Healthcare seeking behaviour for common infectious syndromes among people in three administrative regions of Johannesburg, South Africa, 2015: a cross-sectional study

Introduction Hospital-based surveillance programs only capture people presenting to facilities and may underestimate disease burden. We conducted a healthcare utilisation survey to characterise healthcare-seeking behaviour among people with common infectious syndromes in the catchment areas of two sentinel surveillance hospitals in Johannesburg, South Africa. Methods A cross-sectional survey was conducted within three regions of Johannesburg from August to November 2015. Premises were randomly selected from an enumerated list with data collected on household demographics and selected syndromes using a structured questionnaire. Fisher's exact or chi-square tests were used to determine association of characteristics among different regions. Results Of 3650 selected coordinates, 3358 were eligible dwellings and 2930 (87%) households with 9850 individuals participated. Four percent of participants (431/9850) reported influenza-like illness (ILI) in the last 30 days; equal numbers of participants (0.2%, 20/9850) reported pneumonia or tuberculosis symptoms in the last year and <1% reported diarrhoea or meningitis symptoms. Sixty eight percent (295/431) of participants who reported ILI, 75% (6/8) of children with diarrhoea and all participants who reported pneumonia (20), tuberculosis (20) or meningitis (6) sought healthcare. For all syndromes most sought care at registered healthcare providers. Of these only 10% (24/237) attended sentinel hospitals, predominantly those that lived closer to the hospitals. In contrast, of patients with meningitis, 50% (3/6) sought care at sentinel hospitals. Conclusion Patterns of seeking healthcare differed by syndrome and distance from facilities. Surveillance programs are still relevant in collecting information on infectious syndromes and reflect a proportion of the hospital's catchment area.


Introduction
In South Africa, tuberculosis, influenza and pneumonia were among the top ten causes of death in all age groups in 2014-2016 [1]. In 2016, of 456,612 total deaths reported, tuberculosis was responsible for 29,680 (6.5%), and pneumonia and influenza for 19,634 (4.3%) [1]. In children under 15 years of age, intestinal infectious diseases and central nervous system inflammatory diseases were in the top ten causes of deaths in 2016 [1]. Since 2009, the National Institute for Communicable Diseases (NICD) has conducted active, prospective, hospital-based surveillance for pneumonia in five of South Africa's nine provinces to describe its epidemiology, characterise causative pathogens and prepare for pandemics and outbreaks [2]. From 2012 the NICD conducted prospective surveillance for persons with influenza-like illness (ILI) at outpatient clinics located in the catchment areas of two pneumonia surveillance sites [3]. Standard World Health Organization case definitions were used for ILI and severe acute respiratory illness surveillance, with respiratory samples tested using a multiplex real-time reverse transcription-polymerase chain reaction [4]. The NICD also has active laboratory-based surveillance which monitors several pathogens including those causing diarrhoea and meningitis [5]. The disadvantage of facility-based surveillance programs is that they only capture information for people seeking healthcare at facilities conducting surveillance. This can lead to inaccurate interpretation of surveillance data as other people are missed [6]. South Africa has a health information system but this also captures information at a facility level. A healthcare utilisation survey (HUS) captures information on health seeking behaviour that is missed by surveillance programs [7]. Several HUSs have been conducted in different countries to complement surveillance program data [8][9][10]. As communities differ, it is important to understand healthcare seeking patterns in different settings [9]. Different individual and healthcare provider factors may influence healthcare seeking behaviour [11,12].
Knowledge of healthcare utilisation behaviour within communities is useful when developing health policies and implementing health programs. The study objective was to describe healthcare seeking behaviour for individuals with common infectious syndromes in the catchment areas of two sentinel surveillance hospitals in Johannesburg, South Africa.

Methods
Study design: we performed a cross-sectional community-based HUS using a one-stage cluster design, with households as the sampling unit, similar to previously described methods [9]. We collected data on three respiratory syndromes (tuberculosis, ILI and pneumonia), meningitis and diarrhoea in 27  in the NICD laboratory-based surveillance program [5] and were initiated as pneumonia surveillance sites in 2014 (referred to hereafter as sentinel hospitals). The survey was conducted in specific suburbs of regions A, B and C, which are catchment areas for these hospitals ( Figure 1). The chosen suburbs, based on utilisation of sentinel sites, were mainly low-income areas within these regions.
Sample size: the sample size was calculated using Epi-Info 7 statistical software. It was calculated for pneumonia, the least common respiratory endpoint investigated in this study. Calculations were based on the assumption that 50% of total pneumonia cases would seek healthcare at sentinel surveillance sites [15]. This care seeking rate with a 95% confidence interval (CI) and a 10% precision yielded a minimum sample size of 96 without considering clustering effects. A cumulative incidence of 1% for pneumonia [16]  Definitions: case definitions and recall periods for syndromes were based on previous healthcare utilisation surveys for comparability of data [6,8,9,15,21,22]. ILI was defined as a sudden onset of selfreported fever or a measured temperature of >38ºC with cough and/or sore throat experienced in the last 30 days [9]. Pneumonia was defined as a sudden onset of self-reported fever or worsening fever >38ºC, cough and difficulty breathing that lasted 2-30 days, or diagnosed by a healthcare worker in the last year [23,24]. and there was no significant difference among regions ( Table 2).
Frequency of other reported syndromes: there were six reported cases of meningitis (

