Acute diarrhoea is one of the most common diseases and causes of death in young children in sub-Saharan Africa and in other developing areas [3, 5, 25]. The present study constitutes a comprehensive survey on various enteropathogens in northern Ghana and comprises one of the largest diagnostic approaches using PCR for the detection of enteric viruses in West Africa.
One major finding is the confirmation of rotavirus as the principal cause of paediatric diarrhoea during the dry season in the north of Ghana. Its prevalence among children with acute diarrhoea (55%) is in close accordance with the figure of 53% assessed in the dry season in the near-by Kassena-Nankana district . In contrast, in southern Ghana rotavirus was detected in only 7% of symptomatic preschool children in an all-year survey . The lower prevalence in that survey and other studies [27, 28] is most likely due to the diagnostic means applied in those studies, i.e., agglutination tests and enzyme immuno assays. The PCR method used in the present study, however, is much more sensitive and specific than the assays used in the previous studies [29, 30].
The present study has several limitations. These include the cross-sectional study design, lack of respective data for the rainy season, and the classification of control children. Although the controls did not meet the criteria for acute or chronic diarrhoea  and had not received recent anti-infective treatment they rather reflect a random sample of the paediatric population in the study area than being healthy controls: approximately two thirds of the controls provided watery or loose stools. We decided not to exclude those children from the control group in order to obtain a representative picture of intestinal micro-organisms in the paediatric population and also because of logistical reasons. In fact, potential enteropathogens were detected in more than half of the controls' stool samples, possibly indicating subclinical infections. On the one hand, this high burden of intestinal infections in the paediatric population of the study area corresponds to the poor overall health status as reflected by the proportions of malaria, anaemia, nutritional deficits, and concomitant diseases in the present and preceding studies . On the other hand, the abundance of intestinal micro-organisms hampered the identification of pathogens associated with acute diarrhoea. In any case, the controls of the present study cannot be considered exclusively healthy children, which needs to be taken into account when interpreting our data.
Irrespective of these limitations, children with acute diarrhoea had about eight times higher odds of being infected with rotavirus than children in the control group. Rotavirus infection was particularly frequent in the youngest children. We did not apply a numerical score to grade the severity of episodes as proposed by Ruuska & Vesikari  because of absent or presumably unreliable data, e.g. on the number of vomiting episodes. Attempts to apply a modified, limited score produced no meaningful results. Nevertheless, in the present study, rotavirus infection was found to be associated with febrile disease and watery stools. These findings are consistent with previous results from Africa and elsewhere [8, 27, 32]. For cases of diarrhoea without proof of rotavirus or other pathogens, no associations with clinical or epidemiological features were identified. This is likely due to their comparatively small number. Considering the burden of rotavirus infection in the north of Ghana and elsewhere an early implementation of rotavirus vaccines should have first priority. It is estimated that such vaccines could prevent 5% of all childhood deaths and 40% of all deaths due to diarrhoeal diseases .
Enteric adenoviruses, serotypes 40 and 41, have gained acceptance as important causes of childhood gastrointestinal illness . In the present study, adenovirus was common, however, occurred at similar prevalence in patients (28%) and controls (32%). The high proportion of adenovirus among the latter does not necessarily exclude its pathogenic role as faecal elimination may continue for months after adequate humoral immune response . This may also be true for norovirus, which was observed in some 9% among both patients and controls. Recent data indicate that noroviruses contribute to the pathogenesis of childhood diarrhoea [19, 36]. Its prevalence in our patients corresponds to recent results from Ghana . Astrovirus was rarely identified suggesting a minor role in childhood diarrhoea in northern Ghana. This accords with findings from Botswana  and Malawi .
Protozoa and helminths were comparatively rare. The overall low prevalence of G. lamblia corresponds to its peak occurrence during rainy seasons . However, G. lamblia infected fewer patients (5%) than controls (13%). This could imply limited pathogenicity or asymptomatic carriage. The latter is consistent with studies from Brazil  and Nepal .
Bacteria were the least common potential enteropathogens. In contrast to southern Ghana  and other developing areas [28, 41, 42], the diarrheogenic bacteria Salmonella sp., Campylobacter sp. and Shigella sp. were detected only rarely in patients. An increased number of stool samples (optimally three) per patient and the use of highly sensitive PCR for identification of enteropathogenic Escherichia coli in further investigations might give a more specific picture of bacterial gastrointestinal illnesses in the region. Nevertheless, the overall limited role of bacteria in acute diarrhoea in the study area strongly argues against a widespread, uncritical use of antibiotics.
Simultaneous infections with potential enteropathogens occurred in about a quarter of patients. Rotavirus in combination with adenovirus, astrovirus, or norovirus was primarily seen. Co-infections were significantly less prevalent in control children. In an individual case of co-infection, the relative role of a single enteropathogen for the clinical symptoms cannot be determined, i.e., one enteropathogen might only be excreted while another one is causing the acute disease. Concerning enteric viruses it seems that the severity of diarrhoea does not correspond to the number of pathogens . However, the impact of viral and non-viral co-infections warrants further investigation.
Malaria is hyperendemic in the study area , and, thus, the prevalence of PCR-proven P. falciparum infection of 36% in paediatric diarrhoea patients is rather low. The abundance of home-treatment in this (peri-)urban area is one likely explanation . In a previous study at Bulpeila health centre, 15% of children with uncomplicated falciparum malaria presented with diarrhoea, among other symptoms . In the present study, symptomatic malaria was more frequent among patients without than in patients with identified enteropathogens indicating a potential causative, albeit minor role in acute childhood diarrhoea.