The overall prevalence of parasitic infections in the study population was 33.3% [31.40-35.27], the majority of which were attributable to protozoan species (prevalence = 32.5% [30.62–34.47]), helminth infections in this population being relatively rare (prevalence = 0.79% [0.469–1.249]). The prevalence of each of the ten species of parasitic organisms identified is given in Table 1.
Table 1. Prevalence of protozoan and helminth species.

The most common species was Blastocystis spp. and the rarest was the nematode T. trichiura, which was recorded in only one subject. Because helminths were so rare in this group of subjects, further analysis is focused on infections by protozoan species, but it is of interest to note that of the 18 subjects infected with helminths, 16 were under 14 years in age, and 13 of these were attributable to E. vermicularis. Only two adult subjects were diagnosed with helminths, both in their 30s and both with infections with Taenia spp.
Combined species richness (both protozoa and helminths) was 0.39 ± 0.013 species/host, and mostly attributable to the protozoan species (0.38 ± 0.013). Single species protozoan infections were the most common (27.8% of total population, and 85.6% among the protozoa infected subjects), and multiple species protozoan infections much rarer (two species = 4.1%, three species = 0.57%, and four species 0.44%, of all subjects).
Intrinsic factor affecting prevalence and species richness of combined protozoan parasites
Prevalence of combined protozoan infections varied significantly with host age (Fig. 2A; χ210 = 34.15, P < 0.001), increasing from age class 1 to age class 2 by 11.4% and then falling as host age increased to a low of 23.1% among the oldest age class. Although the fall in prevalence with increasing host age was modest (a reduction of 21.7% from age class 2 to 11), it was nevertheless significant (rs= -0.095, n = 2.277, P < 0.001). Prevalence was very similar and not significantly different between the sexes (Table 2; χ21 = 0.173, P = 0.68).
Protozoan species richness (PSR) also varied significantly between the age classes (Fig. 2B; Kruskal-Wallis test, H10 = 35.42, P < 0.001), and showed declining values with increasing host age (rs= -0.099, n = 2.277, P < 0.001), much as for prevalence. Interestingly, however despite the lower mean values in subjects in age classes 7–11, corresponding to subjects who were older than 29 years of age, eight of the 13 cases of 3-species infections occurred in these age classes. The single case of a 4-species infection was in a 21-year old male. Mean PSR was very similar in both sexes and not significantly different (Table 2; z = 0.781, P = 0.44).
Seasonal, annual and location-specific effects on prevalence and species richness of combined protozoan parasites
Table 2 shows the prevalence of combined protozoan infections in each year, season and area in which the subjects lived. Analysis of each factor in turn revealed that prevalence differed significantly between subjects living in rural and urban areas (χ21 = 26.67, P < 0.001), with prevalence among subjects from rural sites 10.2% higher than that for subjects living in urban sites.
Table 2. Prevalence of combined protozoan infections, by year, season and area.

As Table 2 shows, overall prevalence was similar in each of the years of the study varying only between 28.7% (2015) and 35.7% (2017). Prevalence was even more constant when examined by season, varying only from 30.2–33.9%. Neither of these was significant. However, when all three factors were fitted in a multifactorial model, the highest order interaction (YEAR x SEASON x AREA x INFECTION) could not be simplified (χ29 = 20.673, P = 0.014), indicating significant interactions at this level. These are illustrated in Fig. 3, which shows that whereas prevalence in most data subsets centred around 30%, unusually low prevalence was detected in 2018 in spring and summer among people living in urban areas and a very high prevalence was seen also in 2018 in the spring in rural areas.
Much the same outcome was seen when PSR was analysed by non-parametric tests: there was a highly significant difference in mean PSR between subjects from urban and rural areas (Mann-Whitney-U test, z = 5.406, P < 0.001), but none between the seasons (Kruskal-Wallis test, H3 = 2.19, P = 0.535) or between years (Kruskal-Wallis test, H3 = 5.862, P = 0.5118).
Prevalence and species richness of combined protozoan parasites in relation to sources of drinking water and ownership of animals
Subjects who relied primarily on tap water were more likely to experience infections with protozoan organisms than those who used bottled water (χ21 = 4.04, P = 0.045), but the difference in prevalence, while significant, was not marked between these two groups (Table 2). However, prevalence was more than twice as high among those who had animals compared to those who did not (Table 2; χ21 = 224.6, P < 0.001).
A similar outcome was found for PSR. The difference between subjects relying on bottled and tap water was significant (Table 2; Mann-Whitney U test, z = 2.818, P = 0.005). Mean PSR was also significantly higher among those who had animals compared with those who did not (Table 2; z = 15.52, P < 0.001).
The key factors affecting prevalence of combined protozoan infections
As a final stage to this analysis we fitted a statistical model that incorporated all the key significant factors from stages described above. This included the effects of AGE, AREA, YEAR, WATER, ANIMALS and INFECTION. The minimum sufficient model comprised five terms that included INFECTION. The first term was ANIMALS x AREA X INFECTION (χ21 = 4.81, P = 0.028) and this is illustrated in Fig. 4A, where it is clearly apparent that the possession of animals was a key determinant of prevalence. Whether living in an urban or rural area, prevalence was much higher among subjects that had animals. Although in both cases prevalence was higher among the inhabitants of rural areas, the difference between subjects from rural and urban areas was more marked in those without animals. While possession or contact with animals greatly enhanced the likelihood of infection, the difference in prevalence between subjects who worked with animals and those that did not, also varied between years (ANIMALS x YEAR x INFECTION, χ23 = 27.232, P < 0.001). However, as Table 3 shows, prevalence was always higher in each year of the study among those with animals, but the effect was not totally consistent and the difference in prevalence between the two groups varied in magnitude. The difference between areas also varied between years (Table 3; AREA x YEAR x INFECTION, χ23 = 38.53, P < 0.001), and whilst higher among rural inhabitants in three years, especially in 2018, there was little difference between subjects from these two areas in 2015, and therefore the effect of where the subjects lived was not consistent from year to year.
Table 3. Temporal variation in prevalence of combined protozoan infections in subjects who owned or looked after animals and those that had no contact with animals, those living in rural and urban regions of the province and those that relied on tap or bottled water.

