Beyond Disease Count: Which Chronic Disease Combinations Drive Mortality Risk in Older Adults? Evidence from a Harmonised Analysis of Six Ageing Cohorts

Beyond Disease Count: Which Chronic Disease Combinations Drive Mortality Risk in Older Adults? Evidence from a Harmonised Analysis of Six Ageing Cohorts | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Beyond Disease Count: Which Chronic Disease Combinations Drive Mortality Risk in Older Adults? Evidence from a Harmonised Analysis of Six Ageing Cohorts tao zhou, yudong xia, guohua jiang, ruijinlin hao, yi fang, xuanxuan mao, and 4 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9352419/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 7 You are reading this latest preprint version Abstract Background Multimorbidity affects the majority of older adults worldwide and is a leading driver of mortality in ageing populations. Yet frameworks for measuring multimorbidity almost universally reduce this complexity to a disease count, treating all combinations of the same size as equivalent — an assumption rarely tested across diverse countries and health systems. Methods We conducted a harmonised multicohort analysis of six longitudinal ageing studies from China, England, the United States, continental Europe, Korea, and Mexico, including 177,665 adults and 36,622 deaths. Seven chronic conditions were harmonised across cohorts: hypertension, diabetes, heart disease, stroke, lung disease, cancer, and arthritis. Pure exposures were defined as individuals with exactly the specified conditions and no others among the seven. Eligible exposures comprised 7 single conditions, 13 dyads, and 16 triads. Cohort-specific Cox models were fitted and pooled with random-effects meta-analysis. We quantified count-based gradients, within-count heterogeneity, 10-year Kaplan–Meier absolute risk differences, and cross-region reproducibility using prediction intervals, leave-one-cohort analyses, and descriptive consistency classes. Results Mortality increased stepwise with disease count (pure 1 disease: pooled hazard ratio [HR] 1.28, 95% CI 1.18–1.39; pure 2 diseases: 1.72, 1.51–1.95; pure 3 diseases: 2.29, 1.81–2.88; each versus disease-free participants), but high between-cohort heterogeneity (I² 78–95%) indicated substantial within-count variation. Among dyads, pooled HRs ranged from 1.26 (hypertension plus arthritis) to 5.10 (heart disease plus lung disease), a 4.05-fold range within the same count class. Ten-year absolute risk differences ranged from + 8.9 (hypertension plus arthritis) to + 41.8 percentage points (heart disease plus lung disease). High-risk combinations clustered along cardiometabolic structures compounded by respiratory or cerebrovascular disease. Nineteen of 36 combinations showed HR > 1 in all six cohorts; under a stricter consistency typology incorporating prediction intervals, 3 were consistently high-risk, 28 directionally consistent, and 5 unstable. Conclusions Disease count captures a mortality gradient but obscures substantial variation in which combinations drive population-level burden. Among older adults with identical disease counts, which conditions co-occur is at least as important as how many. These findings support combination-specific epidemiological profiling — rather than count-based summaries alone — to characterise multimorbidity mortality risk across populations and health systems. multimorbidity chronic disease combinations all-cause mortality population health ageing cohorts multicohort study public health surveillance cardiometabolic disease Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Background Multimorbidity—the co-occurrence of two or more chronic conditions in the same person—has become a defining challenge for ageing health systems. Older adults carrying multiple conditions face higher mortality, faster functional decline, greater healthcare use, polypharmacy, and poorer quality of life. As survival after acute illness has improved and life expectancy risen, the number of people in this situation has grown steadily across high-income and middle-income settings alike. [ 1 – 7 ] For all its importance, multimorbidity is still most often operationalised as a simple disease count. Counts capture broad gradients of risk, but they rest on an implicit assumption that rarely holds: that combinations carrying the same number of conditions carry equivalent population-level mortality risk. The prognosis of hypertension plus arthritis bears little resemblance to that of heart disease plus chronic lung disease, and a cardiometabolic triad does not carry the same mortality burden as one that also involves cerebrovascular or respiratory disease. Treating these as interchangeable flattens real, population-level variation in mortality burden — variation that matters for how we measure, monitor, and respond to multimorbidity at scale. [ 1 , 8 , 9 ] There is evidence that specific multimorbidity combinations carry distinct mortality risks, but several gaps remain. Much of the existing work has relied on latent classes or broad pattern groupings rather than direct comparisons between specific, named combinations. Many analyses also used "contains" exposure definitions—classifying someone as exposed whenever they have at least the target conditions, regardless of additional comorbidities—which obscures combination-specific estimates. And because most data come from a single cohort or country, it remains unclear how reproducible these patterns are across populations, health systems, and diagnostic environments. This gap matters: without cross-population evidence, combination-specific mortality risk cannot be used to inform health needs assessment or prioritise public health resources with confidence. [ 8 , 10 – 13 ] What is needed is a study that asks this question explicitly — not in one country but across the full diversity of ageing populations — and quantifies the answer in terms directly usable for population health planning. A multicohort design is well-suited to these questions. It enables direct comparison of cohort-specific estimates, helps separate genuine signal from setting-specific noise, and offers a more credible test of how generalisable combination-specific risk may be. We drew on harmonised data from six ageing cohorts spanning China, England, the United States, continental Europe, Korea, and Mexico to ask whether mortality risk depends not just on how many chronic conditions a person carries, but on which conditions co-occur. The study pursued four questions: how strongly all-cause mortality tracks disease count in this sample; whether combinations within the same count stratum differ meaningfully in risk; whether the most hazardous combinations share recognisable organ-system structures; and how consistently these patterns replicate across regions and health systems. Methods Study design and data sources The analysis drew on six longitudinal ageing studies: CHARLS (China), ELSA (England), HRS (United States), SHARE (continental Europe), KLoSA (Korea), and MHAS (Mexico)—all of which provide comparable data on chronic conditions, relevant covariates, and mortality outcomes in older adults across diverse national settings. [ 14 – 20 ] Disease definitions, inclusion criteria, and the statistical pipeline were harmonised across all six cohorts. For each cohort the earliest available wave with complete ascertainment of all seven target conditions was selected as the analytic baseline. Cohort-specific analyses were run first, with effect estimates subsequently pooled using random-effects meta-analysis. Participants Participants aged 45 years or older at baseline were eligible, provided they had data on chronic disease status and survival. Those with missing information on all seven target conditions or with indeterminate follow-up status were excluded. The final harmonised analytic sample comprised 177,665 participants, with 36,622 deaths observed across follow-up periods ranging from 3 years in MHAS to 12 years in HRS. Harmonised