persisted even after resequencing 31 samples from the second batch (p = 0.0306, paired t - test), with a 2.2% reduction in average coverage. Re - extraction and resequencing of 7 samples failed to improve coverage (Supplemental Fig. 1D). These findings highlight the challenges in reconciling RT-qPCR data with complete viral genomic sequencing and reinforce the critical role of tNGS in understanding and controlling outbreaks in high - risk environments.

Discussion

While this study is based on a single facility in Shenzhen—a high-density urban center with intense post-policy community transmission—and cannot be generalized to all correctional settings, it provides critical evidence that even robust internal NPIs may be insufficient when societal-level containment is relaxed. This contrasts with early-pandemic responses in the U.S. and Europe, where many prisons mitigated risk through population reduction [14, 15], China implemented a “closed-loop management” strategy. This approach included rigorous material stockpiling, zonal isolation (segregating confirmed cases and suspending family visits), a 14-day quarantine for newly admitted individuals, and regular COVID-19 testing for prison staff and visitors [16]. While the prison in this study achieved remarkably low infection rates during the “dynamic zero-COVID” period, the specific drivers of this success remain unclear—whether it was the independent effectiveness of institutional measures or the spillover effects of the broader societal “dynamic zero-COVID” policy. This ambiguity is also observed in other closed environments, such as nursing homes and orphanages, where similar localized protective measures were adopted [17–19]. This underscores a universal principle: the health of closed institutions is inseparable from public health. Sustainable protection requires integrating these settings into community surveillance, ensuring rapid access to diagnostics, and aligning institutional policies with broader epidemic control strategies. Specifically, we recommend establishing formal data-sharing protocols between correctional facilities and local public health authorities to enable real-time monitoring of community transmission trends.

Notably, after the relaxation of China’s “dynamic zero-COVID” policy in December 2022, the prison, despite its stringent internal controls, experienced outbreaks linked to external transmission. This observation highlights the limitations of relying solely on internal measures to isolate closed institutions from external risks. However, the study also reveals that NPIs played a critical role in delaying the spread of the virus within the prison. Compared to the epidemic peak in the general community (December 20–22, 2022) [20], the prison’s overall positivity rate peaked approximately four days later, on December 26, 2022. This delayed peak underscores the efficacy of NPIs in slowing the transmission of COVID-19, even if they could not entirely prevent it. Furthermore, among newly admitted prisoners, prison guards, and the ordinary prisoner population, positivity rates peaked on December 25, December 26, and December 28, respectively, with initial positive cases reported on December 12, December 15, and December 17. These findings indicate that individuals with higher social contact reached the peak earlier, reinforcing the notion that NPIs effectively delayed the epidemic’s progression within the prison.

These results underscore the interdependence of “prison health and public health,” shifting the research focus from the traditional paradigm of “how prisons impact society” [21, 22] to a new perspective of “how society influences closed spaces.” This insight has significant implications for future pandemic responses in high-risk institutions (e.g., prisons, nursing homes), emphasizing the necessity of integrating these closed environments into the broader public health framework. By doing so, it is possible to achieve more sustainable epidemic control while upholding humanitarian principles. While this study is based on a single case and cannot be generalized to represent the entire Chinese context, it provides valuable evidence for understanding the role of NPIs in closed institutions and highlights the importance of balancing internal measures with broader societal strategies to mitigate public health risks.

