A long-term observational study of paediatric snakebite in Kilifi County, south-east Kenya

Introduction Estimates suggest that one-third of snakebite cases in sub-Saharan Africa affect children. Despite children being at a greater risk of disability and death, there are limited published data. This study has determined the: population-incidence and mortality rate of hospital-attended paediatric snakebite; clinical syndromes of snakebite envenoming; and predictors of severe local tissue damage. Methods All children presenting to Kilifi County Hospital, Kenya with snakebite were identified through the Kilifi Health and Demographic Surveillance System (KHDSS). Cases were prospectively registered, admitted for at least 24-hours, and managed on a paediatric high dependency unit (HDU). Households within the KHDSS study area have been included in 4-monthly surveillance and verbal autopsy, enabling calculation of population-incidence and mortality. Predictors of severe local tissue damage were identified using a multivariate logistic regression analysis. Results Between 2003 and 2021, there were 19,606 admissions to the paediatric HDU, of which 584 were due to snakebite. Amongst young children (≤5-years age) the population-incidence of hospital-attended snakebite was 11.3/100,000 person-years; for children aged 6–12 years this was 29.1/100,000 person-years. Incidence remained consistent over the study period despite the population size increasing (98,967 person-years in 2006; and 153,453 person-years in 2021). Most cases had local envenoming alone, but there were five snakebite associated deaths. Low haemoglobin; raised white blood cell count; low serum sodium; high systolic blood pressure; and an upper limb bite-site were independently associated with the development of severe local tissue damage. Conclusion There is a substantial burden of disease due to paediatric snakebite, and the annual number of cases has increased in-line with population growth. The mortality rate was low, which may reflect the species causing snakebite in this region. The identification of independent predictors of severe local tissue damage can help to inform future research to better understand the pathophysiology of this important complication.


Introduction
Snakebite is a neglected tropical disease that affects 5 million people each year, with the greatest burden falling on rural populations of the tropics and sub-tropics [1]. Snakebite disproportionately affects children living in low-income countries, who are likely to be at a greater risk of disability and mortality. In sub-Saharan Africa it has been estimated that 30% of people affected by snakebite are children [2]. Typical activities that may bring children into contact with snakes include outdoor play, agricultural work, and walking to school. The burden of snakebite in sub-Saharan Africa has been estimated at over 1 million DALYs (disabilityadjusted life years) per year [3]. Much of this is accounted for by deaths and limb amputations in the young, who disproportionately contribute to disability adjusted life years and years of life lost [3].
The limited available evidence suggests that children are twice as likely to be administered antivenom following snakebite and have an increased risk of death [4][5][6]. It is hypothesised that children are particularly vulnerable as they receive a higher dose of venom relative to their body weight. Despite this, limited studies have described the burden of paediatric snakebite in sub-Saharan Africa. Two observational studies in South Africa (with samples sizes of 51 and 72 children) identified a substantial burden of painful progressive swelling, with one in four cases undergoing debridement or fasciotomy [7,8]. In a retrospective study of 28 consecutive cases of paediatric snakebite in The Gambia, the mortality rate was 14% [9]. In a community

