Document resource
Introduction Hyperosmolar hyperglycaemic state (HHS) is a potentially fatal complication of diabetes, predominantly affecting individuals with type 2 diabetes, particularly older adults. 1 Mortality rates associated with HHS remain high, ranging from 10% to 20%, with disturbances in electrolyte balance being a strong predictor of adverse outcomes during treatment.2–4 Given the clear influence of electrolyte abnormalities, hyperglycaemia, and hydration status on outcomes, comprehensive data on patients’ biochemical profiles at presentation—alongside fluid and insulin management strategies—are needed to inform and refine future HHS management guidelines.Geographical variations in the incidence of HHS are striking. In high-income regions, HHS is relatively uncommon. A Danish cohort study reported an incidence of 3.9 per 10 000 person-years between 2016 and 2018.5 In contrast, hyperglycaemic crises account for 26–40% of diabetes-related hospital admissions in Africa, with one Nigerian centre reporting that HHS constituted 58% of such cases.6 7 A systematic review further highlighted the wide variability in incidence across African countries, ranging from 2.1% to 46.3%.8Current knowledge of HHS is drawn from case series and small observational cohorts, limiting its applicability to broader clinical practice, and there is a lack of robust, large-scale studies or data exploring the underlying triggers of HHS and the associated mortality. This paucity of reliable health data impedes the development of effective prevention strategies, tailored clinical guidelines, and equitable resource allocation. Therefore, this study aimed to aggregate all previously reported cases of HHS to assess how the precipitating factors, presenting biochemistry and outcomes of the condition vary globally.Methods This systematic review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines. The review protocol was developed a priori and registered with the PROSPERO database (ID CRD42025625923).A comprehensive literature search was performed across seven electronic databases: Ovid MEDLINE, PubMed, EMBASE, Scopus, Web of Science, Cochrane Library, and Ovid EMCARE, from database inception to February 2025. The search strategy combined Medical Subject Headings terms and free-text keywords relevant to ‘hyperosmolar hyperglycaemic state’, ‘HHS’, ‘non-ketotic hyperglycaemia’, and related terms. No language or geographic restrictions were applied. Reference lists of relevant articles and systematic reviews were also manually searched to identify additional eligible studies. Full search strategies for each database are provided in the online supplemental materials.SP110.1136/bmjdrc-2025-005765.supp1Supplementary dataEligible studies included case series, observational studies (cross-sectional, cohort, and case–control), case reports, and systematic reviews reporting on patients with HHS aged 16 years and older. Studies were included only if cases met the diagnostic biochemical criteria defined by the international consensus for HHS: serum osmolality >320 mOsm/kg, serum glucose ≥33.3 mmol/L, absence of significant ketonaemia (serum ketones ≤3.0 mmol/L) and absence of acidosis (arterial/venous pH ≥7.3 and serum bicarbonate ≥15.0 mmol/L), or confirmed diagnosis of HHS using specific International Classification of Diseases codes,9 10 and reported at least one of the following outcomes: precipitating factors for HHS, management of HHS in relation to local guidelines, outcomes, and complications associated with HHS. Studies were excluded if they reported mixed diabetic ketoacidosis/HHS phenotypes without stratified reporting or contained duplicate patient data already reported in other included studies (determined through author correspondence and data cross-checking). We also excluded case studies with less than five episodes to prevent the inclusion of atypical cases that may skew the findings.Two reviewers independently screened titles, abstracts, and full texts using a prepiloted screening form. Discrepancies were resolved through discussion or arbitration by a third reviewer. The full screening process is documented in figure 1. The earliest study that met our inclusion criteria was published in 2006. Data were independently extracted in duplicate using a structured template and included study characteristics, patient demographics and precipitating factors of HHS, biochemical parameters of HHS on admission, management