Discussion
In this study, we describe health-seeking behaviour for common infectious syndromes, including three respiratory syndromes, meningitis and diarrhoea, in three regions of Johannesburg, South Africa. This study identified rates of infectious diseases similar to previous surveys [21,22], and most reported disease was mild. We identified that health seeking behaviour differed by syndrome type and region and that severity of illness usually prompted individuals to seek healthcare. Our study showed that sentinel sites were most likely to detect sicker patients from certain areas. Overall, the two sentinel surveillance hospitals were mainly utilised by participants in regions B and C whose syndromes were severe. This could have been due to ease of access as regions B and C were in close proximity to the hospitals. Even though distance to healthcare provider was not stated as a main barrier in the study, a previous study in South Africa reported distance to healthcare facility as a barrier [27]. (666/4797) [3]. Less than 1% of individuals in our study reported experiencing pneumonia or tuberculosis syndromes; these results are similar to those obtained for a South African HUS conducted in 2013 [21], but slightly lower than the 2% prevalence for pneumonia in another HUS from 2012 [22]. A lower prevalence of tuberculosis and pneumonia syndromes compared to ILI is expected as these syndromes are less common. The prevalence of reported meningitis cases was lower than reported in our previous surveys, possibly due to a more specific case definition [21,22]. In all three  [9,15]. Another study reported hospitals as point of care for pneumonia [23] and this was linked to illness severity.
Tuberculosis diagnosis and treatment in South Africa has been rolled out at primary healthcare facilities; this could explain why most people who reported tuberculosis sought care at these clinics.
Early disease treatment plays a role in reducing morbidity and mortality [29,30].  [36]. The reported pregnancy rate (3.4%) was lower than that reported in the 2015 household survey (5.3%), but the latter included pregnancies in the last 12 months, while we only recorded current pregnancies [18]. Fourth, more households refused participation in region B; this may have biased the representativeness of our findings as region B had a different socio-economic profile. Region B was economically more affluent than other regions and sociodemographic, racial and economic factors have previously been associated with willingness to respond to surveys [37], which may account for the low-response rate in this region. A higher proportion of individuals from Region B accessed private healthcare and were less likely to attend sentinel hospitals for care. Individuals who feel that a survey would benefit them in some way are more likely to respond which could introduce response bias. Fifth, we did not cover all areas served by the sentinel hospitals, which may have resulted in a biased estimate of sentinel site use. Sixth, we only achieved about 20% of our target sample size for pneumonia which limited our ability to conduct sub-analyses. Last, while we made an effort to represent all residential areas, including informal settlements, it is possible that new settlements, transient populations or those living in nonresidential areas, such as commercial districts, may have been systematically excluded from the sample. The strengths of this study include that households were randomly chosen using a list of all enumerated households from Statistics South Africa, the overall response rate was high and teams were monitored to ensure that completion of all syndrome information was standardised. In addition suburbs that were chosen were those identified from the surveillance program to be the source of the majority of patients attending the sentinel sites. The age structure of households was similar to that reported in the national census [19], with 11% of members <5 years and only 3% >64 years of age.

Conclusion
This study highlights healthcare seeking patterns within the catchment areas of a hospital-based pneumonia and laboratory-based surveillance program. Of patients with pneumonia and meningitis who sought care at public hospitals, at least half sought care at sentinel hospitals; this indicates surveillance program relevance in capturing severe disease syndromes, whereas those with milder syndromes mainly sought care at public health clinics. Using HUS data in conjunction with surveillance data will assist in estimating a more precise burden of these syndromes and aid in allocation of resources.
What is known about this topic  In South Africa, public government healthcare facilities are commonly consulted for infectious syndromes;  Different barriers to healthcare exist.

What this study adds
 Describes healthcare-seeking patterns providing context for surveillance data from public healthcare facilities;  Identifies barriers to healthcare access among residents of public hospital catchment areas;  Allows more accurate estimation of the burden of infectious respiratory illnesses to improve targeting of prevention and treatment strategies in these communities.

Competing interests
The authors declare no competing interests.

Acknowledgements
We thank the residents of the communities in which the survey was conducted and the fieldworkers who conducted the interviews; Dorothy L Southern for scientific writing advice and for her critical review of this manuscript and Makatisano Papo for his active contribution to the creation of the study database.        *Some participants sought care at more than one facility; †People who did not seek care for ILI 32% (141/431); †Children who did not receive care for diarrhoea 25% (2/8)