- Top of page
- Materials and methods
Objectives To assess whether longitudinal surveillance in peripheral health facilities could document trends in disease burden, healthcare practice and utilization resulting from large-scale public health interventions made in the decade 1997–2006.
Methods Data were collected from sick child visits (SCVs) among children <5 years attending 14 outpatient facilities in Asembo, rural western Kenya, during 1997–2006. Changes in proportions, and counts and rates were evaluated using chi-square and Poisson regression respectively.
Results During the decade, 64 394 SCVs were made, yielding an average rate of 0.70 SCVs per child-year. The annual number of SCVs stayed constant during 1997–2003, then increased by 74% between 2003 and 2006 (P < 0.01). The time between symptom onset and SCV shortened from 5.6 days in 1997 to 4.4 days in 2006 (P < 0.01). Malaria and upper respiratory tract infection (URTI) were most commonly diagnosed (69% and 36% of SCVs respectively). Between 2003 and 2006, the proportion of SCVs with a malaria diagnosis fell from 77% to 48%, although the rate of malaria SCVs did not. URTI visits increased in 2004–2006. The most frequently prescribed antimalarials changed three times, lagging 1–2 years behind changes in national policy. Treatment of pneumonia with antibiotics varied by year, ranging from 19% to 89%.
Conclusion Surveillance of paediatric SCVs at peripheral health facilities was useful in documenting the timing and penetration of changes in national policies and clinical patterns of drug use for common infections. The surveillance data suggested that improved access to care, rather than disease burden changes, likely led to greater healthcare utilization after 2003.
Objectifs: Déterminer si la surveillance longitudinale dans les services de santé périphériques pouvait documenter les tendances dans la charge de morbidité, la pratique et l’utilisation des soins de santé résultant d’interventions de santé publique à grande échelle appliquées au cours de la décennie 1997–2006.
Méthodes: Des données ont été recueillies au cours de visites d’enfants malades (VEM) chez les enfants de moins de 5 ans fréquentant 14 services pour patients ambulants à Asembo, zone rurale dans l’ouest du Kenya, au cours de 1997–2006. Les changements dans des proportions et les nombres et les taux ont étéévalués à l’aide de chi-carré et de la régression de Poisson, respectivement.
Résultats: Au cours de la décennie, 64394 VEM ont été effectuées, ce qui donne un taux moyen de 0,70 VEM par enfant-année. Le nombre annuel de VEM est resté constant au cours de 1997–2003, puis a augmenté de 74% entre 2003 et 2006 (p < 0,01). Le délai entre l’apparition des symptômes et la VEM s’est raccourcie de 5,6 jours en 1997 à 4,4 jours en 2006 (p < 0,01). La malaria et les infections des voies respiratoires supérieures (IVRS) ont été les plus fréquemment diagnostiquées (69% et 36% des VEM, respectivement). Entre 2003 et 2006, la proportion de VEM avec un diagnostic de malaria a baissé de 77%à 48%, bien que le taux de VEM avec malaria n’a pas baissé. Les visites pour IVRS ont augmenté en 2004–2006. L’antimalarique le plus souvent prescrit a changé 3 fois, avec un retard de 1 à 2 ans derrière les changements dans la politique nationale. Le traitement de la pneumonie avec des antibiotiques variait avec l’année, allant de 19%à 89%.
Conclusion: La surveillance de VEM pédiatrique dans les établissements de santé périphériques a été utile pour documenter les délais et la pénétration des changements dans les politiques nationales et les profils cliniques de l’utilisation des médicaments pour les infections courantes. Les données de surveillance ont suggéré que l’accès amélioré aux soins plutôt que les changements dans la charge de morbidité, a probablement menéà une plus grande utilisaion des soins de santé après 2003.
Objetivos: Evaluar si la vigilancia longitudinal en centros sanitarios periféricos podría documentar las tendencias en la carga de enfermedad, las prácticas en cuidados sanitarios y la utilización que resulta de las intervenciones de salud pública a gran escala realizadas durante la década de 1997–2006.
