Individual and household level factors associated with presence of multiple non-communicable disease risk factors in Kenyan adults

Data source, study participants and sampling

The WHO STEPs survey, carried out in Kenya between April and June 2015, employed a cross-sectional household study design that targeted adult individuals aged 18 years and older. A three-stage cluster sampling was used to, in the first stage, select 100 each of urban and rural clusters from the the fifth national sample surveys and evaluation programme (NASSEP V) sampling frame by the Kenya National Bureau of Statistics. In the second stage, a sample of 30 households in each cluster were identified, while in the third stage, one adult (aged between 18 and 69 years) from each household was randomly selected to participate in the study. Further details regarding the design of the study have been published previously [29].

Data collection, measurements and definitions

Data was collected using personal digital assistants (PDA) loaded with the eSTEPS tool provided by the World Health Organisation [31]. Data was collected on the four main behavioural risk factors for NCDs (tobacco use, harmful use of alcohol, unhealthy diet and insufficient physical activity), as well as on the key physiological risk factors for NCD: overweight and obesity, raised blood pressure/hypertension, raised blood lipids and raised blood glucose/diabetes mellitus. Anthropometric measures for height and weight were also collected.

Blood pressure (BP) was diagnosed using a validated oscillometric automated digital BP device (OMRON™ digital automatic BP monitor). Using appropriate cuff sizes, three readings were taken on the left arm from a respondent in a seated position, at one minute intervals. The mean of the second and the third measurement was recorded. Raised blood pressure/hypertension was based on blood pressure readings cut-off for systolic > = 140 mmHg and/or diastolic > = 90 mmHg or if they reported to have been previously diagnosed and informed to have hypertension by a health care worker and/or were taking medication for raised blood pressure.

For a small number of respondents (84/4066), random blood sugar was taken because they did not fast as instructed. For all other respondents, fasting blood glucose measurements were taken a day after the first contact with the respondent, after they confirmed to have had an overnight fast. A drop of blood from a finger prick was used to test for glucose levels using a digital meter (ACCUCHECK™ glucometer and test strips). An individual was classified to have raised blood glucose/ diabetes mellitus based on random and fasting blood capillary glucose measurements cut-offs of 11.1 mmol/L and > 7.0 mmol/L respectively, or if they reported to have been previously diagnosed and informed to have diabetes by a health care worker and/or were taking medication for diabetes.

Total blood cholesterol levels were categorized either as ideal or high, with a cut off of point of 5.2 mmol/L or if an individual is currently on treatment for raised blood cholesterol. Low blood HDL cholesterol levels were categorized either as ideal or high, with a cut-off point of 1.17 mmol/L. A drop of blood from a finger prick was used to test for cholesterol levels using a digital meter (ACCUCHECK™ glucometer and test strips). A drop of blood from a finger prick was used to test for cholesterol levels using a digital meter (ACCUCHECK™ glucometer and test strips). Sugar intake was self-reported. High sugar intake was defined as regular consumption of excessive sugar in the form of sugary drinks and soda, processed foods high in sugar content on a regular, even daily, basis.

BMI was computed from height and weight measures, collected in centimeters and kilograms using SECA™ height boards and calibrated digital weigh scales respectively. Height was taken while the individual stood in an upright position on a flat surface. Obesity was diagnosed based on international guidelines recommended by the WHO expert consultation for 2008 [32]. General obesity was based on BMI cut-off of ≥30Kg/m² and central obesity on waist circumference of > 94 cm for males and > 80 cm for females.

The other risk factors/variables were self-reported. Insufficient fruits and vegetables consumption is when an individual consumes less than the recommended five servings of fruits or vegetables daily. A serving of fruit/vegetables equates to one small fruit, ½ a cup of raw vegetables, one cup of leady greens or one banana. Insufficient physical activity is less than 150 h/week of moderate-intensive activity or less than 75 h/week of vigorous-intensive physical activities, which may include walking and cycling. An individual that spent 180 min/day sitting or reclining was also classified as physically inactive.

Unhealthy cooking fats referred to the use in cooking of saturated oils and solid fats e.g. lard, margarine, butter and vegetable fat for cooking, instead of unsaturated fats and oils. High salt intake was based on an individual reporting that they consumed salty and processed foods always or often, or if they added salt to cooked food. Excessive alcohol use was defined as the consumption of more than two standard drinks per day for females and more than three standard drinks for males. One standard drink contains about 14 g of pure alcohol, equivalent to 350 ml of beer with 5% alcohol by volume (alc/vol), or 1 glass table wine with 12% alc/vol. Current tobacco use was defined as the daily use of smoked or smokeless tobacco, in the form of cigarettes, cigars, pipes, snuff and other local tobacco products.

Household wealth

A proxy index for wealth was created based on household-level variables including type of dwelling unit; ownership of the dwelling unit; construction materials of the dwelling unit specifically the roof and floor; source of cooking and lighting fuel; several household possessions/goods including a wall clock, electronics like radio and television, refrigerator, and furniture; source of water for drinking and cooking; and type of sanitation facility available for use by the household. A wealth index was then generated using principal component analysis (PCA), multivariate statistical technique. Principal components are weighted averages of the variables used to construct them. The generated index was then used to categorize the households into five categories (quintiles): poorest, second, middle, fourth and richest.

