Epidemiology of Rickettsial Infection in the Municipality of Uramita, Colombia

ABSTRACT.

This study aimed to analyze epidemiological indicators related to seroprevalent and seroincident cases of spotted fever group rickettsiae (SFGR) and to identify housing conditions related to tick infestation. A prospective study (2016–2018) was conducted to examine rickettsial seropositivity in humans, domestic animals, and wild mammals in the municipality of Uramita, Antioquia, Colombia, where a previous Rickettsia rickettsii outbreak was reported from 2014 to 2015. The seroprevalence and its associated factors were estimated at baseline, and the seroincidence and its risk factors for humans were estimated 20 months later. A cross-sectional analysis was performed to evaluate the housing conditions linked to tick infestation. The SFGR seroprevalence was 26.70% (95% confidence interval [CI], 20.79–31.37), and the factors associated with SFGR seropositivity were male sex (adjusted prevalence ratio [PRa], 1.67; 95% CI, 1.19–2.32), age (evaluated in 5-year increments) (PRa, 1.05; 95% CI, 1.01–1.09), and household proximity (PRa_{scattered vs. very near}=3.87; 95% CI, 1.12–8.66). The overall seroincidence was 7.40% (95% CI, 4.71–11.06), and the factors associated with SFGR seroincident cases were the presence of wild animals (adjusted relative risk [RRa], 2.46; 95% CI, 1.06–4.72) and the presence of trees in the peri-domiciliary area (RRa, 0.47; 95% CI, 0.23–0.94). The prevalence of house tick infestation was 27.81% (95% CI, 21.93–34.80), and the factors associated with infestation were dirt floors (PRa, 5.88; 95% CI, 2.28–10.31), fiber cement roofs (PRa, 1.76; 95% CI, 1.07–2.83), and the presence of canines in peri-domiciliary or intra-domiciliary areas (PRa, 5.05; 95% CI, 3.31–7.19). Seropositivity rates for canines and opossums were 35.62% (26/73) and 100% (6/6), respectively. Identification of these factors will help to implement efficient surveillance programs in Colombia.

INTRODUCTION

Spotted fever rickettsiosis (SFR) is a neglected disease without epidemiological surveillance in Colombia. The disease is brought to the attention of public health when deadly and unexpected cases occur. SFR has a high case fatality rate since antibiotic treatment is usually delayed as the disease is not considered in the differential diagnosis of febrile syndromes;^1,2 furthermore, most cases are mistakenly diagnosed as dengue because it is frequent in tropical countries and the symptoms and signs are nonspecific.^3–5

A deadly outbreak of rickettsiosis caused by Rickettsia rickettsii was reported in the municipality of Uramita in northwestern Antioquia (6°53′55″N, 76°10′25″W, 650 m above sea level [msl]) between 2014 and 2015 (Figure 1). This municipality is characterized by the high proximity of houses in its urban area. In 2018, the total population was 6,888 inhabitants, with 2,111 in the urban center and 4,777 in rural areas. The first confirmed case in the outbreak was a 12-year-old child who had an immunoglobulin G (IgG) titer of 8,192 on day 14 after symptom onset, that increased to 16,384 one month later. Both measurements were performed using an indirect immunofluorescence antibody (IFA) assay. The case was nonfatal because of prompt antibiotic treatment. Two additional fatal cases previously reported in May 2014 and July 2014 were classified as “probable” because the patients presented with compatible signs and symptoms and they lived adjacent to the household of the child who was the first confirmed case. In April 2015, the second confirmed case was a 43-year-old male diagnosed by PCR and bacterial culture this patient probably died because of delayed antibiotic treatment.⁶

Figure 1.

Map of the study area localized in western Antioquia, Colombia. This figure appears in color at www.ajtmh.org.

The field study conducted during the outbreak reported that three patients (two who died and one who survived) inhabited the urban center of the municipality; one patient with a fatal case had an occupation cleaning roads between the municipalities of Mutatá (7°14′36″N, 76°26′09″W, 75 msl) and Cañasgordas (6°44′59″N, 76°01′33″W, 1320 msl), which increased the risk of disease. Potential vectors of spotted fever group rickettsiae (SFGR) in other countries are ticks: Riphicephalus sanguineus s.l. in urban areas and Amblyomma cajennense s.l. in rural areas.^7–9 However, during the field study, it was not possible to identify the potential vectors during the outbreak or establish whether SFGR circulated among domestic and wild animals in the area. The main objectives of this study were to analyze epidemiological indicators related to SFGR seroprevalent and seroincident cases, identify housing conditions related to tick infestation, and study SFGR seropositivity in domestic and wild animals in rural and urban areas of the municipality of Uramita, Antioquia, Colombia.

MATERIALS AND METHODS

Study design.

A prospective study was conducted to examine rickettsial seropositivity in the municipality of Uramita, Antioquia, Colombia (6°53′55″N, 76°10′25″W) (Figure 1). The seroprevalence and its associated factors were estimated at baseline (T0; cross-sectional study conducted between August and September 2016). Participants were followed-up 20 months later to estimate the seroincidence and its associated factors factors (T20; follow-up study, conducted in May 2018). Additionally, a cross-sectional study to examine the housing conditions linked to tick infestation was conducted at T0. The study included individuals of both sexes and all ages residing in rural and urban areas of Uramita who agreed to participate and signed the informed consent.

Sampling.

The study included all neighborhoods (N = 10) in the urban center and one village in the rural area of Uramita, where a lethal case of rickettsiosis caused by R. rickettsii was registered during April 2015.⁶ We designed a stratified random sampling based on the number of people and canines inhabiting the households registered in the census. Stratification was performed according to the presence/absence of canines in each household. For both the baseline study (seroprevalence) and follow-up study (seroincidence), the sampling units were the households within neighborhoods and village, and the analysis units were the inhabitants of the households. During the household study conducted to analyze tick infestation, the sampling and analysis units were households. Sampling units were randomly selected proportional to the number of houses in each neighborhood and village and stratified according to the presence of canines in households. The sample size was estimated using the following parameters: expected seroprevalence, 26%; error, 5%; level of confidence, 95%; and design effect, 1.20.¹⁰ A sample size of 234 households was estimated out of a total of 566 households registered in the census. The study included all individuals inhabiting households (defined as individuals staying or sleeping in the household for at least 5 days per week) as well as family pets (canines). The study also included the canines participating in the rabies vaccination campaign in Uramita.

