A national spatial prioritisation framework for the targeted expansion of essential sickle cell disease services across health-facility catchments in Uganda Authors: George Paasi, Grace Ndeezi, Ruth Namazzi, Ezekiel Mupere, Ian Guyton Munabi, Sarah Kiguli, Richard Idro, Peter Olupot-Olupot Published: 2026 ABSTRACT BackgroundSickle cell disease causes substantial childhood morbidity and premature mortality in sub-Saharan Africa, while access to essential services remains uneven. In Uganda, geographic variation in inherited SCD risk supports spatially targeted service planning.MethodsWe conducted a national facility-linked ecological spatial planning analysis using 280 eligible higher-level public facilities. Populated cells were assigned to the nearest eligible facility within a modelled 60-minute travel-time boundary. Epidemiological context was represented by a study-specific co-risk index combining standardised haemoglobin S (HbS) allele frequency and Plasmodium falciparum parasite rate among children aged 2 to 10 years (PfPR2-10). Reported availability was assessed for newborn screening, hydroxyurea, blood transfusion, and specialised SCD clinic care. We derived service-deficit, structural-vulnerability, programme-priority, service-specific priority, and coverage-to-need measures and assessed sensitivity to alternative weighting specifications.ResultsAmong 280 planning zones, 122 (43.6%) had none of the four reported services and 205 (73.2%) had no more than one. Co-risk ranged from 0.003 to 0.934 and service-mediated SCD structural vulnerability from 0.000 to 0.905. Higher programme-priority scores were concentrated mainly in northern, north-eastern, eastern, and parts of central Uganda, while lower-priority tiers were more common in the south-west. Newborn-screening and hydroxyurea priorities were closely aligned, transfusion priorities extended more broadly into northern and central Uganda, and specialised-clinic priorities were more spatially mixed. The overall coverage-to-need gap had a mean of 0.190 and median of 0.205. The 56 zones comprising the national top 20% of programme priority contained approximately 7.0 million people, representing 17.3% of the population within the analytic framework. Priority concentration was strongest in Lango, Teso, Acholi, Karamoja, and Busoga.ConclusionsThe framework identified a concentrated set of candidate service-strengthening nodes under the specified assumptions, mainly in northern, north-eastern, and eastern Uganda. The findings can support further programme assessment and stakeholder review for phased, regionally coordinated, and service-specific expansion of SCD care.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1186/s44263-026-00338-y. Background Sickle cell disease (SCD) remains a major cause of childhood morbidity and premature mortality in sub-Saharan Africa (SSA). In the Global Burden of Disease Study 2021, SCD was estimated to have caused 81,100 deaths among children younger than 5 years globally, underscoring its major contribution to early childhood mortality [1]. The burden is concentrated in SSA, where severe disease presents early and survival is strongly shaped by timely diagnosis and access to comprehensive care [2, 3]. The relationship between haemoglobin S (HbS) and malaria involves two distinct processes. Historically, Plasmodium falciparum malaria selected for the HbAS genotype, helping maintain higher HbS frequencies in malaria-endemic populations, with selection potentially continuing where transmission persists [4, 5]. These allele frequencies influence HbSS births through inheritance, with a 25% probability of HbSS in each pregnancy when both parents have HbAS [6]. Contemporary malaria creates an additional clinical burden because infections can precipitate severe anaemia and increase transfusion needs in people with SCD [7], while malaria-associated mortality has also been documented, although risk varies by setting and access to timely care [7, 8]. Spatial planning may therefore consider inherited sickle-cell risk alongside contemporary malaria transmission as related but distinct contributors to population and health-service burden. Early intervention substantially improves outcomes in SCD. Newborn screening allows presymptomatic diagnosis and early enrolment into penicillin prophylaxis, vaccination, caregiver education, and ongoing follow-up, with studies from Benin, Angola, and Uganda showing that these approaches are feasible at both programme and national scale in African health systems (9–11]. Their clinical value is reinforced by the Prophylactic Penicillin Study, which showed reduced invasive pneumococcal infection in young children with sickle cell anaemia (SCA) [12], and by cohort data from Kilifi and Dar es Salaam showing that mortality is highest in early childhood and lower among children receiving structured care [2, 13]. Disease-modifying therapy is also supported by evidence from Africa, including the multicountry REACH trial, which found hydroxyurea feasible, safe, and associated with fewer vaso-occlusive events, infections, malaria, transfusions, and deaths [14]. Even so, access to screening, preventive care, hydroxyurea, transfusion support, and specialised follow-up remains uneven across the region. Uganda is an important setting for spatial planning of SCD services because it combines substantial geographic variation in inherited SCD risk, heterogeneous contemporary malaria transmission, and uneven access to SCD essential services. National malaria analyses have demonstrated marked spatial heterogeneity in transmission, including persistent clusters of elevated incidence across several northern, eastern, and north-eastern parts of the country [15]. The Uganda Sickle Surveillance Study (US3) similarly demonstrated pronounced geographic variation in inherited risk, reporting a national sickle cell trait prevalence of 13.3%, ranging from 4.6% in the South Western region to 19.8% in East Central, and an SCD prevalence of 0.7%, ranging from 0.2% to 1.5% across these regions [16]. At the same time, access to SCD-related services in Uganda has not historically been uniform. National screening activities were initially targeted towards higher-burden districts, and operational studies have identified geographic and facility-level differences in screening delivery, while facility-based evidence has documented limited hydroxyurea use in some regional SCD clinics [9, 17]. More broadly, national geospatial assessments demonstrate substantial variation in physical accessibility to government health facilities across Uganda [18]. Thus, the geography of inherited SCD risk and the geography of service provision may be spatially mismatched. Identifying where elevated epidemiological need coincides with limited-service availability provides the central rationale for a spatially explicit framework for prioritising SCD service expansion. Despite growing evidence on SCD epidemiology, newborn screening, hydroxyurea, and barriers to care in SSA, national programmes still have limited guidance on how to combine these evidence sources when deciding where specific SCD services should be strengthened [16, 19–21]. Geospatial studies in maternal, newborn, primary, and emergency care have mapped variation in accessibility, facility availability, staffing, and functionality (22–25], but comparable SCD studies have rarely linked epidemiological and service information to existing facility locations for planning across several components of care. We addressed this gap by linking epidemiological context, reported availability of four essential SCD services, routine facility-recorded events, represented population, and modelled geographic access to existing Health Centre IV (HC IV) and higher-level public facilities. The framework was developed to answer a practical planning question: which existing facilities warrant closer assessment for strengthening newborn screening, hydroxyurea, transfusion, or specialised SCD clinic care? It provides an overall comparative ranking alongside separate service-specific rankings, allowing locations with similar priorities across services to be distinguished from those where particular service deficits predominate. The resulting facility-linked outputs are intended to guide subsequent validation, readiness assessment, costing, and stakeholder review before decisions on service expansion are made. Methods Study design We conducted an ecological facility-linked spatial planning analysis. The primary spatial units were modelled planning zones associated with each eligible higher-level public facility defined using a common 60-minute travel-time boundary. The primary objective was to compare existing public-facility nodes as candidates for strengthening four essential SCD service domains: newborn screening, hydroxyurea, blood transfusion support, and specialised SCD clinic care. The analysis did not estimate household-level demand or reproduce observed utilisation since no nationally comparable patient-origin, household care-seeking, affordability, perceived-quality, or realised-utilisation dataset was available for incorporation. The study was reported in accordance with the Spatial Lifecourse Epidemiology Reporting Standards (ISLE-ReSt) statement [26]. Additional methodological details, tables and figures are provided in Supplementary Material 1, and the completed checklist is provided in Supplementary Material 2: ISLE-ReSt checklist. Study setting This study was conducted in Uganda, a landlocked East African country located between latitudes 1°N and 4°S and longitudes 29°E and 35°E with a population of approximately 45 million people, exhibiting marked geographic variation in malaria transmission intensity driven by altitude gradients, rainfall patterns, and vector ecology [27]. Similarly, Uganda has marked spatial heterogeneity in sickle cell trait and SCD prevalence, reflecting geographic variation in inherited population risk [16]. Uganda’s health system is organised across 15 sub regions [28] (Supplementary Material 1: Fig S1), with a referral hierarchy comprising Health Centre II (HC II), Health Centre III (HC III), HC IV, general hospitals, regional referral hospitals, and national referral hospitals [29]. Data sources Routine SCD burden data We used routine facility data from Uganda’s national health management information system (HMIS), implemented through District Health Information Software 2 (DHIS2), an open-source, web-based platform used for routine health data aggregation, validation, and reporting [30]. Uganda adopted DHIS2 nationally in 2011 and completed phased national rollout by 2012 [30]. We extracted monthly records on SCD admissions and deaths from January 1, 2020, to December 31, 2024. The extracted DHIS2 data elements corresponded to the HMIS 108 inpatient monthly report, which records monthly inpatient cases and deaths, and were labelled in the system as “108-NC03a Sickle cell disease (cases)” and “108-NC03a Sickle cell disease (deaths)” [31]. The study team assessed data completeness at monthly resolution for each facility and data element. These data elements represented facility-reported inpatient events rather than unique individuals, and repeated admissions by the same patient could contribute more than once to the cumulative count. Because an event could be recorded only after contact with a reporting facility, counts may vary with diagnosis, care-seeking, referral, service capacity, and reporting completeness as well as underlying morbidity. They were therefore treated as routine-system visibility of severe events, not as independent estimates of community incidence, prevalence, or demand. Health facility selection and analytic eligibility A harmonised national facility master file containing geolinked facilities was used as the reference frame for facility identity, coordinates, and linkage across routine burden and service inventory datasets. From this broader facility frame, 286 public HC IV-and-above facilities were identified as eligible for catchment-based analysis because these levels provide inpatient care. Facilities were excluded if they met any of three prespecified analytic criteria: a supra-catchment national referral role that was not compatible with the facility-linked planning-zone framework, unusable geographic coordinates that prevented spatial delineation, or prolonged non-reporting of the routine SCD burden data required for analysis. Applying these criteria excluded six facilities: Mulago National Referral Hospital because of its supra-catchment referral role, three facilities with unusable geographic coordinates, and two with prolonged non-reporting, leaving 280 facilities in the final analytic facility set. Catchments were delineated around these 280 facilities, yielding 280 catchments for all downstream analyses (Supplementary Material 1: Fig S2). Health facility catchment zone delineation Facility-linked planning zones for eligible higher-level public facilities were delineated using an adapted two-step floating catchment area approach. First, populated raster cells were assigned to their nearest eligible higher-level facility by straight-line distance to generate proximity-based base areas. Second, these base areas were clipped using a modelled 60-minute travel-time threshold derived from a national travel-time surface generated using AccessMod version 5.8 (World Health Organization, Geneva, Switzerland) [32, 33]. AccessMod supports estimation of population accessibility using travel-time thresholds selected according to the service and planning question rather than prescribing a universal threshold [32, 34]. A 60-minute modelled travel-time threshold was selected as an evidence-informed benchmark for potential access to HC IV and higher-level facilities. One-hour accessibility has been used in national health-service accessibility assessments in Uganda [18], while 30- and 60-minute thresholds have also been applied in SSA accessibility studies [35]. The 60-minute threshold provided a consistent national rule across the 280 eligible facilities, with populated cells assigned to the nearest eligible facility and retained only where modelled travel time was 60 min or less. This was not interpreted as an SCD-specific clinical standard or an observed care-seeking boundary; because actual utilisation may vary with transport, referral pathways, facility bypass, cost, perceived quality, and patient preference, the resulting polygons were treated as standardised facility-linked planning zones rather than observed utilisation catchments. The final planning catchment zones were the intersection of nearest-facility base areas and 60-minute travel-time service areas. In addition, facility-linked zones were selected instead of administrative units because the planning question concerned service strengthening at specific existing facilities and both the service inventory and routine-event data were recorded at facility level. Administrative units may contain multiple eligible facilities or no eligible facility and therefore do not map directly to a single service-delivery node. This choice enabled direct linkage of facility and surrounding-area data, but conditioned the analysis on the existing facility network. The primary decision unit was therefore the existing facility node rather than the administrative territory. A sub-county analysis would answer a complementary territorial need question, but would not directly identify the facility at which the four recorded service domains could be strengthened. The zones were not observed utilisation catchments, and the analysis did not incorporate patient-origin, health-seeking, bypassing, or referral-flow data. Represented population and modelled travel time described geographic reach and did not constitute direct measures of demand. Detailed derivation of the planning zones, population assignment, and geometric validation procedures is provided in Supplementary Material 1: Methods S1. Cross-zone referral and facility bypassing could therefore associate recorded events with a facility other than the nearest eligible facility and could extend the relevant service population beyond the modelled boundary. Service inventory data Facility-level service inventory data were used to measure local availability of four core SCD service domains: newborn screening (NBS), hydroxyurea (HU), blood transfusion (TR), and specialised SCD clinic care (CL). These domains were selected because they represent key components of SCD detection, preventive care, acute complication management, and longitudinal follow-up. Service inventory records were harmonised to the national facility master file using facility identifiers, facility names, administrative location, and geographic coordinates, and were then linked to the modelled health-facility catchments used in the analysis. For each service domain, availability was coded as a binary facility-level indicator. NBS was coded as available when a facility was recorded as providing or participating in SCD newborn or early infant screening activities, including collection of samples for SCD screening. HU was coded as available when a facility was recorded as providing hydroxyurea as part of SCD care. TR was coded as available when a facility was recorded as providing blood transfusion services onsite for patients requiring transfusion support. CL was coded as available when a facility was recorded as providing dedicated or structured SCD follow-up care, including regular clinic-based review for patients with known or suspected SCD. Planning-zone reported service-domain breadth was defined separately for each of the four domains. A zone was assigned a value of 1 when the service was reported at the eligible facility defining the zone and 0 otherwise. The reported service-domain breadth score was the unweighted sum of the four binary indicators, with possible values from 0 to 4, and the service-deficit score was its complement. Equal contribution of the four domains was a transparent measure of reported service breadth and not to imply equal clinical importance or equivalent resource requirements. Nationally consistent data on staffing, throughput, commodity continuity, laboratory functionality, service quality, and effective capacity were not incorporated. Consequently, a high-capacity referral hospital and a facility with nominal reported availability could receive the same binary value for a service domain. The indicators therefore described reported local service configuration and could not distinguish service intensity, readiness, reliability, patient uptake, or effective access. Temporal alignment of service inventory and outcome data The primary outcomes were catchment-level counts of recorded SCD admissions and recorded SCD deaths. Routine annual counts were summed across the period 2020 to 2024 to derive cumulative five-year totals for each catchment. We used multi-year aggregation because the study was designed to examine persistent geographic patterning in recorded burden rather than short-term temporal change, and because annual SCD events were sparse in many modelled health-facility catchment zones. Aggregating across five years improved the stability of small-area outcome measures and reduced the influence of transient fluctuation [36]. In keeping with the analytic framework of the study, these outcomes were interpreted as measures of recorded burden, shaped both by underlying epidemiological need and by the ability of local services to detect and document disease. The service inventory captured reported availability of SCD services at the time the inventory was conducted and was treated as a spatial descriptor of local SCD service configuration. Because the service inventory was not available as an annual time-varying measure across the full outcome period, reported service-domain breadth was assumed to represent the relative distribution of SCD reported service configuration across catchments for spatial planning purposes. This assumption was considered reasonable for identifying broad spatial differences in service configuration, but it does not imply that reported service-domain breadth remained unchanged throughout 2020 to 2024. Individual services may have been introduced, interrupted, expanded, or discontinued during the study period. The associations between service deficit and recorded SCD admissions or deaths were therefore interpreted as spatial associations between local service configuration and routine visibility, not as evidence that reported service-domain breadth measured at one time point caused changes in recorded burden across the full five-year period. Temporal mismatch between the service inventory and cumulative outcome data was considered a limitation. Raster-derived epidemiological and population inputs Raster-derived inputs were used to capture epidemiological and demographic context across catchments. These included an HbS allele frequency surface, assembled using two independent national datasets reporting district-level HbS genotypes in Uganda [16, 37], that were modelled using Empirical Bayesian Kriging (EBK) in ArcGIS Pro version 3.7.2 (Esri, Redlands, CA, USA) [38, 39] to a 5 km grid raster clipped to the national boundary, and a malaria transmission surface based on annual modelled rasters of Plasmodium falciparum parasite rate among children aged 2 to 10 years (PfPR2-10) for 2015 through 2024 at nominal 5 km resolution from the Malaria Atlas Project (MAP) [40]. The PfPR2-10 surface was used as an indicator of contemporary malaria transmission intensity and ecological exposure. It was not interpreted as a measure of historical malaria selection pressure or as a proxy for SCD prevalence. Its relevance to the framework was based on the continuing clinical importance of malaria among people living with SCD, including its contribution to acute