Study site
The KHDSS hospital surveillance system has collected data on paediatric admissions (including snakebite cases) to Kilifi County Hospital since 2003. Paediatric care in Kenya is provided to children aged �12-years. To avoid misclassifying cases with delayed onset of clinical envenoming, local policy stipulates that all cases of paediatric snakebite are admitted for �24-hours observation on the paediatric high dependency unit (HDU). This is regardless of disease severity and even applies to cases with no features of envenoming. The paediatric HDU is funded and staffed by the KEMRI-Wellcome Trust Research Programme and a standardised protocol for the management of snakebite is in place. The resources and quality of care available at this paediatric HDU are higher than typical government healthcare facilities in much of Africa, facilitating an approach to snakebite management that aligns with recommended standards in high-income settings. Antivenom is administered in accordance with WHO guidelines [12]. Once paediatric cases with snakebite have been managed on the paediatric HDU for 24-hours, the clinical team decide whether it is appropriate for the child to remain on the paediatric HDU, be stepped down to the Kilifi County Hospital paediatric ward, or to be discharged home. Cases transferred to the paediatric ward continue to have data collected as part of the KHDSS hospital surveillance study, and cases discharged home are followed-up by the KHDSS community surveillance system.
Hospital surveillance data are linked to the KHDSS community surveillance data. The study area is 891 km 2 and Kilifi County Hospital is the only hospital with inpatient paediatric services in the study area [11]. The KHDSS study area was defined based on the lowest number of administrative sublocations that were the site of residence of greater than 80% of the paediatric inpatients at Kilifi County Hospital over a three-year period (1998-2000) [11]. The study area, including all dwellings, has been GPS mapped. In 2021, the KHDSS included 92,063 households and 309,228 residents. Most of the study population reside in rural dwellings and the local economy is predominantly centred on subsistence farming [11].

Identification and eligibility of cases
All cases of snakebite affecting children aged �12 years, attending from January 2003 until December 2021, were eligible for inclusion in this study. These cases were routinely enrolled into the KHDSS hospital surveillance study over this period. At admission and on discharge from Kilifi County Hospital, clinical-research staff prospectively assigned a diagnostic code which was recorded on the KHDSS hospital database. Specific diagnostic codes were in place to classify snakebite, as follows: 'snake venom' and 'snake bite.' As a precaution, to avoid missing cases that may have been incorrectly classified, database search terms also included the following diagnostic codes, which were recorded at admission and on discharge: (1) "snake venom"; (2) "snake bite"; or (3) "acute animal bite." In addition to searching diagnostic codes, free-text sections of the database were searched for the following terms: (1) "snake"; or (2) "venom". All cases identified through this database search were screened by an academic clinician (MA) and non-snakebite cases, for example a dog bites, were excluded. For a case to be included there had to be a specific and clear reference to snakebite being the cause of the admission, in the database or the clinical records. In cases where the diagnosis of snakebite was uncertain, the paper notes were scrutinised by the study team.

Data extraction
The following data were prospectively recorded on the KHDSS database at the time of hospital admission by research staff: demographics; date and time of admission; date and time of discharge; weight; admission vital signs; diagnosis code; mortality; and date of death. The following clinical laboratory results from admission samples were extracted from the KEMRI-Wellcome Trust Research Programme laboratory database: full blood count (including differential); serum sodium; serum potassium; and serum creatinine.
To supplement the above prospective data that were collected in the KHDSS, retrospective data from the paper case notes were extracted by a team of research nurses using a standardised case report form. The following retrospective data were extracted: residence; geographic location of bite; circumstances of bite; date and time of bite; anatomical location of bite; use of traditional treatments; clinical features of local and systemic envenoming; clotting time; antivenom administration; indication for antivenom; antivenom associated adverse events; adjunctive treatments; complications of envenoming; and discharge destination. The antivenom product that was administered was not documented in the medical records. However, from the hospital pharmacy records it was possible to identify the antivenom product available during each study year.
Deaths were identified by searching the KHDSS hospital database, the paper medical records, and the KHDSS community database. Verbal autopsy was routinely conducted for all deaths that occurred within the KHDSS study area. This was conducted using the 2007 World Health Organization (WHO) verbal autopsy tools [13], as described previously [14].
To calculate the population-incidence of hospital-attended snakebite, census data from the KHDSS community database were used. Full details of this surveillance system have previously been published [11]. Community interviewers visited every household in the study area on a 4-monthly basis. A single resident was interviewed, from whom information pertaining to each resident was collected. The identity of all residents was confirmed and any newly born children, in-migrations, deaths, and out-migrations, since the previous enumeration round, were recorded. Person-years of observation were stratified by sex, age and 41 geographic sublocations.

