---
title: "Rising national stroke incidence and deaths in Uganda, 2016–2028 forecasts"
id: "plos-one-5-rising-stroke-burden-in-uganda-a-national-time-series-analysis-and-forecasts"
canonical_url: "https://medichelpline.com/clinical-feed/plos-one-5-rising-stroke-burden-in-uganda-a-national-time-series-analysis-and-forecasts"
content_type: "clinical_feed_article"
specialty: "Neurology"
source_name: "PLOS ONE (Medicine)"
source_url: "https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0360963"
published_at: "2026-10-16T14:00:00.000Z"
evidence_level: "Journal Feed"
license: "CC-BY-NC-4.0 / Informational Use"
---
# Rising national stroke incidence and deaths in Uganda, 2016–2028 forecasts
## Provenance & Clinical Metadata
- **Canonical URL:** https://medichelpline.com/clinical-feed/plos-one-5-rising-stroke-burden-in-uganda-a-national-time-series-analysis-and-forecasts
- **Specialty:** [Neurology](https://medichelpline.com/clinical-feed/neurology.md)
- **Primary Source:** PLOS ONE (Medicine)
- **Source URL:** [Original Journal Publication](https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0360963)
- **Published At:** 2026-10-16T14:00:00.000Z
- **Evidence Rating:** Journal Feed
## Executive GIST (TL;DR)
- Investigators analyzed routine national DHIS2 data from January 2016 through December 2025 to quantify monthly **stroke** cases and in-facility deaths and to forecast future burden through December 2028. - Monthly population estimates were interpolated using exponential growth and included as an offset in regression models; the source did not report disaggregated age or sex data. - The team used Bayesian negative binomial regression to estimate temporal trends, reporting exponentiated time coefficients as **incidence rate ratio** (IRR) to represent average monthly change. - Trend results showed sustained monthly increases: IRR 1.005 for confirmed cases (95% CI 1.004–1.006) and IRR 1.006 for deaths (95% CI 1.005–1.007), both p < 0.001, indicating small monthly rises that accumulate over years. - Decomposition using the Kitagawa method separated observed increases into a population effect and an epidemiological effect; the epidemiological contribution was roughly double the population contribution, per model-derived estimates. - Forecasting methods (ETS chosen on fit criteria) produced December 2028 point forecasts of about 927 monthly confirmed cases (80% PI 709–1,156) and 126 monthly deaths (80% PI 94–160). - Case fatality rate (CFR) remained relatively stable across the period, ranging between 9.9% and 12.4%, despite rising absolute counts. - Data cleaning included ARIMA-based outlier detection and STL decomposition for seasonality; 2016–2024 data trained forecasting models while 2025 was held out for testing. - Investigators noted limitations inherent to DHIS2 routine data, including possible data quality, completeness, and reporting changes; the source emphasizes that stronger primary prevention, acute care capacity, and surveillance systems are needed but did not provide implementation details or intervention effect sizes.
## Clinical Analysis & Structured Key Points
Rising stroke burden in Uganda: A national time-series analysis and forecasts, 2016–2028 | PLOS One Browse Subject Areas ? Click through the PLOS taxonomy to find articles in your field. For more information about PLOS Subject Areas, click here . Article Authors Metrics Comments Media Coverage Reader Comments Figures Figures Abstract Introduction Stroke is a leading cause of mortality and disability globally, with a disproportionate and growing burden in low- and middle-income Countries. However, evidence on national temporal trends and potential impacts in Uganda remains limited. This study examined trends in stroke cases and deaths in Uganda and forecast of future stroke burden through December 2028. Methods We conducted a retrospective time-series analysis of District Health Information System (DHIS2) from 2016 to 2025. Monthly stroke cases, and deaths were analyzed. Monthly populations of Uganda were estimated using exponential growth interpolation and incorporated as an offset term in Bayesian negative binomial regression models to estimate incidence rate ratios (IRR) at 95% confidence interval (CI). Decomposition analysis using the Kitagawa method was applied to derive contributions from population growth and epidemiological change, based on model estimates. Stationarity and autocorrelation were assessed using the unit root and Ljung–Box tests. Forecasting was performed using the exponential smoothing state-space (ETS) model at 80% prediction intervals (PI). Results Trend analysis showed sustained monthly increases in stroke incidence and mortality with IRR; 1.005 (95% CI: 1.004–1.006), and 1.006 (95% CI: 1.005–1.007), respectively (all p < 0.001). The epidemiological effect was approximately twice as large as the population effect, indicating that the majority of the increase