illness, anaemia, hospitalisation, transfusion requirements, and mortality. PfPR2-10 is widely used in malaria mapping and endemicity comparisons, and standardisation of PfPR-based measures has an established epidemiological basis for comparative mapping [41]. Population rasters from WorldPop 2023 [42] were used because gridded demographic surfaces support fine-scale spatial health planning and are widely used for service-access and population-at-risk analyses. Detailed description of the HbS and PfPR2-10 raster inputs is provided in Supplementary Material 1: Methods S2. Prediction uncertainty associated with the HbS interpolation and the modelled PfPR2-10 surfaces was not propagated through standardisation, planning-zone aggregation, or the downstream composite indices. The reported scores and ranks are therefore point estimates conditional on the selected spatial surfaces. Data harmonisation and catchment aggregation Facility identifiers and names were harmonised across routine burden files, service inventory files, and the facility master file. Facilities with valid coordinates were spatially linked to catchment polygons, after which burden and service records were aggregated to catchment level. Raster-derived HbS, PfPR2-10, and population values were also extracted for each catchment polygon. The resulting analytic dataset contained one row per catchment and integrated routine burden, service inventory, epidemiological surfaces, and population estimates into a single national planning dataset (Supplementary Material 1: Table S1). Detailed procedures for facility harmonisation, spatial linkage, burden and service aggregation, raster integration, and quality-control checks are provided in Supplementary Material 1: Methods S2. Construction of planning variables for the prioritisation framework The prioritisation framework was built around related but conceptually distinct facility-linked planning constructs. Co-risk represented relative epidemiological context derived from two proxy surfaces. Service deficit represented the number of four service domains not reported at the facility defining each planning zone. Their combination identified zones where higher co-risk and greater reported service deficit coincided. The overall programme-priority score added a smaller facility-recorded mortality component to produce a comparative national ordering, while the coverage-to-need gap described an arithmetic difference between co-risk and normalised reported service breadth. Service-specific scores identified zones where a focal service was not reported in the presence of higher co-risk. None of these measures independently estimated household demand, effective coverage, or population SCD prevalence. Because these indices were comparative measures, higher values indicated a stronger relative expression of the components used to construct each measure; they did not represent probabilities or percentages of unmet need. The framework was constructed as a deterministic spatial decision-support approach. The composite measures were used to rank catchments relative to one another under explicitly defined planning rules. They were not designed to estimate the probability of SCD, the number of people with SCD, the causal effect of existing services, or the expected health effect of service expansion. The scores should therefore be interpreted as comparative planning indices conditional on the selected data inputs, standardisation procedures, and weighting assumptions. Co-risk score Catchment-level epidemiological need was represented using a co-risk score derived from two raster inputs: an EBK interpolated national HbS allele-frequency surface and a 10-year mean PfPR2-10 surface. To represent epidemiological need, HbS and PfPR2-10 were each rescaled to a common 0 to 1 range using min-max standardisation:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{X}_{i,\mathrm{s}\mathrm{t}\mathrm{d}}=\frac{{X}_{i}-\mathrm{m}\mathrm{i}\mathrm{n}\left(X\right)}{\mathrm{m}\mathrm{a}\mathrm{x}\left(X\right)-\mathrm{m}\mathrm{i}\mathrm{n}\left(X\right)}$$\end{document} where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{X}_{i}$$\end{document} is the catchment-level HbS or PfPR2−10 value for catchment i. After standardisation, the co-risk score was defined as the equal-weighted mean of the two standardised inputs:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\mathrm{C}\mathrm{o}\mathrm{R}\mathrm{i}\mathrm{s}\mathrm{k}}_{i}=\frac{{\mathrm{H}\mathrm{b}\mathrm{S}}_{i,\mathrm{s}\mathrm{t}\mathrm{d}}+{\mathrm{P}\mathrm{f}\mathrm{P}\mathrm{R}}_{i,\mathrm{s}\mathrm{t}\mathrm{d}}}{2}$$\end{document} This measure was used to summarise the joint geography of inherited sickle-related risk and malaria transmission ecology across planning zones. The HbS component reflects inherited-risk geography, not direct disease prevalence, while the PfPR2−10 component reflects surrounding malaria transmission intensity and the associated ecological context of malaria-related clinical vulnerability among people living with SCD, not individual malaria burden among people with SCD. The surfaces are not statistically or biologically independent because historical malaria selection contributed to the population distribution of HbS; equal averaging may therefore reinforce a shared broad-scale pattern rather than combine two independent confirmations of SCD burden. Equal weighting was retained as a transparent planning convention, and the resulting co-risk score was interpreted as a relative, study-specific epidemiological-context index rather than a biologically calibrated prevalence or risk model. The co-risk score was calculated using catchment-level point estimates from the HbS and PfPR2-10 surfaces. Although prediction uncertainty was available for the interpolated HbS surface, it was not propagated through catchment aggregation, standardisation, or calculation of the co-risk score. Uncertainty associated with the modelled PfPR2-10 surface was also not incorporated into the co-risk calculation. Consequently, the co-risk score and the downstream service-mediated structural vulnerability, programme-priority, service-specific priority, and coverage-to-need gap measures were calculated as point estimates without confidence or credible intervals. Numerical differences between closely ranked catchments should therefore not be interpreted as precise differences in underlying epidemiological need. A jointly specified stochastic or Bayesian geostatistical model was not fitted in this study. Such an approach would require specification of probability models and measurement-error structures for the HbS, PfPR2-10, service, burden, and accessibility inputs, together with suitable data for calibration and external validation. The present analysis instead prioritised transparency and reproducibility through a deterministic index in which the contribution of each component could be examined directly. The absence of probabilistic uncertainty propagation remains an important limitation of this approach. Local reported service-domain breadth and service deficit Local reported service-domain breadth was represented using four binary catchment-level indicators: NBS\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{}_{i}$$\end{document}, HU\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{}_{i}$$\end{document}, TR\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{}_{i}$$\end{document}, and CL\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{}_{i}$$\end{document}. For each service, a catchment was assigned a value of 1 if the service was available at one or more linked facilities and 0 otherwise. The reported service-domain breadth score was defined as:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\mathrm{R}\mathrm{e}\mathrm{p}\mathrm{o}\mathrm{r}\mathrm{t}\mathrm{e}\mathrm{d}\:\mathrm{s}\mathrm{e}\mathrm{r}\mathrm{v}\mathrm{i}\mathrm{c}\mathrm{e}\:\mathrm{d}\mathrm{o}\mathrm{m}\mathrm{a}\mathrm{i}\mathrm{n}\:\mathrm{b}\mathrm{r}\mathrm{e}\mathrm{a}\mathrm{d}\mathrm{t}\mathrm{h}\:}_{i}={\mathrm{NBS}}_{i}+{\mathrm{HU}}_{i}+{\mathrm{TR}}_{i}+{\mathrm{CL}}_{i}$$\end{document} with a possible range from 0 to 4. The service-deficit score was then defined as the complement of the local reported service-domain breadth score:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\text{Service deficit}}_{i}=4-{\mathrm{R}\mathrm{e}\mathrm{p}\mathrm{o}\mathrm{r}\mathrm{t}\mathrm{e}\mathrm{d}\:\mathrm{s}\mathrm{e}\mathrm{r}\mathrm{v}\mathrm{i}\mathrm{c}\mathrm{e}\:\mathrm{d}\mathrm{o}\mathrm{m}\mathrm{a}\mathrm{i}\mathrm{n}\:\mathrm{b}\mathrm{r}\mathrm{e}\mathrm{a}\mathrm{d}\mathrm{t}\mathrm{h}\:}_{i}$$\end{document} and rescaled to the unit interval:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\text{Service deficit}}_{i,\mathrm{std}}=\frac{{\text{Service deficit}}_{i}}{4}$$\end{document} Higher values represented weaker availability to essential SCD services. Thus, a value of 0 indicated complete availability of all four service domains, whereas a value of 1 indicated absence of all four domains. This standardised measure was used in the overall prioritisation framework and related planning outputs. Structural vulnerability and service-specific vulnerability measures We operationalised structural vulnerability narrowly as the spatial overlap between epidemiological co-risk and SCD service deficit. This operational definition captures one programmatically relevant dimension of structural vulnerability, but does not fully measure broader social, economic, political, and household-level vulnerability [43, 44]. We refer to this measure as the service-mediated SCD structural vulnerability index (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\mathrm{S}\mathrm{V}}_{i}$$\end{document}) to distinguish the operational index from the broader theoretical construct of structural vulnerability. The composite service-mediated SCD structural vulnerability index was therefore defined multiplicatively as:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\mathrm{S}\mathrm{V}}_{i}={\mathrm{Co-risk}}_{i}\times\:{\text{Service deficit}}_{i,\mathrm{std}}$$\end{document} This formulation assigns the highest values to modelled health-facility catchments in which high co-risk coincides with severe service deficit and constrains the index to settings where both conditions are present. For descriptive presentation, catchments were assigned to four groups according to their relative co-risk and reported service breadth. Higher co-risk was defined as membership in the upper 50% of the national co-risk distribution, using the national median of 0.433 as the cut-point. Lower co-risk was defined as membership in the lower 50% of the distribution. Low reported service-domain breadth was defined as reported availability of zero or one of the four essential SCD service domains. Catchments in this group therefore lacked at least three of the four domains. Higher reported service-domain breadth was