Study design

Our study employed a prevalence-based analysis model, as recommended by Pérez-Ríos and Montes¹² and Pérez-Ríos et al.¹³, for estimating attributable mortality related to tobacco smoking. This model relies on the population attributable fraction (PAF), quantifying the proportion of deaths in the population attributed to a specific risk factor, such as tobacco smoking. Specifically, it calculates age-specific smoke attributable mortality (SAM) rates for individuals aged ≥35 years, factoring in age, sex, and cause-specific mortality rates. This method, widely acknowledged for SAM calculations, was deliberately chosen due to the absence of cancer mortality data among never smokers in Kenya, ensuring a robust estimation of mortality attributable to tobacco smoking.

The prevalence-based approach proves effective in situations where detailed cancer mortality data for non-smokers are unavailable. By employing this model, we could confidently estimate the mortality linked to tobacco smoking in Kenya. The model’s consideration of age-specific SAM rates, along with demographic and cause-specific factors, enhances the precision of our estimations.

This methodological choice offers notable advantages. Firstly, it allows the computation of SAM rates even when comprehensive cancer mortality data for non-smokers are lacking. Secondly, it adheres to international standards for estimating tobacco-related mortality¹³. Thirdly, it facilitates meaningful comparisons with other studies utilizing similar prevalence-based models, thereby enhancing the generalizability of our findings.

By consciously selecting the prevalence-based analysis model, our objective was to provide a comprehensive and dependable assessment of the mortality burden attributed to tobacco smoking in Kenya, thereby contributing to a broader understanding of the impact of tobacco use on public health.

Non-communicable disease causes of death

Tobacco use has been linked to non-communicable diseases such as cancers, cardiovascular diseases, chronic respiratory diseases, tuberculosis^14-16, and diabetes¹⁷. We adopted a similar disaggregation by sex and age-group in our study. Recognizing that the effects of tobacco use manifest later after smoking initiation, we focused on causes of deaths observed in individuals aged ≥35 years in Kenya between 2012 and 2021. Causes of deaths were stratified into two age groups: 35–64 years and ≥65 years. Additionally, causes of deaths were stratified by sex.

Observed all-cause mortality

To determine all-cause mortality, we utilized data from the United Nations website¹⁸ for annual population estimates. Crude death rates for tobacco-related diseases were calculated against national population projections. Joinpoint regression analysis using Joinpoint software 4.9.1.0-April 2022 (Statistic Research and Application Branch, National Cancer Institute) was applied to identify changes in mortality rate trends for every age and sex group. This method determines the year(s) when a trend change occurs based on crude mortality rates. We integrated population data with all-cause mortality data to model deaths attributable to tobacco smoking in Kenya between 2012 and 2021. The mortality data were extracted from the Kenya Health Information System (KHIS), encompassing de-identified, case-based data from health facility-based deaths in Kenya.

All-cause mortality data between 2012 and 2021 were systematically reviewed from hospital medical records. A total of 500 facilities were sampled, including 3 national teaching and referral hospitals and a stratified sample of sub-county and faith-based hospitals. Recertification of the cause of death was performed using WHO-recommended forms, and data were coded using the 10th International Classification of Diseases (ICD-10)¹⁹.

Smoking prevalence in Kenya: mid-year analysis

Data on smoking prevalence were obtained from the STEPS survey 2015⁶, a national cross-sectional household survey covering individuals aged 18–69 years. The GATS survey 2014⁷ was also utilized, which used multistage stratified cluster sampling of 5376 households. Both surveys provided insights into tobacco use and measures of interest. Relative risks of death among smokers and ex-smokers compared to non-smokers were derived from recent systematic reviews and the Cancer Prevention Study II of 1982–1988. For our research, we used the relative risks proposed by the Cancer Prevention Study II and the Royal College of Physicians in 2020²⁰.

Calculation of smoking attributable mortality

The study calculated SAM for each cause of mortality using the formula:

where OM represents observed mortality, and PAF is the population attributable fraction. PAF was determined by the formula:

with p0, p1, and p2 representing the prevalence of non-smokers, current smokers, and former smokers, respectively, and RR1 and RR2 are the risks of dying from any cause for current smokers and ex-smokers, respectively. The SAM formula encapsulated a comprehensive estimation of mortality associated with tobacco smoking, considering the prevalence and risks of different smoking categories. The calculation allowed for the determination of the proportion of deaths attributed to smoking, providing a quantifiable measure of the mortality burden associated with tobacco smoking in the studied population. The methodology integrated data from multiple sources and employed statistical analysis techniques to estimate tobacco-related mortality in Kenya, considering age, sex, and specific causes of death. Ethical considerations, data reliability, and the use of established models contributed to the robustness of the study’s methodology^12,13,21.

Ethical approval

Ethical approval was not sought since our study utilized data from existing medical records rather than human subjects. However, administrative approval was obtained from the Principal Secretary of the Ministry of Health, the Director of Civil Registration Services, and County Directors of Health to permit the use of records.

Observed mortality

Between 2012 and 2021, Kenya experienced 60228 deaths attributed to tobacco-related diseases among adults aged ≥35 years. Median age was 65 years (IQR: 50–77). Males accounted for the majority at 56% (Table 1). The observed mortality demonstrated an annual increase across all cohorts (males and females 35–64 years and ≥65 years) until 2016, with a notable decline in 2017 due to a health-workers strike affecting record-keeping. Subsequently, mortality rates fluctuated, reaching a sharp increase after 2018 (Figures 1 and 2). Joinpoint regression analysis revealed a decline in mortality rates until 2018, followed by a substantial rise.