in stroke cases and deaths is attributable to rising incidence rates rather than population growth alone. ETS model demonstrated that by December 2028, monthly cases and deaths are projected to reach approximately 927 (80% PI: 709–1,156) and 126 (80% PI: 94–160), respectively. Conclusion Stroke burden in Uganda is projected to continue increase further over the next years in the absence of effective prevention measures. Strengthened primary prevention, expanded acute stroke care capacity, and improved surveillance systems are urgently needed to mitigate the growing health system impact. Citation: Nakasendwa S, Mutyoba JN, Ario AR, Omondi E, Bashemera BB, Kakooza F, et al. (2026) Rising stroke burden in Uganda: A national time-series analysis and forecasts, 2016–2028. PLoS One 21(10): e0360963. https://doi.org/10.1371/journal.pone.0360963 Editor: Thien Tan Tri Tai Truyen, Cedars-Sinai Medical Center, UNITED STATES OF AMERICA Received: May 4, 2026; Accepted: October 2, 2026; Published: October 16, 2026 Copyright: © 2026 Nakasendwa et al. This is an open access article distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Data Availability: The data used in this study are derived from Uganda’s District Health Information System 2 (DHIS2) and are subject to data access and governance restrictions by the Uganda Ministry of Health. Requests for access to the underlying data should be directed to the Uganda Ministry of Health, Division of Health Information Management, through hmis@health.go.ug (+256 417 712260). Data access is subject to the Ministry’s data access and sharing procedures. Uganda population data used in the analysis are publicly available ( https://www.worldometers.info/world-population/uganda-population/) [ 23 ]. To facilitate reproducibility and transparency, a data dictionary describing all the data variables, including their definitions and coding, and the R script used for analysis are publicly available on GitHub and archived on Zenodo [ 56 ]. Funding: This research work is a product of the first author’s PhD study at School of Public Health, Makerere University, supported by the Data Science Without Borders (DSWB) program at the African Population and Health Research Centre (APHRC). The funding was received through the Makerere University Centre for Artificial Intelligence (MAK-AI). Competing interests: The authors have declared that no competing interests exist. Introduction Globally, stroke is the second leading cause of death and a major contributor to disability-adjusted life years (DALYs) lost [ 1 ]. Estimates from the Global Burden of Disease Study 2021 indicate that stroke accounted for approximately 7.3 million deaths and over 160 million DALYs globally, reflecting its substantial contribution to both premature mortality and long-term disability [ 2 ]. The burden of stroke is disproportionately concentrated in low- and middle-income countries (LMICs), accounting for nearly 70% of stroke deaths and 87% of stroke-related disability [ 3 ]. Although some high-income countries have experienced declines in age-standardized stroke mortality, many LMICs continue to face increasing absolute numbers of stroke cases and deaths due to a combination of population growth, ageing, and rising exposure to modifiable risk factors such as hypertension, diabetes, and unhealthy lifestyles [ 2 , 4 ]. In Sub-Saharan Africa (SSA), the burden of stroke has escalated dramatically over recent decades, driven by the epidemiological transition, rapid urbanization lifestyle changes, and the increasing prevalence of cardiovascular disease risk factors [ 5 , 6 ]. Population-based studies have documented crude annual incidence rates ranging from 25 per 100,000 in Nigeria, to approximately 260 per 100,000 in Egypt [ 7 ]. These differences highlight the considerable geographical and methodological heterogeneity in estimates of stroke burden across Africa. Despite this growing burden, the epidemiology of stroke in SSA remains incompletely characterized due to limitations in data availability and representativeness [ 8 ]. This phenomenon underscore the importance of strengthening population-based and routine surveillance systems, including District Health Information System 2 (DHIS2), to improve country-specific estimates [ 9 ]. Existing literature, often national planning estimates on stroke in SSA are commonly based on hospital-based cohorts, or on small geographic populations, or cross-sectional studies. In Uganda, stroke prevalence and incidence data remain limited, but available evidence suggests substantial and increasing burden [ 10 , 11 ]. Hospital-based studies have reported high case-fatality rates of 23.3% [ 12 ] and a 6% increased risk of mortality within 30 days of hospital admission [ 13 ]. However, such studies provide limited insight into national-level trends, the temporal dynamics and long-term trajectories of stroke burden [ 14 ]. Routine health