Digital measurement of SARS-CoV-2 transmission risk from 7 million contacts.

How likely is it to become infected by SARS-CoV-2 after being exposed? Almost everyone wondered about this question during the COVID-19 pandemic. Contact-tracing apps1,2 recorded measurements of proximity3 and duration between nearby smartphones. Contacts-individuals exposed to confirmed cases-were notified according to public health policies such as the 2 m, 15 min guideline4,5, despite limited evidence supporting this threshold. Here we analysed 7 million contacts notified by the National Health Service COVID-19 app6,7 in England and Wales to infer how app measurements translated to actual transmissions. Empirical metrics and statistical modelling showed a strong relation between app-computed risk scores and actual transmission probability. Longer exposures at greater distances had risk similar to that of shorter exposures at closer distances. The probability of transmission confirmed by a reported positive test increased initially linearly with duration of exposure (1.1% per hour) and continued increasing over several days. Whereas most exposures were short (median 0.7 h, interquartile range 0.4-1.6), transmissions typically resulted from exposures lasting between 1 h and several days (median 6 h, interquartile range 1.4-28). Households accounted for about 6% of contacts but 40% of transmissions. With sufficient preparation, privacy-preserving yet precise analyses of risk that would inform public health measures, based on digital contact tracing, could be performed within weeks of the emergence of a new pathogen.

Prevalence of persistent SARS-CoV-2 in a large community surveillance study.

Persistent SARS-CoV-2 infections may act as viral reservoirs that could seed future outbreaks1-5, give rise to highly divergent lineages6-8 and contribute to cases with post-acute COVID-19 sequelae (long COVID)9,10. However, the population prevalence of persistent infections, their viral load kinetics and evolutionary dynamics over the course of infections remain largely unknown. Here, using viral sequence data collected as part of a national infection survey, we identified 381 individuals with SARS-CoV-2 RNA at high titre persisting for at least 30 days, of which 54 had viral RNA persisting at least 60 days. We refer to these as 'persistent infections' as available evidence suggests that they represent ongoing viral replication, although the persistence of non-replicating RNA cannot be ruled out in all. Individuals with persistent infection had more than 50% higher odds of self-reporting long COVID than individuals with non-persistent infection. We estimate that 0.1-0.5% of infections may become persistent with typically rebounding high viral loads and last for at least 60 days. In some individuals, we identified many viral amino acid substitutions, indicating periods of strong positive selection, whereas others had no consensus change in the sequences for prolonged periods, consistent with weak selection. Substitutions included mutations that are lineage defining for SARS-CoV-2 variants, at target sites for monoclonal antibodies and/or are commonly found in immunocompromised people11-14. This work has profound implications for understanding and characterizing SARS-CoV-2 infection, epidemiology and evolution.

The epidemiological impact of the NHS COVID-19 app.

The COVID-19 pandemic has seen the emergence of digital contact tracing to help to prevent the spread of the disease. A mobile phone app records proximity events between app users, and when a user tests positive for COVID-19, their recent contacts can be notified instantly. Theoretical evidence has supported this new public health intervention1-6, but its epidemiological impact has remained uncertain7. Here we investigate the impact of the National Health Service (NHS) COVID-19 app for England and Wales, from its launch on 24 September 2020 to the end of December 2020. It was used regularly by approximately 16.5 million users (28% of the total population), and sent approximately 1.7 million exposure notifications: 4.2 per index case consenting to contact tracing. We estimated that the fraction of individuals notified by the app who subsequently showed symptoms and tested positive (the secondary attack rate (SAR)) was 6%, similar to the SAR for manually traced close contacts. We estimated the number of cases averted by the app using two complementary approaches: modelling based on the notifications and SAR gave an estimate of 284,000 (central 95% range of sensitivity analyses 108,000-450,000), and statistical comparison of matched neighbouring local authorities gave an estimate of 594,000 (95% confidence interval 317,000-914,000). Approximately one case was averted for each case consenting to notification of their contacts. We estimated that for every percentage point increase in app uptake, the number of cases could be reduced by 0.8% (using modelling) or 2.3% (using statistical analysis). These findings support the continued development and deployment of such apps in populations that are awaiting full protection from vaccines.

How to avoid a local epidemic becoming a global pandemic.

