Description of study area
Chickens for this study were purchased from local markets of Ambo, Holeta, Guder, Ijaji and Dire Inchini districts of the West Shewa Zone, Oromia Regional State of Ethiopia from January 2016 to April 2017. Ambo is the administrative centre of the Zone, which is located at 114 Kms West of Addis Ababa. The altitude of Ambo is midland. Holeta and Dire Inchini are located 70 Kms East and 40 Kms southwest of Ambo, respectively and both are in highland altitude range. Guder and Ijaji are located 15 Km and 80 Km West of Ambo and both have tropical climate. The chicken population of each district is approximately between 350, 000 to 500,000 .
Study animals and their management
The study animals were backyard chickens that are kept under extensive management system, where chicken scavenge their feed the whole day with a limited supplement and often share same house with humans or other livestock. In this system, chickens are not vaccinated and veterinary service is not well developed. In this study, apparently healthy and clinically sick and culled chickens were purchased from local markets. Chickens of both sexes and local and hybrid breeds managed under backyard system were included. Chickens were categorized as young (≤6 months) and adult (> 6 months) based on their age  and as clinically sick and apparently healthy based on presence or absence of clinical signs of diseases.
Following purchase, chickens were transported in a cage with adequate space and ventilation without exposing to extreme weather condition. Sick and dead chickens were transported separately and slaughtered immediately. Following arrival, chickens were kept in Ambo University in a house with adequate living space and ventilation for a maximum of one day before slaughter. Feed (wheat and cracked corn) and clean water was ad libitum.
Study design, sampling technique and sample collection
A cross-sectional study design was used and a total of 191 chickens were selected consisting of apparently healthy (n = 95) and clinically sick or dead (n = 96). Apparently healthy chickens were selected randomly from local markets, while clinically sick chickens (chickens with diarrhoea, loose appetite, depression etc...) were purchased purposively from traders and farmers. All chickens were physically examined for their health status and subjected to post-mortem examination. Cervical dislocation was used to euthanize chickens in a humane manner. The carcasses were promptly necropsied according to the standard procedures described by Lowenstine . During necropsy, a total of 694 visceral organ samples of liver (n = 191), spleen (n = 191), kidney (n = 191) and ovaries (n = 121) were sampled. About 25 g of each organ sample was collected from the internal portion aseptically in sterile plastic bag (Falconpack, UAE). Samples were kept at + 4 °C for a maximum of 24 until culturing. Bacteriological work was done in the Veterinary Microbiology Laboratory of Ambo University.
Isolation and identification of E. coli
Isolation of E. coli was performed using standard bacteriological methods . Organ samples were crushed by gentle maceration, mixed separately with buffered peptone water (BPW) and incubated at 37 °C for overnight. A loopful of the culture suspension was streaked onto MacConkey agar (HiMedia, Pvt. Ltd., India) and incubated for 24 h at 37 °C aerobically. The next day those pink coloured presumptive E. coli colonies were sub-cultured onto nutrient agar to get a pure colony, followed by sub-culture on Eosin Methylene Blue (EMB) agar (HiMedia, Pvt. Ltd., India). Colonies with metallic green sheen on EMB were later characterized microscopically using Gram’s stain. Putative E. coli colonies were then transferred onto nutrient agar for further identification using biochemical tests. Triple sugar iron (TSI) agar (HiMedia, Pvt. Ltd., India) was used for further characterization. Observation of yellow slant, yellow butt, presence of gas bubbles, and absence of black precipitate in the butt was considered as potentially E. coli isolate. Then the isolates were subjected to different biochemical tests such as indole production, methyl-red, Voges- Proskauer, citrate utilization (IMViC) and motility tests as per Quinn et al. . E. coli ATCC 35218 (obtained from Ethiopian public health institute) was used as a reference organism.
Antimicrobial susceptibility test
The antimicrobial susceptibility testing of E. coli isolates was conducted using Kirby- Bauer disc diffusion method on Mueller-Hinton agar (HiMedia, Pvt. Ltd., India) according to the guidelines of the Clinical and Laboratory Standards Institute . All E. coli isolates were evaluated for antimicrobial susceptibility using 16 antimicrobials (9 antimicrobial classes) commonly used in veterinary and public health sectors in Ethiopia. Accordingly, a McFarland 0.5 standardized suspension of the bacteria in tryptone soya broth (HiMedia, Pvt. Ltd., India) was prepared and incubated for 6–8 h and using sterile cotton swab streaked over the entire surface of Mueller-Hinton agar. A ring of discs containing known concentrations of each antimicrobial drug was then placed onto the inoculum surface using disc dispenser, gently pressed with the point of the forceps for ensuring complete contact with the agar surface and incubated at 37 °C aerobically for 16–18 h. Clear zones of bacterial growth inhibition were measured in mm using a measuring calliper. The antimicrobials and their concentrations used for the susceptibility testing were streptomycin (10 μg), kanamycin (30 μg), gentamicin (30μg), amikacin (30 μg), amoxicillin (20 μg), cloxcillin (5 μg), cefuroxime, ceftazidime (30 μg), cefotaxime (30 μg), chloramphenicol (30μg), ciprofloxacin (5μg), nalidixic acid (30μg), nitrofurantoin (10 μg), tetracycline (30 μg), sulfamethoxazole-trimethoprim (1.25/23.75μg), and norfloxacin (10 μg) (Oxoid Ltd., Cambridge, UK). E. coli ATCC 35218 which is susceptible to all the drugs was used as a quality control. Finally, the findings were recorded as susceptible, intermediate, and resistant according to Clinical and Laboratory Standards Institute  break points.
Data management and analysis
Data collected from questionnaire survey and laboratory study were entered in to Microsoft Excel () Spread sheet and analysed using STATA version 11.0 for windows (Stata corp. College Station, TX, USA). Descriptive statistics was utilized to summarize the data using percentages. The prevalence of E. coli with respect to district, sex, age, and season, health status, and diarrhoea were computed by dividing the number of positive chickens by the number of chickens examined and for organ level prevalence the number of positive organs was divided to the total number of organs examined. The association of potential risk factors with E. coli prevalence was analysed using logistic regression. Stratification method was used for those variables showing significant association to see any difference between the crude and adjusted results. Then, after further checking for collinearity, variables with P-value less than 0.25 during univariable analysis were further analysed using a multivariable logistic regression model. Odds ratio was used to see degree of association and confidence level was held at 95% and significance was at P < 0.05. The percentages of antimicrobial resistance of each pattern (Susceptible, Intermediate and Resistance) were calculated.