defined as reported availability of two to four domains. The resulting groups were: higher co-risk with low reported service-domain breadth, higher co-risk with higher reported service-domain breadth, lower co-risk with low reported service-domain breadth, and lower co-risk with higher reported service-domain breadth. These categories were used as a descriptive summary of the joint distribution of co-risk and service breadth. The classification was not used to calculate the continuous service-mediated structural-vulnerability index, overall programme-priority score, coverage-to-need gap, or service-specific priority scores. Those analyses retained the continuous co-risk and standardised service-deficit values. The median was selected as a distribution-based threshold that divided the national catchment set into two equally sized relative co-risk groups and not to represent a clinically validated threshold for SCD risk or malaria-related vulnerability. The zero-to-one-service threshold was selected to describe catchments lacking most of the four-domain service platform and not a threshold for effective coverage, service quality, or facility readiness. In parallel, service-specific vulnerability measures were derived for each service domain by interacting co-risk with absence of the relevant service:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\mathrm{S}\mathrm{V}}_{i,\mathrm{N}\mathrm{B}\mathrm{S}}={\mathrm{C}\mathrm{o}\mathrm{R}\mathrm{i}\mathrm{s}\mathrm{k}}_{i}\times\:(1-{\mathrm{A}\mathrm{v}\mathrm{a}\mathrm{i}\mathrm{l}\mathrm{a}\mathrm{b}\mathrm{i}\mathrm{l}\mathrm{i}\mathrm{t}\mathrm{y}}_{i,\mathrm{N}\mathrm{B}\mathrm{S}})$$\end{document}\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\mathrm{S}\mathrm{V}}_{i,\mathrm{H}\mathrm{U}}={\mathrm{C}\mathrm{o}\mathrm{R}\mathrm{i}\mathrm{s}\mathrm{k}}_{i}\times\:(1-{\mathrm{A}\mathrm{v}\mathrm{a}\mathrm{i}\mathrm{l}\mathrm{a}\mathrm{b}\mathrm{i}\mathrm{l}\mathrm{i}\mathrm{t}\mathrm{y}}_{i,\mathrm{H}\mathrm{U}})$$\end{document}\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\mathrm{S}\mathrm{V}}_{i,\mathrm{T}\mathrm{R}}={\mathrm{C}\mathrm{o}\mathrm{R}\mathrm{i}\mathrm{s}\mathrm{k}}_{i}\times\:(1-{\mathrm{A}\mathrm{v}\mathrm{a}\mathrm{i}\mathrm{l}\mathrm{a}\mathrm{b}\mathrm{i}\mathrm{l}\mathrm{i}\mathrm{t}\mathrm{y}}_{i,\mathrm{T}\mathrm{R}})$$\end{document}\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\mathrm{S}\mathrm{V}}_{i,\mathrm{C}\mathrm{L}}={\mathrm{C}\mathrm{o}\mathrm{R}\mathrm{i}\mathrm{s}\mathrm{k}}_{i}\times\:(1-{\mathrm{A}\mathrm{v}\mathrm{a}\mathrm{i}\mathrm{l}\mathrm{a}\mathrm{b}\mathrm{i}\mathrm{l}\mathrm{i}\mathrm{t}\mathrm{y}}_{i,\mathrm{C}\mathrm{L}})$$\end{document} These service-specific measures were used to identify catchments in which epidemiological need was high and the focal service was absent. Primary prioritisation framework Overall programme-priority score The primary national prioritisation analysis combined co-risk, standardised reported service deficit, and a smaller contribution from facility-recorded SCD mortality into an overall programme-priority score. The score was designed to order existing facility-linked planning zones where higher epidemiological context coincided with weaker reported service breadth, while allowing recorded mortality to contribute evidence of severe events visible in the routine system. Because all components were linked to the existing facility network, the score was interpreted as a comparative planning index for further assessment rather than an independent estimate of unmet demand. In the base specification, the score was defined as:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\text{Programme priority}}_{i}=0.45\times\:{\mathrm{Co-risk}}_{i}+0.45\times\:{\text{Service deficit}}_{i,\mathrm{s}\mathrm{t}\mathrm{d}}+0.10\times\:{\mathrm{Mortality}}_{i,\mathrm{s}\mathrm{t}\mathrm{d}}$$\end{document} The weighting structure was specified a priori and reflected the planning purpose of the framework. Co-risk and reported service deficit were assigned equal dominant weight because the primary contrast was between epidemiological context and reported service breadth. Facility-recorded mortality was assigned a smaller weight of 0.10 so that severe events visible in the routine system could influence the ordering without dominating it. Mortality was used rather than admissions in the base specification because admissions may be particularly affected by repeated utilisation, referral concentration, diagnostic intensity, and existing reported service-domain breadth. Recorded deaths were also not treated as a direct demand measure because they depend on contact with, diagnosis by, and reporting from the health system. The no-burden specification examined whether top-priority membership depended on this routine-burden component; it did not remove facility anchoring from the planning-zone or service-inventory inputs. Planning zones were ranked nationally according to the composite score. The weighting structure was informed by the stated planning objective rather than by an external clinical effect model. The numerical values were investigator specified and were not derived from WHO guidance, a formal national policy weighting exercise, published effect estimates, statistical optimisation, or structured expert or stakeholder preference elicitation. They should therefore be interpreted as transparent normative choices for this planning application rather than as empirically estimated coefficients. Service-specific rollout priorities Separate service-specific prioritisation outputs were generated for newborn screening, hydroxyurea availability, blood transfusion strengthening, and specialised SCD clinic expansion. Each service-specific output was used to identify catchments in which epidemiological co-risk was high and the focal service was absent. For catchment i and service domain s, the service-specific priority score was defined as:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\mathrm{Priority}}_{i,s}={\mathrm{CoRisk}}_{i}\times\:(1-{\mathrm{Availability}}_{i,s})$$\end{document} where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\mathrm{Availability}}_{i,s}$$\end{document} is a binary indicator taking the value 1 if catchment i had availability to service s and 0 otherwise. Under this formulation, the score equals the co-risk value when the focal service is absent and equals 0 when the service is already available. Thus, the highest service-specific priority values occurred in catchments where inferred epidemiological need was high and the corresponding service was not currently available. Separate national rankings were then generated for each service domain, allowing programme expansion to be tailored to the geography of estimated service-to-need mismatch for newborn screening, hydroxyurea, transfusion support, and specialised clinic care. Secondary analyses Coverage-to-need gap and population-weighted concentration of need To describe the relative mismatch between epidemiological context and reported service breadth, we derived an overall coverage-to-need gap as planning-zone co-risk minus normalised reported service-domain breadth. Parallel service-specific gap measures were constructed for newborn screening, hydroxyurea, transfusion, and specialised clinic care. Positive values indicated that co-risk was numerically higher than reported availability on their common relative scales. The measures did not represent observed service coverage, household demand, or the proportion of need that was unmet. We used these variables to summarise national patterns, produce quintile-based maps, and identify the highest-gap planning zones. Population-weighted analyses then described the number and proportion of people represented within top-priority zones and calculated population-weighted means for co-risk, service deficit, and the coverage-to-need gap. We calculated the proportion of the represented catchment population residing in the national top 20% of overall programme-priority catchments. The top 20% was defined by catchment count and comprised the 56 highest-ranked catchments among the 280 included in the analysis. Because catchments varied in population size, the proportion of the represented population residing in these catchments was not expected to equal 20%. This measure was interpreted as the population share residing within the highest-priority catchments, rather than as service coverage or the number of people with SCD. Population size was not included in the primary programme-priority score because the score was designed to measure the relative intensity of inferred epidemiological context and reported service deficit independently of population reach. Population reach was examined separately through the population-impact measure, which combined the coverage-to-need gap with catchment population. The programme-priority and population-impact measures therefore represented complementary planning perspectives rather than interchangeable rankings. Detailed derivation is provided in Supplementary Material 1: Methods S3. Subregional concentration of priority and rollout needs To translate national catchment-level outputs into implementation-relevant regional groupings, we aggregated results to subregional level. Subregional summaries included catchment counts, population-weighted mean values for co-risk, service deficit, coverage-to-need gap, and overall priority score, the proportion of catchments in the national top 20% of programme priority and the concentration of service-specific top-50 catchments. These summaries were used to identify the subregions in which overall and service-specific rollout needs were most strongly concentrated. Detailed procedures for subregional aggregation, concentration of top-priority catchments, and service-specific top-50 summaries are provided in Supplementary Material 1: Methods S4. Sensitivity to alternative weighting specifications Because the base weights were investigator-specified planning choices rather than fitted coefficients, the overall programme-priority score was recalculated under six prespecified weighting specifications (Table 1, Supplementary Material 1: Table S2). Planning zones were ranked separately under each specification, the top 50 were identified, and pairwise overlap was calculated. For each planning zone, selection frequency was the number of specifications under which it appeared in the top-50 set. This analysis assessed sensitivity to the selected component weights and burden indicator only; it did not quantify uncertainty in the spatial inputs, planning-zone construction, service measurement, or patient flows. Detailed calculations and catchment-level outputs are provided in Supplementary Material 1: Methods S5 and Supplementary Material 1: Tables S20 to S23. Table 1Programme-priority weighting specifications used in the sensitivity analysisSpecificationCo-risk weightService-deficit weightRoutine-burden