Outcomes of SFGR seropositivity in humans and domestic animals.

Serum was obtained from humans and canines to detect IgG against SFGR using an IFA. A seroprevalent case was defined as a human or canine with an IgG titer ≥ 128.¹¹ Seroincident cases were defined as follows: serological conversion of IgG titers (negative at T0 to a positive titer of 128 at T20) or at least a four-fold increase in the IgG endpoint titer at the time of obtaining the second sample (T20) compared with that at the time of obtaining the first sample (T0).

Slides containing antigens of R. rickettsii isolated from a lethal case reported in Uramita were used to detect IgG against SFGR. Slide preparation was performed according to procedures described previously.⁶ Only the seroincident individuals at T20 and seropositive canines at T0 were evaluated with additional slides containing antigens of Rickettsia amblyommatis, Rickettsia parkeri, and Rickettsia bellii to detect the potential exposure to these rickettsial species. The exposure to one Rickettsia species was defined as a four-fold increase in the IgG endpoint titer of the species compared with the titers of the other species evaluated. Each test included negative and positive controls from the research group for Veterinary Sciences Centauro at the laboratory of the University of Antioquia.

Outcomes of household tick infestation.

The presence of ticks inside households was subjectively reported by a survey administered to the head of the household (male or female) according to previously reported protocols.¹²

Exposures and covariates.

The primary exposure to estimate factors associated with seroprevalent cases was recent and previous main occupation (within the past 5 years) and for seroincident cases was the main occupation during the past 20 months. This variable was categorized as outdoors (agricultural workers, farmers, ranchers, day laborers) or indoors (the remaining occupations). Additional variables analyzed were age (years), sex (male or female), time of residence in the municipality (years), travel routes to the place of work, education level, previous exposure to ticks, previous fever episodes, and fever episodes during the study (20 months).

Additionally, household characteristics such as the materials of the floors, walls, and roofs, the presence of vegetation around the house (bushes, trees, grasses, yucca, corn, tomato), the presence of domestic animals (equines, porcine, poultry, canines, felines), the presence of opossums and synanthropic rodents, household proximity (very near, near, scattered, and very scattered), and household location (rural or urban area) were registered.

Finally, some attitudes and practices related to tick infestation were included as covariates, including wearing white clothes when performing outdoor occupations, protection against rodents in the household, pet bathing habits, the use of insect repellents, tick removal after outdoor activity, and family members involved in deforestation and forest fragmentation for agricultural purposes.

Capture, identification, and SFGR seropositivity of opossums.

Opossums were captured in the surroundings of the village using Tomahawk^® traps (length, 26 inches; width, 9 inches; height, 9 inches; Tomahawk Live Trap, Hazelhurst, WI). A mixture of oat flakes, vanilla extract, tuna, and peanut butter was used as bait to capture the wild animals. These traps were georeferenced and maintained in the selected area according to ecological conditions, proximity to households, the presence of poultry, and the proximity of the forest to the household. Animals were captured between June and July 2016 during 4 nights; 18 traps were used each night. The captured opossums were anesthetized with Tiletamina (125 mg) and Zolazepam (125 mg) using a dose of 10 mg per kg of weight. Then, blood samples were obtained by intracardiac puncture in tubes without anticoagulant, and the body weight, tail length (mm), foot length (mm), total length, and sex were registered for age determination.

Serum samples were processed at the Veterinary Sciences Centauro Laboratory to diagnose rickettsial seropositivity by IFA using opossum-specific antibodies conjugated with fluorescein (donation of the Veterinary Parasitology Laboratory of the University of Sao Paulo, Brazil).¹³ Slides with Rickettsia rhipicephali, R. amblyommatis, R. parkeri, R. bellii, and R. rickettsii (the same R. rickettsii antigen used to evaluate human samples) antigens were used to detect exposure to these species. Finally, a molecular diagnosis of acute infection was performed using DNA isolated from blood clots (Thermo Genejet DNA purification) and a real-time PCR specific for the gltA gene of SFGR.¹⁰

Identification of ticks collected from canines and SFGR detection.

Ticks were removed directly from canines during a 3-minute collection effort, and the specimens were placed in containers with 70% ethanol. Stage, sex, and species identification were performed using morphological characteristics.¹⁴ DNA was extracted from each tick or pool (prepared with ≤5 ticks that were not engorged and were of the same sex, stage, morphospecies, and collected from the same host) with a Thermo Genejet DNA purification kit. Engorged ticks were cut in sterile Petri dishes using a sterile scalpel blade, and a small piece was used for DNA extraction.

To detect rickettsial infection in ticks, a real-time PCR specific for the gltA gene from SFGR was used.¹⁰ Positive samples were subjected to amplification and sequencing of the gltA (401 pb)¹⁵ and ompA (631 pb)¹⁶ genes to identify the Rickettsia species. The positive control was DNA of R. rhipicephali donated by the Veterinary Parasitology Laboratory of the University of Sao Paulo, Brazil. Finally, to assess the absence of PCR inhibitors in DNA from ticks, the samples were amplified by conventional PCR of the 12S mtDNA gene.¹⁷

Statistical analysis.

A descriptive analysis of the three outcomes was performed (seropositivity, household infestation by ticks, and seroincidence). The medians and interquartile ranges (IQRs) for quantitative variables, such as age in years and time of residence in the study area, were calculated. In addition, absolute and relative frequencies were used to describe qualitative variables. The SFGR seroprevalence, seroincidence, and prevalence of household tick infestation are presented with their 95% confidence intervals (CIs) adjusted by random effects (neighborhood/village).

Models for seroprevalence and household infestation studies.

