Research web pages
4 varied counties representing urban, peri-urban, rural blended farmers, and rural pastoralist communities in Kenya (Nakuru, Kakamega, Siaya, and Marsabit respectively) had been purposively picked to take part in this study (Fig. 1). These four counties are amongst the 8 in Kenya where by sentinel surveillance internet sites for intense acute respiratory illness (SARI) are situated [14, 15]. In 2009, the approximated populace of Nakuru County was 1,603,325 of which 54% lived in city options . Nakuru county referral clinic (NCRH) is the most important referral community hospital for urban and peri-city inhabitants of Nakuru County. Marsabit County is a semi-arid location in the northern component of Kenya, which is dominated by pastoralists who hold huge herds of cattle and camels. As of 2009, the populace of Marsabit county was 291,166 of which 78% lived in rural configurations . Marsabit CRH (MCRH), a community healthcare facility, is the major referral medical center in the county.
Kakamega CRH (KCRH), a community hospital, is the most important referral medical center for the urban and peri-city inhabitants of Kakamega County in western Kenya. As of 2009, the populace of Kakamega County was 1,660,651 of which 56% lived rurally . The population of Siaya County, also located in western Kenya was 842,304 in 2009 of which 89% lived in rural settings . Siaya CRH (SCRH) is a community medical center and serves as the key referral hospital for the county.
Examine design and sampling methods
We executed a cross-sectional survey to determine healthcare utilization patterns and frequency of ARI and serious pneumonia working with a two-phase cluster sampling process – initially sublocation then house.
Catchment regions and family choice for the survey
We identified that at the very least 80% of the SARI patients who sought healthcare at just about every surveillance clinic resided in 58 sub-locations in Nakuru, 37 in Marsabit, 121 in Kakamega and 61 in Siaya (Supplementary File) . Utilizing a record of all these sub-places, we randomly sampled (irrespective of dimensions) a subset of sub-destinations from each location for inclusion in the study: 39/58 in Nakuru, 29/37 in Marsabit, 63/121 in Kakamega and 39/61 in Siaya. When picked, the amount of homes in each and every of these sub-spots were being allotted proportionate to the population dimension (PPS) based on the 2009 census.
Employing the Geographic Info Process (GIS) program, ArcGIS® application by Esri, we produced random spatial coordinates for the number of homes that had been required from just about every of the picked sub-destinations. The survey teams found the coordinates working with handheld GPS units, and the closest domestic was picked for inclusion in the survey. Only homes with a little one aged < 5 years were included in the survey. In cases where there were no apparent households within a radius of 200 m, or there was no child aged < 5 years, the study team moved on to the next set of coordinates.
Because of the nomadic communities living in Marsabit county, a combination of random geographical coordinates (in areas where there were residential houses that could be visualized on Satellite images) and systematic sampling procedures (in rural areas with make-shift nomadic dwellings) were used to identify households to participate in the survey. For the nomadic settlements, we first generated a list of these communities and their number of households. Settlements that participated in the survey were then randomly selected. In each of the selected settlements, a sampling interval was determined by dividing its size by the number of targeted households (Supplementary File).
The sample size was powered to estimate the proportion of household respondents who had been hospitalized for an episode of pneumonia in the last year. We assumed (a) that 2.1% of the respondents would report an episode of pneumonia in the last 12 months (b) an average of 4 persons per household  (c) an estimated rate of hospitalization for pneumonia of 16%, as determined in the 2005 survey  and (d) a precision of 10%. This led to a sample size of 620 households to yield at least 52 persons with self-reported pneumonia. Further assuming a design effect of two, and allowing for a nonresponse rate of 15%, the effective sample size was determined to be 1,450 households per county.
We defined a case of ARI as those who reported two or more of the following in the last 14 days preceding the survey: cough (new or worsening of chronic cough), difficulty breathing, rapid breathing, runny nose, sore throat, but were not hospitalized or recommended for hospitalization. Severe pneumonia was defined as a participant reporting an episode of respiratory illness in the previous 12 months with (i) cough and difficulty breathing for more than two days, or (ii) a physician-diagnosis of pneumonia [3, 4, 19] and (a) hospitalization or a recommendation for hospitalization by a healthcare worker, or (b) in the case of children aged < 5 years at least one danger sign, or for those > = 5 yrs, a “limitation” in the potential to conduct regime actions. A hazard indicator for children < 5 years was defined as one of inability to breastfeed or drink, persistent vomiting, convulsions or seizures, loss of consciousness [3, 4]. For persons aged ≥ 5 years, a “limitation” in the ability to perform routine activities was assessed using a set of 5 questions (Supplementary File). Briefly, these questions were related to the performance of routine activities such as playing, walking, eating, self-grooming, lifting objects and were scored on a 3-point scale (“not limited”, “limited a little” or “limited a lot”). Those who reported that they were “limited a lot” in a at least two of the five questions were considered as severely ill and thus counted as cases of severe pneumonia.
Outpatient health facilities were defined as all sources of healthcare that did not admit patients for overnight stay and included both public and private ambulatory clinics. Inpatient providers included both public and private hospitals. We defined a household as a persons living together with a common cooking area .
Data collection and management
Using a structured survey instrument electronically loaded on a netbook (Siaya and Kakamega) or on a tablet (Nakuru and Marsabit), trained interviewers collected household and individual data and details of episodes of respiratory disease. An adult proxy was interviewed for children < 18 years, and household members who were not present at the time of interview, including residents who died within the last 12 months. For each household member reporting an episode of ARI or pneumonia, respondents were asked detailed questions about symptoms and healthcare sought, including sources of care and whether the household member was hospitalized, or a healthcare worker had recommended hospitalization. All data were stored in a password-protected SQL database at the KEMRI offices in Kisumu and Nairobi.
For this analysis, only data from the last episode of disease syndrome was included if more than one episode had been reported. Using principal component analysis, specifically using the factor effects derived from the first component of household goods, house construction material, source of water supply, source of cooking fuel and sanitation facility [20, 21], we generated the household wealth index as proxy for socioeconomic status (SES) stratified by county. The wealth index was categorized into quintiles wealthy households in this study were defined as those whose wealth index was in the fourth or fifth quintile.
All analyses were conducted using Stata 15.1 software (StataCorp. 2017. Stata Statistical Software: Release 15. College Station, TX: StataCorp LLC). Multivariable survey logistic regression methods were used in the analyses to identify factors that were independently associated with healthcare seeking behaviors for respiratory illness while accounting for the survey design. Variables that were included in each of the multivariable models assessed were site (county where data were collected), sex, age, SES, and level of education of the head of the household. Other variables that were assessed included religion, household size, childbirth order, and family member status (i.e., part of the nuclear family or other relative).