strategies, and clinical outcomes, including in-hospital mortality, intensive treatment unit (ITU) admission and length of hospital stay. To study variation in presentation by geographic or ethnic background, data on country of publication, patient nationality, and ethnicity were extracted and analysed.Figure 1Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow diagram of study selection. n denotes the number of studies.Summary statistics were stratified by patient region or nationality. Continuous variables were summarised as medians with IQRs, and categorical variables as counts and proportions. All statistical analyses were performed using SPSS V.30.0.For demographics, precipitants, outcomes, and complications, pooled frequencies were calculated by dividing the total number of patients with each characteristic across reporting studies by the total number of patients in those studies. Studies that did not report a given variable were excluded from that calculation.Studies were classified according to whether they included multiple admissions per patient or a single admission per patient, and separate meta-analyses were conducted for each group. As the estimands differ between patient-level and admission-level designs, combining them would yield biased or non-interpretable summary estimates. Where data permitted, we performed a random-effects meta-analysis of mortality proportions using a random-intercept logistic-normal model (generalised linear mixed model) with a logit link. Between-study heterogeneity (tau2) was estimated using maximum likelihood. Hartung-Knapp adjustments were applied to calculate random-effects CIs and the prediction interval. Individual study proportions were displayed on the logit scale and then back-transformed. Study-level 95% CIs were generated using the Clopper-Pearson method. Studies with zero events were included without continuity correction.Figure 3Forest plot demonstrating pooled estimates of mortality.Similarly, where data permitted, we conducted a meta-analysis of median length of stay using the quantile matching estimation method. Between-study heterogeneity was again estimated using maximum likelihood (tau2). Forest plot for All meta-analyses were performed in R. Forest plots for length of stay and mortality can be seen in figure 2 and figure 3 respectively.Figure 2Forest plot demonstrating pooled estimates of length of stay. The primary studies are represented as median (IQR) and the pooled estimate (CI). N denotes the number of patients and I2 indicates test of heterogeneity.Risk of bias and quality assessment Two reviewers independently assessed the methodological quality and risk of bias of eligible studies. Cohort studies were evaluated using the Newcastle-Ottawa Scale (NOS). An adapted NOS version was employed for cross-sectional and case–control designs, in line with previously published methods. 11–13 Discrepancies were resolved through consensus or consultation with a third reviewer.Results 27 studies were deemed eligible for inclusion, with a total of 63 935 cases of HHS, ranging from 2006 to 2024. The median age was 63.8 years (IQR: 56.1–72.0) across 24 studies (s=24). Age varied by region, with the oldest median in Europe (74.5 years, s=4) and the youngest in South America (55.8 years, s=3). While there was no sex preponderance overall (49.4%, n=30 100/60 880), there was a variation of sex ratios worldwide. Most HHS cases reported in Europe were in women (63.7%, n=4529/7106), while North America reported slightly more men having HHS (53.4%, n=266/498). Ethnicity data were almost entirely unrecorded (98.8%, n=63 199/63 935). Among recorded data, 11.6% were black and 0.6% were Asian. The full presenting features of all cases are documented in table 1. Patient medication history for included cases was not reported.Table 1Meta-summary of included HHS cases5 23 29–53ParametersTotal population (s=27)Asia (s=10)Europe (s=4)North America (s=5)South America (s=4)Oceania (s=1)Africa (s=3)Total number of HHS episodes63 93550 9147166343659472313Age, years63.8 (56.1–72.0)(s=24)70.6 (61.2–73.3)(s=9)74.5 (59.2–77.5)(s=4)59.3 (49.3–63.8)(s=4)55.8 (54.5–61.4)(s=3)69.0 (69.0–69.0)(s=1)61.3 (57.6–62.2)(s=3)Gender (male)49.4%(n=30 100/60 880)51.5%(n=26 203/50 904)36.2% (n=2577/7106)53.4%(n=266/498)(s=3)45.8% (n=27/59)(s=4)–44.4%(n=1027/2313)(s=3)Ethnicity Caucasian0.3%(n=9/3436)––0.3%(n=9/3436)(s=2)–––Black11.6%(n=398/3436)––11.6%(n=398/3436)(s=3)–––Asian0.6%(n=300/50 