chronic conditions Seven chronic conditions were selected based on their prevalence across all six cohorts, the comparability of their variable definitions, and their relevance to mortality in older adults: hypertension (HTN), diabetes mellitus (DM), heart disease (Heart), stroke (Stroke), chronic lung disease (Lung), cancer (Cancer), and arthritis (Arthritis). All were ascertained by self-report of a physician diagnosis; cohort-specific harmonisation decisions are described in Additional file 2: Table S1 . A binary indicator was created for each condition, and total disease count was the sum of these seven indicators. Definition of pure combination exposure groups The primary analysis used strict pure combination definitions. A pure exposure group comprised participants whose total disease count equalled exactly the target number and who had exactly those target conditions—no others from the seven-condition set. A pure dyad of hypertension plus diabetes, for instance, included only people with those two conditions and none of the other five. This definition keeps combination-specific comparisons clean by excluding additional conditions that could otherwise confound the estimate. Reassuringly, sensitivity analyses using contains definitions yielded broadly consistent risk rankings (Additional file 2: Figs. S2 and S5), suggesting that while pure definitions improve interpretability, the overall conclusions are robust to exposure coding choice. Participants with none of the seven conditions served as the common reference. Eligible exposures were 7 pure single conditions, 13 pure dyads, and 16 pure triads that met minimum sample size requirements (pooled N ≥ 50 across at least three contributing cohorts; see Additional file 2: Table S2 ); the selection framework is described in the Additional file 2. One triad (heart disease, stroke, and arthritis) was available in only three of the six cohorts, yielding an unstable random-effects pooled estimate (τ² > 100); this combination was retained in descriptive prevalence summaries but treated as exploratory and excluded from the within-count heterogeneity comparison. Fifteen triads were therefore included in the Layer 2 analysis. Sensitivity analyses using "contains" definitions—broader exposure coding that does not exclude additional comorbidities—were run to check whether the more inclusive approach changed direction or magnitude materially. Outcome The primary outcome was all-cause mortality, ascertained by follow-up interview, official death records, or proxy report according to cohort design. Follow-up time was calculated from baseline interview to death, last known follow-up, or the end of the study period, whichever occurred first. Covariates All main models adjusted for age, sex, and educational attainment—covariates that were harmonisable across all six cohorts. Smoking entered the main models for the same reason; alcohol was deferred to sensitivity analyses given less consistent ascertainment. Age- and sex-only models provided a secondary sensitivity check. Restricting main-model adjustment to universally available covariates was a deliberate choice to keep pooled estimates comparable across cohorts; more extensive adjustment was explored where data permitted, in sensitivity analyses. Statistical analysis: primary analysis Within each cohort, Cox proportional hazards models were fitted for each pure exposure group against the disease-free reference. Log hazard ratios and standard errors were pooled using DerSimonian–Laird random-effects meta-analysis. [ 21 ] We report pooled HRs, 95% CIs, and 95% prediction intervals (PIs), with I² as a descriptive measure of between-cohort heterogeneity. [ 22 ] PIs were included because they convey what a comparable new cohort might expect to observe, rather than the uncertainty around a single average. [ 23 , 24 ] The 95% PI was calculated as exp(βpooled ± 1.96×√(τ² + SE²pooled)), where τ² is the DerSimonian–Laird between-study variance and SEpooled is the standard error of the pooled log-HR. The first analytic layer quantified the count gradient by comparing pure 1-disease, 2-disease, and 3-disease groups against disease-free participants. Statistical analysis: structure beyond count The second layer asked whether specific combinations carried risk beyond what their count alone would predict. For each dyad and triad, we summarised the spread of pooled HRs within count class and calculated deviation ratios—each combination's pooled HR relative to the count-class gradient HR (the pooled HR for all combinations at that disease count, as estimated in Layer 1). Sensitivity analyses also used the within-class median HR as an alternative background metric. Kaplan–Meier absolute risk differences (ARDs) at 5 and 10 years were estimated against the disease-free reference within each cohort, then pooled using random-effects meta-analysis with standard errors derived from the Greenwood formula. These ARDs are descriptive absolute-risk summaries, not fully adjusted causal estimates. Core high-risk combinations were identified across four dimensions: risk magnitude, deviation from count-class background, absolute risk burden, and cross-cohort stability. The thresholds were data-adaptive, selected to capture the upper portion of the risk distribution while retaining a sufficient number of combinations for clinical interpretation: (1) pooled HR in the top 50% of its count class; (2) deviation ratio ≥ 0.85 (combination HR divided by count-class gradient HR (the pooled HR for all combinations at that disease count, as estimated in Layer 1)); (3) 10-year ARD in the top 55% of its count class; and (4) I² <88% with directionally stable leave-one-cohort estimates. Combinations meeting at least three of the four were designated core high-risk; full scoring details are in the Additional file 2. The intent was to flag combinations that were not merely statistically elevated but clinically meaningful and reasonably stable across settings. Because the framework required meeting at least three of the four criteria, a combination whose 95% prediction interval crossed the null could still qualify if its pooled HR, absolute risk burden, and deviation ratio were sufficiently high; the wide PI in such cases reflects between-cohort variability in effect size rather than absence of risk elevation. Given the number of combinations tested, p-values for formal deviation tests were adjusted for multiple comparisons using the Benjamini–Hochberg false discovery rate (FDR) procedure (Additional file 2: Table S5 ). The core prioritisation framework was applied to dyads and triads only; single-condition groups were reported separately as reference comparators. Statistical analysis: cross-regional reproducibility Cross-regional reproducibility was assessed through directional consistency across cohorts, prediction intervals, and leave-one-cohort analyses. Combinations were classified descriptively as consistently high-risk, directionally consistent but heterogeneous in magnitude, or unstable and mixed. An exploratory stratified meta-analysis contrasting European and non-European cohorts was also run to probe whether broad regional context shifted the magnitude of associations. These stratified analyses were hypothesis-generating. The direction and magnitude of EU-to-non-EU differences were not consistent across combinations—some combinations showed larger magnitudes in European cohorts and others in non-European cohorts—precluding any simple conclusion about a regional risk gradient. Post hoc structural axes Once core high-risk combinations were identified, we examined their disease composition to see whether any structural patterns kept recurring. Organ-system axes were not defined in advance; they emerged from the data as a post hoc, descriptive summary of repeatedly appearing disease structures. Seven descriptive axes were identified post hoc: (1) cardiometabolic combinations compounded by respiratory disease (CM + Respiratory); (2) cardiometabolic combinations compounded by