information systems, such as the DHIS2 have emerged as the predominant health management information system platform in LMICs, currently deployed at national scale in over 30 Countries and supporting health service delivery [ 15 , 16 ]. DHIS2 enables systematic collection, aggregation, and analysis of routine health facility-based data across multiple health programs [ 16 ]. In Uganda, DHIS2 has been utilized for disease surveillance and health program monitoring, including for COVID-19 response, demonstrating the platform’s flexibility and local capacity for rapid adaptation [ 9 ]. However, challenges related to data quality, completeness, and timeliness have been documented across DHIS2 implementations in SSA, necessitating careful preprocessing and validation when utilizing these data for epidemiological analysis and forecasting [ 17 , 18 ]. Despite the potential of DHIS2 data to provide population-level insights, these data remain underutilized for longitudinal analyses and forecasting applications [ 19 , 20 ]. Time-series analysis and forecasting methods have been successfully applied to predict healthcare utilization and disease burden across various conditions [ 21 , 22 ]. Despite the growing recognition of stroke as a priority non-communicable disease in Uganda, comprehensive analyses of national trends and future projections remain limited. This study examined trends in stroke cases, and stroke-related deaths, using routinely collected national data between January 2016 and December 2025. We further projected stroke events beyond the year 2025, through December 2028 using time-series forecasting methods. Materials and methods Study design and setting This study employed a retrospective study design using routinely collected stroke data in Uganda, to examine temporal trends using time-series analysis, and to forecast stroke events. Data were obtained from the national District Health Information System (DHIS2) of Uganda, which compiles routine service delivery data reported by health facilities throughout the country. Because annual reporting of stroke events in DHIS2 is available from the year 2016 onwards, monthly counts for each indicator were extracted for the period January 2016 to December 2025. Annual population data of Uganda from the year 2015–2026 were obtained from Worldometer, which reports population estimates derived from the United Nations Department of Economic and Social Affairs, Population Division [ 23 ]. Stroke events and definitions Data were extracted from the DHIS2 platform and included three routine service indicators related to stroke. The variable “ 105-CV01. Stroke/Cerebrovascular Accident (CVA)” represents the total number of patients recorded with a clinical diagnosis of stroke (cerebrovascular accident) during the reporting period, reflecting overall stroke-related service utilization or admissions. The variable “ 108-CV03a. Stroke/Cerebrovascular Accident (CVA) - Cases” captures the number of confirmed stroke cases documented within the same period, as reported by health facilities through routine surveillance registers. The variable “08-CV03b. Stroke/Cerebrovascular Accident (CVA) - Deaths” denotes the number of in-facility deaths among patients diagnosed with stroke during the reporting period. All variables were aggregated at the reporting-unit level and represent routinely collected administrative data. For clarity and consistency in the manuscript, these indicators were relabeled as suspected stroke cases , confirmed stroke cases , and stroke deaths , respectively, while preserving their original operational definitions as defined in DHIS2. Data management All data were cleaned, prepared and analyzed using R programming language, version 4.5.2. We extracted stroke monthly data per year per stroke event, creating a dataset of five variables (Year, month, suspected stroke cases, confirmed stroke cases, stroke deaths). Variables; year and month were combined to create a continuous monthly date variable “period” with values 2016-Jan to 2025-Dec. Annual population figures as of December 2015 to December 2025 were obtained. Because stroke events data was extracted on a monthly basis, annual population figures were interpolated to generate monthly population estimates. Interpolation was performed assuming exponential population growth, reflecting a gradual and continuous demographic change over time. The annual population as of December 2015 was only used to generate the population exponential growth rate from December 2015 to December 2016. This growth rate was used to interpolate monthly populations from January 2016 to November 2016. However, the annual population in 2015 was not used in the trends analysis. A similar exponential interpolation strategy was implemented for the years 2016–2025 to generate monthly populations. This set of data consisted of the date variable “period” and monthly population. The stroke, and