Here, we combine international air travel passenger data with a standard epidemiological model of the initial 3 mo of the COVID-19 pandemic (January through March 2020; toward the end of which the entire world locked down). Using the information available during this initial phase of the pandemic, our model accurately describes the main features of the actual global development of the pandemic demonstrated by the high degree of coherence between the model and global data. The validated model allows for an exploration of alternative policy efficacies (reducing air travel and/or introducing different degrees of compulsory immigration quarantine upon arrival to a country) in delaying the global spread of SARS-CoV-2 and thus is suggestive of similar efficacy in anticipating the spread of future global disease outbreaks. We show that a lesson from the recent pandemic is that reducing air travel globally is more effective in reducing the global spread than adopting immigration quarantine. Reducing air travel out of a source country has the most important effect regarding the spreading of the disease to the rest of the world. Based upon our results, we propose a digital twin as a further developed tool to inform future pandemic decision-making to inform measures intended to control the spread of disease agents of potential future pandemics. We discuss the design criteria for such a digital twin model as well as the feasibility of obtaining access to the necessary online data on international air travel.

Viral burden is associated with age, vaccination, and viral variant in a population-representative study of SARS-CoV-2 that accounts for time-since-infection-related sampling bias.

In this study, we evaluated the impact of viral variant, in addition to other variables, on within-host viral burden, by analysing cycle threshold (Ct) values derived from nose and throat swabs, collected as part of the UK COVID-19 Infection Survey. Because viral burden distributions determined from community survey data can be biased due to the impact of variant epidemiology on the time-since-infection of samples, we developed a method to explicitly adjust observed Ct value distributions to account for the expected bias. By analysing the adjusted Ct values using partial least squares regression, we found that among unvaccinated individuals with no known prior exposure, viral burden was 44% lower among Alpha variant infections, compared to those with the predecessor strain, B.1.177. Vaccination reduced viral burden by 67%, and among vaccinated individuals, viral burden was 286% higher among Delta variant, compared to Alpha variant, infections. In addition, viral burden increased by 17% for every 10-year age increment of the infected individual. In summary, within-host viral burden increases with age, is reduced by vaccination, and is influenced by the interplay of vaccination status and viral variant.

Correction: Viral burden is associated with age, vaccination, and viral variant in a population-representative study of SARS-CoV-2 that accounts for time-since-infection-related sampling bias.

[This corrects the article DOI: 10.1371/journal.ppat.1011461.].

Lineage replacement and evolution captured by 3 years of the United Kingdom Coronavirus (COVID-19) Infection Survey.

The Office for National Statistics Coronavirus (COVID-19) Infection Survey (ONS-CIS) is the largest surveillance study of SARS-CoV-2 positivity in the community, and collected data on the United Kingdom (UK) epidemic from April 2020 until March 2023 before being paused. Here, we report on the epidemiological and evolutionary dynamics of SARS-CoV-2 determined by analysing the sequenced samples collected by the ONS-CIS during this period. We observed a series of sweeps or partial sweeps, with each sweeping lineage having a distinct growth advantage compared to their predecessors, although this was also accompanied by a gradual fall in average viral burdens from June 2021 to March 2023. The sweeps also generated an alternating pattern in which most samples had either S-gene target failure (SGTF) or non-SGTF over time. Evolution was characterized by steadily increasing divergence and diversity within lineages, but with step increases in divergence associated with each sweeping major lineage. This led to a faster overall rate of evolution when measured at the between-lineage level compared to within lineages, and fluctuating levels of diversity. These observations highlight the value of viral sequencing integrated into community surveillance studies to monitor the viral epidemiology and evolution of SARS-CoV-2, and potentially other pathogens.

Frequency matters: comparison of drug resistance mutation detection by Sanger and next-generation sequencing in HIV-1.