cerebrovascular disease (CM + Cerebrovascular); (3) cancer-containing cardiometabolic combinations (Cancer + CM); (4) pure cardiometabolic structures (Pure CM); (5) musculoskeletal-dominant combinations, typically involving arthritis and vascular or pulmonary disease (MS-dominant); (6) multi-axis combinations spanning more than two organ systems; and (7) cancer-dominated combinations where cancer is the sole or predominant condition. Axes are used for descriptive visualisation only and do not constitute predefined exposure categories or latent classes. Sensitivity analyses Prespecified sensitivity analyses examined alternative exposure definitions (contains rather than pure), alternative reference groups, exclusion of deaths occurring within the first year of follow-up and exclusion of deaths within the first two years (two cutoffs examined), reduced-adjustment models, subgroup analyses by age and sex, and leave-one-cohort robustness. Proportional hazards assumptions were evaluated using Schoenfeld residuals. Exploratory increment analyses were undertaken for selected core combinations to contextualise structural risk escalation relative to simpler combinations and component conditions; these analyses were not interpreted as mechanistic interaction tests. Software and statistical thresholds All analyses were run in Python 3.10 using pandas, numpy, scipy, and lifelines. Tests were two-sided. Given the descriptive and exploratory nature of several analyses, the emphasis was placed on effect sizes, intervals, and reproducibility rather than dichotomous significance cutoffs. This study is reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines (Additional file 1). Use of AI-assisted technology During the preparation of this study, the authors used Claude (Anthropic) to assist with reviewing and refining statistical analysis code. All analytical code and outputs were reviewed and verified by the authors, who take full responsibility for the integrity of the data analysis and the content of the manuscript. Results Cohort characteristics and prevalence structure The six cohorts contributed 177,665 participants in total, with 36,622 deaths during follow-up—an overall mortality of 20.6%. That figure varied substantially across cohorts, driven by differences in follow-up length, age distribution, and national context. Hypertension was the most prevalent condition in all cohorts; cancer and chronic lung disease were generally less common. The extent of cross-cohort variation in disease prevalence underscored why reproducibility needed to be tested rather than assumed (Fig. 1 ; Table 1 ). Table 1 Overview of the six ageing cohorts included in the analysis Cohort Region Total N Deaths Combinations (n) HR (1 vs 0 disease) Notes CHARLS (China) East Asia 17,517 – 34 – Community-dwelling adults aged ≥ 45 ELSA (England) Europe 18,492 – 36 – Community-dwelling adults aged ≥ 45 HRS (USA) North America 33,838 – 36 – Community-dwelling adults aged ≥ 45 SHARE (Europe) Europe 89,290 – 36 – Community-dwelling adults aged ≥ 45 KLoSA (Korea) East Asia 8,967 – 32 – Community-dwelling adults aged ≥ 45 MHAS (Mexico) Latin America 9,561 – 30 – Community-dwelling adults aged ≥ 45 ARD = absolute risk difference; HR = hazard ratio; CI = confidence interval. Total N = total analytic sample per cohort (including disease-free participants). Cohort-specific deaths and single-condition HRs are reported in Fig. 1 and Additional file 2: Table S1 . Dashes (–) indicate values presented in Fig. 1 to avoid misinterpretation of unweighted totals across cohorts with different follow-up durations. Table 2 Pooled prevalence and mortality risk by disease-count class and single conditions Category N exposed (pooled) Pooled HR (95% CI) I² (%) Cohorts (k) 10-yr ARD, % (95% CI) Pure 1 disease (any) 54,033 1.28 (1.18–1.39) 78% 6 – Pure 2 diseases (any) 30,684 1.72 (1.51–1.95) 90% 6 – Pure 3 diseases (any) 13,008 2.29 (1.81–2.88) 95% 6 – Lung disease (single) 2,327 1.89 (1.49–2.40) 79% 6 + 12.7 (6.0–19.4) Cancer (single) 2,887 1.83 (1.52–2.20) 65% 6 + 12.9 (7.7–18.1) Stroke (single) 977 1.80 (1.61–2.02) 0% 6 + 27.6 (22.9–32.4) Diabetes (single) 4,092 1.56 (1.27–1.90) 77% 6 + 7.8 (1.5–14.0) Heart disease (single) 4,122 1.43 (1.24–1.65) 58% 6 + 12.8 (6.1–19.4) Hypertension (single) 23,992 1.21 (1.07–1.37) 83% 6 + 5.5 (2.7–8.2) Arthritis (single) 15,636 1.09 (1.02–1.16) 29% 6 + 4.0 (0.8–7.1) All HRs are adjusted for age, sex, educational attainment, and smoking, and are relative to the disease-free group (0 of 7 conditions). ARD = 10-year Kaplan–Meier absolute risk difference versus disease-free reference, pooled across cohorts. I² = between-cohort heterogeneity statistic. HR = hazard ratio; CI = confidence interval; k = number of cohorts contributing. Figure 1 and Table 1 summarise cohort structure, follow-up, mortality, and the prevalence of the seven harmonised chronic conditions. The prevalence of core dyads and triads is shown in Table 2 . Layer 1: disease-count gradient in mortality risk A clear count gradient in all-cause mortality was observed. Relative to disease-free participants, pooled hazard ratios were 1.28 (95% CI 1.18–1.39) for pure single-condition groups, 1.72 (1.51–1.95) for pure dyads, and 2.29 (1.81–2.88) for pure triads. Ten-year absolute risk differences also increased stepwise across count classes (Fig. 2 ; Table 3 ). Table 3 Disease-count gradient and single-condition mortality risk Section Combination N exposed HR (95% CI) I² (%) 10-yr ARD, % Prediction interval Consistency type Count gradient 1 disease (any), pure 54,033 1.28 (1.18–1.39) 78% – – – Count gradient 2 diseases (any), pure 30,684 1.72 (1.51–1.95) 90% – – – Count gradient 3 diseases (any), pure 13,008 2.29 (1.81–2.88) 95% – – – Single conditions Lung disease 2,327 1.89 (1.49–2.40) 79% + 12.7 (6.0–19.4) 1.11–3.21 T2: High-risk heterogeneous Single conditions Cancer 2,887 1.83 (1.52–2.20) 65% + 12.9 (7.7–18.1) 1.27–2.64 T2: High-risk heterogeneous Single conditions Stroke 977 1.80 (1.61–2.02) 0% + 27.6 (22.9–32.4) 1.61–2.02 T1: Consistent high-risk Single conditions Diabetes 4,092 1.56 (1.27–1.90) 77% + 7.8 (1.5–14.0) 1.00–2.42 T2: High-risk heterogeneous Single conditions Heart disease 4,122 1.43 (1.24–1.65) 58% + 12.8 (6.1–19.4) 1.09–1.88 T2b: Moderate stable Single conditions Hypertension 23,992 1.21 (1.07–1.37) 83% + 5.5 (2.7–8.2) 0.92–1.60 T3: Mixed/unstable Single conditions Arthritis 15,636 1.09 (1.02–1.16) 29% + 4.0 (0.8–7.1) 0.98–1.20 T2b: Moderate stable HRs are adjusted for age, sex, educational attainment, and smoking versus the disease-free group. ARD = 10-year absolute risk difference (pooled). Prediction interval = 95% PI for the pooled HR. Consistency type: T1 = consistently high-risk in all cohorts; T2 = directionally consistent but heterogeneous; T3 = unstable or mixed direction. All dyad and triad estimates are shown in Additional file 2: Table S2 . Even so, I² values of 78–95% at the disease-count level signalled that count alone was not fully capturing the prognostic landscape—substantial variation remained within each count stratum. This count-level pattern served as the background against which combination-specific heterogeneity was evaluated. Among individual single conditions, stroke showed the most stable risk elevation (I² = 0%, Type 1 consistency), whereas hypertension—the most prevalent single condition—showed the highest between-cohort heterogeneity (I² = 83%) with a prediction interval that crossed the null (0.92–1.60, Type 3), likely reflecting cross-cohort variation in hypertension management and ascertainment practices. Layer 2: within-count heterogeneity among dyads and triads Within count strata, the spread in mortality risk was striking. Among the 13 dyads, pooled hazard ratios spanned from 1.26 (hypertension plus arthritis) to 5.10 (heart disease plus lung disease) — a more than four-fold difference in relative mortality risk within the same disease-count class. The 15 analysed triads showed a narrower but still substantial spread, from 1.62 to 