Data analysis

The analysis was done using Stata 14.1 (Stata Corporation, College Station, TX). The outcome variable for this study is the number (count) of NCD risk factors in an individual in any combination out of the twelve described NCD risk factors. The number of NCD risk factors in an individual were categorised into three groups: 1–3, 4–6 and 7+ risk factors. Proportions and bar-graphs were used to summarize frequency/counts of NCD risk factors. Ordered logistic regression was used to assess the association between individual and household level sociodemographic factors with the presence of one or more NCD risk factors in an individual. Proportional odds assumption was tested using brant command method in Stata, with non-proportional odds model fitted wherever the assumption is violated.

We first fitted bivariate (unadjusted) ordered logistic regression models for each of the sociodemographic factors: age, sex, level of education, occupation, marital status, place of residence, ethnic group and household wealth-status with the outcome variable. We then fitted a multivariate model to assess the strength of association of each of the sociodemographic factors with the number of NCD risk factors in an individual. We used the likelihood ratio test to compare the goodness of fit, comparing a null model which is a special case of the alternative model with one additional variable being tested. Odds ratios are reported together with their 95% confidence intervals. Stratified analysis by sex was done to check for differences in the magnitude and pattern of the NCD risk factors between males and females, after confirming that there existed no statistical interactions between the sociodemographic variables in question with sex. Analysis was weighted at the individual level to account for study design, and to allow for generalizability of the findings.

Ethical considerations

The study protocol was reviewed and approved by the Scientific and Ethics Review Unit at Kenya Medical Research Institute (Kemri) SSC No. 2607. Informed consent was sought from each and every participant prior to enrolement in the study. The study team was introduced to concepts of research ethics and encouraged to protect the rights of human research participants. Researchers upheld justice, regard for welfare and respect for study participants at all times during the study. During analysis and reporting, personal identifiers were delinked from the data.

Characteristics of study participants

The weighted prevalence of the twelve NCD risk factors in this population is shown in Fig. 1. Overall (both sexes), the prevalence of insufficient consumption of fruits and vegetables, high salt consumption and insufficient physical activity were 99.8, 89.5, and 80.3% respectively. One in four individuals had high blood pressure while the overall prevalence of diabetes was 2.6%. Majority (80%) of individuals with raised blood pressure were new cases diagnosed in this survey, while about 50% of the individuals with raised blood glucose did not know their status before being diagnosed in this study.

Harmful alcohol use, daily tobacco use, hypertension, low high density lipoprotein (HDL) cholesterol, bad fat intake, and high sugar intake were more prevalent in males than females, 24.5% vs 3.7%, 17.9% vs 2.9%, 25.9% vs 23.7%, 66.5% vs 49.3%, 44.4% vs 35.4%, 16.5% vs 13.3%, while diabetes, obesity, and raised blood total cholesterol were more common among females, 2.9% vs 2.3%, 56.5% vs 22.4%, 12.8% vs 7.2%, respectively.

Occurence of non-communicable disease risk factors in the population

The highest number of risk factors recorded in an individual was ten out of the possible 12 NCD risks factors considered in this study. The bulk of the population (75.8%) possesed between four and six NCD risk factors. The biggest proportion, 30.4%, possessed five; 25.3% four; and 20.0% six NCD risk factors. A little less than 12% possesssed three, while only about 2 % possessed two or fewer NCD risk factors. About 10% possessed seven and more NCD risk factors. Figure 2 summarises the occurrence of the NCD risk factors, showing the differences between the sexes.

Factors associated with multiple non-communicable disease risk factors in Kenyan adult population

The factors associated with an increase in NCD risk count in the population were mainly driven by the female sex, and thus there existed some differences in socio-demographic factors for a high count of NCD risk factors observed between men and women. Men were one and half times more likely to have higher counts for NCD risk factors compared to women. Being in a marital union was also associated with higher odds (1.27) of possessing higher counts of NCD risk factors. Women in a marital union were one and half times likely to have a higher count of NCD risk factors, but not so among men.

Overall (both sexes), compared to the Kalenjin ethnic group, the Kamba, Kisii, and the Luo had lower odds of having a higher counts of NCD risk factors (0.45, 0.51, 0.38 odds ratio respectively), while the Meru had almost twice as high odds (1.92) of having a higher count of NCD risk factors compared to the Kalenjin. Among men, being a Kisii or Luo was associated with lower odds (0.48 and 0.25 respectively) of having a higher counts of NCD risk factors. However, while a woman of the Akamba ethnic group had lower odds (0.43) of having a high NCD rik factors count, a women from the Meru ethnic group had 3.58 odds of possessing higher NCD risk factor counts when compared to the Kalenjin women.

Strengths and limitations

Our findings suffer from one main limitation. Data on behavioural risk factors was based on self-reports and may be affected by potential under-reporting, especially on the socially discouraged practices like smoking/tobacco use and harmful alcohol consumption. That notwithstanding, our study applies a standard chronic disease risk surveillance approach (WHO STEPs) that confers comparability of our findings with those from other setttings and countries. Our findings provide new insights on the patterns and the distribution of multiple NCD risk factors together with their sociodemographic determinants nationally.