BACKGROUND: Next-generation sequencing (NGS) is gradually replacing Sanger sequencing (SS) as the primary method for HIV genotypic resistance testing. However, there are limited systematic data on comparability of these methods in a clinical setting for the presence of low-abundance drug resistance mutations (DRMs) and their dependency on the variant-calling thresholds. METHODS: To compare the HIV-DRMs detected by SS and NGS, we included participants enrolled in the Swiss HIV Cohort Study (SHCS) with SS and NGS sequences available with sample collection dates ≤7 days apart. We tested for the presence of HIV-DRMs and compared the agreement between SS and NGS at different variant-calling thresholds. RESULTS: We included 594 pairs of SS and NGS from 527 SHCS participants. Males accounted for 80.5% of the participants, 76.3% were ART naive at sample collection and 78.1% of the sequences were subtype B. Overall, we observed a good agreement (Cohen's kappa >0.80) for HIV-DRMs for variant-calling thresholds ≥5%. We observed an increase in low-abundance HIV-DRMs detected at lower thresholds [28/417 (6.7%) at 10%-25% to 293/812 (36.1%) at 1%-2% threshold]. However, such low-abundance HIV-DRMs were overrepresented in ART-naive participants and were in most cases not detected in previously sampled sequences suggesting high sequencing error for thresholds <3%. CONCLUSIONS: We found high concordance between SS and NGS but also a substantial number of low-abundance HIV-DRMs detected only by NGS at lower variant-calling thresholds. Our findings suggest that a substantial fraction of the low-abundance HIV-DRMs detected at thresholds <3% may represent sequencing errors and hence should not be overinterpreted in clinical practice.

Limited emergence of resistance to integrase strand transfer inhibitors (INSTIs) in ART-experienced participants failing dolutegravir-based antiretroviral therapy: a cross-sectional analysis of a Northeast Nigerian cohort.

BACKGROUND: Due to the high prevalence of resistance to NNRTI-based ART since 2018, consolidated recommendations from the WHO have indicated dolutegravir as the preferred drug of choice for HIV treatment globally. There is a paucity of resistance outcome data from HIV-1 non-B subtypes circulating across West Africa. AIMS: We characterized the mutational profiles of persons living with HIV from a cross-sectional cohort in North-East Nigeria failing a dolutegravir-based ART regimen. METHODS: WGS of plasma samples collected from 61 HIV-1-infected participants following virological failure of dolutegravir-based ART were sequenced using the Illumina platform. Sequencing was successfully completed for samples from 55 participants. Following quality control, 33 full genomes were analysed from participants with a median age of 40 years and median time on ART of 9 years. HIV-1 subtyping was performed using SNAPPy. RESULTS: Most participants had mutational profiles reflective of exposure to previous first- and second-line ART regimens comprised NRTIs and NNRTIs. More than half of participants had one or more drug resistance-associated mutations (DRMs) affecting susceptibility to NRTIs (17/33; 52%) and NNRTIs (24/33; 73%). Almost a quarter of participants (8/33; 24.4%) had one or more DRMs affecting tenofovir susceptibility. Only one participant, infected with HIV-1 subtype G, had evidence of DRMs affecting dolutegravir susceptibility-this was characterized by the T66A, G118R, E138K and R263K mutations. CONCLUSIONS: This study found a low prevalence of resistance to dolutegravir; the data are therefore supportive of the continual rollout of dolutegravir as the primary first-line regimen for ART-naive participants and the preferred switch to second-line ART across the region. However, population-level, longer-term data collection on dolutegravir outcomes are required to further guide implementation and policy action across the region.

Frequency-dependent selection can forecast evolution in Streptococcus pneumoniae.

Predicting how pathogen populations will change over time is challenging. Such has been the case with Streptococcus pneumoniae, an important human pathogen, and the pneumococcal conjugate vaccines (PCVs), which target only a fraction of the strains in the population. Here, we use the frequencies of accessory genes to predict changes in the pneumococcal population after vaccination, hypothesizing that these frequencies reflect negative frequency-dependent selection (NFDS) on the gene products. We find that the standardized predicted fitness of a strain, estimated by an NFDS-based model at the time the vaccine is introduced, enables us to predict whether the strain increases or decreases in prevalence following vaccination. Further, we are able to forecast the equilibrium post-vaccine population composition and assess the invasion capacity of emerging lineages. Overall, we provide a method for predicting the impact of an intervention on pneumococcal populations with potential application to other bacterial pathogens in which NFDS is a driving force.

EpiBeds: Data informed modelling of the COVID-19 hospital burden in England.

The first year of the COVID-19 pandemic put considerable strain on healthcare systems worldwide. In order to predict the effect of the local epidemic on hospital capacity in England, we used a variety of data streams to inform the construction and parameterisation of a hospital progression model, EpiBeds, which was coupled to a model of the generalised epidemic. In this model, individuals progress through different pathways (e.g. may recover, die, or progress to intensive care and recover or die) and data from a partially complete patient-pathway line-list was used to provide initial estimates of the mean duration that individuals spend in the different hospital compartments. We then fitted EpiBeds using complete data on hospital occupancy and hospital deaths, enabling estimation of the proportion of individuals that follow the different clinical pathways, the reproduction number of the generalised epidemic, and to make short-term predictions of hospital bed demand. The construction of EpiBeds makes it straightforward to adapt to different patient pathways and settings beyond England. As part of the UK response to the pandemic, EpiBeds provided weekly forecasts to the NHS for hospital bed occupancy and admissions in England, Wales, Scotland, and Northern Ireland at national and regional scales.