While overall reliance on tap rather than bottled water resulted in a higher prevalence of protozoan infections, the difference between the two groups was dependent on host age (WATER x AGE x INFECTION, χ210 = 22.83, P = 0.011). As Fig. 4B shows the earlier finding of declining prevalence with age was mostly driven by subjects who relied on tap water and among them there was a significant negative correlation between age and prevalence of infection with combined protozoa (rs= -0.176, n = 1.201, P < 0.001). Among those that used bottled water, prevalence was much the same in all age classes and did not correlate significantly with host age (rs= -0.029, n = 1.076, P = 0.34). There was also a marked difference between years in the extent of the difference in prevalence between those that used bottled or tap water (Table 3; WATER x YEAR x INFECTION, χ23 = 14.431, P = 0.002). In two years (2015 and 2018) prevalence was higher among those relying on tap water, and in the other two prevalence was higher among bottled water users.
The key factors affecting protozoan species richness
As a final step in exploring the factors that influenced PSR, we fitted a GLM (with Poisson errors, main effects and all 2-way interactions) with all the factors (SEX, AGE, AREA, SEASON, YEAR, WATER and ANIMALS). This confirmed the significance of differences in mean PSR between age classes, although with all other factors and their 2-way interactions taken into account significance was only marginal (Wald χ210 = 18.11, P = 0.05). The difference between areas in which subjects lived was also significant (Wald χ21 = 19.11, P < 0.001), as was the hugely significant effect of ownership of animals (Wald χ21 = 119.25, P < 0.001). As in the earlier analysis the difference between the sexes was not significant (Wald χ21 = 0.051, P = 0.82), nor was that between seasons (Wald χ23 = 3.34, P = 0.34). However, with all other factors and their 2-way interactions taken into account there was no independent effect of quality of water (Wald χ21 = 1.58, P = 0.21), and the difference between years gained significance (Wald χ23 = 11.912, P = 0.008).
This analysis identified five significant interactions, which are illustrated in Fig. 5. The most marked interaction was that between AREA and YEAR (Wald χ23 = 26.212, P < 0.001). While mean PSR was generally higher among subjects living in rural compared with urban areas, when the data were broken down by year of the study, it became apparent that this effect was not consistent from year to year (Fig. 5A). In 2015 there was no difference, in 2016 and 2017, there was only a marginal difference while in 2018 there was a huge difference with mean PSR being almost four times higher among subjects in rural areas. Ownership of animals (Fig. 5C) was another key factor. While mean PSR was considerably higher in all four years of the study among subjects that owned animals, compared to those that did not, the extent of the difference between the two groups varied from year to year (Wald χ23 = 17.960, P < 0.001), but was always in the same direction. Therefore, the effect of ownership of animals showed some temporal consistency and was clearly an important risk factor predisposing people to infection with protozoan parasites. While the earlier univariate analysis indicated that people who relied on tap water had a higher mean PSR than those who used bottled water (see above), when other factors had been taken into account the difference was no longer significant, but there was a significant interaction between WATER and YEAR (Wald χ23 = 16.498, P = 0.001; Fig. 5B). Subjects relying on tap water had a higher mean PSR in 2015 and there was a marked difference 2018, but not in 2016 and 2017, so the difference between people relying on tap versus bottled water was not consistent from year to year. The two remaining significant interactions were weaker (Figs. 5D and 5E). Although there was no overall significant difference between the sexes, and this was particularly the case in summer and autumn, in spring mean PSR was higher in female subjects while in winter it was higher in males (Wald χ23 = 11.727, P = 0.008). Finally, the difference in mean PSR between subjects who lived in urban versus rural areas was greater when they relied on bottled water compared with tap water (Wald χ21 = 6.361, P = 0.012).
Prevalence and species richness of combined protozoan parasites in relation to symptoms experienced by the subjects
Perhaps as expected, prevalence of combined protozoan infections was considerably higher among subjects who showed some symptoms (79.5% [76.55–82.12]; χ21 = 849.3, P < 0.001) compared to asymptomatic subjects (14.0% [13.16–16.59]). Among the 623 subjects with symptoms, no evidence of protozoan infection was found in 128, but 18 of these subjects harboured helminth infections.
Table 4 shows the breakdown of symptoms reported by the subjects and/or confirmed at clinical inspection, and the associated prevalence of combined protozoan infections. In all symptomatic categories prevalence was much higher than in symptomless subjects, although sample size was small in some cases.
Table 4. Prevalence of combined protozoan infection among asymptomatic subjects and those showing a range of symptoms.