population datasets were then merged into the analyzable dataset of five variables (period, suspected stroke cases, confirmed stroke cases, stroke death, and population), using “period” as the unique identifier. Statistical analysis Data pre-processing. Time-series plots were explored to identify temporal patterns in suspected stroke cases, confirmed stroke cases, and stroke deaths. Outliers were identified using the tso() function, with autoregressive integrated moving average (ARIMA) model parameters automatically selected via auto.arima() from the forecast. The procedure detects statistically significant additive outliers based on residual diagnostics from the fitted ARIMA model. Identified outliers were replaced with ARIMA model adjusted values, and the resulting cleaned data was used for subsequent trend analyses and forecasting. The seasonal and Trend decomposition using Loess (STL) was applied to assess the strength of seasonality and trend. Seasonal subseries plots were used to assess within-year variation by displaying monthly stroke counts across all years. Seasonality was evaluated by visually examining whether specific months consistently showed higher or lower values and whether recurring monthly patterns were evident. Descriptive analysis of stroke rates. Suspected stroke cases were retained for descriptive purposes and to assess trends in case detection. However, all primary analyses including incidence, mortality, decomposition and forecasting were based on confirmed stroke cases. For each month, we calculated the incidence rate as the number of confirmed stroke cases per 100,000 population, and the mortality rate as the number of stroke deaths per 100,000 population. The case fatality rate (CFR) was computed as the proportion of confirmed stroke cases resulting in death. To illustrate temporal trends, we generated line graphs of incidence, mortality, and CFR over the study period. These visualizations facilitated the identification of long-term trends, seasonal patterns, and potential fluctuations in reporting or disease occurrence. Graphical inspection of the trends also informed subsequent modeling and decomposition analyses. Trends analysis. Formal trend analysis was conducted using Bayesian negative binomial regression models to account for over-dispersion in monthly stroke counts, while enabling probabilistic inference through posterior distributions and uncertainty quantification. Time, measured in months, was included as a continuous predictor to estimate temporal trends. To account for changes in population size over time, monthly interpolated population estimates were included as an offset term in all regression models. The natural logarithm of the population was specified as the offset, allowing model coefficients to be interpreted as rates rather than crude counts. Exponentiated time coefficients are reported as incidence rate ratios (IRR) representing the average monthly change in stroke rates. Model adequacy was assessed through residual diagnostics using autocorrelation function plot. Statistical significance was evaluated at the 5% level, and results were reported with 95% confidence intervals. Decomposition of stroke burden. To quantify the drivers of changes in stroke burden over time, we performed a decomposition analysis using model-based estimates. Model predicted counts were aggregated to annual totals, and corresponding rates per 100,000 population were computed using annual population estimates. These model-based estimates were then used in a decomposition framework to separate the observed changes in stroke cases and deaths into two components: (1) a population effect, representing changes attributable to population growth, and (2) an epidemiological effect, representing changes in stroke incidence and mortality independent of population size. We applied the Kitagawa decomposition method [ 24 ], a symmetric approach widely used in epidemiology and demography [ 25 , 26 ], which decomposes year-to-year changes by averaging the rates and population sizes of consecutive years. This approach minimizes bias associated with selecting a single baseline year and provides a balanced estimate of the contributions of population growth and changes in incidence and mortality. Decomposition results were expressed both as absolute changes in counts and as rates per 100,000 population to facilitate comparison with incidence and mortality trends. To aid interpretation, the contributions of population and epidemiological effects were visualized using line graphs, illustrating their relative roles in driving temporal changes in stroke burden over the study period. This approach allowed us to distinguish whether increases in stroke burden were primarily driven by demographic expansion or by changes in the underlying risk of stroke. Forecasting. Prior to forecasting model fitting, stationarity was evaluated using the unit root test. Ljung-Box