Human Immunodeficiency Virus (HIV) Genetic Diversity Informs Stage of HIV-1 Infection Among Patients Receiving Antiretroviral Therapy in Botswana.

BACKGROUND: Human immunodeficiency virus (HIV)-1 genetic diversity increases during infection and can help infer the time elapsed since infection. However, the effect of antiretroviral treatment (ART) on the inference remains unknown. METHODS: Participants with estimated duration of HIV-1 infection based on repeated testing were sourced from cohorts in Botswana (n = 1944). Full-length HIV genome sequencing was performed from proviral deoxyribonucleic acid. We optimized a machine learning model to classify infections as < or >1 year based on viral genetic diversity, demographic, and clinical data. RESULTS: The best predictive model included variables for genetic diversity of HIV-1 gag, pol, and env, viral load, age, sex, and ART status. Most participants were on ART. Balanced accuracy was 90.6% (95% confidence interval, 86.7%-94.1%). We tested the algorithm among newly diagnosed participants with or without documented negative HIV tests. Among those without records, those who self-reported a negative HIV test within <1 year were more frequently classified as recent than those who reported a test >1 year previously. There was no difference in classification between those self-reporting a negative HIV test <1 year, whether or not they had a record. CONCLUSIONS: These results indicate that recency of HIV-1 infection can be inferred from viral sequence diversity even among patients on suppressive ART.

Effect of Universal Testing and Treatment on HIV Incidence - HPTN 071 (PopART).

BACKGROUND: A universal testing and treatment strategy is a potential approach to reduce the incidence of human immunodeficiency virus (HIV) infection, yet previous trial results are inconsistent. METHODS: In the HPTN 071 (PopART) community-randomized trial conducted from 2013 through 2018, we randomly assigned 21 communities in Zambia and South Africa (total population, approximately 1 million) to group A (combination prevention intervention with universal antiretroviral therapy [ART]), group B (the prevention intervention with ART provided according to local guidelines [universal since 2016]), or group C (standard care). The prevention intervention included home-based HIV testing delivered by community workers, who also supported linkage to HIV care and ART adherence. The primary outcome, HIV incidence between months 12 and 36, was measured in a population cohort of approximately 2000 randomly sampled adults (18 to 44 years of age) per community. Viral suppression (<400 copies of HIV RNA per milliliter) was assessed in all HIV-positive participants at 24 months. RESULTS: The population cohort included 48,301 participants. Baseline HIV prevalence was 21% or 22% in each group. Between months 12 and 36, a total of 553 new HIV infections were observed during 39,702 person-years (1.4 per 100 person-years; women, 1.7; men, 0.8). The adjusted rate ratio for group A as compared with group C was 0.93 (95% confidence interval [CI], 0.74 to 1.18; P = 0.51) and for group B as compared with group C was 0.70 (95% CI, 0.55 to 0.88; P = 0.006). The percentage of HIV-positive participants with viral suppression at 24 months was 71.9% in group A, 67.5% in group B, and 60.2% in group C. The estimated percentage of HIV-positive adults in the community who were receiving ART at 36 months was 81% in group A and 80% in group B. CONCLUSIONS: A combination prevention intervention with ART provided according to local guidelines resulted in a 30% lower incidence of HIV infection than standard care. The lack of effect with universal ART was unanticipated and not consistent with the data on viral suppression. In this trial setting, universal testing and treatment reduced the population-level incidence of HIV infection. (Funded by the National Institute of Allergy and Infectious Diseases and others; HPTN 071 [PopArt] ClinicalTrials.gov number, NCT01900977.).

Epidemiological dynamics of bacteriocin competition and antibiotic resistance.