Statistical analysis
Clinical data were described using summary statistics including means, medians, and proportions. Population-incidence of hospital-attended snakebite was calculated with 95% confidence intervals. Population incidence was calculated separately for young children (0-5 years inclusive) and older children (6-12 years inclusive). The hospital surveillance system did not consistently capture all admissions until 2006, thus, to avoid underestimation, incidence estimates have only been calculated for the period of 2006-2021.
A logistic regression analysis was conducted to identify variables associated with severe local tissue damage. Severe local tissue damage was defined as any case developing local skin necrosis or requiring surgical intervention-criteria that are congruent with the recently established snakebite core outcome measurement set [15]. The following variables were included in a univariate analysis, and those with a significance value of p �0.10 were selected for inclusion in the multivariate analysis: age, site of bite (dichotomised as upper limb and lower limb), elapsed time from bite to admission, MUAC (mid-upper arm circumference, which was measured away from the bite site)-for-age z-score (using the zscorer package in R), vital signs on admission (pulse rate, respiratory rate, systolic blood pressure, capillary refill time, axillary temperature, and oxygen saturations), admission full blood count (haemoglobin, white cell count, granulocyte count, lymphocyte count, and platelet count), and admission serum biochemistry (sodium, potassium, and estimated glomerular filtration rate [eGFR]). The eGFR was calculated using the Schwartz equation [16]. Multiple imputation, using the mice package in R, was undertaken to replace missing values [17]. R version 4.2.2 (R Foundation for Statistical Computing) was used for all analyses.

Incidence of hospital-attended paediatric snakebite
During the study period there were 78,038 paediatric admissions to Kilifi County Hospital (children aged �12-years), and 19,606 admissions to the paediatric HDU. The diagnostic code search of the KHDSS hospital database identified 724 potential paediatric snakebite cases. Following manual review of clinical data, 72 were excluded as they were not cases of snakebite (these were predominantly bites by other animals). Further exclusions are detailed in the CONSORT diagram (Fig 1). There were 584 children aged 12-years or under that presented with snakebite to Kilifi County Hospital between January 2003 and December 2021. Snakebite thus represented 2.98% of all admissions to the paediatric HDU over the study period. Details of the proportion of admissions to Kilifi County Hospital for snakebite by year are available in Table A in S1 Text. The median age was 8 years (IQR 5-10 years) and 47.6% were female. Clinical records were available for 472 (80.8%) participants, and most were resident in the KHDSS study area (N = 399; 68.3%).
The population-incidence of hospital-attended snakebite was calculated using the number of admissions per year amongst children that resided in the KHDSS study area (numerator) and annual age-specific census data from the KHDSS (denominator). As young children had a substantially lower risk of snakebite, population-incidence was stratified between the ages 0-5-years and 6-12-years. For children aged �5-years, the average population-incidence between 2006 and 2021 was 11.3/100,000 person-years; for children aged 6-12-years, the average PLOS NEGLECTED TROPICAL DISEASES population-incidence was 29.1/100,000 person-years. Fig 2 demonstrates the annual population-incidence for young children and older children, with 95% confidence intervals. Although there is variability between study years, the incidence remained broadly consistent until 2020, with a decline in 2021. As there has been a substantial increase in the number of people residing in the KHDSS study area, absolute numbers presenting with snakebite have increased over the study period. In 2006 there were 98,967 person-years of follow-up amongst children �12 years of age; by 2021 this had increased to 153,453 person-years.
Population-incidence of hospital-attended snakebite was calculated by year of age, as shown in Fig 3. There was a substantial increase in incidence with age: from 3.6/100,000 person-years at age 1-year, to 35.9/100,000 person-years at age 9-years. With increasing age above 9-years incidence fell, reaching 30.0/100,000 person-years by age 12-years.