Métodos: Se recolectaron datos de visitas de niños enfermos (VNE) menores de 5 años que acudieron a 14 centros de consultas externas en Asembo, Kenia rural del oeste, entre 1997–2006. Se evaluaron cambios en proporciones, y conteos y tasas utilizando el chi-cuadrado y la regresión de Poisson, respectivamente.
Resultados: Durante la década que duró el seguimiento, se realizaron 64,394 VNEs, las cuales dieron una tasa de 0.70 VNEs por niño-año. El número anual de VNEs se mantuvo constante durante 1997–2003, y aumentó un 74% entre 2003 y 2006 (p < 0.01). El tiempo entre el comienzo de los síntomas y la VNE se acortó de 5.6 días en 1997 a 4.4 días en 2006 (p < 0.01). La malaria y las infecciones del tracto respiratorio superior (ITRS) eran las patologías más comúnmente diagnosticadas (69% y 36% de VNEs, respectivamente). Entre 2003 y 2006,la proporción de VNEs con diagnóstico de malaria cayó del 77% al 48%, aunque la tasa de VNE por malaria no lo hizo. Las visitas por ITRS aumentaron en el período 2004–2006. El antimalárico más prescrito cambió tres veces, con un retraso de 1–2 años con respecto a los cambios en la política nacional. El tratamiento de neumonías con antibióticos varió cada año, con un rango de entre 19% y 89%.
Conclusión: La vigilancia de VNEs pediátricas en centros sanitarios periféricos fue útil para documentar la coordinación y el alcance de los cambios en políticas nacionales, así como los patrones clínicos en el uso de medicamentos para infecciones comunes. Los datos de vigilancia sugieren que una mejora en el acceso a los cuidados, más que cambios en las cargas de enfermedad, podrían llevar a una mejor utilización de los cuidados sanitarios después del 2003.
- Top of page
- Materials and methods
To address the unacceptably high burden of morbidity and mortality among African children several large-scale interventions and policy changes have been implemented over the last decade (Black et al. 2003; Jones et al. 2003; Bryce et al. 2005). The most powerful measure of an intervention’s impact is a reduction in child mortality (Bryce et al. 2003). Reductions in child mortality have been documented through large-scale surveys following implementation of Integrated Management of Childhood Illness (IMCI), insecticide-treated bednets (ITNs) and vitamin A (Schellenberg et al. 2001, 2003; Phillips-Howard et al. 2003b; Binka et al. 2007). Mortality, however, is a distal measure of population-level health status changes and often is difficult to measure accurately in rural African settings where most deaths occur outside health facilities. Data gathered from hospitalizations underestimate the true burden of disease, and are biased towards severe disease and populations that live closer to the hospital, limiting their representativeness (Weber et al. 2002; Schellenberg et al. 2004a; Tornheim et al. 2007).
Tracking visits of sick patients to peripheral health centres is another way to document temporal changes in disease burden, healthcare practice and utilization. Rural African children more often use peripheral health facilities than hospitals and a wider spectrum of common diseases is seen there. Since July 1996, the Kenya Medical Research Institute (KEMRI) and the US Centers for Disease Control and Prevention (CDC) have maintained continuous surveillance of paediatric visits to peripheral health facilities in Asembo, western Kenya. Since 1996, there have been major changes in health policy in Kenya, as well as public health interventions within the surveillance site (Table 1). This study’s objective was to assess whether longitudinal surveillance in peripheral health facilities could document trends in disease burden and healthcare practice resulting from these changes in the decade 1997–2006.