3.22 (approximately 1.98-fold) (Fig. 3 ; Table 3 ; Additional file 2: Fig. S9 ). The population-level scale of this difference was substantial. Heart disease plus lung disease—the highest-risk dyad—carried a 10-year pooled ARD of + 41.8 percentage points relative to disease-free participants. Hypertension plus arthritis, the most common dyad, had an excess 10-year risk of + 8.9 percentage points. Combinations with the same disease count differed not modestly but by over 30 percentage points in absolute mortality burden. Formal within-count summaries confirmed that dyads showed the widest relative spread, and that only a minority of combinations deviated markedly upward from the count-class background (Additional file 2: Fig. S9 and Table S3 ). Knowing the number of conditions is not enough—which conditions co-occur matters too. Layer 3: core high-risk combinations and priority structure Using the prespecified four-dimensional prioritisation framework, 17 combinations were identified as core high-risk combinations: 7 dyads and 10 triads. Among dyads, the most hazardous combinations included heart disease plus lung disease (pooled HR 5.10), diabetes plus heart disease (2.50), hypertension plus lung disease (2.03), hypertension plus cancer (2.03), and hypertension plus stroke (1.97). Among triads, the highest-risk combinations included hypertension plus diabetes plus lung disease (3.22), hypertension plus diabetes plus stroke (3.10), hypertension plus heart disease plus lung disease (2.83), and hypertension plus heart disease plus stroke (2.67) (Fig. 4 ; Table 4). This figure displays pooled hazard ratios with 95% confidence intervals and prediction intervals for the core single conditions, dyads, and triads retained in the final main analysis. The figure shows seven single conditions (Panel A), the six highest-risk dyads (Panel B), and the five highest-risk triads (Panel C), selected from the 17 core combinations by pooled HR ranking. Complete estimates for all 17 core combinations are in Table 4. Table 4. Core high-risk multimorbidity combinations: dyads and triads Combination N exposed HR (95% CI) I² (%) 10-yr ARD, % Prediction interval Direction consistency Consistency type Organ axis Heart disease + Lung disease 380 5.10 (2.55–10.22) 92% +41.8 (32.5–51.0) 1.09–23.79 100% Type 2 CM + Resp Diabetes + Heart disease 559 2.50 (1.57–3.99) 81% +26.9 (17.2–36.6) 0.90–6.95 83% Type 2 Pure CM Hypertension + Cancer 1,330 2.03 (1.56–2.64) 68% +18.3 (10.1–26.5) 1.21–3.41 100% Type 2 Cancer + CM Hypertension + Lung disease 1,072 2.03 (1.40–2.93) 83% +18.5 (8.3–28.6) 0.88–4.70 100% Type 2 CM + Resp Hypertension + Stroke 1,145 1.97 (1.58–2.46) 63% +24.9 (15.9–33.8) 1.27–3.07 100% Type 2 CM + CV Stroke + Arthritis 415 1.93 (1.58–2.35) 21% +25.9 (10.6–41.2) 1.45–2.56 100% Type 1 MS-dominant Lung disease + Arthritis 1,281 1.92 (1.53–2.42) 67% +20.2 (15.0–25.3) 1.21–3.05 100% Type 2 MS-dominant HTN + Diabetes + Lung disease 264 3.22 (1.66–6.23) 79% +25.7 (17.1–34.2) 0.76–13.67 83% Type 2 CM + Resp HTN + Diabetes + Stroke 383 3.10 (2.30–4.19) 52% +44.7 (38.0–51.4) 1.78–5.41 83% Type 2 CM + CV HTN + Heart disease + Lung disease 374 2.83 (1.94–4.12) 66% +42.0 (24.9–59.1) 1.34–5.99 83% Type 2 CM + Resp HTN + Heart disease + Stroke 466 2.67 (1.71–4.16) 79% +35.8 (23.1–48.6) 1.06–6.71 83% Type 2 CM + CV HTN + Cancer + Heart disease 281 2.56 (1.70–3.86) 59% +43.1 (35.1–51.1) 1.24–5.30 80% Type 2 Cancer + CM HTN + Diabetes + Cancer 326 2.56 (2.13–3.08) 0% +30.8 (17.4–44.3) 2.12–3.08 100% Type 1 Cancer + CM HTN + Diabetes + Heart disease 1,169 2.36 (1.57–3.56) 87% +23.8 (10.7–36.8) 0.95–5.85 83% Type 2 Pure CM Diabetes + Lung disease + Arthritis 105 2.27 (1.09–4.76) 73% +29.8 (13.1–46.5) 0.58–8.89 100% Type 2 CM + Resp Heart disease + Lung disease + Arthritis 315 2.25 (1.60–3.16) 67% +31.3 (12.2–50.5) 1.21–4.19 100% Type 2 CM + Resp HTN + Stroke + Arthritis 652 1.96 (1.55–2.48) 49% +35.4 (29.3–41.5) 1.28–2.99 83% Type 2 CM + CV Abbreviations: HR, hazard ratio; CI, confidence interval; PI, prediction interval; ARD, absolute risk difference; CM, cardiometabolic; CV, cerebrovascular; Resp, respiratory. Axis labels in this table are simplified from the full seven-category post-hoc classification; see Additional file 2: Fig. S10 and Table S16 for the complete classification scheme. Core high-risk combinations meeting ≥3 of 4 prioritisation criteria. Dyads (top 7 rows) and triads (bottom 10 rows) are listed in descending order of pooled HR within each group. HTN = hypertension; CM = cardiometabolic; Resp = respiratory; CV = cerebrovascular; MS = musculoskeletal. HRs adjusted for age, sex, educational attainment, and smoking versus disease-free reference. ARD = 10-year absolute risk difference (pooled). Direction consistency = proportion of cohorts showing HR > 1. Type 1 = consistently high-risk; Type 2 = directionally consistent but heterogeneous in magnitude. The highest-risk combinations were not the most prevalent ones. Combinations carrying the greatest absolute mortality burden were often uncommon, while many of the most frequently occurring combinations were only moderately hazardous. This systematic mismatch between prevalence and mortality burden challenges frameworks that prioritise multimorbidity management by condition frequency alone — and points toward a population health approach that distinguishes common combinations from genuinely dangerous ones (Fig. 5; Table 4). Ranking combinations by frequency alone would systematically miss those carrying the highest mortality burden. Layer 4: post hoc structural axes and cross-regional reproducibility Looking at the composition of the 17 core combinations, a clear structural pattern emerged: high-risk combinations were not scattered randomly. They congregated along two main axes—cardiometabolic disease compounded by respiratory disease, and cardiometabolic disease compounded by cerebrovascular disease. Pure cardiometabolic combinations also appeared in the core set but generally carried lower average risk than these more complex extensions. Cancer-containing combinations contributed in selected instances but were less consistently recurrent (Fig. 5; Additional file 2: Figs. S10 and S11). Across the 36 analysed exposures, 19 showed directionally consistent risk elevation in all six cohorts, and 25 had prediction intervals that excluded the null. Under the descriptive consistency typology, 3 combinations were consistently high-risk, 28 directionally consistent but variable in magnitude, and 5 unstable or mixed. These two tallies reflect different standards. Nineteen combinations showed HR > 1 in every individual cohort (strict directional consistency). Using the three-type consistency typology—which additionally considers prediction intervals and leave-one-cohort stability—3 were classified as consistently high-risk, 28 as directionally consistent but heterogeneous in magnitude, and 5 as unstable or mixed; the first two categories sum to 31. Directional reproducibility was far more common than strict quantitative homogeneity (Fig. 6; Table 4). Heart disease plus lung disease illustrates the trade-off between magnitude and stability—it was the highest-risk dyad, yet also the most heterogeneous, with cohort-specific estimates spanning a wide range. Several cardiometabolic combinations sat lower on the risk scale but maintained their direction more reliably across settings. Sensitivity analyses using contains definitions, alternative references, early-death exclusions, and reduced adjustment did not materially change the overall picture (Additional file 2: Figs. S2, S3, S5, S6, S8, and S12). Discussion Principal findings Disease count and disease structure each shape mortality risk, but in different ways. Across 177,665 older adults in six countries, count was associated with a clear stepwise gradient—yet combinations at the same count level differed markedly in both relative and absolute risk. The highest-risk dyads and triads fell along recognisable structural axes rather than being scattered at