914)0.6% (n=300/50914) (s=2)–––––Other0.8% (n=29/3495)––0.6% (n=20/3436)(s=2)15.3% (n=9/59)(s=1)––Not recorded98.8%(n=63 199/63 935)99.4%(n=50 614/50 914)(s=8)100% (n=7166/7166)(s=4)87.6% (n=3009/3436)(s=2)84.7% (n=50/59)(s=3)100.0% (n=47/47)(s=1)100% (n=2313/2313)(s=3)Biochemistry Glucose, mmol/L46.8 (39.4–53.8)(s=12)50.6 (40.1–53.0)(s=3)48.8 (44.6–53.0)(s=2)75.6 (44.4–107.0)(s=2)38.0 (37.0–43.5)(s=3)54 (54–54)(s=1)43.5 (43.5–43.5)(s=1)Glucose, mg/dL842.4 (709.2–968.4)(s=12)912.8 (721.8–954)(s=3)878.4 (802.8–954.0)(s=2)1360.8 (799.2–1926.0)(s=2)684.0 (666.0–783.0)(s=3)972.0 (972.0−972.0)(s=1)783.0 (783.0–783.0)(s=1)Osmolality, mOsm/kg342 (331–353)(s=10)353 (342–355)(s=2)339 (328–349)(s=2)340 (336–340)(s=3)342 (342–342)(s=1)370 (370–370)(s=1)321 (321–321)(s=1)pH7.38 (7.35–7.39)(s=9)7.39 (7.39–7.40)(s=3)7.37 (7.35–7.38)(s=2)7.32 (7.29–7.35)(s=2)7.37 (7.37–7.37)(s=1)7.39 (7.39–7.39)(s=1)–Bicarbonate, mmol/L23.2 (20.9–23.7)(s=10)23.5 (21.3–25.1)(s=3)23.4 (23.4–23.4)(s=1)23.0 (21.9–23.4)(s=3)21.2 (21.2–21.2)(s=1)23.8 (23.8–23.8)(s=1)21.0 (21.0–21.0)(s=1)Ketones, mmol/L–––––––Lactate, mmol/L3.2 (3.2–3.2)(s=1)3.2 (3.2–3.2)(s=1)–––––Potassium, mmol/L4.6 (4.2–4.9)(s=10)4.0 (3.7–4.4)(s=2)4.6 (4.6–4.6)(s=1)5.1(5.0–5.2)(s=3)4.6 (4.5–4.7)(s=3)–3.8 (3.8–4.1)(s=1)Sodium, mmol/L140 (136–153)(s=11)150 (142–153)(s=4)144 (144–144)(s=1)134 (132–136)(s=2)140 (134–142)(s=3)–138 (138–138)(s=1)Urea, mmol/L13.4 (8.5–27.3)(s=4)7.5 (7.5–7.5)(s=1)––21.3 (16.4–26.3)(s=2)–15.3 (15.3–15.3)(s=1)Blood urea nitrogen, mmol/L37.5 (23.8–76.4)(s=4)21.0 (21.0–21.0)(s=1)––59.8 (45.9–73.6)(s=2)–42.9 (42.9–42.9)(s=1)HbA1c, mmol/mol96.7 (86.7–108.0)(s=9)100.0 (85.5–113.0)(s=4)91.3 (91.3–91.3)(s=1)95.0 (81.0–109.0)(s=2)102 (99–104)(s=2)––HbA1c, %11.0 (10.1–12.0)(s=9)11.3 (10.0–12.5)(s=4)10.5 (10.5–10.5)(s=1)10.8 (9.6–12.1)(s=2)11.5 (10.8–11.7)(s=2)– Parameters are reported as % (n) and median (IQR). Ethnicity and gender frequency were calculated using cases only where that parameter was stated, therefore do not total 100% of cases.n denotes the number of HHS cases and s denotes the number of studies.HHS, hyperosmolar hyperglycaemic state.The most commonly reported precipitants for HHS were intercurrent non-infective illness (49.5%, s=2, n=58/117), peaking at 100% in one Asian study (n=38/38), followed by infection (44.0%, s=7, n=20 471/46 420), reaching 72.2% in a North American study (n=57/79). New diagnosis of diabetes was identified as a precipitant in 31.7% of cases (s=3, n=45/142), with the highest rate reported in South America (44.0%, s=1, n=11/25). Additional precipitating factors and their regional variations are documented in table 2.Table 2HHS case precipitants, outcomes and complications5 23 29–53ParametersTotal population (s=27)Asia (s=10)Europe (s=4)North America (s=5)South America (s=4)Oceania (s=1)Africa (s=3)Total number of HHS episodes63 93550 9147166343659472313PrecipitantInfection44.0%(n=20 471/46 420)(s=7)44.0%(n=20 360/46 244)(s=3)73.3% (n=44/60)(s=1)72.2% (n=57/79) (s=1)*–58.3% (n=7/12)(s=1)Intercurrent non-infective illness49.5% (n=58/117)(s=2)100.0% (n=38/38)(s=1)–25.3% (n=20/79) (s=1)––New diagnosis of diabetes31.7% (n=45/142)(s=3)31.6% (n=12/38)(s=1)–27.8% (n=22/79) (s=1)44.0%(n=11/25)(s=1)––Suboptimal treatment compliance27.4% (n=128/467)(s=8)29.4% (n=83/282)(s=3)10.0% (n=6/60)(s=1)33.0% (n=26/79) (s=1)32.4% (n=11/34)(s=2)–*Acute coronary syndrome22.8%(n=10 559/46 244) (s=3)22.8%(n=10 559/46 244) (s=3)–––––Steroid induced6.2% (n=18/286)(s=2)6.2% (n=14/226)(s=1)*––––Trauma3.5% (n=8/226)(s=1)3.5% (n=8/226)(s=1)–––––Medication associated2.6% (n=6/226)(s=1)2.6% (n=6/266)(s=1)–––––Other21.4% (n=27/126)(s=4)––25.8% (n=23/89) (s=2)*–*OutcomesLength of stay (days)7.5 (5.2–14.2)(s=14)17.5 (12.7–19.2)(s=4)8.0 (7.3–12.0)(s=4)4.0 (3.9–4.1)(s=3)4.0 (4.0–4.0)(s=1)–6.7 (6.4–6.9)(s=2)Mortality21.1%(n=2765/13 120) (s=17)18.2% (n=585/3202)(s=7)17.4% (n=1246/7166)(s=4)4.8% (n=21/431)(s=2)*17.0% (n=8/47) (s=1)40.0% (n=904/2261) (s=2)ITU admission40.8% (n=100/245)(s=3)39.4% (n=89/226)(s=1)–70.0% (n=7/10)(s=1)*––ComplicationAcute kidney injury7.6% (n=32/421)(s=1)––7.6% (n=32/421)(s=1)–––Hypoglycaemia14.5% (n=64/440)(s=3)––14.6% (n=63/431)(s=2)*––Rhabdomyolysis3.3% (n=14/421)(s=1)––3.3% (n=14/421)(s=1)–––Acute coronary syndrome3.9% (n=18/459)(s=2)*–3.6% (n=15/421)(s=1)–––Arrhythmia2.8% (n=52/1844)(s=3)2.8% (n=50/1784)(s=2)*––––Cerebrovascular accident1.7% (n=11/647)(s=2)*–1.6% (n=7/421)(s=1)–––Deep vein thrombosis0.9% (n=451/49 544)(s=5)0.9% (n=413/46 264) (s=3)–1.2% (n=38/3280)(s=2)–––Pulmonary embolism*–*––––Pulmonary oedema4.8% (n=14/286)(s=2)4.4% (n=10/226)(s=1)*––––Sepsis*–*––––Shock*–*––––Parameters are reported as % (n) and median (IQR). Frequencies were calculated based on the number of cases reporting each parameter relative to the stated sample size. Totals may exceed 100% where multiple parameters were reported in each study.n denotes the number of patients and s denotes the number of studies.