Unlike earlier pandemics characterized by limited mobility and clear transmission routes, such as the 1918 influenza outbreak in San Quentin Prison, California, where infections were introduced by newly admitted prisoners and spread to other facilities [23], recent outbreaks have often involved mixed infections. For example, the April 2020 outbreak in two London care homes featured a mix of 614D and 614 G viral strains [24]. This shift underscores the complexity of modern epidemics and the challenges they pose for containment strategies. The prison population studied here, confined to a uniform environment with shared contact histories, resembles a superspreading event, which plays a significant role in the transmission dynamics of SARS-CoV-2 [25]. However, our findings reveal that, despite the shared setting, multiple viral strains were present from the outset of the epidemic. This highlights the critical importance of tNGS in outbreak surveillance and source tracing. By identifying and tracking distinct viral strains, tNGS provides essential insights into transmission pathways, enabling more targeted and effective containment measures. This approach is particularly vital in closed institutions like prisons, where the interplay of internal and external factors can complicate epidemic control efforts. What is more, previous research has often presumed a direct correlation between the propensity for SARS-CoV-2 release and infectiousness, a hypothesis that remains to be conclusively validated. Studies, including those by Zhou et al. [26], have reported on the phenomenon of superspreading in the context of airborne transmission of SARS-CoV-2. It is critical to note that gene copy numbers may not accurately reflect viral load. In our study, despite high levels of RT-qPCR reactions in the second round of samples, most did not contain entire viral particles. Moreover, despite similar levels of social contact, individuals within the same unit were infected with distinct strains. Thus, it is proposed that the focus of “superspreading” should pivot from viral load to the number of individuals infected. Although tNGS implementation in resource-limited prisons is currently constrained by cost, infrastructure, and expertise, its critical role in outbreak resolution—coupled with steadily declining sequencing costs—suggests it will become increasingly feasible and widely adopted in the near future.

The COVID-19 pandemic has exposed significant vulnerabilities within global health infrastructure, economic systems, and societal constructs, highlighting the imperative for ongoing viral monitoring and the reassessment of intervention strategies. Correctional facilities, characterized by dense living conditions and the constrained autonomy of their populations, emerge as critical venues for the implementation of outbreak prevention and control measures. This study contributes to the understanding of NPI effectiveness in a prison environment following the relaxation of stringent COVID-19 policies. It advocates for a heightened emphasis on tNGS in epidemic surveillance, facilitating the precise identification and control of the spread of multiple viral strains within and across communities.

Limitation

This study is limited to a single correctional facility in Shenzhen, a megacity with high population density and intense post-December 2022 community transmission. Regional factors—including local mobility patterns, healthcare infrastructure, and policy implementation fidelity—may influence the observed dynamics. Therefore, the findings should not be extrapolated to prisons in rural areas or countries with different incarceration or pandemic response models.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Abbreviation

COVID-19

Coronavirus Disease 2019

NPIs

Non-Pharmaceutical Interventions

tNGS

Targeted Next-Generation Sequencing

RT-qPCR

Real-time Quantitative Polymerase Chain Reaction

Authors’ contributions

Dachuan Lin conceived the study, designed the methodology, and drafted the manuscript. Shannan Wu, Hao Yang and Letian Zhou performed laboratory analyses; Qi Chen and Zhongyao Xu conducted statistical analysis; Yuzhong Xu coordinated field data collection and Critical revisions. All authors reviewed and approved the final manuscript.

Funding

This study was supported by the Guangdong Basic and Applied Basic Research Foundation (2024A1515030195), the Key Scientific and Technological Project of Shenzhen Science and Technology Innovation Committee (JSGG20210713091811036 and KCXFZ202002011006190), the National Natural Science Foundation of China (NO. 82341060) and Bao’an District Medical Association (BAYXH2023007).

Data availability

Sequence reads used in this paper were deposited in the Sequence Read Archive (SRA), National Genomics Data Center (NGDC), under BioProject accession number PRJCA021110.

Declarations

Ethics approval and consent to participate

This study exclusively utilized irreversibly anonymized residual pathogen samples obtained from routine clinical diagnostics. All samples were stripped of all personal identifiers prior to research use and were entirely non-linkable to any identifiable individuals. Consequently, the requirement for informed consent from participants was waived for this research. This waiver was granted in accordance with Article 39 (Clause 4) of the Ethical Review Measures for Biomedical Research Involving Human Subjects (National Health and Family Planning Commission Order No. 11) (https://www.gov.cn/zhengce/2016-10/12/content_5713806.htm), owing to the use of anonymous data/information obtained through legal channels with no additional risks to the subjects. The study protocol, including the informed consent waiver, was reviewed and approved by the Medical Ethics Committee of Shenzhen University Health Science Center (SUHSC-MEC), with approval number PN-202400026. This handling of anonymized biospecimens aligns with the ethical principles of minimizing risks and protecting participant privacy, which are central to the Declaration of Helsinki. Clinical trial number: not applicable.

Consent for publication

Not applicable (data were anonymized and non-identifiable).

Competing interests

The authors declare no competing interests.