Bacteriocins, toxic peptides involved in the competition between bacterial strains, are extremely diverse. Previous work on bacteriocin dynamics has highlighted the role of non-transitive 'rock-paper-scissors' competition in maintaining the coexistence of different bacteriocin profiles. The focus to date has primarily been on bacteriocin interactions at the within-host scale (i.e. within a single bacterial population). Yet in species such as Streptococcus pneumoniae, with relatively short periods of colonization and limited within-host diversity, ecological outcomes are also shaped by processes at the epidemiological (between-host) scale. Here, we first investigate bacteriocin dynamics and diversity in epidemiological models. We find that in these models, bacteriocin diversity is more readily maintained than in within-host models, and with more possible combinations of coexisting bacteriocin profiles. Indeed, maintenance of diversity in epidemiological models does not require rock-paper-scissors dynamics; it can also occur through a competition-colonization trade-off. Second, we investigate the link between bacteriocin diversity and diversity at antibiotic resistance loci. Previous work has proposed that bacterial duration of colonization modulates the fitness of antibiotic resistance. Due to their inhibitory effects, bacteriocins are a plausible candidate for playing a role in the duration of colonization episodes. We extend the epidemiological model of bacteriocin dynamics to incorporate an antibiotic resistance locus and demonstrate that bacteriocin diversity can indeed maintain the coexistence of antibiotic-sensitive and -resistant strains.

The mutational spectrum of SARS-CoV-2 genomic and antigenomic RNA.

The raw material for viral evolution is provided by intra-host mutations occurring during replication, transcription or post-transcription. Replication and transcription of Coronaviridae proceed through the synthesis of negative-sense 'antigenomes' acting as templates for positive-sense genomic and subgenomic RNA. Hence, mutations in the genomes of SARS-CoV-2 and other coronaviruses can occur during (and after) the synthesis of either negative-sense or positive-sense RNA, with potentially distinct patterns and consequences. We explored for the first time the mutational spectrum of SARS-CoV-2 (sub)genomic and anti(sub)genomic RNA. We use a high-quality deep sequencing dataset produced using a quantitative strand-aware sequencing method, controlled for artefacts and sequencing errors, and scrutinized for accurate detection of within-host diversity. The nucleotide differences between negative- and positive-sense strand consensus vary between patients and do not show dependence on age or sex. Similarities and differences in mutational patterns between within-host minor variants on the two RNA strands suggested strand-specific mutations or editing by host deaminases and oxidative damage. We observe generally neutral and slight negative selection on the negative strand, contrasting with purifying selection in ORF1a, ORF1b and S genes of the positive strand of the genome.

PopART-IBM, a highly efficient stochastic individual-based simulation model of generalised HIV epidemics developed in the context of the HPTN 071 (PopART) trial.

Mathematical models are powerful tools in HIV epidemiology, producing quantitative projections of key indicators such as HIV incidence and prevalence. In order to improve the accuracy of predictions, such models need to incorporate a number of behavioural and biological heterogeneities, especially those related to the sexual network within which HIV transmission occurs. An individual-based model, which explicitly models sexual partnerships, is thus often the most natural type of model to choose. In this paper we present PopART-IBM, a computationally efficient individual-based model capable of simulating 50 years of an HIV epidemic in a large, high-prevalence community in under a minute. We show how the model calibrates within a Bayesian inference framework to detailed age- and sex-stratified data from multiple sources on HIV prevalence, awareness of HIV status, ART status, and viral suppression for an HPTN 071 (PopART) study community in Zambia, and present future projections of HIV prevalence and incidence for this community in the absence of trial intervention.

OpenABM-Covid19-An agent-based model for non-pharmaceutical interventions against COVID-19 including contact tracing.

SARS-CoV-2 has spread across the world, causing high mortality and unprecedented restrictions on social and economic activity. Policymakers are assessing how best to navigate through the ongoing epidemic, with computational models being used to predict the spread of infection and assess the impact of public health measures. Here, we present OpenABM-Covid19: an agent-based simulation of the epidemic including detailed age-stratification and realistic social networks. By default the model is parameterised to UK demographics and calibrated to the UK epidemic, however, it can easily be re-parameterised for other countries. OpenABM-Covid19 can evaluate non-pharmaceutical interventions, including both manual and digital contact tracing, and vaccination programmes. It can simulate a population of 1 million people in seconds per day, allowing parameter sweeps and formal statistical model-based inference. The code is open-source and has been developed by teams both inside and outside academia, with an emphasis on formal testing, documentation, modularity and transparency. A key feature of OpenABM-Covid19 are its Python and R interfaces, which has allowed scientists and policymakers to simulate dynamic packages of interventions and help compare options to suppress the COVID-19 epidemic.