componentBase0.450.45Recorded SCD deaths, 0.10Co-risk emphasis0.600.30Recorded SCD deaths, 0.10Service-deficit emphasis0.300.60Recorded SCD deaths, 0.10No burden0.500.50None, 0.00Death heavy0.400.40Recorded SCD deaths, 0.20Admission heavy0.400.40Recorded SCD admissions, 0.20Note: All components were standardised to the interval from 0 to 1 before combination, and weights within each specification summed to 1. The specifications were prespecified normative decision rules and were not statistically fitted models or uncertainty intervals Table 1Programme-priority weighting specifications used in the sensitivity analysisSpecificationCo-risk weightService-deficit weightRoutine-burden componentBase0.450.45Recorded SCD deaths, 0.10Co-risk emphasis0.600.30Recorded SCD deaths, 0.10Service-deficit emphasis0.300.60Recorded SCD deaths, 0.10No burden0.500.50None, 0.00Death heavy0.400.40Recorded SCD deaths, 0.20Admission heavy0.400.40Recorded SCD admissions, 0.20Note: All components were standardised to the interval from 0 to 1 before combination, and weights within each specification summed to 1. The specifications were prespecified normative decision rules and were not statistically fitted models or uncertainty intervals Programme-priority weighting specifications used in the sensitivity analysis Note: All components were standardised to the interval from 0 to 1 before combination, and weights within each specification summed to 1. The specifications were prespecified normative decision rules and were not statistically fitted models or uncertainty intervals All analyses were done in R version 4.6.1 (R Foundation for Statistical Computing, Vienna, Austria), and cartographic outputs were produced using QGIS version 4.2.2 (QGIS Development Team, QGIS Association, Switzerland) [45]. Results Analytic dataset construction and descriptive planning inputs The final national dataset comprised 280 catchments, with a cumulative recorded burden of 65,800 SCD admissions and 1,010 SCD deaths over the study period. The complete catchment-level analytic dataset is provided in Supplementary Material 1: Table S1. Data completeness was high, with no missing HbS values, five catchments missing PfPR2-10 values, and no missing population estimates. The co-risk score showed marked geographic heterogeneity, ranging from 0.003 to 0.934, with a median of 0.433 (interquartile range [IQR], 0.298 to 0.656) and a mean of 0.457. Essential SCD reported service-domain breadth was sparse. 122 catchments (43.6%) had availability to none of the four essential services, 83 (29.6%) had availability to one, 32 (11.4%) had availability to two, 20 (7.1%) had availability to three, and only 23 (8.2%) had availability to all four. In total, 205 catchments (73.2%) had availability to zero or one essential service, indicating that weak reported service-domain breadth was a widespread feature of the national catchment landscape. Detailed summaries of analytic coverage, raster completeness, co-risk, and service availability are presented in Supplementary Material 1: Table S3. Service-mediated SCD structural vulnerability and operational planning classes Service-mediated SCD structural vulnerability showed marked heterogeneity across the 280 catchments, ranging from 0.000 to 0.905, with a median of 0.310 (IQR 0.097 to 0.487) and a mean of 0.317. These values indicate substantial variation in the extent to which elevated co-risk coincided with weak essential SCD reported service-domain breadth across the national catchment system. The descriptive four-cell co-risk and reported service-domain breadth classification showed that 95 catchments (33.9%) were classified as high co-risk and low reported service-domain breadth, 45 (16.1%) as high co-risk and high reported service-domain breadth, 110 (39.3%) as low co-risk and low reported service-domain breadth, and 30 (10.7%) as low co-risk and high reported service-domain breadth. Service-specific vulnerability was greatest for hydroxyurea and newborn screening. Mean vulnerability scores were 0.380 for hydroxyurea, 0.352 for newborn screening, 0.280 for transfusion, and 0.278 for specialised clinic care (Table 2). Table 2Distribution of service-mediated SCD structural vulnerability, descriptive four-cell co-risk and reported service-domain breadth classes, service-specific vulnerability, and programme typology categories across catchmentsMeasureMetricValueService-mediated SCD structural vulnerabilityMinimum0.00025th percentile0.097Median0.310Mean0.31775th percentile0.487Maximum0.905Descriptive four-cell co-risk and reported service-domain breadth classificationHigh co-risk + low reported service-domain breadth, n (%)95 (33.9)High co-risk + high reported service-domain breadth, n (%)45 (16.1)Low co-risk + low reported service-domain breadth, n (%)110 (39.3)Low co-risk + high reported service-domain breadth, n (%)30 (10.7)Service-specific vulnerabilityHydroxyurea, mean0.380Newborn screening, mean0.352Transfusion, mean0.280Specialised clinic care, mean0.278 Table 2Distribution of service-mediated SCD structural vulnerability, descriptive four-cell co-risk and reported service-domain breadth classes, service-specific vulnerability, and programme typology categories across catchmentsMeasureMetricValueService-mediated SCD structural vulnerabilityMinimum0.00025th percentile0.097Median0.310Mean0.31775th percentile0.487Maximum0.905Descriptive four-cell co-risk and reported service-domain breadth classificationHigh co-risk + low reported service-domain breadth, n (%)95 (33.9)High co-risk + high reported service-domain breadth, n (%)45 (16.1)Low co-risk + low reported service-domain breadth, n (%)110 (39.3)Low co-risk + high reported service-domain breadth, n (%)30 (10.7)Service-specific vulnerabilityHydroxyurea, mean0.380Newborn screening, mean0.352Transfusion, mean0.280Specialised clinic care, mean0.278 Distribution of service-mediated SCD structural vulnerability, descriptive four-cell co-risk and reported service-domain breadth classes, service-specific vulnerability, and programme typology categories across catchments The highest service-mediated SCD structural vulnerability catchments were tightly concentrated within the high co-risk, low reported service-domain breadth group. All 20 top-ranked catchments were classified as high co-risk and low reported service-domain breadth. The first five ranked catchments had no availability to any of the four essential SCD service domains and a standardised service deficit of 1.00, while the remaining 15 had availability to only one service domain and a standardised service deficit of 0.75 (Supplementary Material 1: Table S4). This concentration indicates that the highest-vulnerability settings were characterised by simultaneous elevation of inferred need and profound weakness of the local service platform. National prioritisation framework and service-specific rollout priorities The overall programme-priority surface showed marked spatial clustering across the national catchment network. The highest-priority catchments were concentrated mainly in the north, north-east, east, and parts of the central belt, while much of the south-west was dominated by lower-priority tiers (Fig. 1). The full list of top 50 -ranked overall programme-priority catchments is presented in Supplementary Material 1: Table S5. Fig. 1Spatial distribution of the base-case programme-priority score across facility-linked planning zones in Uganda. Planning zones are grouped into quintiles, with P1 indicating the highest priority and P5 the lowest. Areas outside the represented planning zones are unclassified. Subregional and national boundaries, water bodies, a north arrow, and a scale bar are shown for geographic referenceThe map presents the spatial distribution of programme-priority scores across planning zones in Uganda, grouped into five priority levels from P1, the highest priority, to P5, the lowest priority. Several planning zones in the central and northern regions are classified as P1, indicating the highest priority. Surrounding these are zones with P2 and P3 priority levels, while many zones in the western and southwestern regions fall into the lower priority categories P4 and P5. Water bodies, including large lakes and rivers, are marked and do not have priority scores. Subregional and national boundaries are delineated to provide geographic context. Spatial distribution of the base-case programme-priority score across facility-linked planning zones in Uganda. Planning zones are grouped into quintiles, with P1 indicating the highest priority and P5 the lowest. Areas outside the represented planning zones are unclassified. Subregional and national boundaries, water bodies, a north arrow, and a scale bar are shown for geographic reference Service-specific rollout maps showed substantial but incomplete overlap across intervention domains. Newborn screening and hydroxyurea priorities were closely aligned and concentrated strongly in the north-east, east, and south-east, suggesting a shared geography of estimated service-to-need mismatch for early diagnosis and disease-modifying therapy. Transfusion strengthening priorities had a somewhat broader footprint, extending across the east, north-east, north, and parts of the north-central belt, indicating that deficits in emergency and supportive care affected a wider set of catchments. Specialised clinic priorities were more selective, though still concentrated mainly in the east and north, suggesting a more targeted geography for expansion of higher-level continuity and coordination of care (Fig. 2). Fig. 2Service-specific top-50 priority planning zones in Uganda for newborn screening (A), hydroxyurea (B), transfusion strengthening (C), and specialised SCD clinic expansion (D). Red indicates the 50 highest-ranked zones for each service; blue indicates all other represented zones. Areas outside the planning-zone framework are unclassified. Boundaries, water bodies, north arrows, and scale bars are shown for referenceThe image presents four maps of Uganda, each identifying the top 50 priority planning zones for different health services related to sickle cell disease. Each map highlights specific zones ranked highest for one service: newborn screening expansion, hydroxyurea expansion, transfusion strengthening, and specialized sickle cell disease clinic expansion. The maps show that the highest-ranked zones for all services are concentrated mainly in the central and eastern regions of Uganda, with fewer priority zones in the northern and western areas. Water bodies, sub-region boundaries, and national boundaries are included for geographic context. Service-specific top-50 priority planning zones in Uganda for newborn screening (A), hydroxyurea (B), transfusion strengthening (C), and specialised SCD clinic expansion (D). Red indicates the 50 highest-ranked zones for each service; blue indicates all other represented zones. Areas outside the planning-zone framework are