Clinical features
The circumstances of the snakebite were available in the clinical records in 307 (52.6%) cases. Most snakebites occurred outdoors and near to the child's home (131 cases; 42.7%) or in the The bands represent the 95% confidence intervals. Population incidence is stratified by age category.
Traditional therapies that were sought prior to admission included application of a 'black stone' in 110 (23.3%) cases, application of a torniquet in 48 (10.2%) cases, and cutting the skin at the bite site in 47 (10.0%) cases (Table 1). Two or more types of traditional therapy were sought prior to admission in 56 (11.9%) cases (S1 Fig). The median elapsed time from bite until admission was 6-hours and 45-minutes (IQR 3-15-hours; range 10 minutes-17 days). The median elapsed time from admission until antivenom administration was 2-hours and 50-minutes (IQR 1-9 hours). Children who had received traditional therapies took a median of 3.6 hours longer to present to hospital, and this difference was statistically significant (median 9.4 hours and 5.8 hours; p = 0.003).
Most children had local swelling at presentation, being present in 399 (84.5%) cases (Table 1). There were six cases with systemic bleeding, and two with neurotoxic envenoming. There were no features of envenoming in 51 cases (10.8%). Age-adjusted tachycardia, hypotension, and tachypnoea were present in 311 (53.3%), 21 (4.6%), and 345 (59.3%) cases, respectively. The 20-minute whole blood clotting test (20-WBCT) was documented in only 18 cases. Many were conducted incorrectly and followed a procedure akin to the Lee-White clotting time (with repeated checks of the sample before 20-minutes had elapsed). Two of the 18 cases where a bedside clotting test was documented were prolonged over 20-minutes.
Full blood count and serum biochemistry were routinely undertaken for all children presenting with snakebite. In cases where insufficient blood sample volumes were obtained, due to challenging access, the full blood count was performed in preference to biochemistry. These  demonstrated anaemia (haemoglobin <8.2 g/dL) in 51 (9.6%) cases, leukocytosis in 314 (59.5%) cases, and reduced eGFR in 12 (5.0%) cases (Table B in S1 Text). The clinical laboratory results, stratified by severe local tissue damage, have been depicted in Fig 5. The numbers of children with age-adjusted abnormal clinical laboratory results have been summarised in Table B in S1 Text.
The mean duration of hospital stay was 6.3 days (SD 17.8 days). There was an average of 195 bed-days occupied per year due to paediatric snakebite admissions. The mean duration of hospital stay was significantly prolonged (55.0 days vs 4.2 days; p<0.001) in cases with severe local tissue damage (defined as developing skin necrosis or undergoing local surgery). Amongst the 25 cases with severe local tissue damage, 20 (80.0%) were admitted for �7-days.

Predictors of severe local tissue damage
Severe local tissue damage developed in 25 cases (4.3%). Necrosis at the site of the bite developed in 22 cases (3.8%), and 19 (3.3%) required surgery. Ten cases underwent debridement, seven had a fasciotomy, four underwent skin grafting, and three had an amputation. The three cases fulfilling the criteria for severe local tissue damage that did not have skin necrosis had all undergone fasciotomy. All the cases that underwent amputation (n = 3) were preceded by the development of skin necrosis, and none were preceded by surgical fasciotomy.
Following multiple imputation, the following covariates were assessed in a univariate logistic regression analysis to identify potential predictors of severe local tissue damage: age, MUAC-for-age z-score, site of bite (upper vs lower limb), time from bite to admission, vital signs on admission (axillary temperature, pulse rate, respiratory rate, systolic blood pressure, capillary refill time in seconds, and oxygen saturations), serum sodium, serum potassium, eGFR, white blood cell count, granulocyte count, lymphocyte count, platelet count, and
Lymphocyte count and granulocyte count were omitted from the multivariate logistic regression model, as each were positively associated with severe local tissue damage and, therefore, the total white blood cell count was selected (to avoid multicollinearity). The rate of severe local tissue damage was similar between bites to the arm and the hand, although the event rate was small. Two participants (13.3%) with arm bites developed severe local tissue damage, five (11.9%) with hand bites, six (4.3%) with leg bites and seven (3.3%) with foot bites (Fig 4). Therefore, the covariate of upper limb bite (hand or arm) was entered into the multivariate analysis. The following statistically significant predictors were identified from the multivariate analysis: upper limb bite site (OR 3.27; 95% CI 1.17-9.17; p = 0.03); white cell count (OR 1.14; 95% CI 1.06-1.22; p<0.01); systolic blood pressure (OR 1.03; 95% CI 1.00-1.07; p = 0.04); serum sodium (OR 0.9; 95% CI 0.82-0.99; p = 0.03); and haemoglobin (OR 0.72; 95% CI 0.56-0.92; p = 0.01).

