This was a collaborative project with the ministries of health in Tanzania and Kenya, funded by the UK Department for International Development and aimed at conducting detailed situation appraisal, policy development and implementation in both countries [1,2]. As part of the situation appraisal, two epidemiological surveys were conducted in urban Tanzania and rural Kenya, aiming to establish the prevalence of common mental disorders, psychosis and substance use and misuse in both settings. The results of the Tanzanian survey have been reported elsewhere [3,4,5]. This paper reports the results of the Kenyan survey in relation to the prevalence of psychotic symptoms in rural Kenya and their relationship with demographic, socio-economic and other risk factors. The results for common mental disorders (CMD) and substance use will be reported elsewhere.

Surveys of psychosis have been conducted using clinician administered instruments [6] which can establish both psychotic symptom severity and diagnostic category; using family reports [7], and using systematic assessment of psychotic symptoms by detailed interviews administered by non-medical interviewers, leading to enumeration of symptom frequency and severity, and estimate of probable psychosis [8,9].

The response rate was 87.6%. The prevalence of at least one psychotic symptom in this group was 8.1% with only 0.6% having two symptoms and none having three or more symptoms (Table 1). Thought insertions and paranoia were the commonest psychotic symptoms at 3.6 and 3.5%. The prevalence of CMD in this sample was 10.8%, largely comprising mixed anxiety depression (6.1%), panic disorder (2.6%), generalised anxiety disorder (1.6%) and depressive episodes (0.7%) (see Table 1). Table 2 shows that the presence of psychotic symptoms was highly associated with the presence of CMD, but was not associated with demographic or socio-demographic variables in this sample (Table 3), although type of home and presence of poor physical health were associated at the P < 0.1 level of significance.

There has been a paucity of systematic community surveys of psychosis using standardised instruments in Africa [15]. This rural study in Kenya using the PSQ has found an annual psychotic symptom rate of 8.1%. No relationship was found between prevalence of psychotic symptoms and those background variables studied.

The annual prevalence of one or more psychotic symptoms reported here was somewhat higher than findings from Ethiopia, where a prevalence of 6.0% was observed for psychotic symptoms using the Self-Reporting Questionnaire in rural Ethiopia [16], and in Tanzania using the PSQ where an annual prevalence of psychotic symptoms was 3.9% in an urban sample. Although population based surveys in the United States [17], the Netherlands [18], and New Zealand [19], have found somewhat higher prevalence rates for psychotic symptoms (28%, 17.5% and 20.1% respectively) in Britain, where the PSQ was used, the prevalence of psychotic symptoms was 5.5%, a figure more comparable to the current results [20].

The PSQ is an instrument to screen for psychotic symptoms and does not of itself confirm the presence of psychosis, and we would expect rates of one or more psychotic symptoms to be much higher than rates of three or more psychotic symptoms. Studies attempting to estimate rates of actual psychosis have been conducted in Ethiopia, Mozambique and Zanzibar. In Ethiopia, past urban month rates of combined schizophrenia and schizoaffective disorder were 0.7% using the CIDI [6], and psychotic illness was 0.3% based on clinical psychiatric interview [21]. The Mozambique study used key informants (the first person found in the randomly selected household able to answer on behalf of others) to identify disordered behaviour via vignette. The authors found higher lifetime prevalence of psychoses (4.4%) in the poorer rural area compared to 1.6% in Maputo city [7]. Using a similar methodology in Zanzibar, the rate of chronic psychosis was found to be 2.6/1000 and acute psychotic episodes 0.6/1000 [22].

The association between psychotic symptoms and CMD has been previously reported in the British national surveys [20], and in Tanzania [3]. Unlike most other studies–but similar to the recent Tanzania study [3]–age, gender and marital status were not found to be significantly associated with psychotic symptoms.

The current paper also investigated household status, housing type, education, income and ethnicity, given their previously reported associations with psychotic symptoms in Britain [23], but none of these relationships was significant at the p = 0.5 level after adjustment for other variables, although there was an association with housing type (significant at the 0.9 level) and with presence of physical illness (significant at the 0.7 level). The association with physical illness has been found more strongly elsewhere [24]. It may be that a larger sample size is needed to find socio-demographic associations, especially as this was a relatively homogeneous population in terms of the above variables.

The strengths of this study are the relatively good response rate, the use of a random sample of households, the use of instruments which provide comprehensive assessments of common mental disorders, and a systematic approach to screening for psychosis. Several limitations should be noted. While the sampling frame was well defined and based on the 2000 Kenya census which had enumerated all dwellings, adequate supervision of the implementation of the survey was difficult for financial and logistical reasons. Over-representation of heads of household indicates that the random sampling of individuals within each household was inadequately implemented.

The PSQ was not originally designed for Sub-Saharan Africa. It was therefore carefully scrutinised by local clinicians for content validity within the local cultural context, but was not tested against a gold standard interview. As the prevalence of psychotic symptoms in general population samples is commonly low, the overall sample size was not large enough to yield a great number of people reporting past-year psychotic symptoms, and therefore the power to detect associations was limited. We did not arrange to confirm probable psychosis with a follow-up clinical interview by trained psychiatrists, due to the high opportunity cost of such an exercise in a low income country with few psychiatrists (there is only one psychiatrist in Nyanza Province for around 5 million population). As always, the potential for measurement error when using screening instruments should be acknowledged, given self-reported experiences may be subject to recall or social desirability response bias. In addition the lay volunteer community health workers were not experienced survey interviewers. We do not know whether some of the psychotic symptoms here were alcohol or drug related, as the data on substance abuse was insufficiently complete to analyse. Finally, the current findings are specific to Maseno division in Nyanza Province and are not necessarily applicable to other parts of Kenya, particularly urban areas.

Single psychotic symptoms in rural Kenya are relatively common (8%) in the adult population, but with only 0.6% having two symptoms, and in this sample size none were found with three or more symptoms. Symptoms are evenly distributed across this relatively poor rural population, and are significantly associated with CMD, and to a lesser extent with poor physical health and housing type. Health and social care systems in Kenya should take psychotic symptoms into account when planning for both physical and mental disorders.