Additionally, a multilevel complementary log-log regression model to estimate factors associated with SFGR seroprevalent cases and household infestation by ticks was performed. Bivariate analyses were performed to select the main variables for inclusion in the multivariate model (variables with P ≤ 0.20 in the bivariate analysis). Three levels were included in the regression model for the seroprevalence study: individuals within households, and households within neighborhoods/village (random effect). The model to study tick infestation in households included two levels: households and households within neighborhoods/village (random effect). Furthermore, a variance component correlation matrix was used in the model. Finally, the multivariate analysis was performed using the stepwise approach driven by the researchers’ criteria, and the regression for the seroprevalence study was weighted by the inverse probability of individual selection within the household to adjust the model for the loss of individuals during the sampling process.

Potential confounders and effect modifiers were evaluated, and the best model explaining the outcomes was selected according to Bayesian Information Criteria. The prevalence ratios were calculated as follows¹⁸:

$$\frac{P_1}{P_0}=(1-e^{-\text{e}(b_0+b_1+\sum _{j=2}^{k-1}{b}_j{X}_j})/(1-e^{-\text{e}(b_0+\sum _{j=2}^{k-1}{b}_j{X}_j})$$

where b₀ is the model intercept, b represents the regression parameters (b_1; b_2;…; b_j), X represents covariates (X1; X2;…; Xk-1), and e is the base of the Napierian logarithms.

Models for the seroincidence study.

The factors associated with SFGR seroincident cases were estimated using a multilevel logistic regression model with the neighborhood/village as a random effect and using a variance component correlation matrix. The methodology used to select variables and model building was the same as that described previously. The transformation of the odds ratio to the relative risk (RR) was performed using the following function¹⁹:

$$RR=\left(1+e^{\left(-b_0\right)}\right)/\left(1+e^{\left(-b_0-b_1\right)}\right)$$

where e is the base of Napierian logarithms, b₀ is the regression model intercept, and b₁ is the regression model parameter. SAS 9.04.01 statistical software (SAS, Cary, NC) and the PROC GLIMMIX program (SAS) were used to perform all analyses.

Ethics statement.

The Committee for Animal Experimentation (meeting on June 10, 2014) and the Committee of Ethics in Research (meeting on May 22, 2014) of the University of Antioquia approved all procedures performed during the present study.

RESULTS

SFGR seroprevalence and associated factors.

The study initially included 232 randomly selected households (571 inhabitants). However, 54 additional households (73 inhabitants) were included because their inhabitants volunteered to participate actively in the study. A total of 286 households (644 inhabitants) were analyzed. The sensitivity analysis did not show differences whether the randomly selected houses or the total number of houses were included in the study; therefore, the results presented include all households (N = 286).

Individual and household sampling rates were 74.80% (644/861) and 100% (286/286), respectively. The SFGR seroprevalence rate was 26.70% (172/644) (95% CI, 20.79–31.37). The proportions of people inhabiting urban and rural areas among seropositive cases were 81.98% (141/172) and 18.02% (31/172), respectively. The proportions of seropositive cases among neighborhoods and the village ranged from 16.00% (8/50) to 37.50% (30/80), with the highest proportion estimated in the village (rural area). In addition, we found that the proportions of seropositive cases according to the distance between households were 60.00% (6/10), 37.08% (46/148), and 24.69% (120/486) in areas with disperse houses (disperse but visible houses), near houses (close but not attached), and very near houses (attached houses), respectively.

Seropositive individuals had a median age of 39.38 years (IQR, 25.15–56.21 years). Approximately half were males (51.16%; 88/172), and the median time of residence in the study area was 10.50 years (IQR, 3.00–20.00 years). Additionally, 23.26% (40/172) of the seropositive individuals had worked outdoors (recently or within the past 5 years), and 12.79% (22/172) had been bitten by ticks previously (days or years ago). Additional sociodemographic, household characteristics, and practices of the study population are shown in Table 1.

Table 1.

Descriptive, bivariate, and multivariable mixed effects complementary log-log regression analysis of seroprevalent cases of rickettsial infection