random, and those axes showed directional reproducibility across regions even when effect sizes varied. Disease count is, in short, useful but insufficient. Among older adults with the same number of conditions, prognosis depends substantially on which conditions co-occur—a distinction that matters particularly given evidence that existing mortality prediction models relying on standard morbidity summaries often perform only modestly across external validation settings. [9,25] Comparison with previous studies Our findings extend the disease-count literature by demonstrating what count-based frameworks leave unanswered: whether all dyads or all triads of the same size are prognostically equivalent. They are not. Our results sit alongside, rather than against, prior work on multimorbidity patterns. Latent class and cluster-based approaches have established that multimorbidity is structured, but they produce groupings that are difficult to compare across datasets and do not yield combination-specific effect estimates. Working with strict pure combinations allowed us to compare named dyads and triads directly — a more tractable framing for both clinical risk assessment and population-level epidemiological surveillance. [10-13] There is single-cohort evidence pointing toward particularly dangerous dyads—combinations of heart failure and COPD being the most studied. Our analysis situates those observations within a broader framework: such combinations are not isolated findings but recurring structural patterns in the epidemiology of multimorbidity mortality across diverse ageing populations worldwide — a finding with implications for how international health surveillance systems characterise this burden. [8,26,27] What sets this study apart from prior pattern-based work is the direct quantification of same-count heterogeneity. Rather than inferring that structure matters, we show it explicitly: within identical disease-count strata, risk varies substantially, and that variation is structured. Interpretation and implications The gap between prevalence and danger has practical consequences. Risk stratification tools anchored to condition count or disease frequency may systematically under-prioritise less common but genuinely hazardous combinations—heart disease with chronic lung disease being the clearest example. Common combinations such as hypertension plus arthritis may drive the bulk of service demand while contributing far less to mortality. [28,29] The cardiometabolic–respiratory axis that emerged from this analysis is mechanistically plausible. Cardiovascular and chronic respiratory diseases share pathways through systemic inflammation, hypoxia, pulmonary vascular changes, reduced exercise tolerance, and exacerbation cycles. The cardiometabolic–cerebrovascular axis is grounded in shared vascular risk, impaired post-stroke recovery, frailty, and downstream complications. Our observational design cannot establish mechanism, but the concentration of high-risk combinations along biologically coherent axes—rather than scattered at random—supports the structural interpretation. [26,27,30] These findings suggest that multimorbidity profiling in older adults could be sharpened by asking not just how many conditions an individual carries, but which ones co-occur. Condition count alone misses the structure. [29,31] From a population health standpoint, these findings point toward three concrete implications. First, surveillance systems that aggregate multimorbidity burden by disease count alone may systematically underestimate the mortality concentration in specific combination subgroups. A person with heart disease plus lung disease faces more than four times the hazard of one with hypertension plus arthritis — yet count-based metrics assign them equal weight. Second, health needs assessments — which guide resource allocation across health systems — could be refined by profiling the prevalence–danger landscape documented here: the most prevalent combinations (hypertension plus arthritis, 14% of pure dyad participants) are not the most lethal, while the most lethal (heart disease plus lung disease, +41.8 percentage points in ten-year absolute risk) are comparatively rare. Prioritising by frequency will systematically underinvest in the highest-risk groups. Third, integrated care programmes for multimorbidity are expanding across health systems; the cardiometabolic–respiratory and cardiometabolic–cerebrovascular axes identified here offer empirically grounded structural targets for coordinated care design — a framework that goes beyond managing individual conditions in isolation. Heterogeneity and external validity The substantial between-cohort heterogeneity was expected—baseline mortality rates, disease ascertainment practices, healthcare access, treatment patterns, and follow-up duration all differ across six countries. High I² does not invalidate pooled estimates, but it does change how they should be read: as average signals in heterogeneous populations, not portable, fixed-point risk multipliers. Prediction intervals and leave-one-cohort analyses communicate this uncertainty more honestly than confidence intervals alone. Directional consistency was more common than quantitative homogeneity, and we regard it as the more appropriate standard. Risk elevation in the same direction across settings spanning China, Europe, and Mexico—even with wide prediction intervals—is consistent with genuine structural risk that regional context modulates in magnitude. Strengths and limitations The study's main strengths are its scale and analytical design. Six geographically diverse cohorts with harmonised analyses offer a breadth of evidence not achievable in single-country work. Strict pure-combination definitions reduce contamination from additional conditions. Reporting both hazard ratios and absolute risk differences makes findings interpretable at both the individual and population level — the former supporting prognostic reasoning, the latter directly relevant to public health burden estimation. Heterogeneity is addressed directly through prediction intervals, leave-one-cohort analyses, and descriptive consistency classes rather than being treated as a nuisance. Several limitations warrant acknowledgement. All six cohorts recruited community-dwelling participants, potentially underrepresenting institutionalised or very frail older adults in whom multimorbidity burden may be higher and combination-specific risks may differ. All seven conditions relied on self-reported physician diagnosis, leaving recall bias, under-diagnosis, and cross-country differences in healthcare access as plausible sources of misclassification. [32] In particular, “heart disease” encompasses a heterogeneous category whose scope may vary across cohorts and respondents—potentially including coronary heart disease, heart failure, valvular disease, and arrhythmias—and cross-country differences in lay interpretation of this term may contribute to between-cohort heterogeneity for heart-disease-containing combinations. Restricting the analysis to conditions available across all six cohorts understates the real complexity of multimorbidity—chronic kidney disease, dementia, and depression are conspicuously absent. Strict pure-combination definitions, while analytically cleaner, reduce sample sizes for rarer combinations; triads with fewer than four contributing cohorts should be interpreted with particular caution. The DerSimonian–Laird random-effects estimator used here is known to underestimate between-study variance when the number of studies is small (k = 6) [21], which may yield confidence intervals that are too narrow; alternative estimators such as REML or the Hartung–Knapp adjustment could be considered in future work. Residual confounding—from disease severity, medication use, and unmeasured lifestyle factors—cannot be excluded. ARDs from cohort-specific Kaplan–Meier estimates are descriptive, not causal. And the structural axes identified are post hoc patterns in