Neurological impairments are an important cause of disability and premature mortality.^1-3 At least 291 million (95% uncertainty interval 250–335) or 11.2% of all children and adolescents globally have epilepsy, intellectual disability, or sensory impairments, with the burden being higher in older children living in low- and middle-income countries (LMIC).⁴ Sub-Saharan Africa bears the highest prevalence of neurological impairments both in children under⁵ and above 5 years of age.⁴ Improved child survival^{6, 7} and increased incidence of known risk factors, such as adverse prenatal, perinatal, neonatal, and postnatal events, may have added to the prevalence of neurological impairments in older children in this region.^{8, 9} Africa is undergoing an epidemiological transition from infectious to non-communicable diseases;^{10, 11} however, no current studies have investigated changes in the prevalence and risk factors of neurological impairments in older children over the period when control measures for infectious diseases were introduced.

There is a paucity of data on the prevalence of neurological impairments in children surviving the first 5 years of life in LMIC.¹⁰ A community survey conducted in rural Kenya in children aged 6 to 9 years in 2001 reported a prevalence of neurological impairments of 61 per 1000.¹² This survey was done when malaria transmission was high,^{13, 14} before the introduction of vaccines against organisms causing meningitis, and when the health system was not as well developed as a decade later. Malaria control since 2001 has reduced the prevalence of malaria-related seizure disorders including epilepsy;^{15, 16} however, the rate of admissions with neonatal risk factors, which may influence the prevalence of neurological impairments in the long term, increased during this period.¹⁷ Changes in the rate of hospital admissions with adverse neonatal events may reflect a general improvement in the utilization of maternal and childcare services in the community. However, the relative changes in prevalence due to the differential effects of these changing risk factors are yet to be examined in LMIC.

We hypothesized that the prevalence of neurological impairments in children aged 6 to 9 years would increase over a period of improved child survival characterized by a decline in mortality rates in children aged under 5 years, while recognizing that prevention of meningitis and malaria may decrease the neurological disability attributable to these infections. We also hypothesized that changes over time in the prevalence of neurological impairments from surveys conducted in a defined rural area might be associated with (directly or indirectly) or be consistent with secular trends of adverse neonatal events admitted to a hospital linked to the demographic surveillance system that produced the survey samples from 2005 to 2016. Therefore, we used the same methodology to conduct two successive epidemiological surveys in a defined demographic surveillance area, one in 2001 and another in 2015, to investigate geographical changes over time in the prevalence of and risk factors for neurological impairments and disability. We estimated the rates of neonatal admissions with preterm birth, neonatal encephalopathy, and low birthweight (LBW) using paediatric surveillance data from a rural referral hospital, which draws most of its admissions from the same defined area. Results from this study will provide evidence for setting research and policy priorities and planning public health interventions.

METHOD

Study setting

We conducted epidemiological surveys and hospital surveillance of paediatric admissions in the Kilifi Health and Demographic Surveillance System (KHDSS), a defined geographical area covering approximately 900km² in Kenya (Fig. 1). The KHDSS is re-enumerated once every 4 to 6 months by a household survey to update vital statistics and indicators of socioeconomic status since 2000. The total resident population increased from 198 063 in 2001 to 279 158 in 2012, an increase mainly driven by high birth rates.¹⁸ Neonatal and under-5 years mortality per 1000 live births are 17 and 41 respectively.¹⁸ Over 80% of inpatient admissions to Kilifi County Hospital originate from the KHDSS. This closely linked community, clinical, and hospital surveillance has been used to define the incidence and prevalence of various infectious and non-infectious diseases.^{13, 19}

We obtained approval for the epidemiological studies from the National Ethical Review Committee of the Kenya Medical Research Institute; permission to use data from the KHDSS was granted by the Data Governance Committee at the Kenya Medical Research Institute-Wellcome Trust Research Programme in Kilifi. Written informed consent was sought from the parents, guardians, and caregivers of the participants in all phases before data collection.

