Igabi Local Government Area (LGA) with headquarters at Turunku, is one of the 23 LGAs in Kaduna State, North-west geopolitical zone, Nigeria. It has a population of 557,624 and is sub-divided into five urban districts, which are, Afaka, Birnin-yero, Kwarau, Rigachikun and Rigasa, and seven rural districts namely, Fanshanu, Gwaraji, Igabi, Kerawa, Sabo-Birnin, Turunku, and Zangon Aya. The LGA is located 650 m above sea level, between latitude 10o 47/ 0// N and longitude 7o 46/ 0// E in the tropical Sahel to Sudan Savannah with annual rainfall varying from 1000–1500 mm and highest precipitation of 72% in August. The rainy season and period of high malaria transmission is usually from June to October. The annual mean temperature is 34oC. This can rise to 41oC at the peak of the dry season in April and drop to as low as 12oC in January during severe harmattan. The prevalence of malaria cases diagnosed microscopically among under-five children in Kaduna state is 36.7%  This study employed a cross-sectional design.
Study Design And Population
This was a part of a larger study that has been published and addressed the characteristics of women of child bearing age (WCBA) associated with Long-lasting insecticidal nets (LLIN) ownership and utilization . Moreover, the current analysis focused on rural - urban disparity in fever prevalence, mothers’ malaria knowledge, and associated factors with delayed care seeking for fever in U5 children. Mothers and their under-five children with at least 12 months of residence in the selected households in the community were included. Anyone with cognitive deficits, or any chronic and debilitating illness that hinders effective participation in an interview was excluded.
Sample Size Determination
Using the formula, n = z2pq/d2 at 95% confidence interval (1.96), the percentage for children with fever for whom advice or treatment was sought in North-Western Nigeria is 63.6%  with a confidence limit of +/- 5%, and a non-response rate of 10%. To adjust the required sample size for cluster sampling design, a correction factor of 1.5 design effect was used as a less heterogenous communities in this district was assumed. Therefore, the calculated minimum sample size was 615. However, a minimum sample size of 630 was used to accommodate the sample size for the WCBA study .
Participants were selected using multi-stage sampling technique (Fig. 1).
The sampling technique has been described elsewhere . In Stage 1, Igabi LGA was stratified into rural and urban wards and ten wards were randomly selected. These included six rural wards - Fanshanu, Gwaraji, Igabi, Sabo-Birnin, Turunku, Zangon Aya and four urban wards - Afaka, Birnin-yero, Rigachikun, and Rigasa. For the selection of a cluster or village/settlement in stage 2, we used the household enumeration data generated by WHO/UNICEF during micro-planning for the mass long-lasting insecticidal nets (LLIN) campaign in these settlements as the sampling frame. With a sample size of 630, to improve the validity and precision of estimates and considering a wide community representation (population variance), using a probability proportional to size; 63 villages or clusters (i.e., 30 rural and 33 urban) were randomly selected with a cluster size of 10. Using the list of the households in each village/settlement generated during micro planning for mass LLIN distribution, ten households were systematically selected from each village/settlement at stage 3. One mother each was interviewed in a household. In a selected household with multiple eligible mothers of under-five children, one was randomly selected by the ballot and recruited into the study. The last childbirth order of the eligible mothers was recruited into the study and if the last child was a set of twins, one of the twins was randomly selected by balloting.
Data was collected by ten trained female community health extension workers and nurses who speak English and Hausa languages fluently and reside within the district from September to October 2015 as described elsewhere . They were supervised by five undergraduate medical students. The data collection tool was a pretested, structured questionnaire adapted from the Malaria Indicator Survey [24, 25] and other literature [26, 27]. This was used to collect information on demographic profile, mother’s knowledge of malaria, history of fever onset in under-five children and mothers’ fever treatment seeking behaviour within 48 hours of fever onset. The malaria knowledge items were questions from the literature, malaria indicator surveys and demographic health surveys. The questionnaire’s face and content validity, item’s accuracy, relevance, and clarity has been described elsewhere . The knowledge of malaria was assessed using six thematic areas including cause of malaria, mode of malaria transmission, mosquito feeding time, malaria symptoms, malaria diagnosis and malaria prevention. Participants who correctly responded that mosquito bites caused malaria, malaria transmissions to humans is through the bite of mosquito, malaria mosquito feeding time was night-time, malaria symptoms are fever and any other symptoms, and malaria preventions by using either the Long Lasting Insecticidal Nets, Indoor Residual Spray or mosquito repellent coils were given “1” and incorrect responses scored “0”. The individual score was calculated by finding a percentage of total score obtained from maximum allowable score of 6 and knowledge of malaria was categorized using percentile scores. Good knowledge of malaria, if final score falls at 75th percentile or more, score between 50th to 74th percentile as average knowledge and poor knowledge if score was < 50th percentile.
The outcome or dependent variables for this study was mothers’ care/treatment seeking advice for fever within 48 hours of onset. Independent and explanatory variables were under-five children and respondents’ characteristics and mother’s knowledge of malaria items. Fever was assumed by mothers after tactile palpation of the children’s skin and felt the hotness beyond normal. However, any fever episode in under-five children two weeks prior to the survey was taken as history of fever from which questions were asked from the mothers to know if advice or treatment was sought. Delayed care seeking for fever was defined as inability of the mothers to seek care for U5 children withing 48 hours of fever onset .
Data processing and analysis
Data were entered and analysed using Epi-Info version 7 statistical software. We calculated frequency and proportions for socio-demographic characteristics, knowledge of malaria, presence of fever in under-five children in last two weeks, and mothers seeking advice or treatment for fever within 48 hours of onset. We calculated median (range) for continuous descriptive variables, frequency and proportions for categorical variables. Chi squared test was used to test for association between dependent outcome and independent categorical variables; results were presented in odd ratios at 95% confidence interval (C.I). To understand the associations between explanatory variables and delayed care seeking for fever, a rural-urban stratification analysis was done. Any associated factors with p value ≤ 0.2 were selected for upward loading into multivariable logistic regression model to identify predictors of delayed care seeking for fever by the mothers. Results of all statistical analyses were considered significant at p-value of < 0.05.