Variables	Total, n = 644, n (%)	Seropositive, n = 172, n (%)	Seronegative, n = 472, n (%)	Crude PR (95% CI)	Adjusted PR (95% CI)
Individuals
Occupation (outdoors)	116 (18.01)	40 (23.26)	76 (16.10)	1.47 (1.05–2.02)	0.97 (0.62–1.49)
Male sex	259 (40.22)	88 (51.16)	171 (36.23)	1.61 (1.22–2.09)	1.67 (1.19–2.32)
Age, years, median (IQR)	34.80 (18.28–52.21)	39.38 (25.15–56.21)	32.31 (17.29–50.50)	1.01 (1.001–1.02)	1.01 (1.001–1.02)
Tick infestation	73 (11.34)	22 (12.79)	51 (10.81)	1.16 (0.73–1.78)
History of fever	318 (49.38)	78 (45.35)	240 (50.85)	0.87 (0.64–1.16)
Time of residence in the area	10.00 (2.00–20.00)	10.50 (3.00–20.00)	9.00 (2.00–19.00)	1.01 (0.99–1.02)
Household characteristics
Roof material
Zinc	246 (38.20)	60 (34.88)	186 (39.41)	0.82 (0.49–1.31)
Cement	113 (17.55)	34 (19.77)	79 (16.74)	1.35 (0.81–2.17)
Tile	27 (4.19)	6 (3.49)	21 (4.45)	1.16 (0.51–2.42)
Wood	5 (0.78)	2 (1.16)	3 (0.64)	1.73 (0.26–4.99)
Fiber cement (Eternit)	263 (40.84)	74 (43.02)	189 (40.04)	1.03 (0.66–1.56)
Floor material
Dirt soil	25 (3.88)	6 (3.49)	19 (4.03)	0.91 (0.30–2.40)
Cement	497 (77.17)	138 (80.23)	359 (76.06)	1.31 (0.83–2.00)
Wood	5 (0.78)	1 (0.58)	4 (0.85)	1.17 (0.19–4.28)
Tile	165 (25.62)	38 (22.09)	127 (26.91)	0.77 (0.49–1.18)
Wall material
Wood	6 (0.93)	2 (1.16)	4 (0.85)	1.35 (0.20–4.67)
Brick wall	621 (96.43)	166 (96.51)	455 (96.40)	1.26 (0.44–3.14)
Tapia	17 (2.64)	3 (1.74)	4 (2.97)	0.33 (0.07–1.36)
Soil	3 (0.47)	1 (0.58)	2 (0.42)	1.51 (0.12–5.15)
Household location
Urban	558 (86.65)	141 (81.98)	417 (88.35)	1.00
Rural	86 (13.35)	31 (18.02)	55 (11.65)	1.57 (0.79–2.84)
Household proximity
Very near	486 (75.47)	120 (69.77)	366 (77.54)	1.00	1.00
Near	148 (22.98)	46 (26.74)	102 (21.61)	1.34 (0.82–2.09)	1.29 (0.78–2.09)
Scattered	10 (1.55)	6 (3.49)	4 (0.85)	3.34 (1.12–5.43)	3.87 (1.12–8.66)
Characteristics of peri-domiciliary area
Vegetation
Bushes	458 (75.78)	137 (79.65)	3.41 (74.36)	1.50 (0.96–2.28)
Trees	461 (71.58)	123 (71.51)	338 (71.61)	1.09 (0.72–1.64)
Grasses	96 (14.91)	25 (14.53)	71 (15.04)	1.42 (0.70–2.65)
Presence of wild animals in peri-domiciliary area
Opossum	124 (19.25)	30 (17.44)	94 (19.92)	0.82 (0.49–1.33)
Wild animals	70 (10.87)	15 (8.72)	55 (11.65)	0.75 (0.40–1.34)
Presence of domestic and synanthropic animals in intra-domiciliary or peri-domiciliary area
Canines	235 (36.49)	61 (35.47)	174 (36.86)	0.91 (0.62–1.31)
Rats	267 (41.46)	64 (37.21)	203 (43.01)	0.78 (0.53–1.13)
Practices common among families in the study area
Practices to prevent tick infestation
Wearing white clothes during outdoor activity	23 (3.57)	7 (4.07)	16 (3.39)	1.03 (0.39–2.39)
Wearing long-sleeve shirts when working outdoors	615 (95.50)	165 (95.93)	450 (95.34)	0.99 (0.44–2.09)
Tick removal after outdoor activity	19 (295)	3 (1.74)	16 (3.39)	0.67 (0.18–2.15)
Forest fragmentation and deforestation	19 (2.95)	4 (2.33)	15 (3.18)	0.62 (0.16–2.11)
Roden elimination measures	38 (5.90)	11 (6.40)	27 (5.72)	1.15 (0.49–2.44)
Hunting practices	6 (0.93)	2 (1.16)	4 (0.85)	1.84 (0.16–5.18)

CI = confidence interval; IQR = interquartile range; PR = prevalence ratio.

Outdoor occupations, age (in years), the presence of rodents in the intra-domiciliary area, household location, household proximity, “tapias” or mud walls (partially or entirely), and the presence of bushes in the peri-domiciliary were included in the multivariable analysis of rickettsial seropositivity. The final multivariable model showed that male sex (adjusted prevalence ratio [PRa], 1.67; 95% CI, 1.19–2.32) and age (evaluated using 5-year intervals) (PRa, 1.05; 95% CI, 1.01–1.09) were risk markers, and household proximity (PRa _{scattered versus very near}, 3.87; 95% CI, 1.12–8.66) was the main risk factor for seropositivity. Importantly, sex confounded the association between outdoor occupation and SFGR seropositivity (PRa _{outdoor occupation}, 1.46; 95% CI, 1.04–2.01; PRa _{outdoor occupation adjust by sex}, 1.13; 95% CI, 0.75–1.67) (Table 1).

Tick infestation in households and associated factors.

We studied 284 of the 286 households (data of two households were missing). Most were located in urban areas (85.92%; 244/284). The prevalence of tick infestation in households was 27.81% (79/284) (95% CI, 21.93–34.80). The proportion of tick infestation varied among neighborhoods and the village from 11.11% (3/27) to 45.83% (11/24).

The households infested with ticks had partially or entirely tiled roofs (5.06%; 4/79), fiber cement roofs (53.16%; 42/79), and dirt floors (6.33%; 5/79). Canines were present in peri -domiciliary and intra-domiciliary areas in 69.62% (55/79) of the houses. Table 2 shows additional characteristics related to household materials, the social practices of participants, vegetation, and wild, domestic, and synanthropic animals in the peri-domiciliary area. A total of 40 ticks were collected from 11 canines. All the tick specimens were Rhipicephalus sanguineus (21 females and 19 males), and all were negative for the detection of rickettsial DNA.

Table 2.

Descriptive, bivariate, and multivariable mixed effects complementary log-log regression analysis of tick infestation in households