*Small data suppression has been carried out for all values where n<5.HHS, hyperosmolar hyperglycaemic state; ITU, intensive treatment unit.Biochemically, patients presented with a median glucose level of 46.8 mmol/L (IQR: 39.4–53.8, s=12), with the highest reported value in North America (75.6 mmol/L). Serum osmolality (median 342 mOsm/kg, IQR: 331–353, s=10), blood pH (median 7.38, IQR: 7.35–7.39, s=9), and bicarbonate levels (median 23.2 mmol/L, IQR: 20.9–23.7, s=10) were within criteria for inclusion in the study. The elevated median glycated haemoglobin (HbA1c) of 11.0% (IQR: 10.1–12.0, s=9) indicates a heightened risk of HHS with poorly controlled diabetes.ITU admission was frequent overall (40.8%, s=3), notably in North America (70.0%, s=1). No consistent data on specific management strategies were reported across studies.The median length of hospital stay for the meta-summary was 7.5 days (IQR: 5.2–14.2, s=14). Regional variation was marked, with Asia reporting the longest stay (median 17.5 days, s=4) and both North (s=3) and South (s=1) America reporting the shortest (median 4.0 days). Four studies were sufficiently homogeneous in terms of outcome definition and study design to permit quantitative pooling. This yielded a mean length of stay of 6.35 days (95% CI 4.47 to 8.23), with significant between-study heterogeneity (tau2=3.33).In-hospital mortality was 21.1% for the studies included in the meta-summary (s=17), but regional differences were stark, highest in Africa at 40.0% (s=2), followed by Asia (18.2%, s=7) and Europe (17.4%, s=4), and lowest in North America (4.8%, s=2). Pooled analysis of the five studies eligible for meta-analysis yielded a mortality summary estimate of 9.6% (95% CI 3.9% to 21.9%) with a prediction interval of 1.3–46.3% and moderate heterogeneity (tau2=0.45). The summary of meta-analytic estimates for mortality and length of stay can be seen in table 3.Table 3Summary of meta-analytic estimates for mortality and length of stayOutcomeAnalysisNumber of studies (number of observations)Events, n (%)Summary estimate (95% CI)Prediction intervalTau2MortalityOne admission per patient5 (3321)538 (16.2)0.0958(0.0385 to 0.2191)0.0129 to 0.46270.4464Multiple admissions per patient1 (394)74 (18.7)–––Length of stayOne admission per patient4 (943)6.3514(4.4726 to 8.2302)3.3277Multiple admissions per patient1 (394)–––Random-effects meta-analysis of mortality and length of stay data using a random-intercept logistic-normal model (GLMM) with a logit link. Between-study heterogeneity was again estimated using maximum likelihood (tau2).GLMM, generalised linear mixed model.Among the most common complications found in our study were acute kidney injury (7.6%), pulmonary oedema (4.8%), and acute coronary syndrome (3.9%). These, alongside several other rarer complications, could be antecedent conditions or precipitating factors rather than sequelae arising after the diagnosis of HHS.14 In contrast, hypoglycaemia (overall 14.5%, s=3; 14.6% in North America) and cardiac arrhythmia (overall 2.8%, predominantly reported in Asia) could be associated with HHS management.Discussion This systematic review provides the most comprehensive epidemiological portrait of HHS to our knowledge. Our analysis reveals considerable regional variation in demographic characteristics, precipitants, biochemical presentation, complications, and outcomes, highlighting significant disparities in recognition and care delivery.The median age of 63.8 years aligns with the traditional view of HHS as a condition of older adults with type 2 diabetes.1 However, the regional age disparities—most notably the younger median age in South America (55.8 years) and the markedly older median in Europe (74.5 years)—warrant exploration. This may reflect regional differences in healthcare access, diabetes duration at diagnosis, or the demographic profile of diabetes populations. Prior studies have described similar trends, with HHS more