RESULTS

A total of 10 218 and 11 223 children aged 6 to 9 years were screened using the TQQ in 2001 and 2015 respectively; 810 (84.8%) out of 955 (9.3%) children with a positive TQQ result and 766 (89.0%) of 861 (9.3%) randomly sampled controls were assessed in phase 2 in 2001. In 2015, 70% of all children screening positive on the TQQ and the randomly selected controls were assessed in phase 2. A summary of survey characteristics comparing the 2001 and 2015 surveys is shown in Table 1. Similarly, adverse neonatal events between 2005 and 2016 in Kilifi County Hospital are shown in Table S2 (online supporting information).

Three hundred and six (3.0%) children out of 10 218 assessed in 2001 had neurological impairments; in 2015, 251 (2.2%) out of 11 223 children aged 6 to 9 years who were assessed had neurological impairments. The overall adjusted prevalence for any neurological impairment, adjusted for TQQ sensitivity and specificity and attrition between phases, decreased from 61.0 per 1000 (95% CI 48.0–74.0) in 2001 to 44.7 (95% CI 40.9–48.6) in 2015 (26.7% reduction; p<0.001). The prevalence of lifetime epilepsy was reduced from 41.0 per 1000 (95% CI 31.0–51.0) in 2001 to 17.2 (95% CI 15.0–19.9) in 2015 (p<0.001). Similarly, the prevalence of hearing impairments was reduced from 14.0 per 1000 (95% CI 9.0–18.0) in 2001 to 3.0 (95% CI 2.1–4.2) in 2015 (p<0.001). However, the prevalence of cognitive and visual impairments did not significantly change over the 14 years (Table 1).

There was a significant decline in the prevalence of neurological impairments in children aged 6 to 9 years at four locations in the KHDSS, namely Kilifi township, Mtwapa, Tezo, and Ziani (Fig. 2). Actual data for changes in location-level prevalence are provided in Table S3 (online supporting information). Generally, there was little evidence of geographical clustering at both survey points. In summary, four high and three low clusters and 10 high and four low clusters were identified in 2001 and 2015 respectively (Fig. S1, online supporting information). In 2001, only one low cluster was statistically significant (p=0.010) while one high cluster was statistically significant in 2015 (p=0.033).

Adverse perinatal events were significant predictors of epilepsy and cognitive and motor impairments in both surveys and for hearing impairment in 2015 (Tables 2 and 3); however, a significant increase over time in the strength of the association was observed only for hearing impairment by 2015 (exponentiated coefficient for the interaction=5.94, p=0.03) (Table 4). Higher weight-for-age and height-for-age z-scores were inversely associated with motor and hearing impairments in 2001 and with cognitive impairments in both surveys (Tables 2 and 3). Notably, there was a significant decrease over time in the association of weight for age with cognitive and motor impairments (exponentiated coefficient for the interaction=0.49 and 0.95 respectively) and height for age with cognitive and motor impairments (exponentiated coefficient for the interaction=0.44 and 0.89 respectively; Table 4). Home delivery, atypical delivery, neonatal jaundice, incomplete immunization, previous hospitalization, and a family history of seizures were positively associated with one or more impairments in either survey (Tables 2 and 3). However, the odds of previous hospitalization were significantly higher only in children with epilepsy in 2001 (exponentiated coefficient for the interaction=2.69, p=0.04). Similarly, the odds of home delivery in children with cognitive impairment was higher in 2001 (exponentiated coefficient for the interaction=1.33, p=0.02).

