A hospital-based case-control study was conducted to identify the risk factors for MESBDs. The study participants were recruited as they presented to the child welfare clinics, neonatal/paediatric units and occupational clinics for care during data collection period from May 31 ^st 2018 to and July 31 ^st 2019. A case-control design was the optimal design for this study considering its suitability for the investigation of rare outcomes, as is the case with MESBDs. Even though a population-based design would have been more preferable, the ease of recruiting case and control subjects within the hospital settings disproportionately favoured the hospital-based design. This was an observational study, therefore was reported as per the STROBE guidelines ²⁵ .

The study was conducted in 13 hospitals comprising three county referral hospitals (Kiambu, Gatundu, and Thika), eight sub-county hospitals (Karuri, Kihara, Wangige, Nyathuna, Lari, Tigoni, Lussigetti, and Kigumo), and two faith-based hospitals (Presbyterian Church of East Africa Kikuyu Orthopedic and African Inland Church Cure International) situated within Kiambu County, Kenya. Notably, neither population-based or hospital-based surveillance systems for MESBDs existed in the county nor the study hospitals. Nonetheless, cases detected by primary health providers during childbirth and in neonatal care were recorded for the compilation of monthly hospital reports and subsequent entry into the District Health Information System (DHIS). The cases were drawn from Kiambu, Thika, Gatundu, Tigoni, Kikuyu, and Cure hospitals, which provided occupational and rehabilitative health services to children with MESBDs. The controls, on the other hand, were drawn from Kiambu, Gatundu, Thika, Karuri, Kihara, Wangige, Nyathuna, Lari-Rukuma, Tigoni, Lussigetti, and Kigumo hospitals, which provided child welfare services to the under-fives. Kiambu is the second-most densely inhabited county with an estimated population of 2.4 million people out of an estimated national population of 47.5 million ²⁶ . Its economic mainstay is largely agriculture, comprising tea, coffee, and dairy farming ²⁶ . Of the county’s total estimated population, approximately 2.2% aged ≥5 years are living with lifelong disabilities ²⁶ . A study carried out in the county between 2014 and 2018 observed defects of the musculoskeletal system as the most prevalent single system defects followed by central nervous, orofacial clefts genital, ocular, and anal organ defects ²⁷ .

The study population consisted of children aged ≤5 years old seeking health services at the study hospitals during the study period spanning from May to July 2019. All children whose mothers consented to participate in the study were recruited.

Cases were defined as children aged ≤5 years born with at least one MESBD to resident women of Kiambu County and seeking health care services at the neonatal units, paediatric wards, child welfare clinics and/or occupational therapist clinics of the study hospitals during the three-month study period. The Research Assistants (RAs) liaised with team leads of the departments listed above to identify cases of MESBDs. The team leads had been working in these departments, thus were conversant with the cases seeking services. The team lead invited the mothers of the children who met the case definition to comfortable private rooms within the departments where informed consent was sought and interviews conducted by the RAs. All cases that met this definition and whose carers consented to participate were prospectively recruited into the study until the required sample was attained (see Sample size determination).

Controls were children aged ≤5 years born without any forms of birth defects to resident women of Kiambu County and attending routine child-welfare clinics at the study hospitals during the same three-month study period. The Research Assistants liaised with team leads of the child welfare clinics to identify the children without any form of birth defects and were seeking routine immunization, and growth monitoring services. The team leads had been working in these clinics, hence were familiar with most of the under-fives seeking the services. These services are provided between 8.00 am and 5.00 pm from Monday to Friday; the team leads introduced the RAs who then briefed the potential participants on the study objectives. Because of the relatively large number of controls available, they were selected by simple randomization using sealed envelopes upon definition of the sample population and frequency-matched to the cases by the day of presentation.