Table 1. Major policy changes and public health interventions in Kenya and in KEMRI/CDC surveillance site, Asembo, Kenya, 1997–2006
|1997||Free ITNs for half of the children under 5 years in Asembo as part of study of impact of ITNs|
|1998||Policy change from chloroquine to sulfadoxine–pyrimethamine as first-line treatment for malaria|
|1999||Free ITNs for all the children under 5 years in Asembo|
|2001||Early implementation phase of Integrated Management of Childhood Illness by Kenya MOH, not fully implemented in Asembo|
|2002||Large-scale social marketing of ITNs in Kenya begins|
|2002||Introduction of Haemophilus influenza type b and hepatitis B vaccines into Kenya EPI programme|
|2004||Interim policy change from sulfadoxine–pyrimethamine to amodiaquine as first-line treatment for malaria|
|2004||Study of intermittent preventive treatment of malaria in infants begins in four clinics, offering free care to participants|
|2005||Surveillance project begins in villages around Lwak clinic offering free care to all participating children|
|2006||Sulfadoxine–pyrimethamine is replaced by artemether–lumefantrine, as a first-line treatment for malaria (second half of year)|
- Top of page
- Materials and methods
Between 1997 and 2006, 64 394 SCVs were made to 14 peripheral health facilities. On average, there were 6439 visits per year with a minimum of 4341 visits in 2002 and a maximum of 10 392 visits in 2005 (Figure 1). There was no change in the number of SCVs during the first 7 years, but a 74% increase occurred between 2003 and 2006 (P < 0.01). Younger children accounted for a disproportionate number of SCVs; children <1 year accounted for 40% of SCVs and children <2 years for 66%. During the decade the percentage of SCVs made by children under 1 year decreased from 48.4% in 1997 to 35.6% in 2006 (P < 0.001, chi-square test for trend). The percentage of SCVs made by boys (51%) and girls (49%) was similar.
The average rate of SCVs was 0.70 visits per child-year and ranged between 0.47 visits per child-year in 2002 and 1.11 visits per child-year in 2005 (Table 2). In 1997–1998, 46% of children making a SCV were reported to have slept under an ITN the previous night (half the population was residing in control villages). From 1999 to 2006, ITN use the previous night ranged from 80% (1999) to 91% (2006).
Table 2. Yearly rate of SCV per child-year and percentage of total SCVs due to the most common diagnoses, Asembo, Kenya, 1997–2006
|Year||Malaria||Pneumonia||URTI||Gastro-intestinal infections||Total rate SCV per child-year|
|SCV per child-year||% of all SCV||SCV per child year||% of all SCV||SCV per child-year||% of all SCV||SCV per child year||% of all SCV|
Among the 55 410 SCVs made in the six health facilities included during the full decade, there was variability of individual clinic attendance over time (Figure 2). For example, Ongielo, the busiest clinic, varied the most annually in the number of SCVs from a low of 622 SCVs in 2002, when no nurse or clinical officer was present, to a high of 2761 SCVs in 2005. Three of the four busiest facilities had significant increases in SCVs between 2003 and 2006 (Figure 2).
Figure 2. Number of sick child visits (SCVs) to the four busiest peripheral health facilities in Asembo western Kenya, 1997–2006.
Download figure to PowerPoint
Over the decade, the number of days between symptom onset and attending a peripheral health facility declined by 0.15 days per year (P < 0.01). In 1997, people waited on average 5.6 days from symptom onset to SCV. This declined to 4.4 days in 2006.
The number of SCVs varied by month from a high in June (n = 620) to a low in October (n = 363, Figure 3). Each increase in SCVs followed 1–2 months after a peak in rainfall, but not every peak in rainfall was followed by an increase in SCVs; the association between monthly rainfall and SCVs was not statistically significant when using a 1–2 month lag-time.
Figure 3. Monthly trends in rainfall and sick child visits (SCVs) at six peripheral health facilities in Asembo, western Kenya, 1997–2006.
Download figure to PowerPoint
Fever and cough were the most common symptoms, reported, on average, in 86% (range: 81–89%) and 67% (range: 62–72%) of SCVs (Table 3). During the first 8 years of surveillance, the percentage of SCVs in which fever and cough were reported remained constant. In the last 2 years, there was a statistically significant decrease in the proportion of SCVs reporting fever (to 81% in 2006) and an increase in cough (to 72% in 2006). The proportion of SCVs reported with diarrhoea did not change over time (P = 0.33). Twenty per cent of diarrhoea was bloody.