Estimating SARS-CoV-2 variant fitness and the impact of interventions in England using statistical and geo-spatial agent-based models.

The SARS-CoV-2 epidemic has been extended by the evolution of more transmissible viral variants. In autumn 2020, the B.1.177 lineage became the dominant variant in England, before being replaced by the B.1.1.7 (Alpha) lineage in late 2020, with the sweep occurring at different times in each region. This period coincided with a large number of non-pharmaceutical interventions (e.g. lockdowns) to control the epidemic, making it difficult to estimate the relative transmissibility of variants. In this paper, we model the spatial spread of these variants in England using a meta-population agent-based model which correctly characterizes the regional variation in cases and distribution of variants. As a test of robustness, we additionally estimated the relative transmissibility of multiple variants using a statistical model based on the renewal equation, which simultaneously estimates the effective reproduction number R. Relative to earlier variants, the transmissibility of B.1.177 is estimated to have increased by 1.14 (1.12-1.16) and that of Alpha by 1.71 (1.65-1.77). The vaccination programme starting in December 2020 is also modelled. Counterfactual simulations demonstrate that the vaccination programme was essential for reopening in March 2021, and that if the January lockdown had started one month earlier, up to 30 k (24 k-38 k) deaths could have been prevented. This article is part of the theme issue 'Technical challenges of modelling real-life epidemics and examples of overcoming these'.

Evaluation of Phylogenetic Methods for Inferring the Direction of Human Immunodeficiency Virus (HIV) Transmission: HIV Prevention Trials Network (HPTN) 052.

BACKGROUND: Phylogenetic analysis can be used to assess human immunodeficiency virus (HIV) transmission in populations. We inferred the direction of HIV transmission using whole-genome HIV sequences from couples with known linked infection and known transmission direction. METHODS: Complete next-generation sequencing (NGS) data were obtained for 105 unique index-partner sample pairs from 32 couples enrolled in the HIV Prevention Trials Network (HPTN) 052 study (up to 2 samples/person). Index samples were obtained up to 5.5 years before partner infection; partner samples were obtained near the time of seroconversion. The bioinformatics method, phyloscanner, was used to infer transmission direction. Analyses were performed using samples from individual sample pairs, samples from all couples (1 sample/person; group analysis), and all available samples (multisample group analysis). Analysis was also performed using NGS data from defined regions of the HIV genome (gag, pol, env). RESULTS: Using whole-genome NGS data, transmission direction was inferred correctly (index to partner) for 98 of 105 (93.3%) of the individual sample pairs, 99 of 105 (94.3%) sample pairs using group analysis, and 31 of the 32 couples (96.9%) using multisample group analysis. There were no cases where the incorrect transmission direction (partner to index) was inferred. The accuracy of the method was higher with greater time between index and partner sample collection. Pol region sequences performed better than env or gag sequences for inferring transmission direction. CONCLUSIONS: We demonstrate the potential of a phylogenetic method to infer the direction of HIV transmission between 2 individuals using whole-genome and pol NGS data.

On the evolutionary ecology of multidrug resistance in bacteria.

Resistance against different antibiotics appears on the same bacterial strains more often than expected by chance, leading to high frequencies of multidrug resistance. There are multiple explanations for this observation, but these tend to be specific to subsets of antibiotics and/or bacterial species, whereas the trend is pervasive. Here, we consider the question in terms of strain ecology: explaining why resistance to different antibiotics is often seen on the same strain requires an understanding of the competition between strains with different resistance profiles. This work builds on models originally proposed to explain another aspect of strain competition: the stable coexistence of antibiotic sensitivity and resistance observed in a number of bacterial species. We first identify a partial structural similarity in these models: either strain or host population structure stratifies the pathogen population into evolutionarily independent sub-populations and introduces variation in the fitness effect of resistance between these sub-populations, thus creating niches for sensitivity and resistance. We then generalise this unified underlying model to multidrug resistance and show that models with this structure predict high levels of association between resistance to different drugs and high multidrug resistance frequencies. We test predictions from this model in six bacterial datasets and find them to be qualitatively consistent with observed trends. The higher than expected frequencies of multidrug resistance are often interpreted as evidence that these strains are out-competing strains with lower resistance multiplicity. Our work provides an alternative explanation that is compatible with long-term stability in resistance frequencies.