Variables	Total, n = 284, n (%)	Infested, n = 79, n (%)	Noninfested, n = 205, n (%)	Crude PR (95% CI)	Adjusted PR (95% CI)
Household characteristics
Roof material
Zinc	104 (36.62)	25 (31.65)	79 (38.54)	0.78 (0.48–0.81)
Fiber cement (Eternit)	120 (42.25)	42 (53.16)	78 (38.05)	1.55 (1.07–2.17)	1.76 (1.07–2.83)
Wood	2 (0.70)	1 (1.27)	1 (0.49)	0.52 (0.30 –3.59)
Tile	16 (5.63)	4 (5.06)	12 (5.85)	0.92 (0.33–2.14)
Cement	49 (17.25)	10 (12.66)	39 (19.02)	0.69 (0.37–0.80)
Floor material
Dirt soil	8 (2.82)	5 (6.33)	3 (1.46)	2.33 (1.17–3.43)	5.87 (2.28–10.31)
Cement	217 (76.41)	61 (77.22)	156 (76.10)	1.03 (0.64–1.60)
Wood	3 (1.06)	1 (1.27)	2 (0.98)	1.21 (0.19–3.43)
Tile	77 (27.11)	22 (27.85)	55 (26.83)	1.04 (0.67–1.57)
Wall material
Wood	3 (1.06)	1 (1.27)	2 (0.98)	1.18 (0.18–3.42)
Blocks	273 (96.13)	78 (98.73)	195 (95.12)	3.14 (0.49–10.03)
Tapia	9 (3.17)	1 (1.27)	8 (3.90)	0.39 (0.06–2.02)
Housing location
Urban	244 (85.92)	68 (86.08)	176 (85.85)	1.00
Rural	40 (14.08)	11 (13.92)	29 (14.15)	0.94 (0.47–0.53)
Distance between housing
Very near	218 (76.76)	59 (74.68)	159 (77.56)	1.00
Near	61 (21.48)	17 (21.52)	44 (21.46)	1.02 (0.62–1.60)
Disperse	5 (1.76)	3 (3.809	2 (0.98)	2.21 (0.87–3.53)
Characteristics of peri-domiciliary area
Vegetation
Bush	203 (71.48)	55 (69.62)	148 (72.20)	0.90 (0.58–1.36)
Trees	192 (67.61)	54 (68.35)	138 (69.32)	1.03 (0.67–2.82)
Grass	35 (12.32)	11 (13.92)	24 (11.71)	1.14 (0.63–1.89)
Presence of wild animals in peri-domiciliary area
Opossum	50 (17.61)	13 (16.46)	37 (18.05)	0.92 (0.53–1.50)
Wild animals	31 (10.92)	11 (13.92)	20 (9.76)	1.39 (0.78–2.27)
Presence of domestic and synanthropic animals in intra-domiciliary or peri-domiciliary area
Canines	101 (35.56)	55 (69.62)	46 (22.44)	4.33 (3.01–5.77)	5.05 (3.31–7.19)
Rats	113 (39.93)	34 (43.04)	79 (38.73)	0.97 (0.88–1.08)
Practices common among families in the study area
Wearing white clothes during outdoor activity	11 (3.87)	5 (6.33)	6 (2.93)	1.67 (0.78–2.89)
Wearing long-sleeve shirts during outdoor activity	267 (94.01)	75 (94.94)	192 (93.66)	1.17 (0.46–2.52)
Tick removal after outdoor activity	8 (2.82)	3 (3.80)	5 (2.44)	1.39 (0.49–2.88)
Forest fragmentation and deforestation	6 (2.11)	3 (3.80)	3 (1.46)	2.10 (0.72–3.40)
Rodent elimination measures	15 (5.28)	1 (1.27)	14 (6.83)	0.22 (0.03–1.36)
Hunting practices	1 (0.35)	1 (1.27)	0 (0.00)

CI = confidence interval; PR = prevalence ratio.

The following variables were selected from the bivariate analysis to be included in the multivariable analysis for tick infestation in households: fiber cement roofs (partially or entirely); dirt floors (partially or entirely); wearing white clothes to perform outdoor occupations; forest fragmentation and deforestation; protection against rodents; and the presence of canines in peri-domiciliary or intra-domiciliary areas. The final multivariable model showed that dirt floors (partially or entirely) (PRa, 5.88; 95% CI, 2.28–10.31), fiber cement roofs (partially or entirely) (PRa, 1.76; 95% CI, 1.07–2.83), and the presence of canines in peri-domiciliary or intra-domiciliary areas (PRa, 5.05; 95% CI, 3.31–7.19) were risk factors for household tick infestation (Table 2).

SFGR seroincidence in humans and associated factors.

Of the 644 people included at the baseline (T0), 473 were followed-up 20 months later (T20). The cumulative SFGR seroincidence was 7.40% (35/473) (95% CI, 4.71–11.06). The median age of individuals with seroincident cases was 34.21 years (IQR, 19.41–51.02 years). Individuals with seroincident cases who had recent outdoor occupations or in the past (20.00%; 7/35) reported at least one episode of fever during the follow-up period (11.76%; 4/34), had canines in peri-domiciliary or intra-domiciliary areas (28.57%; 10/35), and reported tick infestation during the follow-up period (5.71%, 2/35). Additional sociodemographic, household, and peri-domiciliary characteristics of seroincident cases are shown in Table 3.

Table 3.

Descriptive, bivariate, and multivariable mixed effects logistic regression analysis of seroincident cases of rickettsial infection