The sample size was estimated as per the Kelsey JL et al. ²⁸ formula specified for case-control studies as follows: -

n 1 = ( Z α + Z β ) 2 p q ¯ ( r + 1 ) r ( p 1 − p 2 ) 2

q ¯ = 1 − p ¯

n 2 = r n 1

p 1 = p 2 O R 1 + p 2 ( O R − 1 )

p ¯ = p 1 + r p 2 r + 1

Where: n ₁ is the number of cases and n ₂ is the number of controls; p ₁ is the proportion of cases whose caregivers did not begin prenatal care in the first trimester (primary exposure), p ₂ is the proportion of controls whose care-givers did not begin prenatal care in the first-trimester set at 57% ^{22, 23} . Remarkably, Z _α/2 (1.96) and Z _β (-0.84) are the values specifying the desired two-tailed confidence level (95%) and statistical power (80%), respectively. The odds ratio (OR) for the effect of the primary exposure (cases whose caregivers did not begin prenatal care in the first trimester) was hypothesized to be 3.0 ^{22, 23} . The ratio (r) of unexposed to exposed individuals was set at 3, and given the estimates, a total sample size of 408 participants was derived (102 cases, and 306 controls).

Before data collection, four nursing graduate interns were recruited and trained as RAs on sound interviewing techniques, and information derivation/validation from antenatal care (ANC) booklets. This was to ensure the data collection process spanning three months (May 31 ^st to July 31 ^st, 2019) was conducted in a standardized manner. The ANC booklet contains maternal profile, medical/surgical history, previous pregnancy history, clinical notes, and physical examination findings on ANC visits among others. Maternal profile includes name, age, parity gravidity, height, weight, last menstrual period (LMP), expected date of delivery (EDD) and date of first ANC. Face-to-face structured questionnaires (see Extended data) were administered to the mothers of the study participants by RAs in comfortable secluded rooms within neonatal units and occupational therapy clinics for cases and child welfare clinics for the controls. Data were gathered retrospectively on exposures to environment-teratogens (pesticides and teratogenic medicines proxied by chronic illnesses), multifactorial inheritance (parity, nature of pregnancy, history of siblings with MESBDs and sex of the lastborn child) and sociodemographic-environmental factors (maternal age, education level, occupation, and adequate prenatal care proxied by gestational age and preconception folic acid intake). The predictors were assessed as shown in Table 1.

A conceptual framework depicting the predictor-outcome relationship is displayed in Figure 1. The flow chart of the simple-random systematic sampling strategy is shown in Figure 2.

Ethical approval for this study was obtained from the Kenyatta National Hospital [KNH]-University of Nairobi [UoN] Ethics Review Committee [Ref. No: KNH-ERC/A/44]. The purpose of the study was explained to participants and written informed consent was obtained from the mothers of the study subjects before engaging in the study.

Considering potential biases inherent in case-control studies that were likely to invalidate the study results, deliberate attempts were made to minimize their occurrence. First and foremost, the research assistants were trained on sound interviewing techniques and information derivation/validation from ANC booklets to minimize interviewer and minimize information biases, respectively. In a bid to minimize recall bias, gestational age (weeks) at the first ANC were estimated from the dates of the last menstrual period and dates of the first ANC obtained from the ANC booklets.