A separate analysis of admissions to hospital showed that there was a significant increase in the admission rate of preterm birth (rate ratio 1.08, 95% CI 1.07–1.09, p<0.001), LBW (rate ratio 1.08, 95% CI 1.06–1.10, p<0.001), and neonatal encephalopathy (rate ratio 1.08, 95% CI 1.06–1.09 p<0.001) between 2005 and 2016 as assessed from admissions to Kilifi County Hospital (Fig. S2, online supporting information). The annual average percentage change in admission rate with preterm birth was +8.36% (95% CI 7.07–9.67); it was +8.55% (95% CI 7.02–10.10) for neonatal encephalopathy and +6.49% (95% CI 4.40–8.62) for LBW. The cumulative risk of preterm birth among newborn infants comparing the 4-year periods between 2005 and 2016 is provided in Table S4 (online supporting information).

DISCUSSION

The overall adjusted prevalence of neurological impairments did not increase as hypothesized but rather decreased from 61 per 1000 in 2001 to 44.7 per 1000 in 2015. This reduction was mainly driven by a decline in the prevalence of epilepsy, hearing, and motor impairments; the prevalence of cognitive and visual impairments did not significantly change over the 14-year period. Adverse perinatal events and undernutrition were significant risk factors for neurological impairment in both surveys; other risk factors, such as place of delivery (2001 survey), immunization status (2001 survey), and a family history of seizures (2015 survey) were significant only in one survey. There was a significant change in the association of adverse perinatal events with hearing impairments (an increase) and undernutrition with motor impairments (a decrease) when we compared the 2001 and 2015 surveys. Generally, there was limited evidence of geographical variation in the prevalence of neurological impairments in both surveys. The rate of hospital admissions with adverse neonatal events increased significantly between 2005 and 2016, which could be related to improved utilization of hospitals and increased births in health facilities over the years.

The overall prevalence of neurological impairments was significantly reduced from 61 per 1000 to 44.7 per 1000 in 2015. This decline was driven, in part, by a decline in the prevalence of epilepsy (from 41.0 per 1000 to 17.2 per 1000). The 2001 Neurological Impairment Survey was conducted when malaria transmission was high^{13, 14} and the reduced prevalence of epilepsy may be attributed to a decline in seizures attributable to malaria and meningitis, which subsequently increased the risk for epilepsy.^{15, 16} The decline in the prevalence of epilepsy may also be attributed to sensitizations and prompt management of acute seizures, which reduced the risk of subsequent unprovoked seizures.²⁸

The introduction of Haemophilus and pneumococcal vaccines might have reduced the overall burden of neurological impairments, such as hearing and epilepsy, by reducing the incidence of childhood meningitis. Second, the prevalence of hearing impairments decreased between the two surveys possibly due to a shift from the use of gentamicin (ototoxic effects²⁹) to the use of cephalosporins/cefotaxime instead of aminoglycosides to treat bacterial infections in children. However, empirical studies are required to determine if changes in antibiotic use and the introduction of these vaccines are associated with a reduction in admissions to hospitals for children with neurological impairments. The decline in the prevalence of motor impairments is likely explained by the health benefits of an increased number of hospital deliveries in the years preceding the latter survey,³⁰ although some would also reduce due to the control measures for infectious diseases described earlier, particularly Plasmodium falciparum malaria.³¹

The prevalence of cognitive and vision impairments did not significantly change between the two surveys; this could mean that known (and unknown) risk factors for these impairments, such as adverse perinatal and neonatal events, are unaddressed. Admissions with neonatal illnesses from a separate data set increased over the years;¹⁷ some were associated with individual domains of neurological impairment in either surveys in this and other studies.^{32, 33} Notably, the increased rate of hospital admissions in this data set might reflect a general improvement in health utilization in the community that may: (1) reduce the long-term risk for neurological impairments from gains made by hospital management of these risk factors; and (2) indirectly improve the prognosis of neurological impairment if older children continue using hospital services as is the case for those with adverse perinatal events. However, a direct inference might be impossible because we are comparing different cohorts of children from different studies conducted at different time points. Visual impairments were associated with incomplete immunization in 2001, possibly because children with incomplete vaccine schedules did not receive vitamin A supplements or had vaccine-preventable infections like measles in childhood;^{34, 35} however, other risk factors for neurological impairment driving high prevalence may have emerged.