Table 3. The most common symptoms reported for paediatric sick visits in Asembo, western Kenya, 1997–2006
|Symptom||Average (%†)||Minimum (%†) (year)||Maximum (%†) (year)|
|Fever*||86||81 (2006)||89 (2004)|
|Cough*||67||62 (1998)||72 (2006)|
|Not feeding||59||50 (1998)||67 (2003)|
|Vomiting*||38||31 (2006)||42 (1997)|
|Loss of weight*||33||19 (1998)||43 (2002)|
|Diarrhoea||28||24 (1997)||32 (1998)|
|Difficulty breathing*||24||18 (2006)||30 (1997)|
|Rash*||21||16 (2000)||26 (2006)|
Malaria and upper respiratory tract infection (URTI) were the most common diagnoses over time and represented on average 69% and 38% of SCVs per year (Table 2). In the last few years, the percentage of SCVs resulting in a malaria diagnosis decreased (77% in 2003 vs. 48% in 2006). However, the rate of SCVs with a malaria diagnosis did not decline noticeably in these years (Table 2). In contrast, during these years the percentage and rate of diagnoses of URTI increased. Co-diagnosis of malaria and URTI occurred in 26% of SCVs. Gastro-intestinal infections (diarrhoea, dysentery and gastroenteritis) accounted for an annual average of 13% of diagnoses (annual range: 11–16%), while pneumonia accounted for an annual average of 7% of diagnoses (annual range: 5–11%). Although the percentage of SCVs with these diagnoses did not change significantly over the whole decade, both increased slightly between 2003 and 2006 (Table 2). The distribution of diagnoses did not vary significantly by clinic.
Overall 1–5% of the SCVs led to hospital referral. Children diagnosed with malaria, pneumonia and malnutrition were referred to the hospital in <0.5%, 8% and 6% of their visits respectively. There was no trend in referrals by season or year.
Analgesics (mostly paracetamol) were the most commonly prescribed medications throughout the decade, prescribed in 78% of SCVs. Malaria treatment changed substantially over the decade (Figure 4). Until 1999, chloroquine was the most prescribed medicine for malaria. By 2001, sulfadoxine–pyrimethamine (SP) was the most commonly prescribed antimalarial, and the use of chloroquine at government health clinics for first-line antimalarial treatment disappeared in 2003. SP use peaked at 80% of malaria SCVs in 2003. Its use declined as amodioquine use increased, surpassing SP use in 2006. In June 2006, artemether–lumefantrine was introduced as first-line treatment for malaria and accounted for at least 7% of antimalarial prescriptions thereafter, although this is likely an underestimate as artemether–lumefantrine had to be written in the ‘other antimalarial’ category on the form. Of children receiving an antimalarial, on average 42% also received an antibiotic, varying from 27% to 63% by year.
Figure 4. Malaria treatment at six peripheral health facilities in Asembo, western Kenya, 1997–2006. SP, sulfadoxine–pyrimethamine.
Download figure to PowerPoint
Use of antibiotics among SCVs with pneumonia decreased from 66% in 1997 to 20% from 1999 to 2002, when the country experienced a penicillin shortage (Figure 5). It increased to over 60% again after 2002. The most commonly used antibiotic changed over time from septrin (trimethaprim–sulfamethoxazole) in 1997–2002, to penicillin V in 2003, to amoxicillin after 2003. In addition to the most common antibiotics shown in Figure 5, gentamicin was given for pneumonia treatment in <10% of pneumonia SCVs throughout the decade, and tetracycline was given in <5% of pneumonia SCVs; use of neither antibiotic changed significantly during the decade. On average, 63% of URTI SCVs resulted in antibiotic prescriptions, ranging from 8% in 2001 to 89% in 2004.
Figure 5. The use of the most common antibiotics for pneumonia treatment at six peripheral health facilities in Asembo, western Kenya, 1997–2006. Pen V is penicillin V and septrin is trimethoprim–sulfamethoxazole.