unclassified. Boundaries, water bodies, north arrows, and scale bars are shown for reference The complete top-50 service-specific rankings are presented separately for newborn screening in Supplementary Material 1: Table S6, hydroxyurea in Supplementary Material 1: Table S7, transfusion in Supplementary Material 1: Table S8, and specialised SCD clinic care in Supplementary Material 1: Table S9. Coverage-to-need gaps and population-weighted concentration of need Nationally, the overall coverage-to-need gap had a mean of 0.190 and a median of 0.205, with values ranging from − 0.750 to 0.905. Service-specific mean gaps were 0.189 for hydroxyurea, 0.168 for newborn screening, 0.146 for transfusion, and 0.138 for specialised clinic care. Median values were 0.183 for newborn screening, hydroxyurea, and transfusion, and 0.098 for specialised clinic care. Maximum values were 0.934 for newborn screening, hydroxyurea, and transfusion and 0.905 for the overall and specialised-clinic gaps; the minimum was − 0.750 for all five measures (Table 3). The equal medians for three service-specific gaps were empirical summaries of their observed distributions. Table 3National summary of overall and service-specific coverage-to-need gaps, population concentration, and population-weighted mean planning metrics across Ugandan health facility catchmentsDomainMetricValueOverall gapMean overall coverage-to-need gap0.190Median overall coverage-to-need gap0.205Minimum overall coverage-to-need gap-0.750Maximum overall coverage-to-need gap0.905Service-specific gapsMean newborn-screening gap0.168Median newborn-screening gap0.183Minimum newborn-screening gap-0.750Maximum newborn-screening gap0.934Mean hydroxyurea gap0.189Median hydroxyurea gap0.183Minimum hydroxyurea gap-0.750Maximum hydroxyurea gap0.934Mean transfusion gap0.146Median transfusion gap0.183Minimum transfusion gap-0.750Maximum transfusion gap0.934Mean specialised-clinic gap0.138Median specialised-clinic gap0.098Minimum specialised-clinic gap-0.750Maximum specialised-clinic gap0.905Population concentrationTotal population represented40.4 millionPopulation in top 20% priority catchments, n (%)7.0 million (17.3)Population-weighted meansCo-risk0.472Service deficit0.709Coverage-to-need gap0.181 Table 3National summary of overall and service-specific coverage-to-need gaps, population concentration, and population-weighted mean planning metrics across Ugandan health facility catchmentsDomainMetricValueOverall gapMean overall coverage-to-need gap0.190Median overall coverage-to-need gap0.205Minimum overall coverage-to-need gap-0.750Maximum overall coverage-to-need gap0.905Service-specific gapsMean newborn-screening gap0.168Median newborn-screening gap0.183Minimum newborn-screening gap-0.750Maximum newborn-screening gap0.934Mean hydroxyurea gap0.189Median hydroxyurea gap0.183Minimum hydroxyurea gap-0.750Maximum hydroxyurea gap0.934Mean transfusion gap0.146Median transfusion gap0.183Minimum transfusion gap-0.750Maximum transfusion gap0.934Mean specialised-clinic gap0.138Median specialised-clinic gap0.098Minimum specialised-clinic gap-0.750Maximum specialised-clinic gap0.905Population concentrationTotal population represented40.4 millionPopulation in top 20% priority catchments, n (%)7.0 million (17.3)Population-weighted meansCo-risk0.472Service deficit0.709Coverage-to-need gap0.181 National summary of overall and service-specific coverage-to-need gaps, population concentration, and population-weighted mean planning metrics across Ugandan health facility catchments The combined ranking of overall and service-specific coverage-to-need gaps is provided in Supplementary Material 1: Table S10. Separate top-50 rankings are provided for the newborn-screening gap in Supplementary Material 1: Table S11, hydroxyurea gap in Supplementary Material 1: Table S12, transfusion gap in Supplementary Material 1: Table S13, and specialised-clinic gap in Supplementary Material 1: Table S14. The spatial distribution of coverage-to-need mismatches was strongly clustered. Catchments in the highest overall gap quintile were concentrated mainly in the north, north-east, east, and south-east, broadly overlapping the main programme-priority corridor. This pattern indicates that the areas of greatest estimated service-to-need mismatch were also the areas in which current service provision fell furthest short of inferred need (Fig. 3). Fig. 3Overall coverage-to-need gap across facility-linked planning zones in Uganda. Planning zones are grouped into quintiles, with Q1 indicating the highest gap and Q5 the lowest. The gap is the co-risk score minus normalised reported service breadth and is a relative arithmetic difference rather than a measure of realised coverage or the proportion of need unmet. Areas outside the represented planning zones are unclassifiedThe map presents the relative coverage-to-need gap across facility-linked planning zones in Uganda, grouped into five categories from highest gap to lowest gap. Zones with the highest gap, labeled Q1, are concentrated mainly in the northern and eastern regions, indicating a larger difference between co-risk scores and reported service breadth. Zones with the lowest gap, labeled Q5, are primarily located in the southwestern and some central areas, showing smaller differences. Intermediate gap levels, Q2 through Q4, are distributed between these extremes. The map also outlines national and sub-region boundaries and marks water bodies, for geographical context. Overall coverage-to-need gap across facility-linked planning zones in Uganda. Planning zones are grouped into quintiles, with Q1 indicating the highest gap and Q5 the lowest. The gap is the co-risk score minus normalised reported service breadth and is a relative arithmetic difference rather than a measure of realised coverage or the proportion of need unmet. Areas outside the represented planning zones are unclassified Service-specific gap patterns showed substantial, but not complete, overlap across intervention domains. Newborn screening and hydroxyurea gaps were closely aligned, with large clusters of highest-gap catchments in the east, north-east, and south-east. Transfusion gaps showed a broader geographic footprint, extending more clearly into the north and central belt. By contrast, specialised clinic care gaps were more spatially mixed, with high-gap catchments still present in the east and north but showing a less continuous national pattern. The four service-specific gap maps are presented in Supplementary Material 1: Fig S3. The 56 catchments comprising the national top 20% of overall programme priority contained approximately 7.0 million people, equivalent to 17.3% of the 40.4 million people represented within the analytic catchment framework. The top-priority group was defined by catchment rank rather than cumulative population, and its population share was therefore not expected to equal 20%. The mean represented population was approximately 125,000 people among the top-priority catchments, compared with approximately 144,000 people across all 280 catchments. Thus, the top-priority catchments were descriptively about 13% less populous on average than the full catchment set. No formal comparison of the catchment-population distributions was undertaken. This result indicates that the primary priority score did not preferentially select the most populous catchments. Population-weighted national means were 0.472 for co-risk, 0.709 for service deficit, and 0.181 for the overall coverage-to-need gap (Table 3). The spatial distribution of planning zones in the national top 20% of programme priority and the distribution of catchment population are shown in Supplementary Material 1: Fig S4. The separate population-impact analysis identified catchments where larger coverage-to-need gaps coincided with greater population reach. Population-weighted concentration summaries and the top 50 catchments ranked by population-impact priority are presented in Supplementary Material 1: Table S15. Subregional concentration of priority and rollout needs Subregional analysis showed that overall programme priority was concentrated most strongly in Lango, Teso, Acholi, Karamoja, and Busoga, which consistently emerged among the highest-priority subregions across the aggregated planning metrics. The subregional distribution of the population-weighted mean overall programme-priority score is shown in Fig. 4. Fig. 4Population-weighted mean base-case programme-priority by subregion (Q5 highest; Q1 lowest)The map displays subregional programme priorities ranked from lowest (Q1) to highest (Q5). The northern and northeastern subregions, including Acholi and parts of Karamoja, have the highest priority level, Q5. Central and some eastern subregions, such as parts of Bugisu and Teso, are assigned Q4 priority. The central area around Kampala and some surrounding subregions fall into Q3. The southwestern and western subregions, including Ankole and parts of Toro, have the lowest priority levels, Q1 and Q2. Water bodies and national boundaries are also marked for geographic context. Population-weighted mean base-case programme-priority by subregion (Q5 highest; Q1 lowest) The highest population-weighted mean priority scores were observed in Lango (0.667), Teso (0.640), Acholi (0.633), Karamoja (0.632), and Busoga (0.618) (Table 4). These same subregions also had among the highest population-weighted co-risk values, particularly Lango (0.835), Acholi (0.826), and Busoga (0.766), indicating that the strongest programme priorities were concentrated mainly in northern, north-eastern, and eastern Uganda. The complete subregional summary of programme priority, recorded burden, co-risk, service deficit, and coverage-to-need gap is provided in Supplementary Material 1: Table S16. The distribution of programme typology categories by subregion is presented in Supplementary Material 1: Table S17, and the subregions with the strongest concentration of overall programme priority are summarised in Supplementary Material 1: Table S18. The distribution of planning zones in the national top 20% priority group followed a similar pattern. Acholi had 10 of 15 zones (66.7%) in the top-priority group, followed by Busoga with 13 of 21 (61.9%), Lango with 8 of 14 (57.1%), Teso with 8 of 18 (44.4%), and Karamoja with 2 of 7 (28.6%). Greater Kampala, included in Table 4 as a high-population comparator, had 0 of 26 top-20% zones and a population-weighted mean priority score of 0.496. In the complete subregional results, every subregion containing at least one top-20% zone had a higher population-weighted mean priority score than Greater Kampala. Service-specific rollout priorities were also concentrated in a relatively small number of subregions. Busoga had the largest combined number of service-specific top-50 appearances across newborn screening, hydroxyurea, transfusion, and specialised clinic care [50], followed by Teso [41], Acholi [38], Lango [34], and