Snakebite associated mortality
Nine of the children in this study have died. Four of these deaths were unrelated to the snakebite and occurred years later during separate hospital episodes. The cause of deaths in these cases were epilepsy, accidental fall, acute respiratory infection associated with HIV/AIDS, and seizures secondary to previous bacterial meningitis, and these deaths occurred 13-, 5-, 5-, and

PLOS NEGLECTED TROPICAL DISEASES
Paediatric snakebite in Kilifi County, Kenya 4-years after the snakebite incident, respectively. Of the five snakebite associated deaths, two were due to neurotoxic envenoming, one had cardiovascular instability (hypotension, bradycardia and respiratory distress), one developed antivenom associated anaphylaxis, and one infant had a general deterioration of an uncertain nature, which culminated in cardio-pulmonary arrest ( Table 3). The five snakebite associated deaths occurred within one day of the hospital admission.

Discussion
This study represents one of the most comprehensive analyses of paediatric snakebite in Africa and demonstrates the substantial burden of this disease. The concerning trend of rising cases of paediatric snakebite in Kilifi, in parallel with population growth, underscores the need for strengthened targeted prevention strategies, improved training of healthcare providers, and increased availability of antivenom treatments. There is also an urgent need for similar studies on the epidemiology of paediatric snakebite envenoming to be conducted, particularly in sites with access to established health and demographic surveillance systems (HDSS) in Africa [18].