this data, not predefined subtypes; they need replication in independent cohorts before they can be treated as validated risk categories. Conclusion In 177,665 older adults across six countries, disease count predicted mortality in a stepwise gradient — but it concealed far more than it revealed about which combinations actually drive population-level mortality burden. Among people with the same number of conditions, mortality risk differed substantially depending on which conditions co-occurred. The highest risks clustered in cardiometabolic combinations compounded by respiratory or cerebrovascular disease, and these structural patterns held directionally across diverse settings even when effect sizes varied. The evidence supports a shift toward combination-specific epidemiological characterisation of multimorbidity — one that identifies not just how many conditions co-occur but which structural patterns carry the greatest population-level mortality burden. Such an approach can strengthen multimorbidity surveillance, inform health needs assessment, and provide an empirical basis for prioritising prevention and integrated care for the highest-risk combination structures across diverse health systems. Abbreviations ARD Absolute risk difference CHARLS China Health and Retirement Longitudinal Study CI Confidence interval DM Diabetes mellitus ELSA English Longitudinal Study of Ageing HR Hazard ratio HRS Health and Retirement Study HTN Hypertension KLoSA Korean Longitudinal Study of Ageing KM Kaplan–Meier LOC Leave-one-cohort MHAS Mexican Health and Aging Study PH Proportional hazards PI Prediction interval SHARE Survey of Health, Ageing and Retirement in Europe Declarations Ethics approval and consent to participate This study used de-identified secondary data from six established longitudinal ageing cohorts. Each original cohort study obtained ethics approval from the relevant institutional review board or ethics committee, and all participants provided informed consent in accordance with local study procedures. The current work is a secondary analysis of harmonised cohort data and did not require new ethics approval. Ethics approvals for the parent cohort studies are described in their cohort profile publications. [14-19] Consent for publication Not applicable. No individually identifiable data are included in this manuscript. Availability of data and materials The datasets analysed in this study are derived from established longitudinal ageing studies. Data access is available through each study’s data portal: CHARLS (http://charls.pku.edu.cn), ELSA (https://www.elsa-project.ac.uk), HRS (https://hrs.isr.umich.edu), SHARE (https://share-eric.eu), KLoSA (https://survey.keis.or.kr), MHAS (http://www.mhasweb.org). Derived summary statistics and supplementary outputs are provided with this manuscript. Competing interests The authors declare that they have no competing interests. Funding This work was supported by the Jiangsu Province Frontier Technology R&D Program (BF2025613) and the Jiangsu Provincial Key Research and Development Program (BE2023818). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Authors’ contributions TZ and YX contributed equally to this work. TZ, LL, and JW conceived and designed the study. TZ and YX performed data harmonisation and statistical analysis. GJ, RH, YF, and XM contributed to data processing and quality control. XT and XZ assisted with interpretation of results. TZ drafted the manuscript. LL and JW supervised the study and critically revised the manuscript. All authors read and approved the final manuscript. Acknowledgements The authors thank the participants and research teams of CHARLS, ELSA, HRS, SHARE, KLoSA, and MHAS for making these data available for research. References Nicholson K, Liu W, Fitzpatrick D, Hardacre KA, Roberts S, Salerno J, et al. Prevalence of multimorbidity and polypharmacy among adults and older adults: a systematic review. Lancet Healthy Longev. 2024;5(4):e287–96. 10.1016/S2666-7568(24)00007-2 . Chowdhury SR, Chandra Das D, Sunna TC, Beyene J, Hossain A. Global and regional prevalence of multimorbidity in the adult population in community settings: a systematic review and meta-analysis. EClinicalMedicine. 2023;57:101860. 10.1016/j.eclinm.2023.101860 . Ferris JK, Fiedeldey LK, Kim B, Clemens F, Irvine MA, Haji Hosseini S, et al. A systematic review and meta-analysis of disease clusters in multimorbidity. Nat Commun. 2026;17(1):634. 10.1038/s41467-025-67372-6 . Han S, Chen W, Shen M, Shao R, Yang W, Wang C, et al. Multimorbidity progression and the heterogeneous impact of healthy ageing risk factors: a multicohort study. BMJ Public Health. 2025;3(2):e002474. 10.1136/bmjph-2024-002474 . The Academy of Medical Sciences. Multimorbidity: a priority for global health research. London: Academy of Medical Sciences. 2018. Available from: https://acmedsci.ac.uk/policy/policy-projects/multimorbidity . Accessed 15 Mar 2026. Barnett K, Mercer SW, Norbury M, Watt G, Wyke S, Guthrie B. Epidemiology of multimorbidity and implications for health care, research, and medical education: a cross-sectional study. Lancet. 2012;380(9836):37–43. 10.1016/S0140-6736(12)60240-2 . Makovski TT, Schmitz S, Zeegers MP, Herzig S, Besi C, Schneider S, et al. Multimorbidity and quality of life: systematic literature review and meta-analysis. Ageing Res Rev. 2019;53:100903. 10.1016/j.arr.2019.04.005 . Damiano C, Dam V, Marengoni A, Ferrara N, Rizzuto D, Grande G, et al. Multimorbidity patterns and mortality in older adults: a two-cohort pooled analysis. Aging Clin Exp Res. 2025;37(1):250. 10.1007/s40520-025-03150-0 . Nunes BP, Flores TR, Mielke GI, Thumé E, Facchini LA. Multimorbidity and mortality in older adults: a systematic review and meta-analysis. Arch Gerontol Geriatr. 2016;67:130–8. 10.1016/j.archger.2016.07.008 . Ogaz-González R, Zou Q, Maroto-Rodriguez J, Gutiérrez-Robledo LM, Escamilla-Santiago R, López-Cervantes M, et al. Mortality among older adults across multimorbidity categories and lifestyle patterns. Aging Dis. 2026. 10.14336/AD.2025.1012 . Online ahead of print. Joseph JJ, Bennett A, Ojo O, Ogunmoroti O, Melissa AE, Durkalski-Mauldin V, et al. Associations of cardiometabolic multimorbidity with all-cause and coronary heart disease mortality among Black adults in the Jackson Heart Study. JAMA Netw Open. 2022;5(10):e2237552. 10.1001/jamanetworkopen.2022.37552 . Prados-Torres A, Calderón-Larrañaga A, Hancco-Saavedra J, van den Poblador-Plou B. Multimorbidity patterns: a systematic review. J Clin Epidemiol. 2014;67(3):254–66. 10.1016/j.jclinepi.2013.09.021 . Garin N, Koyanagi A, Chatterji S, Tyrovolas S, Olaya B, Leonardi M, et al. Global multimorbidity patterns: a cross-sectional, population-based, multi-country study. J Gerontol Biol Sci Med Sci. 2016;71(2):205–14. 10.1093/gerona/glv128 . Zhao Y, Hu Y, Smith JP, Strauss J, Yang G. Cohort profile: the China Health and Retirement Longitudinal Study (CHARLS). Int J Epidemiol. 2014;43(1):61–8. 10.1093/ije/dys203 . Steptoe A, Breeze E, Banks J, Nazroo J. Cohort profile: the English Longitudinal Study of Ageing. Int J Epidemiol. 2013;42(6):1640–8. 10.1093/ije/dys168 . Sonnega A, Faul JD, Ofstedal MB, Langa KM, Phillips JWR, Weir DR. Cohort profile: the Health and Retirement Study (HRS). Int J Epidemiol. 2014;43(2):576–85. 10.1093/ije/dyu067 . Börsch-Supan A, Brandt M, Hunkler C, Kneip T, Korbmacher J, Malter F, et al. Data resource profile: the Survey of Health, Ageing and Retirement in Europe (SHARE). Int J Epidemiol. 2013;42(4):992–1001. 10.1093/ije/dyt088 . Lee J. KLoSA-Korean Longitudinal Study of Aging. Korean J Fam Med. 2020;41(1):1–2. 10.4082/kjfm.19.0160 . Wong R, Michaels-Obregón A, Palloni A. Cohort profile: the Mexican Health and Aging Study (MHAS). Int J Epidemiol. 2017;46(2):e2. 10.1093/ije/dyu263 . Lee J, Ailshire J, Crimmins E, et al. J Gerontol B Psychol Sci Soc Sci. 2021;76(Suppl 1):S5–16. 10.1093/geronb/gbab050 . Gateway to Global Aging Data: resources for cross-national comparisons of aging. DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials. 1986;7(3):177–88. 