Variables	Total, n = 473, n (%)	Positives, n = 35, n (%)	Negatives, n = 438, n (%)	Crude RR (95% CI)	Adjusted RR (95% CI)
Individuals
Occupation (outdoors)	79 (16.70)	7 (20.00)	72 (16.44)	1.42 (0.62–3.02)
Male	187 (39.53)	14 (40.00)	173 (39.50)	1.15 (0.58–2.17)
Age, years, median (IQR)	38.53 (20.10–54.99)	34.21 (19.41–51.02)	38.59 (20.10–55.29)	0.99 (0.98–1.01)
Tick infestation*	17 (3.65)	2 (5.71)	15 (3.48)	1.59 (0.38–5.18)
History of fever (past 20 months)†	106 (22.84)	4 (11.76)	102 (23.72)	0.45 (0.16–1.22)
Time of residence in the area	11.66 (4.66–21.66)	13.66 (5.66–23.66)	11.66 (4.66–21.66)	1.00 (0.98–1.03)
Household characteristics
Roof material
Zinc	194 (41.01)	13 (37.14)	81 (41.32)	1.06 (0.52–2.06)
Cement	80 (16.91)	7 (20.00)	77 (16.67)	1.12 (0.46–2.59)
Tile	18 (3.81)	0 (0.00)	18 (4.11)
Wood	4 (0.85)	0 (0.00)	4 (0.91)
Fiber cement (Eternit)	185 (39.11)	15 (42.86)	170 (38.81)	1.01 (0.48–1.97)
Floor material
Dirt soil	19 (4.02)	2 (5.71)	17 (3.28)	1.44 (0.29–5.50)
Cement	375 (79.28)	27 (77.14)	348 (79.45)	1.00 (0.43–2.19)
Wood	4 (0.85)	0 (0.00)	4 (0.91)
Tile	116 (24.52)	10 (28.57)	106 (24.20)	1.09 (0.50–2.26)
Wall material
Wood	6 (1.27)	0 (0.00)	6 (1.37)
Brick wall	458 (96.83)	35 (100.00)	423 (96.58)
Tapia	12 (2.54)	0 (0.00)	12 (2.74)
Soil
Household location
Urban	408 (86.26)	31 (88.57)	377 (86.07)	1.00
Rural	65 (13.74)	4 (11.43)	61 (13.93)	0.72 (0.23–2.03)
Household proximity
Very near	354 (74.84)	29 (82.86)	325 (74.20)	1.00
Near	112 (23.68)	5 (14.29)	107 (24.43)	0.54 (0.21–1.35)
Scattered	7 (1.48)	1 (2.86)	6 (1.37)	1.74 (0.23–7.21)
Characteristics of peri-domiciliary area
Vegetation
Bushes	359 (75.90)	22 (62.86)	337 (76.94)	0.62 (0.30–1.23)
Trees	340 (71.88)	19 (54.29)	321 (73.29)	0.54 (0.27–1.05)	0.49 (0.29–0.94)
Grasses	70 (14.80)	2 (5.71)	68 (15.53)	0.32 (0.07–1.33)
Presence of wild animals in peri-domiciliary area
Opossum	92 (19.45)	7 (20.00)	85 (19.41)	0.94 (0.39–2.16)
Wild animals	46 (9.73)	6 (17.14)	40 (9.13)	2.28 (0.86–5.26)	2.46 (1.06–4.72)
Presence of domestic and synanthropic animals in intra-domiciliary area
Canines	183 (38.69)	10 (28.57)	173 (39.50)	0.69 (0.39–1.18)
Rats	193 (40.89)	14 (40.00)	179 (40.96)	0.94 (0.52–1.66)
Practices common among families in study area
Practices to prevent tick infestation
Wearing white clothes during outdoor activity	15 (3.17)	1 (2.86)	14 (3.20)	0.94 (0.11–4.92)
Wearing long-sleeve shirts during outdoor activity	457 (96.62)	33 (94.29)	424 (96.80)	0.61 (0.12–2.50)
Tick removal after outdoor activity	15 (3.17)	1 (2.86)	14 (3.20)	2.65 (0.72–7.16)
Forest fragmentation and deforestation	16 (3.38)	0 (0.00)	16 (3.65)	1.59 (0.31–5.94)
Rodent elimination measures	26 (5.50)	2 (5.71)	24 (5.48)	0.89 (4.48–7.72)
Practices of hunting	5 (1.06)	0 (0.00)	5 (1.14)

CI = confidence interval; IQR = interquartile range; RR = relative risk.

^*Seven had missing data.

^†Nine had missing data.

Levels of antibody titers in seroincident cases were similar for several rickettsiae antigens (Rickettsia, R. amblyommatis, R. parkeri, and R. bellii). However, the highest titers were detected against Rickettsia and R. amblyommatis (Table 4). In addition to the 35 seroincident cases, we found 40 patients with equal IgG titers at T0 and T20 (128), 11 participants with an increase of one dilution of IgG titers at T20 (128 to 256, n = 10; 256 to 512, n = 1), four participants with a decrease of one dilution in IgG titers at T20 (256 to 128, n = 3; 512 to 256, n = 1), and one participant with a four-fold decrease in IgG titers (1024 to 128); 54 participants with positive cases (128) had negative cases at T20.

Table 4.

IgG titers of SFGR seroincident cases at T0 and T20

Seroincident cases	Titers at T0*	Titers at T20
		R. rickettsii	R. amblyommatis	R. parkeri	R. bellii
55	Negative	256	256	Negative at 128	Negative at 128
99	Negative	128	256	128	Negative at 128
110	Negative	128	128	128	Negative at 128
113	Negative	256	256	128	Negative at 128
115	128	1,024	512	Negative at 256	Negative at 128
117	Negative	128	Negative at 128	ND	ND
123	Negative	128	128	Negative at 128	Negative at 128
127	Negative	128	128	Negative at 128	Negative at 128
129	Negative	128	128	Negative at 128	Negative at 128
167	Negative	128	256	128	Negative at 128
232	Negative	256	256	128	128
233	Negative	128	128	128	Negative at 128
266	Negative	256	256	128	Negative at 128
280	Negative	256	256	128	Negative at 128
302	128	512	512	Negative at 256	Negative at 256
308	Negative	128	128	128	Negative at 128
318	Negative	256	256	Negative at 128	256
320	Negative	256	256	128	128
377	Negative	512	256	Negative at 128	Negative at 256
390	Negative	128	128	128	Negative at 128
397	Negative	128	ND	ND	ND
398	Negative	128	256	128	128
399	Negative	256	256	128	Negative at 128
418	Negative	128	128	128	Negative at 128
442	Negative	128	Negative at 256	Negative at 128	Negative at 128
462	Negative	256	256	128	Negative at 128
499	Negative	128	256	Negative at 128	Negative at 128
509	Negative	512	512	256	256
511	Negative	256	256	128	Negative at 128
520	128	1,024	256	Negative at 128	512
522	128	512	512	Negative at 256	256
584	Negative	128	256	128	Negative at 128
594	Negative	256	256	128	Negative at 128
610	Negative	512	512	512	512
636	Negative	1,024	1,024	Negative at 512	Negative at 512

IgG = immunoglobulin G; SFGR = spotted fever group rickettsiae; T0 = baseline; T20 = follow-up at 20 months.

^*Titers detected using R. rickettsii antigen.