test and autocorrelation functions were examined to characterize temporal dependence and guide model specification. Data from 2016 to 2024 were used for model training, while 2025 was held out as a testing period to evaluate out-of-sample forecasting performance. Forecasts were then generated for 2025 and the subsequent period, 2026–2028. This three-year horizon was selected to capture multiple seasonal cycles while maintaining reasonable predictive reliability, thereby supporting strategic health service planning under the assumption of stable trends and reporting patterns. Forecasting models were then fitted using Naive model as a baseline model, ARIMA and exponential smoothing state-space frameworks. Model selection was based on minimization of Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE), with preference given to parsimonious models that adequately captured the underlying temporal structure of the data. Fitted values and forecasts, with prediction intervals were used to quantify uncertainty around the projected values. Forecast uncertainty was expressed using 80% prediction intervals to provide a narrower and more operationally relevant range for short-term planning. The 80% interval captures the most probable forecast values while avoiding excessively wide bounds that may reduce interpretability in health service planning contexts. Forecast performance and plausibility were evaluated through comparison of fitted values with observed data in the training period, and through inspection of forecast trajectories to ensure epidemiological coherence. Results Population, stroke events and rates in Uganda from January 2016 to December 2025 The population in Uganda has steadily increased from approximately 37.5 million in 2015 to over 50 million in 2025. Between 2016 and 2025, the burden of stroke in Uganda increased substantially ( Table 1 ). The annual number of confirmed stroke cases rose from 3,089 in 2016–8,449 in 2025, representing nearly a threefold increase over the study period. This corresponds to the increase in incidence rate per 100,000 population from 8.08 to 16.65. Download: PNG larger image TIFF original image Table 1. Annual population, stroke events and rates in Uganda from January 2016 to December 2025. https://doi.org/10.1371/journal.pone.0360963.t001 Stroke related deaths also increased from 342 to 908, with the mortality rate rising from 0.90 to 1.79 per 100,000 population ( Table 1 ). Despite these increases, the case fatality rate remained relatively stable, fluctuating between 9.9% and 12.4%. Trends in stroke events from January 2016 to December 2025 Fig 1 panel A illustrates monthly trends in suspected and confirmed cases of Stroke over the study period. Suspected cases displayed marked short-term variability with recurrent peaks, but with an overall increasing trajectory from 2016 to 2025. Confirmed cases similarly showed a steady upward trend, with lower magnitude and less variability. Download: PNG larger image TIFF original image Fig 1. Time series plots of stroke events and rates from January 2016 to December 2025. https://doi.org/10.1371/journal.pone.0360963.g001 Fig 1 p
## Related Clinical Research

- [Increasing stroke risk in younger adults requires targeted prevention and updated risk assessment](https://medichelpline.com/clinical-feed/plos-medicine-0-stroke-risk-in-younger-age-groups-is-increasing-and-requires-targeted-prevention.md)
- [Trends in stroke incidence and modifiable risk factors in England (2018–2024): age-specific change](https://medichelpline.com/clinical-feed/plos-medicine-0-time-trends-in-stroke-incidence-and-modifiable-risk-factor-prevalence-in.md)
- [Estrogen-only menopausal therapy associated with reduced Alzheimer’s biomarkers](https://medichelpline.com/clinical-feed/medical-news-today-0-estrogen-only-therapy-after-menopause-linked-to-reduced-alzheimer-s-biomarkers.md)
- [Avoiding smoking, diabetes, and high blood pressure in midlife may add up to 13 dementia-free years](https://medichelpline.com/clinical-feed/medical-news-today-0-avoiding-3-risk-factors-in-midlife-could-add-13-dementia-free-years.md)
- [Wealth-related inequality in excessive red meat consumption among hypertensive adults in Iran](https://medichelpline.com/clinical-feed/plos-one-23-assessment-and-decomposition-of-excessive-red-meat-consumption-among.md)

## Navigation
- [← Back to Neurology Feed](https://medichelpline.com/clinical-feed/neurology.md)
- [← All Clinical Specialties](https://medichelpline.com/clinical-feed.md)
## Medical & Regulatory Disclaimer

> [!CAUTION]
> MedicHelpline content is structured for research, educational, and professional discovery purposes. It does not constitute individual medical advice, clinical diagnosis, or treatment recommendations.
> Always verify dosing, contraindications, and regulatory alerts against official product labeling and primary regulatory sources before clinical decision-making.