common in older adults from high-income countries and in younger, newly diagnosed individuals from low- and middle-income settings where type 2 diabetes diagnosis is often delayed.15 16A modest male minority (49.4%) was observed globally. However, the striking sex distribution differences—ranging from 36.2% males in Europe to 53.4% in North America—suggest that sociocultural, biological, or health-seeking behavioural factors are at play. These differences are underexplored in existing literature and merit targeted investigation, especially given the implications for diagnosis and care engagement.The scarcity of ethnicity data underscores a persistent blind spot in HHS research. Where reported, black individuals represented 11.6% of cases, a figure likely under-representative given the disproportionate type 2 diabetes burden in many black communities. This aligns with wider calls for improved ethnicity data collection in diabetes research and care audits.17The triad of infection, intercurrent non-infective illness, and a new diagnosis of diabetes as the most common triggers is well recognised in classical descriptions of HHS pathophysiology, where stress-induced counter-regulatory hormone surges exacerbate insulin deficiency and hyperglycaemia.9 15 However, the unusually high infection prevalence in North America (72.2%) raises questions about diagnostic thresholds and coding practices or possibly reflects higher rates of concurrent sepsis presentations in tertiary centres.Nearly one-third of HHS cases globally are precipitated by newly diagnosed diabetes—peaking at 44.0% in South America—highlighting persistent diagnostic delays. This is nearly twice the rate of new diagnoses of diabetes triggering HHS cases than previously reported in the literature.18 Globally, type 2 diabetes incidence is increasing, and the population is ageing—notably in developed nations.19 20 HHS may also represent the first presentation of diabetes, a phenomenon especially noted in immigrant or underserved populations.9 This represents missed prevention opportunities and suggests an urgent need for enhanced screening strategies in high-risk groups.The median admission glucose and osmolality are consistent with classical diagnostic criteria, confirming that, despite regional variation, the biochemical phenotype of HHS remains largely uniform.9 The overall median HbA1c of 11.0% supports chronic suboptimal glycemic control as a key antecedent. Elevated HbA1c may serve as a predisposing factor rather than a marker of acute severity and has been associated with increased mortality in HHS.21 22While our meta-analytic estimate provides a pooled global benchmark for mortality (9.6%, 95% CI 3.9% to 21.9%), it only includes studies meeting strict eligibility criteria for meta-analysis. Consequently, relying solely on these data would underestimate the overall mortality observed across all reported cases. Our meta-summary approach incorporates the broader set of studies, capturing the full range of reported outcomes and reflecting a median mortality of approximately 21%, consistent with existing literature.23 However, the stark range in mortality—from 4.8% in North America to 40.0% in Africa—is a glaring signal of global inequity. This is further underscored by the wide prediction interval (1.3–46.3%) found in our meta-analysis. Factors such as delayed presentation, limited access to insulin or fluids,24 and under-resourced intensive care infrastructure may underpin this disparity. Regional case series from sub-Saharan Africa often cite high lethality due to delayed diagnosis and lack of electrolyte correction protocols.25 These findings call for urgent investment in acute metabolic emergency management in low-resource settings.Notably, ITU admission rates were high overall (40.8%), but particularly high in North America (70.0%), and significantly lower in Asia (39.4%). These discrepancies likely reflect not only disease severity but also differing thresholds for ITU referral, availability of intermediate care, and triaging practices. The apparent paradox of high ITU admission yet shorter stays in North America (4.0 days) further underscores systemic differences in care delivery and discharge planning.26 Our pooled estimate of mean hospital stay (6.35 days, 95% CI 4.47 to 8.23) aligns broadly with the median durations reported regionally and in the meta-summary (7.5 days), suggesting that prolonged hospitalisation remains common