Download figure to PowerPoint
Of SCVs with gastroenteritis, 74% received antibiotics in the first 6 years, declining to 43% in the last 2 years. Septrin was prescribed most commonly (29–62% of gastroenteritis), followed by metronidazole (14–68% of gastroenteritis). Oral rehydration solution was given in 60% of gastroenteritis cases from 2003 to 2006, the only years its use was recorded.
- Top of page
- Materials and methods
This study provides one of the longest continuous periods of passive surveillance of paediatric visits to peripheral health facilities in Africa. The data demonstrated several temporal trends related to the evolving policies of the Kenya MOH, drug supply fluctuations and changes in clinical practice during the decade. The most evident trend was the change in antimicrobial use. The programmatic implementation of a new first-line antimalarial in the peripheral health facilities lagged 1–2 years behind the MOH’s policy change. It took 2 years from the time the national malaria policy changed in 1998 for SP to replace chloroquine as the most commonly used drug, and chloroquine use was still over 10% in 2001, the year the Kenya MOH prohibited chloroquine use (Shretta et al. 2000; Phillips-Howard et al. 2003c). Likewise, in 2004 when the Kenya MOH changed its recommendations for first-line treatment from SP to amodioquine as the interim first-line antimalarial drug until artemether–lumefantrine became available, it took until 2006 for amodiaquine to replace SP as the most commonly prescribed antimalarial. Treatment of pneumonia also changed dramatically over the decade. The percentage treated with any antibiotic ranged from 20% to 90% and the type of antibiotic most used to treat pneumonia also changed in different years. Antibiotics were likely overused in cases of acute non-dysenteric gastroenteritis and URTI, with over half of children with these diagnoses receiving antibiotics.
Another salient trend was an increase in the number of SCVs in 2004–2006. We believe this increase was not due to a true increase in disease burden. There was no documentation of increases in any diseases from other surveillance sources in Bondo District during that period (Bondo District MOH data). It is unlikely that there was a sudden increase in the numbers of HIV-infected children in these years because no jump in HIV prevalence among pregnant women was noticed and PMTCT services had just become more available in 2003. Lastly, child mortality decreased markedly between 2003 and 2006 in the DSS area, which would not have been expected if there were a real increase in major infectious diseases (unpublished CDC data).
Most likely, the increase in SCVs in the last 3 years was due to changes in access to care. This is supported by the decreasing lag-time during the decade between symptom onset and visiting the clinic. Second, the substantial increase in a non-severe illness, such as URTI, suggests that parents were bringing their children in for milder illnesses. Third, the large increase in SCVs in 2004–2006 was observed disproportionately in some clinics. In 2004, a study of intermittent preventive treatment of malaria in infants, in which free health care and attendance by a study-hired clinical officer was available to all enrolled children, was started in four clinics – Ongielo, Lwak, Saradidi and Abidha. These were the clinics that saw the greatest increase in these years. Moreover, an enhanced morbidity surveillance project that offered free care by study-hired clinical officers in villages near Lwak Hospital began in late 2005; Lwak had a 114% increase in SCVs between 2004 and 2006. Not only free care, but improvements in drug supply and clinical staff, can lead to improvements in clinic utilization, as exemplified by the increased attendance in Saradidi after 2002 when community health workers were replaced by MOH nurses. Another factor possibly improving access was the paving of a major road into Asembo in 2004–2005 that passes very near to Ongielo clinic, which might have led to an increase in SCVs there. Access alone is increasingly seen as a key component of health utilization and improved health for children in developing countries (Hayes 1990). Access to care includes several variables like distance, cost, availability of drugs and clinical staff, whereas healthcare utilization refers to the actual visiting of the clinic. Along with preventive measures like immunizations and ITNs, access to care, which includes minimizing distance to clinics, making care affordable and providing high quality of services, is a key component of efforts to decrease child mortality (Schellenberg et al. 2003; Bryce et al. 2005). In Tanzania, both the introduction of IMCI, with a reliable supply of drugs and trained clinicians, and the introduction of effective artemisinin-based combination antimalarial therapy led to augmentation in care-seeking among children (Schellenberg et al. 2004b; Bhattarai et al. 2007).