Some risk factors were significantly associated with neurological impairments in both surveys. For instance, adverse perinatal events were positively associated with epilepsy and cognitive, motor, and hearing impairments. This is not a surprising finding because adverse perinatal events predispose newborn infants to hypoxia, which is an established cause of encephalopathy and other neurological sequelae.^36-38 Also, the association between malnutrition and neurological impairment and disability is well established;³⁹ thus, it is not surprising that higher weight-for-age and height-for-age z-scores were protective against cognitive, motor, and hearing impairments. However, the effect sizes for the association between nutritional status and neurological impairment were reduced by 2015, which might reflect changes in the nutritional status of these children in the community. The other observed interactions by year of survey; for example, adverse perinatal events might be explained by changes in the utilization patterns of maternal and child health care services in Kilifi.

Some risk factors were significant in either of the two surveys, with others such as previous hospitalization being associated with most domains of neurological impairment in both surveys. Neonatal jaundice and incomplete vaccination were significantly associated with neurological impairment more in 2001 than in 2015, which may suggest that there was improved management of these conditions in subsequent years, including the 2015 survey period. Increased access and utilization of health care services might explain the importance of atypical deliveries in 2015. Family history of seizures was important in 2015, when there may have been better characterization and understanding of the phenotype, while lack of association in 2001 may underpin the increasing role of environmental–genetic interaction in later years. Some risk factors in both surveys could be inherently related to the nature of neurological impairments, for example, hospitalization where individuals with severe neurological impairment are expected to visit hospitals regardless of the survey period. Hospital admissions might be a consequence of neurological impairment since children with neurological impairments may seek out services, for example, patients with epilepsy seeking antiseizure medication, or related to neonatal conditions, such as jaundice and sepsis or encephalopathy.⁴⁰

The rate of hospital admissions with preterm birth, LBW, and neonatal encephalopathy increased between 2005 and 2015. The observed increase may reflect improved utilization of health services or better reporting in the latter years and not necessarily an increase in the incidence of these conditions. While it is recognized that adverse perinatal events predispose to neurological impairment, we hypothesize that the increase seen in our hospital admissions does not directly affect the prevalence of reported neurological impairments because: (1) the hospital admission study did not start in 2001 when the baseline survey was conducted; and (2) participants in the hospital admission study are different from those participating in the surveys. It may require interventional and/or longitudinal studies to make a meaningful conclusion of the impact of such changes on the burden of neurological impairments. Besides, our hospital estimates may grossly underestimate the incidence of preterm birth and LBW, which are lower than elsewhere in sub-Saharan Africa⁴¹ because a significant number of deliveries still occur at home, there is inadequate monitoring of deliveries occurring in hospitals in most LMIC,⁸ and more than half of the infants are not weighed at birth. However, our estimate for the annual admission with neonatal encephalopathy rate in 2010 (17.0 per 1000) is comparable with an estimated incidence rate of 14.9 per 1000 from studies conducted in similar settings.⁸

In the spatial analysis, the location-level estimates of prevalence reflected the general decline observed in the larger KHDSS. However, the decline was significant in 4 out of 15 locations (Kilifi township, Mtwapa, Tezo, and Ziani). The observed change in Mtwapa might be a chance observation because there were very few participants from this location in 2001. The decline in the other three predominantly semi-urban settings might represent locations that can easily access hospitals and are within reach of public health interventions. We also sought to identify areas with high or low clustering of neurological impairments in 2001 and 2015. Neurological impairments were generally homogeneously distributed in the KHDSS in both surveys except for one hotspot located towards the south of the KHDSS in 2015. Since the current study observed less geographical heterogeneity in the geographical distribution of neurological impairments, appropriate interventions on the significant risk factors may be applied evenly throughout the study area.