Design and implementation of the study
Our study area was a small rural town of Ingwavuma, located in the north of KwaZulu-Natal province in South Africa, near the borders of eSwatini (to the north) and Mozambique (to the east). Ingwavuma is under traditional authority and is considered the poorest area in KwaZulu-Natal with an estimated annual income of R32,812 ( p. 17). Most people in this area rely heavily on social grants estimated at R13,090 per year (less than $1,000) as their main source of income. . The prevalence of Schistosoma haematobium among school-age children (over 10 years old) in Ingwavuma is 37.5%. However, in the 1-5 year age group, the prevalence of both S.mansoni and S.haematobium is 2% . Risk factors for schistosomiasis in young children include age of caregiver, type of household head, poor sanitation, access to a water source, and knowledge about schistosomiasis .
The study was an analytical cohort study that described and compared the language skills of uninfected and infected children with schistosomiasis and soil-transmitted helminthiasis using clinical assessments and observational data. This was an ancillary study to the Fighting infections for the benefit of Africa (TIBA SA) (http://tiba-partnership.org/about/what-is-tiba). More than 700 preschool children were tested for soil-transmitted helminthiasis and schistosomiasis in the TIBA study between 2017 and 2020. Schistosoma haematobium was diagnosed using the filtration technique on urine samples ; S.mansoni and STH (Taenia, Ascaris Lumbricoides, Trichuris Trichiura) were diagnosed from stool samples using the Kato Katz technique .
Deliberate random sampling was used. To be eligible for the study, a child had to be between 4.0 and 6.11 years old; attend an isiZulu Intermediate Preschool or ECD in the target area, be developmentally free, be a monolingual isiZulu speaker, and pass a hearing screening test. Study participants were categorized by age (4-6 years), gender (50% of each gender), and inclusion of participants positive for infection at a rate of 2 negative cases for 1 positive case. The principal investigator did not know the status or the nature of the infection of the study participants. Children were tested in two phases; phase 1 was tested immediately after parasitological screening and phase 2 was tested at least 12 weeks later and after treatment of children with schistosomiasis and STH with oral praziquantel. Some children participated in both phases while others only participated in phase 1 and could not be traced for repeat testing. Data for children who participated in both phases were treated separately as the phases were 3 months apart (affecting the maturity level of the children) and the language test was adapted for phase 2 to reduce the bias of contents.
The distribution of children in phase 1 and 2 varied in terms of age, sex and infectious agents as shown in Table 1. The mean age was 4 years 9 months in phase 1 and 5 years 9 months in phase 2. The total number of children positive for schistosomiasis (23%) was lower than the number of children positive for soil-transmitted helminthiasis (33%), the infectious agents not necessarily being coexisting. The spectrum of infections in phases 1 and 2 indicated that infection with soil-transmitted helminthiasis resulted mainly from A. Lumbricoides (16%).
Seventeen kindergartens participated in the study; all had pit latrines, one had a rainwater harvesting tank in the facility, and all provided one free meal a day for children. During our test dates, we observed that school lunches had no meat and vegetables other than beans. All of the pre-school teachers involved had the minimum secondary education ECD qualification and an R-level teaching qualification. All 17 preschools had R-level books provided by the government, but none had a computer, TV or internet access and generally not all of them had adequate access. educational resources such as puzzles and board games.
Data collection procedures
Given the socioeconomic profile of the study area, children were given a peanut butter sandwich and orange juice before undergoing tests to ensure they had a healthy breakfast and had the energy needed to participate in the tests. The tests began with a hearing screening which included an otoscopic examination and tympanometry to rule out ear infections and its contribution to poor speech scores . A nutritionist calculated body mass index for age (weight in kg) and height for age (height and arm circumference in cm) to determine stunting classified as mild (1), moderate ( 2) or severe (3) . The prevalence of stunting was 26%, showing that the majority of participants had adequate nutrition and no correlation was found between test scores and stunting in either study phase.
All the children were monolingual and spoke isiZulu at home and at school. The developmental language test , a non-standardized test developed for a research project, was adapted for this study following a pilot study and observations by research assistants. Based on these observations, the test vocabulary was adjusted to include the local dialect (north coast of KwaZulu-Natal) and the sequence of test items was formatted into 5 sections for ease of scoring and interpretation, as shown in Table 2. New test illustrations were developed using images taken in Ingwavuma with local community members for clarity and to fit the adjusted format of the test (see Supplementary File 1: Appendix A for illustrations of the test and B for the test form).
The adapted language development test showed a sensitivity of 89.7% in phase 1 and 81.3% in phase 2, indicating that the children who tested positive were true positives, while the specificity was 10, 3 and 18.8%, respectively. Cronbach’s Alpha was determined to be 0.869 (SD=5.1) in phase 1 and 0.813 (SD=7.7) in phase 2 demonstrating adequate internal consistency and suggesting that all items measured the same construct.
The test was administered to one child at a time by the principal researcher who is a speech therapist and an Isizoulou speaker, familiar with the dialect and culture of the region. Point scoring ranged from 0 to 4 points per target, depending on the extent of response as described in the test form (Supplementary File 1: Appendix B). Scoring was immediate and automated through the kobo collect app, an open source platform used to collect and analyze data .
Quantitative data analysis for both phases including descriptive frequency analysis, independent paired samples t-tests, ANOVA and bivariate correlations on SPSS (Statistical Package for Social Scientists, v. 25, IBM, Chicago, IL, United States). Post hoc tests were performed using Bonferroni corrections to measure the specific contribution of variables such as age, gender, school and infection on language categories and the time taken to complete the test . The overall error rate was controlled by using adjusted significance levels (α = 0.05). Information Processing Model for Cognitive Skills  and Vygotsky’s sociocultural theory for development and learning guided data analysis and score interpretation .