This surveillance system had limitations. The diagnoses were based on non-standardized clinical impressions. IMCI had not been formally implemented in the area. Turnover of clinic staff during the decade likely also led to variability in diagnosis. Moreover, none of the clinics except Lwak had the capacity to do blood smears for malaria; therefore history of fever in a child resulted in malaria diagnosis, which was consistent with national guidelines although likely led to overdiagnosis of malaria. Supporting this assertion is the finding that in Lwak in 2005–2007, when blood smears were done consistently, only 34% of children with reported fever had malaria parasitaemia. Community prevalence of malaria parasitaemia in children in this area has decreased by about 50% from the early 1990s to 2006, yet the rate of SCVs with a malaria diagnosis did not decline appreciably over this time period, although the proportion of SCVs with malaria did [(Ter Kuile et al. 2003), CDC unpublished data]. As malaria transmission is lowered as a result of ongoing malaria control scale-up efforts, continued presumptive malaria treatment will waste costly antimalarial drugs, create missed opportunities to diagnose and treat other causes of fever, potentially contribute to antimalarial drug resistance, and limit ability to track the impact of malaria interventions. Malaria treatment based on laboratory diagnosis with rapid diagnostic tests or microscopy where feasible will be needed to strengthen malaria programs and general health services.
Another limitation is that no facilities had x-ray capacity nor did lumbar punctures or blood cultures, limiting the accuracy of diagnosing pneumonia and meningitis. This limited the ability to document changes resulting from introduction of the Hib vaccine because Hib presents as pneumonia or meningitis, syndromes caused by many aetiologies. Moreover, because of the limited ability to diagnosis HIV in the outpatient setting during these years, the role of HIV in the trends in SCVs during the decade is unclear. Lastly, the morbidity and care-seeking behaviours of the Asembo population might not be representative because of the long-history of public health research in the area.
Because of these limitations, we believe this surveillance system was not sensitive or specific enough to accurately document changes in disease burden over the decade. To monitor temporal trends in disease burden, a passive outpatient surveillance system would need to be in an area with stable access to care and with clinics that provide consistent and specific diagnostic testing. Few rural areas in Africa today meet these criteria. Yet, passive surveillance in peripheral health centres has utility in some situations when monitoring disease burden. Recent efforts to implement the Integrated Disease Surveillance and Response model in Africa have focused on laboratory diagnostics for most of the diseases under surveillance (Anonymous 2003). The resultant increase in specificity of diagnosis is important when identifying outbreaks of disease and when monitoring progress towards diseases for eradication. In addition, passive surveillance has been used to document the impact of large-scale interventions in study settings. The introduction of ITNs into some villages in Asembo in the 1990s was shown to decrease malaria SCVs in those villages, a finding corroborated by more specific malaria indicators, like parasitaemia and anaemia (Phillips-Howard et al. 2003b; Ter Kuile et al. 2003).
In summary, surveillance at peripheral health facilities in rural Kenya from 1997 to 2006 was useful in documenting the impact of several important health policy changes. These data allow the Kenya MOH to monitor the penetration of changes in drug policy and usage, the impact of drug shortages, and areas where enhanced clinical training is needed. The surveillance also suggested that improvements in access to care and health services led to greater healthcare utilization, which is important information for the Kenya MOH as they continue to implement health sector reforms (Ministry of Health 1997). We believe though that this surveillance could not adequately document changes in disease burden over the decade because of changes in access to care and lack of consistency and specificity of diagnostic criteria. To monitor widespread changes in disease burden, an essential aspect of most large-scale interventions in Africa, different and more comprehensive models of surveillance based on diagnostic criteria rather than syndromic definitions will likely be needed.