The variables included in the multivariable analysis of SFGR seroincident cases were the presence of bushes, trees, and grasses in the peri-domiciliary area, the presence of wild animals (opossum, iguanas, snakes, and, lizards), the practice of tick removal after working outdoors, and at least one episode of fever during the follow-up period. The final model showed that the presence of wild animals in the peri-domiciliary area was a risk factor (adjusted relative risk [RRa], 2.46; 95% CI, 1.06–4.72), and that the presence of trees in the peri-domiciliary area was a protective factor for SFGR seroincident cases (RRa, 0.47; 95% CI, 0.22–0.94) (Table 3).

Rickettsiae seropositivity in canines and opossums.

The study included 73 canines; of these, 26 had seropositive cases (35.62%). Among the most frequent seropositive breeds were the mixed breed (16.43%; 12/73), Labrador retriever (4.09%; 3/73), Pitbull (4.09%; 3/73) and Fino Colombiano (2.74%; 2/73). The proportion of seropositive canines in rural areas was 30.00% (3/10) and 36.51% (23/63) in urban areas. Overall antibody titers in seropositive canines were highest against R. rickettsia (Table 5).

Table 5.

IgG antibody titers against rickettsial antigens in seropositive canines and opossums

	R. rickettsii	R. amblyommatis	R. parkeri	R. rhipicephali	R. bellii
Canines
Springer Spaniel	2,048	512	Negative at 256	ND	ND
Labrador Retriever	8,172	2,048	1,024	ND	ND
French Poodle	256	1,024	Negative at 512	ND	ND
Fino Colombiano	4,096	256	512	ND	ND
Opossums
Male 1	4,096	1,024	1,024	Negative at 1,024	Negative at 1,024
Male 2	256	256	256	Negative at 128	2,048

IgG = immunoglobulin G; ND = not done.

This study also included six opossums, Didelphis marsupialis (four males and two females with offspring); all were captured in the rural area of Uramita near to the site where the lethal case of rickettsiosis was reported. All the opossums were seropositive against rickettsiae (6/6; 100%). The potential rickettsial species in contact with opossums was identified in two individuals: one male was exposed to R. rickettsia and another male was exposed to R. bellii (Table 5). All clot samples of opossums were negative for SFGR-PCR.

DISCUSSION

During this study, the SFGR seroprevalence was 26.70%, and the factors associated with seropositivity were male sex, age (in years), and household proximity. The prevalence of tick infestation was 27.78%, and the factors associated with tick infestation were dirt floors, fiber cement roofs, and the presence of canines in peri-domiciliary or intra domiciliary areas. The estimated seroincidence was 7.39%, and the factors associated with seroincident cases were the presence of wild animals and the presence of trees in the peri-domiciliary area. These results are consistent with previous findings in the Urabá region,^10,20,21 where two deadly outbreaks of rickettsiosis were reported in 2006 and 2008. Conversely, additional factors that increase the risk of transmission, such as land use, low species diversity, and social practices (use of insect repellents and wearing white clothes to detect tick infestation),²² were not observed during this work.

Surprisingly, the SFGR seroprevalence was similar to estimates for the Urabá region (25.62%; 95% CI, 22.11–29.12), as were associated factors such as age (in years), male sex, and the confounding relationship between sex and occupation.¹⁰

Scattered households were associated with seropositivity in the municipality of Uramita. This type of housing layout is found mainly in rural areas, where there could be an increased risk of infection because of the frequent interaction among reservoirs, vectors, pathogens, and humans, as was previously reported in the Urabá region.¹⁰ Accordingly, the patient who died of rickettsiosis during the outbreak of 2014 lived in one of the rural areas included in this study. However, no new cases have been reported in this area since that outbreak. Additional characteristics of rickettsial endemic and nonendemic zones are also found in Uramita, such as urbanization, sporadic cases of disease, potential vectors infesting humans and domestic animals, and lack of epidemiological surveillance.^23–25

Households were infested with R. sanguineus, a tick commonly found parasitizing canines in urban areas. Similarly, 94.23% (98/104) of households with fiber cement roofs were located in this area, suggesting a high risk of infestation with R. sanguineus s.l. because of its link with urban zones.^9,26 Additionally, households with dirt floors were at increased risk of tick infestation, probably because eggs or different stages of ticks could not be eliminated by detergents or products used for floor cleaning, and ticks could have a greater chance of remaining in dirt than in other types of floors. Furthermore, 87.50% (7/8) of households with dirt floors were located in the urban area, which is also a potential risk for R. sanguineus s.l. infestation.

Labruna²⁷ described R. sanguineus as a “… tick with nesting habits spending its life’s stages in rooms or resting sites of its hosts. In the case of canines, resting sites can be the dog’s house, a room of a house or the back yard.” The free stages of this species (ecdysis, posture, and egg incubation) persist in grooves or holes in walls and ceilings to survive in the environment; they also detach from its hosts (canines) inside houses or shelters to continue the cycle in the environment and to find a new host.²⁷

Unlike this study, in the Urabá Antioqueño, domestic canines are infested with A. ovale and A. cajennense s.l. in addition to R. sanguineus.^10,13 The rural areas of Urabá are characterized by greater household dispersion, crops in peri-domiciliary areas (cassava, yam, corn, rice, banana), and different house materials, such as wooden walls and thatched roofs.¹⁰ These characteristics are less frequently found in Uramita.