Following data collection, filled questionnaires were manually checked daily for accuracy and completeness and subsequently entered into a Microsoft Excel spreadsheet (Microsoft Office Professional Plus 2019) by two independent data managers to reduce potential errors. The excel dataset was validated and exported to Stata software version 14.0 (Stata Corporation, Texas, USA) for further cleaning, coding, and analyses. Descriptive analyses (means, medians, standard deviations, and ranges) were used to summarize continuous variables, whereas proportions and percentages for categorical variables were generated and presented in frequency tables. Afterward, the effect of each predictor on the odds of MESBDs was assessed using univariable logistic regression models at a liberal P-value (P≤0.20) ²⁹ . Gestational age (weeks) at first ANC as a continuous variable was categorized into groups (≤8 weeks and ≥9 weeks) for evaluation in the univariable analyses ^{1, 4, 22– 24} . Additionally, parity as a continuous variable was grouped into two groups; primiparous or multiparous categories for assessment in the univariable analyses ^{30, 31} . However, maternal age as a continuous variable was insignificant in the univariable analyses, thus, recategorized into two groups; ≤34 years, and ≥35 and reassessed for statistical significance; women aged at least 35 years have previously been reported to have an increased likelihood of giving birth to children with MESBDs ³² . Variables found statistically significant in the univariable analyses were fitted to a multivariable model where a backward stepwise approach was used to eliminate variables from the model at P-value >0.05. To minimize the confounding effects, elimination of non-significant predictors was only considered when their exclusion from the model did not yield more than a 30% change in the effects of the remaining variable ²⁹ . Two-way interactions were fitted between the remaining variables of the final model and assessed for significance. A Hosmer-Lemeshow test was used to assess the goodness of fit of the logistic model, with a P-value of >0.05 being suggestive of a good fit.

Sociodemographic-environment: The median age of the study respondents was 26 years with a mean of 27.31 years (SD=5.73, R; 17-47) ( Table 2). The median age of mothers in the case group was 28 years with a mean of 28.73 (SD=5.95, R; 19-47), whereas the median age of mothers in the control group was 26 years with a mean of 26.84 (SD=5.58, R; 17-42) ( Table 2). Of the 408 study participants, 184 (45.10%) had attained a secondary level of education; 38 (37.25%) and 146 (47.71%) in the case and control groups, respectively ( Table 2). Environmental-teratogens: Of the 408 study respondents, 15 (3.68%) were exposed to farm-sprayed pesticides, of which four (3.92%) were in the case group and 11 (3.59%) were in the control group ( Table 2).

Multifactorial inheritance: Of the 408 study respondents, 404 (98.77%) had single gestations for the current child, of which 99 (97.06%) and 304 (99.35%) were in the case and control groups, respectively ( Table 2). Of the study participants, 15 (3.68%) had given birth to children with birth defect in previous gestations, with 9 (8.82%) in the case group and 6 (1.96%) in the control group ( Table 2).

Notably, all the factors assessed for statistical significance in the univariable analyses were associated with MESBDs at P≤0.20; age, education, occupation, sex of the lastborn child, history of siblings with birth defects, preconception folic acid intake, nature of pregnancy, pesticide exposure, chronic illnesses, parity, gestational (age) weeks at first ANC, and ANC beginning eight weeks post-conception ( Table 3). Subsequently, these variables were fitted to the multivariable model for the final analysis, except gestational age at first ANC, education, occupation, and prenatal care beginning eight weeks post-conception being distal relative to pesticide exposure and preconception folic acid intake ( Figure 1).

In the multivariable analysis, only maternal age, and history of siblings with MESBDs were shown to be significant predictors at a 5% significance level ( Table 4). Compared to women aged ≥35 years old, holding all factors constant, women aged ≤34 years old were 59% less likely to give birth to children with MESBDs (adjusted odds ratio (aOR): 0.41; 95% CI: 0.18-0.91; P=0.03) ( Table 4). Additionally, compared to the history of siblings without MESBDs holding all factors constant, siblings with history of MESBDs were 5.21 times likely to have MESBDs (aOR: 5.21; 95%CI; 1.35-20.12; P =0.02) ( Table 4).

Figshare: Risk factors for major external structural birth defects among children in Kiambu County, Kenya: a case-control study. https://doi.org/10.6084/m9.figshare.13614548.v1 ³³ .

Harvard Dataverse: Risk factors for major external structural birth defects among children in Kiambu County, Kenya: a case-control study. https://doi.org/10.7910/DVN/ZDEOZ5

This project contains the following extended data:

Data are available under the terms of the Creative Commons Zero "No rights reserved" data waiver (CC0 1.0 Public domain dedication).