despite consistent treatment practices. The substantial heterogeneity between studies may further reflect regional differences in discharge criteria or disease severity.The relatively high incidence of hypoglycaemia (14.5%)—a known risk in overcorrection of hyperglycemia—signals a need for caution in fluid and insulin titration protocols. Yet, without consistent data on management strategies across studies, it is difficult to assess adherence to treatment guidelines or the reasons behind iatrogenic complications. Given the management of HHS has remained relatively unchanged since the 1990s, temporal variation in HHS management was not considered to be of significant impact, given the first study included in this review is from 2006.16Most of the reported complications were either rare or, more plausibly, pre-HHS in origin. For instance, pulmonary oedema/embolism, sepsis, and acute coronary syndrome may represent either precipitants or consequences, but current data lack the granularity to distinguish timing. Notably, arrhythmias were disproportionately reported in Asian studies, potentially reflecting population-specific electrolyte abnormalities or reporting bias.The 2024 international consensus addresses the significant gaps in HHS care identified through this systematic review by standardising diagnosis, promoting data collection, and providing adaptable treatment protocols for varied resource settings.9 Complementing this, the Digital Evaluation of Ketosis and Other Diabetes-related Emergencies (DEKODE) project can play a crucial role by implementing multicentre digital quality improvement initiatives that enhance real-time monitoring, support protocol adherence, and facilitate data sharing across hospitals.27 28 DEKODE’s focus on structured care pathways and technology-enabled audit can help reduce treatment variability, improve patient outcomes, and bridge regional disparities—particularly in under-resourced areas—thus advancing the global aims of the consensus for safer, equitable HHS management.Our study has several strengths. This study provides a detailed meta-summary of both biochemical and clinical parameters with direct relevance to clinical practice. By stratifying the data by demographic and regional variables, our analysis reveals global variations in HHS characteristics that have not been previously examined. However, several limitations should be considered. Heterogeneous reporting, inconsistent definitions, and varying diagnostic thresholds restrict causal inference. Insufficient data on management protocols made it difficult to determine their impact on outcomes. Regional representation was also uneven, with fewer studies from Oceania, Africa, and South America, potentially underestimating true global variation. The small number of studies also precluded meaningful subgroup analyses or further statistical tests, so the study focused on overall summary estimates and descriptive contextualisation. This also limited the applicability of funnel plots, which are not recommended with fewer than 10 studies due to underpowered and potentially misleading asymmetry tests. The generally low-to-moderate methodological quality of included studies may affect the certainty of pooled estimates, reflecting the retrospective observational nature of most available HHS data. Publication bias was not formally assessed, as fewer than 10 studies contributed to individual meta-analyses. However, as the evidence base consisted predominantly of cohort studies rather than small interventional trials, the risk of substantial small-study publication bias is likely low.Future work focusing on evaluating outcomes from protocolised HHS care bundles in low- and middle-income countries and establishing an international registry to track HHS incidence, treatment patterns, and outcomes over time is recommended.Conclusion This systematic review reaffirms that while HHS remains biochemically uniform across regions, its clinical presentation and outcomes are deeply shaped by systemic factors. Efforts to reduce HHS mortality must prioritise improved diabetes diagnosis pathways, particularly in regions with high proportions of new-onset cases. Moreover, harmonising diagnostic and management protocols, with appropriate contextualization for resource availability, could reduce regional mortality gaps. This will ultimately serve to inform local and global guidelines on the prevention and effective management of HHS.