Although all collected R. sanguineus s.l. were negative for SFGR infection, this species is a recognized vector of R. rickettsia in northern Mexico and southern North America.^9,22 Consequently, canines are important links in rickettsial transmission, not only because they are the main host of R. sanguineus but also because they are potential amplifying hosts and can maintain a constant load of ectoparasites in peri-domiciliary and intra-domiciliary areas, which is a phenomenon that increases the risk of infection in humans.^28–30

During the present study, the seropositivity rate for dogs was 35.61%, indicating that these animals are in contact with SFGR. Similarly, it has been reported that areas where domestic animals (canines and horses) had seropositivity rates of 20% to 40% or > 40% are associated with higher human seropositivity.¹⁰ Therefore, canines and horses can be good sentinels of rickettsial infection in particular areas.³¹ Also, seropositivity in opossums was highly similar to that in Brazil, Panamá, and northwestern Colombia, where seropositivity rates were 43%, 100%, and 89% respectively.^13,32,33

The estimated seroincidence in Uramita was 7.39% (95% CI, 5.10–10.40), which is very close to the seroincidence of the Urabá region of 6.23% (95% CI, 3.67–9.78).³⁴ In this study, seroincident cases were likely caused by rickettsiae of low pathogenicity, such as R. amblyommatis (Table 4), which has the highest titer corresponding to R. rickettsia and R. amblyommatis. Because fatal or severe cases were not reported during the study period, seroincident cases were probably caused by R. amblyommatis. Associated factors in this study were the presence of wild animals (RRa, 2.46; 95% CI, 1.06–4.72) and the presence of trees (RRa, 0.47; 95% CI, 0.23–0.94) in the peri-domiciliary area. Fever during follow-up (RRa, 4.26; 95% CI, 1.15–9.31) and the practice of removing ticks after returning from the field (RRa, 0.26; 95% CI, 0.08–0.84) were identified as risk factors in the Urabá region.³⁴

It is important to recognize the two infection cycles of SFGR (urban and rural) and the mammalian hosts and tick species intervening in those cycles. Accordingly, epidemiological studies of rickettsiae should also focus on all of the links in the cycle to incorporate areas, time, and individuals at risk. In the municipality of Uramita, it was clear that humans and canines were in contact with SFGR. Although R. sanguineus s.l. is a potential vector of infection, ticks collected in this study were not infected with Rickettsia. These findings did not rule out the possibility that this species is a potential vector in the area because other tick species such as Amblyomma patinoi (potential vector in Colombia)³⁵ could not be found in rural zones of Uramita. It is also possible that there were other mammalian hosts, especially wild hosts, acting as potential bacteria amplifiers in the study region.^7,13

The design and implementation of an epidemiological surveillance program for emerging and re-emerging diseases must be a priority in Colombia, especially in the areas most affected by those diseases. One of the main problems with accurately diagnosing rickettsiosis is that febrile syndromes are frequently classified as other diseases commonly found in the region (dengue, malaria, etc.). Without a confirmed diagnosis or treatment, this could lead to an increased risk of fatal outcomes. To avoid fatalities caused by rickettsial agents, it is important to draw the attention of public health authorities and healthcare personnel to the diagnosis and timely treatment using doxycycline. The design of the epidemiological program should include assessing acute-phase diagnostic tools, evaluating the management and treatment of cases, and controlling factors that trigger the disease. Such evaluations can be designed through mixed approaches based on the implementation of research in the communities most affected by the disease. Training programs for neglected diseases occurring in the country (clinical findings, differential diagnoses, treatments, and therapeutic management), such as SFR, need to be implemented for healthcare personnel (physicians, nurses, nursing assistants, microbiologists, and others) in the regional and central hospitals of the country. The ability of the program to successfully establish efficient approaches to prevent fatal outcomes will depend on transdisciplinary views and the participation of social, basic, and biomedical professionals.

Finally, according to national and international researchers of rickettsiosis, the most important aspects of the study of tick-borne diseases depend on direct efforts toward research focused on the following: determining the serological status of individuals living in potential areas at risk for transmission to identify exposed and nonexposed populations and the determinants of infection; understanding the tick species infesting individuals and households in potential areas of rickettsial transmission and estimating the proportion of ticks infected with rickettsiales agents; developing acute, sensitive, and specific diagnostic tests; conducting longitudinal studies to determine the incidence of rickettsiosis, its geographical distribution, seasonality of clinical cases, socioeconomic risk factors, clinical manifestations, and severity of the disease; determining the pathophysiology of rickettsiosis, the complications and mechanisms of tissue damage, the identification of biomarkers of disease severity, virulence factors in circulating Rickettsia strains, and immunological protection mechanisms against rickettsiae; and evaluating the environmental conditions related to the transmission of rickettsial infection using risk and predictive maps to propose targeted interventions focused on key drivers of infection or disease.

The main limitation of this seroprevalence study was the failure to reach 100% coverage of the individuals selected for the sampling. In addition, household level variables (those shared by all inhabitants of households) need to be confirmed by future studies through questionnaires with particular attention to age. Nonetheless, we are aware that customs and habits are common in homes and regions; in fact, people share risk perceptions of healthcare and farming based on the region where they have been raised or where they have lived. An additional limitation is the small sample size in the rural area compared to the urban area, which may have affected the magnitude and precision of estimates. During the tick infestation study, an important limitation was the subjective measurement of tick infestation in homes. However, this subjective measurement has allowed meaningful progress in the recognition of areas at risk for infestation with vectors of Chagas disease.¹² Additional objective measures, such as tick quantification during a given period of time, would help to validate these findings. During this prospective study, the main limitation was the loss of participants to follow-up. A loss of participants can introduce selection bias in the estimates of cumulative incidence and associated factors. Additionally, the sensitivity and specificity of IFA need to be considered to provide more accurate estimates. Finally, information biases are likely present in the study, such as memory biases associated with the information regarding episodes of fever and tick infestation during the 20-month follow-up.

CONCLUSIONS

The current prospective study addressed the potential problem of SFGR transmission in the municipality of Uramita, where lethal outbreaks of rickettsiosis have occurred. The estimated SFGR seroprevalence was 26.70%, the prevalence of tick infestation in households was 27.78%, and the SFGR seroincidence was 7.32%. Different factors were associated with seroprevalent cases (male sex, quantitative age, household proximity), tick infestation in households (dirt floors, fiber cement roofs, the presence of canines in peri-domiciliary or intra-domiciliary areas), and seroincident cases (the presence of wild animals and trees in peri-domiciliary areas). Finally, seropositivity against rickettsiae in canines was 35.62%, and that in opossums was 100%. The identification of factors associated with seroprevalence and seroincidence will help us understand the exposure routes to this agent and implement efficient surveillance and control programs for SFR in Colombia.

Note: Supplemental materials appear at www.ajtmh.org.