Bukedi [17]. The full subregional distribution of service-specific top-50 priority catchments is presented in Supplementary Material 1: Table S19. These findings indicate that both overall programme priority and intervention-specific rollout needs were concentrated in a limited set of subregions rather than being diffusely distributed across the country. Table 4Subregions with the six highest population-weighted mean programme-priority scores, with Greater Kampala included as a high-population comparator, and service-specific top-50 countsSubregionPW mean priority scoreTop 20% priority catchments, n (%)Total service-specific top-50 countsLango0.6678 (57.1)34Teso0.6408 (44.4)41Acholi0.63310 (66.7)38Karamoja0.6322 (28.6)5Busoga0.61813 (61.9)50Bukedi0.5675 (33.3)17Greater Kampala*0.4960 (0.0)0* Greater Kampala was included separately because it contained the largest share of the represented population; it was not among the six highest-ranked subregions. The population-weighted mean incorporates all planning-zone scores weighted by represented population, whereas the top-20% column counts zones among the 56 highest-ranked nationally Table 4Subregions with the six highest population-weighted mean programme-priority scores, with Greater Kampala included as a high-population comparator, and service-specific top-50 countsSubregionPW mean priority scoreTop 20% priority catchments, n (%)Total service-specific top-50 countsLango0.6678 (57.1)34Teso0.6408 (44.4)41Acholi0.63310 (66.7)38Karamoja0.6322 (28.6)5Busoga0.61813 (61.9)50Bukedi0.5675 (33.3)17Greater Kampala*0.4960 (0.0)0* Greater Kampala was included separately because it contained the largest share of the represented population; it was not among the six highest-ranked subregions. The population-weighted mean incorporates all planning-zone scores weighted by represented population, whereas the top-20% column counts zones among the 56 highest-ranked nationally Subregions with the six highest population-weighted mean programme-priority scores, with Greater Kampala included as a high-population comparator, and service-specific top-50 counts * Greater Kampala was included separately because it contained the largest share of the represented population; it was not among the six highest-ranked subregions. The population-weighted mean incorporates all planning-zone scores weighted by represented population, whereas the top-20% column counts zones among the 56 highest-ranked nationally Stability of top-50 membership across weighting specifications Across the six weighting specifications, 50 planning zones appeared in the top 50 under at least five specifications, 23 appeared under one specification, and 207 were not selected under any specification. No planning zone appeared under two to four specifications. These selection frequencies describe stability to the evaluated weighting choices only. The overall top-50 selection-frequency summary is provided in Supplementary Material 1: Table S20, with catchment-level lists presented in Supplementary Material 1: Tables S21 and S22. Agreement between specifications was highest among the base, no-burden, death-heavy, and admission-heavy specifications, which selected the same 50 planning zones (50/50, 100.0%). Overlap with the base specification was lower under co-risk emphasis (41 of 50, 82.0%) and lower still under service-deficit emphasis (36 of 50, 72.0%). The smallest overlap was observed between the co-risk-emphasis and service-deficit-emphasis specifications (27/50, 54.0%). The identical base and no-burden top-50 sets indicate that removing the routine-burden component did not change membership of the selected top set. This limits concern that the small mortality component alone determined those selections, but it does not address facility anchoring in the service inventory or planning-zone construction. The complete pairwise overlap structure is shown in Supplementary Material 1: Table S23. Discussion We developed a national facility-linked framework that compared existing higher-level public-facility nodes for potential strengthening of essential SCD services in Uganda. Under the specified inputs and assumptions, higher comparative scores were concentrated in eastern, north-eastern, northern, and parts of central Uganda, while service-specific priorities overlapped only partly. Newborn screening and hydroxyurea priorities were more closely aligned, transfusion priorities had a broader geographic footprint, and specialised-clinic priorities were more selective. These findings identify candidate facility nodes and relative planning contrasts rather than precise estimates of community demand or a definitive sequence for implementation. The contribution of this study lies in linking national epidemiological, service, routine health-system, population, and accessibility data to existing facility nodes within a single planning framework. Previous Ugandan studies have separately examined SCD epidemiology [16], malaria transmission[15], national screening activity [9], clinic-based treatment experience [17], and geographic access to government health facilities [18]. By bringing these data together at the level of 280 existing HC IV and higher-level facilities, the present analysis addresses a different planning problem: where should programmes focus when considering expansion of particular SCD services? The service-specific rankings also show that the geographic pattern for one component of care cannot be assumed to apply to the others, which is relevant when planning screening, hydroxyurea, transfusion, and specialised follow-up as distinct service functions. The concentration of higher scores in eastern and northern Uganda broadly corresponds with previously reported geographic patterns of SCD in the country. In the US3, sickle cell trait and SCD prevalence were highest in East Central Uganda and were also high in the Mid-Northern region, whereas the South Western region had the lowest reported prevalence [16]. Subsequent targeted screening similarly documented SCD prevalence above 1% in East Central and Mid-Northern districts [46], while earlier district-level data reported substantially higher sickle cell trait prevalence in eastern Uganda than in Mbarara and Ntungamo in the south-west [47]. These comparisons provide epidemiological context for the present spatial pattern but should not be interpreted as independent validation because the co-risk index incorporates an HbS surface derived partly from the same national screening evidence. The spatial agreement is therefore expected in part and demonstrates internal coherence rather than external validation. In addition, the HbS and PfPR2-10 inputs reflect related historical processes because malaria selection contributed to the distribution of HbS. Their combination was used to describe inherited-risk geography together with contemporary malaria transmission ecology, not as two independent estimates of SCD prevalence. The priority surface is a composite facility-linked planning signal produced by the overlap of epidemiological proxy values, reported service configuration, and a small routine-mortality component. Primary geospatial studies from SSA have shown that national averages can conceal subnational differences in service availability and accessibility [48, 49], but the present analysis remains conditioned on the existing higher-level public-facility network. The identical top-50 membership under the base and no-burden specifications indicates that the routine-burden component did not determine the selected top set. It does not, however, remove supply-side dependence arising from the facility-linked service inventory and the construction of planning zones around existing facilities. Lower-priority values, including those in much of south-western Uganda, should therefore be interpreted comparatively and not as evidence of absent demand or absent need. Demand-side data on patient origins, care-seeking, bypassing, referral flows, and effective access would be required to evaluate how closely the rankings correspond to community utilisation and demand. Service priorities differed across domains, indicating that SCD expansion in Uganda cannot follow a single rollout model. Newborn screening and hydroxyurea priorities overlapped most strongly in eastern, north-eastern, and south-eastern Uganda, which is clinically coherent because screening provides the entry point for early diagnosis and longitudinal care, while hydroxyurea has proven feasible and beneficial in African settings. Uganda’s national screening programme demonstrated the feasibility of large-scale routine screening, and the REACH trial showed that hydroxyurea reduced vaso-occlusive pain, infections, malaria, transfusions, and hospitalisations in African children with sickle cell anaemia [9, 21]. The observed spatial overlap suggests that some catchments could be considered for joint facility-readiness assessment for newborn screening and hydroxyurea. Whether these services should be expanded together would depend on confirmatory diagnostic capacity, medicine availability, laboratory monitoring, staffing, referral pathways, costs, and patient follow-up systems. Transfusion priorities were more geographically extensive, particularly in northern and north-central Uganda, suggesting that supportive and emergency care needs extend beyond the areas prioritised for screening and hydroxyurea. This is consistent with the wider systems requirements of safe transfusion care, including blood availability, compatibility testing, and immunohaematology capacity. Evidence from Uganda showing red-cell alloimmunisation among patients with SCD further underscores the need for laboratory and systems strengthening alongside wider transfusion access [50]. Priorities for specialised SCD clinics were more selective, although still concentrated mainly in the east and north. This suggests that continuity-oriented services such as longitudinal monitoring, counselling, and coordinated follow-up may require more targeted expansion. Clinic-based evidence from eastern Uganda supports this interpretation, showing concentration of specialised care in few facilities and limited hydroxyurea use among attendees [17]. The coverage-to-need analyses provided an additional descriptive view of where epidemiological proxy values were high relative to reported service breadth. The highest overall gap quintiles were concentrated mainly in eastern and northern Uganda and broadly overlapped the main programme-priority corridor. Because the gap was calculated as a difference between two standardised measures, it should be interpreted as a relative arithmetic mismatch rather than observed coverage, household demand, or the proportion of need that was unmet. Service-specific gaps were not distributed uniformly: newborn screening and hydroxyurea gaps were closely aligned, transfusion gaps were broader, and specialised-clinic gaps were more mixed. This variation is