PLOS NEGLECTED TROPICAL DISEASES
There was a substantial fall in hospital-attended snakebite incidence in 2021, which may have been caused by the SARS-CoV-2 pandemic altering health seeking behaviour [19]. One quarter of paediatric snakebite cases were given antivenom, which was most frequently indicated for local envenoming. Most recently (years 2019-2022), Inoserp Pan-Africa (Inosan Biopharma) polyvalent antivenom has been used, although, in 2022 it was withdrawn from the Kenyan market after failing a risk-benefit assessment conducted by the World Health Features of envenoming: Bite site swelling, and neurotoxicity Clinical narrative: Snakebite to the lower limb was followed by difficulty in swallowing and blurred vision. At hospital developed respiratory arrest which was managed with mechanical ventilation. The following day spontaneous breathing returned, and mechanical ventilation was ceased. Subsequently developed copious haematemesis followed by cardiac arrest. Managed with cardiopulmonary resuscitation and adrenaline but died shortly after. Organization [20]. Since 2016, intermittent stocks of SAIMR (South African Vaccine Producers) polyvalent antivenom have been available through charitable donation by the Bio-Ken Snake Farm in Watamu, which tends to be reserved for more severe cases, given its evidence of pre-clinical efficacy [21]. Although the frequency of severe allergic reactions was low, there was one case that died as a direct result of antivenom induced anaphylaxis. Despite local envenoming being the most frequent indication for administering antivenom in much of Africa, its effectiveness for this indication is unproven, particularly if it is given late, and clinical trials are urgently needed [22]. Novel oral small molecule therapeutics may hold promise, particularly if they can be administered in rural clinics and thus reduce the time to treatment [23][24][25].
Bleeding was the most common sign of systemic envenoming. Despite this, measures of coagulopathy, such as the 20WBCT, were rarely documented in the case files. Early detection It was not possible to describe the predominant biting species in this study. It is believed that the puff adder (Bitis arietans), spitting cobras (Naja spp.), and burrowing asps (Atractaspis spp.) are the predominant medically important species in this region, but the relative contribution of these, and other less medically important species, is unknown. Mambas (Dendroaspis spp.) and non-spitting cobras (Naja haje, N. subfulva) are habitual to this region of Kenya. Although there were only two cases of neurotoxic envenoming in this study, both were fatal.
Delayed presentation to hospital was frequent and often prolonged. As most cases resided within the KHDSS study area, which is near to Kilifi County Hospital, it is likely that there is a delay in the decision to attend hospital. It is concerning that a large proportion of children received traditional therapies prior to presenting to hospital, particularly as this was associated with a statistically significant prolongation of the bite to admission time. The most frequently sought traditional therapy was application of a 'black stone,' which has been used in many geographic settings despite its lack of efficacy [26].
Most children in this study received antimicrobials. The majority had cloxacillin, although broader spectrum agents such as ceftriaxone and gentamicin were also used. Unlike other animal bites, snakebite rarely results in infection and routine antimicrobial prophylaxis is not recommended [27,28].
Severe local tissue damage developed in 4.4% of cases and was often associated with admissions that were weeks or even months long. Low haemoglobin was associated with severe local tissue damage. The direction of causality is uncertain, and it is feasible that children with anaemia may have other comorbidities that put them at risk of local tissue damage. Snakebite can cause anaemia as a result of thrombotic microangiopathy, although this is usually associated with thrombocytopaenia, which was uncommon in this study [29]. A raised white cell count on admission was also associated with severe local tissue damage, which has been demonstrated in other settings [30,31]. This is likely to be a bi-directional process, with activation of the innate immune system causing collateral damage at the bite site, and damage of local tissues triggering an immune response. Children that had sustained a snakebite to the upper limb were more likely to develop severe local tissue damage, the reason for which is uncertain. It is regarded that children are at a greater risk of envenoming, compared to adults, as they receive a higher dose of venom relative to their body weight; therefore, it may follow that the small upper limb of a child is particularly at risk. Increasing systolic blood pressure and lower serum sodium were associated with severe local tissue damage, although the small effect size and borderline statistical significance make the clinical relevance of these associations uncertain. Ultimately, further studies of local envenoming in sub-Saharan Africa are needed to confirm whether the predictors of severity identified in this single site study are reproducible.
A limitation of this study was that paediatric snakebite cases that did not attend hospital were missed, and therefore the true burden of disease has been under-estimated. A household survey is needed to further define the epidemiology of snakebite in Kilifi. Although the KHDSS study enabled reliable identification of consecutive cases of paediatric snakebite, with routine data collection and clinical laboratory analyses, the KHDSS study was not specifically designed to study snakebite. Thus, many important datapoints, such as whether antivenom was administered, needed to be retrospectively collected from the hospital records. Nevertheless, documentation on the paediatric HDU tended to be detailed and accurate, with standardised admission and discharge case report forms and contemporaneous daily documentation during admission. All cases were managed on a paediatric HDU which is supported and staffed by the KEMRI-Wellcome Trust Research Programme. There were missing data, particularly for biochemistry laboratory results and for items with variable documentation in the clinical records, such as the use of traditional therapies. The risk of bias due to missing data was partially mitigated using multiple imputation.
In conclusion, this study demonstrates the substantial burden of snakebite envenoming amongst children in rural Kenya. This is traumatic for children, interrupts schooling and development, is disruptive for families, places a substantial burden on healthcare facilities, and can lead to permanent disability or death. There is an urgent need for improved community awareness, with particular focus on preventative strategies, appropriate first aid, and the importance of early presentation to hospital. Many children in Kilifi receive antivenom for local envenoming, and it is important to assess whether this is effective.
Supporting information S1 Text. Table A