10.1016/0197-2456(86)90046-2 . Higgins JPT, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency in meta-analyses. BMJ. 2003;327(7414):557–60. 10.1136/bmj.327.7414.557 . IntHout J, Ioannidis JPA, Rovers MM, Goeman JJ. Plea for routinely presenting prediction intervals in meta-analysis. BMJ Open. 2016;6(7):e010247. 10.1136/bmjopen-2015-010247 . Riley RD, Higgins JPT, Deeks JJ. Interpretation of random effects meta-analyses. BMJ. 2011;342:d549. 10.1136/bmj.d549 . Ho L, Pugh C, Seth S, Arakelyan S, Lone NI, Lyall MJ, et al. Performance of models for predicting 1-year to 3-year mortality in older adults: a systematic review of externally validated models. Lancet Healthy Longev. 2024;5(3):e227–35. 10.1016/S2666-7568(23)00264-7 . Axson EL, Sundaram V, Bloom CI, Bottle A, Cowie MR, Quint JK. Hospitalisation and mortality in patients with comorbid COPD and heart failure: a systematic review and meta-analysis. Respir Res. 2020;21(1):54. 10.1186/s12931-020-1312-y . Chen W, Thomas J, Sadatsafavi M, FitzGerald JM. Risk of cardiovascular comorbidity in patients with chronic obstructive pulmonary disease: a systematic review and meta-analysis. Lancet Respir Med. 2015;3(8):631–9. 10.1016/S2213-2600(15)00241-6 . Nüesch E, Dieppe P, Reichenbach S, Williams S, Iff S, Jüni P. All cause and disease specific mortality in patients with knee or hip osteoarthritis: population based cohort study. BMJ. 2011;342:d1165. 10.1136/bmj.d1165 . National Institute for Health and Care Excellence. Multimorbidity: clinical assessment and management. NICE guideline [NG56]. London: NICE. 2016. Available from: https://www.nice.org.uk/guidance/ng56 . Accessed 15 Mar 2026. Di Angelantonio E, Kaptoge S, Wormser D, Willeit P, Butterworth AS, Banber N, et al. Association of cardiometabolic multimorbidity with mortality. JAMA. 2015;314(1):52–60. 10.1001/jama.2015.7008 . Dumbreck S, Flynn A, Nairn M, Wilson M, Treweek S, Mercer SW, et al. Drug-disease and drug-drug interactions: systematic examination of recommendations in 12 UK national clinical guidelines. BMJ. 2015;350:h949. 10.1136/bmj.h949 . Okura Y, Urban LH, Mahoney DW, Jacobsen SJ, Rodeheffer RJ. Agreement between self-report questionnaires and medical record data was substantial for diabetes, hypertension, myocardial infarction and stroke but not for heart failure. J Clin Epidemiol. 2004;57(10):1096–103. 10.1016/j.jclinepi.2004.04.005 . Additional Declarations No competing interests reported. Supplementary Files Additionalfile1STROBEchecklist.docx Additional file 1: STROBE checklist for cohort studies (DOCX). Additionalfile2Supplementarymaterials.docx Additional file 2: Supplementary tables (Tables S1–S18) and supplementary figures (Figs. S1–S13) (DOCX). Contains variable harmonisation details, combination selection criteria, within-count heterogeneity summaries, deviation ratios, full meta-analysis results, absolute risk differences, leave-one-cohort sensitivity, organ-system axis classification, and all supplementary figures. High-resolution versions of Figs. S1–S13 are also provided individually as Additional files 3–15. Additionalfile3FigS1allcombosprevalence01.jpg Additionalfile4FigS2containssensitivity01.jpg Additionalfile5FigS3altrefearlydeath01.jpg Additionalfile6FigS4stratified01.jpg Additionalfile7FigS5purevscontains01.jpg Additionalfile8FigS6leaveonecohort01.jpg Additionalfile9FigS7comboincrement01.jpg Additionalfile10FigS8locheatmap01.jpg Additionalfile11FigS9withincountformal01.jpg Additionalfile12FigS10organaxis01.jpg Additionalfile13FigS11diseasecooccurrence01.jpg Additionalfile14FigS12stratifiedmeta01.jpg Additionalfile15FigS13coreincrement01.jpg Cite Share Download PDF Status: Under Review Version 1 posted Reviews received at journal 24 Apr, 2026 Reviewers agreed at journal 23 Apr, 2026 Reviewers invited by journal 20 Apr, 2026 Editor invited by journal 10 Apr, 2026 Editor assigned by journal 08 Apr, 2026 Submission checks completed at journal 08 Apr, 2026 First submitted to journal 08 Apr, 2026 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9352419","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":629040231,"identity":"11218317-711a-4b27-92aa-581c5b8e308d","order_by":0,"name":"tao zhou","email":"","orcid":"","institution":"The First Affiliated Hospital of Nanjing Medical University","correspondingAuthor":false,"prefix":"","firstName":"tao","middleName":"","lastName":"zhou","suffix":""},{"id":629040235,"identity":"1c6b49b4-347a-4655-a6c4-8ffccc2c946e","order_by":1,"name":"yudong xia","email":"","orcid":"","institution":"The First Affiliated Hospital of Nanjing Medical University","correspondingAuthor":false,"prefix":"","firstName":"yudong","middleName":"","lastName":"xia","suffix":""},{"id":629040237,"identity":"c472620a-7b21-41cf-a996-e83750d00b65","order_by":2,"name":"guohua jiang","email":"","orcid":"","institution":"The First Affiliated Hospital of Nanjing Medical University","correspondingAuthor":false,"prefix":"","firstName":"guohua","middleName":"","lastName":"jiang","suffix":""},{"id":629040238,"identity":"0d6dccbc-d2d1-4c29-a6aa-201aecc02c4f","order_by":3,"name":"ruijinlin hao","email":"","orcid":"","institution":"The First Affiliated Hospital of Nanjing Medical University","correspondingAuthor":false,"prefix":"","firstName":"ruijinlin","middleName":"","lastName":"hao","suffix":""},{"id":629040239,"identity":"908d0015-0a48-46e7-ad1c-3cb142f61ba0","order_by":4,"name":"yi fang","email":"","orcid":"","institution":"The First Affiliated Hospital of Nanjing Medical University","correspondingAuthor":false,"prefix":"","firstName":"yi","middleName":"","lastName":"fang","suffix":""},{"id":629040240,"identity":"5cee56f9-040a-4f6a-a24c-2da6192f2e88","order_by":5,"name":"xuanxuan mao","email":"","orcid":"","institution":"Nanjing Normal University","correspondingAuthor":false,"prefix":"","firstName":"xuanxuan","middleName":"","lastName":"mao","suffix":""},{"id":629040241,"identity":"a8ff46a5-7a06-4cca-8195-6f76e8645c87","order_by":6,"name":"xin tan","email":"","orcid":"","institution":"Nanjing Normal University","correspondingAuthor":false,"prefix":"","firstName":"xin","middleName":"","lastName":"tan","suffix":""},{"id":629040242,"identity":"d12c8e76-01af-4046-8a01-29c86634a237","order_by":7,"name":"xiaojin zhang","email":"","orcid":"","institution":"The First Affiliated Hospital of Nanjing Medical University","correspondingAuthor":false,"prefix":"","firstName":"xiaojin","middleName":"","lastName":"zhang","suffix":""},{"id":629040244,"identity":"d2a83788-6ab9-455c-a1c7-fdf5ae80ed9f","order_by":8,"name":"li liu","email":"","orcid":"","institution":"The First Affiliated Hospital of Nanjing Medical University","correspondingAuthor":false,"prefix":"","firstName":"li","middleName":"","lastName":"liu","suffix":""},{"id":629040247,"identity":"951bb78b-829e-4a44-9001-b9e71f30796d","order_by":9,"name":"jun wu","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA5klEQVRIiWNgGAWjYHCCBAaGChs5fvYGBgMw/wBRWs6kGUv2HCBeCwMDY9vhxA03EqA8QloMbiQ8ky5sO2xscPPtgaKbbQxyfDcSGD8X4NeSJj3jXLqc5O28BOPcNgZjyRsJzNIz8GgxA2nhKbM25rudYwDSAnIhGzMPQS1szIkNN8+AtdQTqaXNOXHCDR6wlgQDQlrszzxItuYBBzLQYTnnJAxnnnnYLI1Pi2R7TuJtHnBUnjEzzimzkec7nnzwMz4tDAw8CTAWGzAqJYA0YwNeDQwM7AdgLOYHBJSOglEwCkbBCAUA0fBNPVfncu8AAAAASUVORK5CYII=","orcid":"","institution":"The First Affiliated Hospital of Nanjing Medical University","correspondingAuthor":true,"prefix":"","firstName":"jun","middleName":"","lastName":"wu","suffix":""}],"badges":[],"createdAt":"2026-04-08 06:25:19","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9352419/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9352419/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":108183323,"identity":"d8577463-6bbe-4a28-88f6-26dc5fa67ec7","added_by":"auto","created_at":"2026-04-30 09:00:41","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":197713,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eStudy overview across six ageing cohorts.