consistent with evidence that different measures of spatial access can produce different representations of service reach [23]. The gap outputs can therefore help identify facility-linked areas for further assessment, but decisions would also require information on capacity, service quality, affordability, referral pathways, and care-seeking. The subregional summaries translate facility-linked point estimates into broader implementation groupings. Concentration in Lango, Teso, Acholi, Busoga, and Karamoja identifies areas in which detailed demand, readiness, referral, feasibility, and equity assessments may be especially informative. These descriptive rankings do not establish a first wave of rollout or demonstrate that lower-ranked subregions lack SCD service need. Greater Kampala was retained as a high-population comparator, and its absence from the national top-20% set should not be interpreted as absence of need. The national top 20% of programme-priority catchments contained 17.3% of the represented population because the priority group was defined by the number of catchments rather than by cumulative population. Catchments varied in population size, and the 56 top-priority catchments contained approximately 125,000 people on average, compared with approximately 144,000 across the full catchment set. They were therefore descriptively less populous on average, although no formal test of the population distributions was undertaken. This finding reflects the distinction between the intensity of relative priority and the scale of population reach. The primary score ranked catchments according to co-risk, reported service deficit, and a smaller routine mortality component. It did not include population size as an additional weight. This avoided conflating severe relative service-to-need mismatch with the number of people living within a catchment. The population-impact analysis provided a complementary planning perspective by identifying catchments where a substantial service-to-need gap coincided with a larger represented population. Decision-makers may therefore consider both measures: programme priority to identify the intensity of relative need and service weakness, and population impact to consider the potential scale of population reach. The alternative specifications assessed sensitivity to normative weighting choices, not general robustness of the framework. The base, no-burden, death-heavy, and admission-heavy specifications selected the same top-50 planning zones, indicating that membership of this set was unchanged under the evaluated modifications to the routine-burden component. Membership changed more when the relative emphasis assigned to co-risk and reported service deficit was altered. Such variation is best interpreted as reflecting transparent planning trade-offs rather than methodological weakness. This view accords with the broader priority-setting literature, which emphasises that structured allocation frameworks are most credible when weighting decisions are explicit and their effects on rankings are visible to decision makers [51, 52]. However, these results do not address uncertainty in the epidemiological surfaces, planning-zone geometry, travel-time threshold, service indicators, or supply-dependent routine data. The findings can inform subsequent programme appraisal rather than prescribe resource allocation. The overlap of newborn-screening and hydroxyurea priorities may support joint feasibility assessment in some planning zones, while the broader transfusion pattern and more selective specialised-clinic pattern identify different systems questions for further investigation. Final sequencing should incorporate direct demand-side evidence, functional readiness and quality, referral relationships, costs, implementation feasibility, equity priorities, and structured stakeholder deliberation. The framework should therefore be used as one decision-support input alongside these additional considerations. Before operational adoption, the relative weights could be revisited through structured stakeholder elicitation or multi-criteria decision analysis; data-driven estimation would require an independent outcome against which candidate weights could be fitted and validated. However, this study has several limitations. First, the framework was predominantly supply-side and facility anchored: planning zones, service availability, and routine admissions and deaths were all linked to existing eligible public facilities, while direct demand-side measures such as patient origin, care-seeking, affordability, bypassing, referral flows, and realised utilisation were unavailable. Although the base and no-burden specifications produced identical top-50 sets, this does not remove supply dependence from the planning-zone and service-inventory components. Second, the composite weights were normative, investigator-specified choices rather than values derived from statistical optimisation, formal stakeholder elicitation, or an established external standard; the alternative specifications therefore assess sensitivity to weighting choices, not overall statistical robustness. Third, the binary service score measured reported breadth across four domains rather than functionality, capacity, readiness, staffing, continuity, quality, or effective coverage, and facilities with very different capacities could receive the same score. Fourth, the descriptive four-cell classification depended on selected cut-points. Higher co-risk was defined relative to the national median of 0.433 and low service availability as zero or one reported service domain. These thresholds were not clinically validated or tested against alternatives, so group membership may vary under different definitions, although the continuous structural-vulnerability, programme-priority, coverage-to-need, and service-specific measures were independent of these cut-points. Fifth, facility-recorded admissions and deaths are shaped by diagnosis, access, referral, service capacity, reporting completeness, and repeated utilisation and therefore are not independent measures of community-level SCD demand or burden. Sixth, the epidemiological inputs were spatial proxies rather than direct measures of planning-zone SCD prevalence. HbS represented inherited-risk geography and PfPR2−10 contemporary malaria transmission ecology; because historical malaria selection contributed to the geographic distribution of HbS, the two surfaces may share spatial structure, and averaging them may reinforce areas where both are elevated. The analysis did not isolate the independent contribution of PfPR2−10 after accounting for HbS or compare the equal-weight formulation with an HbS-only alternative, so the co-risk score should be interpreted as a study-specific relative epidemiological-context index rather than a biologically calibrated measure of prevalence, case numbers, or demand. Seventh, the source data introduce additional sampling, spatial, and temporal uncertainty. The HbS surface was interpolated from district-level programmatic screening among HIV-exposed infants and may smooth within-district variation, while PfPR2−10 was a modelled 2015 to 2024 mean among children aged 2 to 10 years and did not measure malaria exposure specifically among people living with SCD. Eighth, uncertainty in the HbS and PfPR2−10 surfaces was not propagated through the composite indices or rankings, so the resulting measures are conditional point estimates without confidence or credible intervals. Ninth, the 60-minute travel-time threshold was an evidence-informed planning specification rather than an SCD-specific clinical or utilisation threshold. Alternative 30- and 90-minute specifications were not evaluated, so zone extent, represented population, and downstream rankings remain conditional on the selected threshold. Tenth, the facility-linked planning zones were analytic units rather than observed utilisation catchments and did not capture patient choice, cross-zone care-seeking, bypassing, referral networks, or service-specific differences in accessibility. Eleventh, the prioritisation outputs were not externally validated against independent spatial data on confirmed SCD prevalence, newborn-screening yield, registry cases, patient residence, or utilisation, and their concordance with observed SCD burden therefore remains uncertain. Future research should incorporate geocoded patient-origin, referral, utilisation, affordability, and facility-readiness data; assess alternative or service-specific travel-time thresholds; validate rankings against independent prevalence, screening, and registry data; quantify the spatial association between HbS and PfPR2−10; compare HbS-only and alternative co-risk formulations; and examine stakeholder-informed weighting. Probabilistic geostatistical or Bayesian approaches could further propagate measurement and spatial uncertainty across epidemiological, service, accessibility, and utilisation inputs into planning-zone and regional priority estimates. Conclusions This study developed a national facility-linked approach for comparing where essential SCD services may warrant further strengthening within Uganda’s existing higher-level public health system. Overall priorities were concentrated mainly in eastern and northern Uganda, but the patterns differed across newborn screening, hydroxyurea, transfusion, and specialised clinic care, indicating that service expansion is unlikely to follow a single geographic pattern. The rankings provide a starting point for identifying facilities and surrounding populations for more detailed assessment of demand, readiness, referral arrangements, costs, feasibility, and equity. They do not provide a direct estimate of household demand, observed patient flows, effective service capacity, or a definitive territorial allocation rule and should be considered alongside demand-side, readiness, feasibility, and equity information before implementation. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1: Supplementary methods, tables and figures. Contains detailed methods for planning-zone construction, analytic dataset development, coverage-to-need and population analyses, subregional aggregation, and weighting sensitivity analyses; catchment- and subregion-level supplementary tables; and supplementary maps and figures. Supplementary Material 1: Supplementary methods, tables and figures. Contains detailed methods for planning-zone construction, analytic dataset development, coverage-to-need and population analyses, subregional aggregation, and weighting sensitivity analyses; catchment- and subregion-level supplementary tables; and supplementary maps and figures. Supplementary Material 2: ISLE-ReSt checklist. Completed Spatial Lifecourse Epidemiology Reporting Standards checklist for this study. Supplementary Material 2: ISLE-ReSt checklist. Completed Spatial Lifecourse Epidemiology Reporting Standards checklist for this study.