Figure 4

Figure 4. Malaria outpatients, inpatients, and deaths, versus time.

Figure 3

Figure 3. Positive slides, rainfall, and average temperature, per month.

Central Hospital illness and death data were available from 1990 through 1997, except for outpatient and death data from October to December 1992. Tea company temperature data were available from 1984 through 1997, and rainfall data from 1971 through 1997 (Figures 1,2). The retrospective illness data clearly show the nearly annual occurrence of epidemic malaria (Figures 2,3) and the importance of known factors in highland malaria transmission; rainfall >150 mm per month in this area promotes epidemic malaria, and average temperature <18°C appears to extinguish epidemics. In the Kericho area, malaria epidemics occur during May to July, after the long rains and before the temperature drops in July (Figure 3). Figure 4 shows outpatient and inpatient malaria and malaria deaths from 1991 through 1997. For outpatient malaria, the mean annual rate was 48,112 per 100,000 (range 33,914 to 56,040 per 100,000), i.e., each year malaria was diagnosed in approximately 50% of company health-care beneficiaries; for malaria admissions the mean annual rate was 857 per 100,000 (range 361 to 1,058 per 100,000); and for malaria deaths the mean annual rate was 42 per 100,000 (range 14 to 64 per 100,000). In hospitalized patients malaria diagnosis was laboratory confirmed, but in most outpatients malaria was diagnosed purely on a clinical basis. Of the 764 inpatient deaths documented from 1991 through 1997, 254 (32%) were caused by malaria, the leading cause of death in this highland population. Of the 254 malaria deaths, 37% were recorded as being due to severe anemia, 31% cerebral malaria, and 32% malaria with gastroenteritis or pneumonia.

Questionnaire results were available for 244 (95%) of the 257 patients admitted for malaria in June 1997. The remaining 13 patients were either discharged before interview and could not be traced to their houses on the estates, died before interview and the family was not available for questioning, or were minors whose parents were not available for interview. These malaria hospitalized patients were 48% female, 52% male, with mean age 13.2 years (range 1 month to 58 years); 20% were tea company employees; 80% were dependents. Only 16% of malaria hospitalized patients belonged to ethnic groups originating from areas with holo/hyperendemic malaria, versus 32% of all company employees (relative risk = 0.4 [95% confidence interval 0.29 to 0.57] for hospitalization for malaria if ethnic origin was from a holoendemic-disease area versus a nonendemic disease area). Forty-three (18%) of the patients hospitalized with malaria had traveled away from the Kericho area in the 30 days before admission, with an average of 19 days away (range 1 to 90 days). Of the 43 hospitalized patients who had traveled, 19 had been to malarious areas; therefore, only 8% of the 244 malaria hospitalized patients had recently traveled to an area with known malaria transmission.

These data document the reemergence of recurrent epidemic malaria due to local transmission in the highlands of western Kenya. Although malaria in the highlands was a serious problem after World War II, because of improved transportation and increased immunity, recurrent epidemic malaria had not been seen in a generation. Several pertinent factors have remained stable and do not appear to explain the increase in malaria in this area. Firstly, climate data from the Kericho area show no obvious change in average temperature or rainfall over the last 10 to 20 years that would explain the present almost annual epidemics in the highlands. Although the climate data used in this study may not be sensitive enough to detect small changes in environmental temperature, global warming would also not explain the epidemics seen in the 1940s to 1950s. Deforestation may be relevant, providing more breeding sites for vector mosquitoes in sunlit pools and perhaps leading to localized changes in weather patterns and increased microenvironmental temperatures (1).

Secondly, travel of infected persons from adjacent malaria-endemic areas is not new. The main road from the malaria-holoendemic Lake Victoria basin to this highland area was built in the 1950s; in fact, people had been making the same journey using ox wagons decades before the road was paved (5,6). According to the tea company's central office, the tribal/ethnic composition of the employees has remained stable at least since the 1970s, as has the total number of health-care system beneficiaries. Therefore, the increase in malaria does not seem to be explained by an increase in the number of nonimmune relative to semiimmune/immune persons nor simply by an increase in the total population.

A third factor that could lead to an impression of increasing malaria would be the degradation of the health-care infrastructure: if the system became overwhelmed by other diseases and lack of funding, it would be unable to cope with the same number of malaria cases as before, leading to increased hospitalizations and deaths. Indeed, the public health sector in much of sub-Saharan Africa has been overwhelmed by the AIDS epidemic, lack of funding, and other institutional problems; however, the tea company in this study continues to provide excellent care.

One hypothesis for the increase in highland malaria is the failure of the drugs used to cure malaria infections. The period during which epidemic malaria was absent from the highlands corresponds to the time when pyrimethamine and chloroquine were still effective malaria treatments. In 1953, pyrimethamine was used successfully as the first step of a large malaria control campaign in the Kenya highlands, but by 1954, rapid development of resistance was reported in several areas (9,10). Nevertheless, widespread use of chloroquine as well as incorrect use of other malaria medications continue. Clinical control of symptoms (but not cure) creates a large number of semiimmune gametocyte carriers, who can infect mosquitoes when transmission is favored at midyear. In this way, the failure to cure falciparum malaria infections leads to an increase in the human gametocyte-carrier pool, resulting in the rapid spread of epidemic malaria among the largely nonimmune highland population. Studies are under way to further investigate this hypothesis. While increased drug resistance may be the major factor in the documented increase of epidemic malaria, the causes are undoubtedly multifactorial and include vector, host, and environmental components.

Several possible interventions to stop the cycle of epidemic malaria and the associated high costs may be worth exploring. Spraying of houses on the tea estates with residual insecticide could be effective against the endophilic and endophagic A. gambiae (11), which appears to be the primary vector in these highlands (5,7). Indoor spraying campaigns in the past in western Kenya with DDT (12) and Dieldrin (7) have effectively combated malaria epidemics. Even now, worker acceptability would likely not be an issue as houses are regularly sprayed for other public health reasons. Insecticide-impregnated bed nets may also be effective with proper education and follow-up. Finally, improved treatment regimens to cure malaria infections, gametocytocidal drugs to prevent transmission, and chemoprophylaxis of workers during travel to malaria-holoendemic areas should be evaluated. Protection of a well-defined, economically important, and less immune population may make effective certain strategies of drug use, such as intermittent chemoprophylaxis, which would not normally be considered in areas with higher levels of malaria endemicity.