Figure 1

Absolute mortality numbers from medical records of persons aged ≥35 years who died due to smoking-related diseases in Kenya, 2012–2021

Figure 2

Mortality rate trends for 60228 persons aged ≥35 years who died due to tobacco-related diseases in Kenya, 2012–2021

The major contributors to mortality included respiratory diseases (36%), diabetes mellitus (18%), malignant cancers (17%), tuberculosis (16%), and cardiovascular diseases (13%). Among respiratory infections, pneumonia and influenza (86%) predominated, surpassing COPD (13%) and bronchitis/emphysema (1%). The year 2020 witnessed a significant two-fold increase in pneumonia and influenza deaths, attributed to COVID-19. Noteworthy cancer causes included esophagus cancer (37%), cervical cancer (19%), stomach cancer (16%), and pancreatic cancer (9%). Cerebrovascular diseases (87%) emerged as the primary cardiovascular cause of death, followed by ischemic heart diseases (10%) and other arterial diseases (3%) (Table 1).

Smoking prevalence

Mid-decade prevalence of tobacco smoking (2012–2021), including persons aged <35 years, indicated that 17.4% of men, 0.9% of women, and 9% overall were current smokers; former smokers constituted 10.6% of men, 1.4% of women, and 5.9% overall. Among individuals aged ≥35 years (our study population), 24.1% of men, 1.4% of women, and 12.9% overall were current smokers, while 17.3% of men, 2.4% of women, and 10% overall were former smokers (Table 2).

Smoking-attributable mortality

Out of the 60228 observed deaths from respiratory diseases, diabetes mellitus, malignant cancers, tuberculosis, and cardiovascular diseases between 2012 and 2021, 16.5% (9943) were attributed to tobacco smoking (Table 3). This included 40.5% from respiratory diseases, 31.4% from malignant cancers, 13% from tuberculosis, 8.9% from cardiovascular diseases, and 6.1% from diabetes mellitus. Within respiratory diseases, pneumonia and influenza contributed 59%, while chronic airway obstruction accounted for 39%. Noteworthy cancer causes included esophageal cancer (56%) and trachea, lungs, and bronchi combined cancers (14%). Of the cardiovascular tobacco-attributable deaths, 83% were from cerebrovascular diseases, with a notable distribution of 58% and 25% among 35–64 years and ≥65 years, respectively. The primary causes for smoking-attributable deaths were pneumonia and influenza (24%), esophageal cancer (18%), chronic airway obstruction (16%), and tuberculosis (13%), constituting 70% of all SAM (6987/9943).

Magnitude of smoking-attributable mortality

In this study, the examination of all-cause mortality data between 2012 and 2021 revealed a substantial impact of smoking, contributing to 16.5% of deaths among adults aged ≥35 years in Kenya. These findings align with a growing body of evidence underscoring the profound health implications associated with smoking²⁰.

Trends in mortality and global comparisons

The observed trend of escalating mortality across the studied conditions throughout the decade mirrors patterns observed in analogous studies. Comparable trends were noted in a study in China examining cancer mortality attributable to tobacco smoking over a ten-year period²². However, disparities emerge when comparing our findings with a study in Morocco²³, where 9.7% of all deaths were attributed to tobacco smoking, notably lower than the 16.5% observed in Kenya.

Disease-specific contributions to smoking-attributable deaths

Respiratory diseases

Respiratory diseases, particularly pneumonia and influenza, emerged as the predominant contributors to deaths attributable to tobacco smoking, followed closely by malignant cancers. Although local data on lung cancer were limited, the similarities with Australia’s 2018 observations reinforce the consistency of these findings^11,24.

Chronic respiratory diseases

Chronic respiratory diseases demonstrated a distinctive pattern in our study, with pneumonia and influenza overshadowing chronic obstructive pulmonary disease (COPD). This contrasts with the Global Burden of Disease (GBD) Study 2019⁵, which highlighted COPD as the primary cause of death in chronic respiratory diseases (CRDs).

Cancers

Cancer-related deaths attributed to smoking revealed esophageal cancer as the leading cause, consistent with Chinese findings where lung, liver, esophageal, and stomach cancers were frequently associated with smoking-associated cancer mortality^25,26.

Cardiovascular diseases

Within cardiovascular deaths, cerebrovascular diseases dominated among men, followed by ischemic heart diseases, aligning with established evidence of tobacco smoking’s pervasive impact across the cardiovascular system²⁷. Males exhibited a higher risk of cardiovascular disease (CVD) mortality than females.

Diabetes

The study also revealed the association between smoking and diabetes-related mortality, consistent with existing epidemiological studies²⁸. This underscores the importance of addressing smoking cessation as a crucial aspect of managing diabetes-associated mortality risks^29-31.

Tuberculosis

Moreover, our study illuminated the role of smoking in tuberculosis (TB) mortality in Kenya. While our findings are consistent with the meta-analysis of Bates et al.¹⁴, attributing 31% of TB cases and deaths to smoking, the percentage in our study was lower (13.4%). The gender disparity in TB deaths attributable to smoking mirrored findings from South Korea, India, and Bangladesh^32-35, with higher proportions in males.

Strengths and limitations

The study utilized nationally representative datasets that directly measured smoking status to determine tobacco smoking prevalence averaged over a decade. This robust approach provides a comprehensive overview despite potential fluctuations in smoking prevalence due to various interventions. The adoption of revised relative risks from recent systematic reviews and meta-analyses enhances the accuracy of the prevalence-based model, overcoming previous inconsistencies.

Despite these valuable insights, the study has limitations. It focused exclusively on smoking, neglecting other forms of tobacco use, such as smokeless tobacco, which may contribute significantly to morbidity and mortality. The challenge of ill-defined causes of death and the need for improved death certification processes highlight the study’s limitations. Efforts to enhance the quality of cause-of-death certification, including periodic reviews and training for healthcare providers, are imperative.