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003ePanel A summarises the analytic sample and deaths in each cohort. Panel B summarises median follow-up time and cohort-level mortality. Panel C shows the prevalence of the seven harmonised chronic conditions across cohorts. Together, these panels illustrate both the scale of the study and the substantial cross-cohort variation in baseline disease structure.\u003c/p\u003e","description":"","filename":"fig1studyoverview.jpg","url":"https://assets-eu.researchsquare.com/files/rs-9352419/v1/f432f47e5419481632e5db8f.jpg"},{"id":108183321,"identity":"46f603c9-a765-4599-ae7e-96f5493388cd","added_by":"auto","created_at":"2026-04-30 09:00:41","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":113276,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eCount-based background gradient in all-cause mortality.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003ePanel A shows pooled hazard ratios for pure single-condition, pure dyad, and pure triad groups relative to disease-free participants. Panel B shows 10-year Kaplan–Meier absolute risk differences for the same count classes. This figure establishes the background count gradient against which combination-specific heterogeneity was evaluated.\u003c/p\u003e","description":"","filename":"fig2countgradient.jpg","url":"https://assets-eu.researchsquare.com/files/rs-9352419/v1/d7af6919b461ad179bb95c37.jpg"},{"id":108183303,"identity":"7b200fd2-fc75-430c-a0a7-85a23ad10565","added_by":"auto","created_at":"2026-04-30 09:00:37","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":360110,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eWithin-count heterogeneity in mortality risk.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003ePanels A and B show pooled hazard ratios for all pure dyads and triads ordered within count class; parenthetical percentages indicate combination prevalence among the covariate-complete analytic sample (denominators vary by combination owing to covariate availability). Panel C shows deviation ratios from the count-class gradient HR for each combination. Panel D displays within-count HR spread across disease-count strata as a boxplot and dot plot, illustrating that dyads have the widest relative spread. Combinations within the same disease-count stratum differ substantially in mortality risk, particularly among dyads.\u003c/p\u003e","description":"","filename":"fig3withincount.jpg","url":"https://assets-eu.researchsquare.com/files/rs-9352419/v1/4b0ccc0d07d3e1c9c4f852b6.jpg"},{"id":108183733,"identity":"217f82e4-1ec4-4603-9b9a-83aedf1c756f","added_by":"auto","created_at":"2026-04-30 09:02:35","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":713036,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eCore high-risk conditions and combinations.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis figure displays pooled hazard ratios with 95% confidence intervals and prediction intervals for the core single conditions, dyads, and triads retained in the final main analysis. The figure shows seven single conditions (Panel A), the six highest-risk dyads (Panel B), and the five highest-risk triads (Panel C), selected from the 17 core combinations by pooled HR ranking. Complete estimates for all 17 core combinations are in Table 4.\u003c/p\u003e","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-9352419/v1/af2dfa89d2fed40853a158a9.png"},{"id":108183304,"identity":"2fa9dbb7-54aa-48ac-a78b-2186ef8636a6","added_by":"auto","created_at":"2026-04-30 09:00:37","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":752843,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eClinical priority map combining prevalence and danger.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003ePanels show the relationship between pooled prevalence and 10-year absolute risk difference for core dyads and triads. The figure highlights the mismatch between the most common combinations and the most hazardous combinations, thereby supporting prioritisation beyond prevalence alone. Quadrant boundaries are drawn at the median pooled prevalence and median 10-year absolute risk difference across all displayed dyads (Panel A) and triads (Panel B), respectively.\u003c/p\u003e","description":"","filename":"floatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-9352419/v1/f5fa7ed0523bfb1d801ee26f.png"},{"id":108184924,"identity":"947745e6-7d39-43fc-b267-ee07cc1a2c88","added_by":"auto","created_at":"2026-04-30 09:05:04","extension":"jpg","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":232262,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eCross-regional reproducibility and heterogeneity.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003ePanel A shows selected core combinations (three representative dyads and three representative triads) with pooled hazard ratios, 95% confidence intervals, and 95% prediction intervals. Panel B classifies combinations into consistent high-risk, high-risk but heterogeneous, and unstable or mixed categories. Panel C presents exploratory stratified results by broad regional grouping. The figure emphasises that directional reproducibility is more common than full homogeneity of effect size.\u003c/p\u003e","description":"","filename":"fig6crossregion.jpg","url":"https://assets-eu.researchsquare.com/files/rs-9352419/v1/6f94d7c2fffcfb28fb6dc598.jpg"},{"id":108491863,"identity":"1fb21cab-094a-4fc4-90e5-b7ba3f2558bb","added_by":"auto","created_at":"2026-05-05 09:56:00","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2891491,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9352419/v1/9fba62ee-614c-4465-8fca-c5d013270870.pdf"},{"id":108183487,"identity":"84dd6b7a-6bc2-4e3a-9697-e92a45bf2da9","added_by":"auto","created_at":"2026-04-30 09:01:38","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":38222,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eAdditional file 1:\u003c/strong\u003e STROBE checklist for cohort studies (DOCX).\u003c/p\u003e","description":"","filename":"Additionalfile1STROBEchecklist.docx","url":"https://assets-eu.researchsquare.com/files/rs-9352419/v1/3cf13e560888797f56ed2a10.docx"},{"id":108183305,"identity":"508394f5-8fdf-4de7-a151-59eca6922425","added_by":"auto","created_at":"2026-04-30 09:00:37","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":237200,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eAdditional file 2: \u003c/strong\u003eSupplementary tables (Tables S1–S18) and supplementary figures (Figs. S1–S13) (DOCX). Contains variable harmonisation details, combination selection criteria, within-count heterogeneity summaries, deviation ratios, full meta-analysis results, absolute risk differences, leave-one-cohort sensitivity, organ-system axis classification, and all supplementary figures. High-resolution versions of Figs. S1–S13 are also provided individually as Additional files 3–15.\u003c/p\u003e","description":"","filename":"Additionalfile2Supplementarymaterials.docx","url":"https://assets-eu.researchsquare.com/files/rs-9352419/v1/680947816f1bb8eb53f98cd2.docx"},{"id":108183764,"identity":"7af2fc22-59c6-4d28-9db9-417e5534147e","added_by":"auto","created_at":"2026-04-30 09:02:42","extension":"jpg","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":561734,"visible":true,"origin":"","legend":"","description":"","filename":"Additionalfile3FigS1allcombosprevalence01.jpg","url":"https://assets-eu.researchsquare.com/files/rs-9352419/v1/496e95bf6577f5200cd65a59.jpg"},{"id":108183774,"identity":"f2dd0b98-5b0d-4969-95a8-b06ec1f7300f","added_by":"auto","created_at":"2026-04-30 09:02:43","extension":"jpg","order_by":4,"title":"","display":"","copyAsset":false,"role":"supplement","size":529490,"visible":true,"origin":"","legend":"","description":"","filename":"Additionalfile4FigS2containssensitivity01.jpg","url":"https://assets-eu.researchsquare.com/files/rs-9352419/v1/3236b20c1529ddfdefa4b13c.jpg"},{"id":108183320,"identity":"ff0fd13f-2e9a-4528-b98f-66ae17c8bc25","added_by":"auto","created_at":"2026-04-30 09:00:41","extension":"jpg","order_by":5,"title":"","display":"","copyAsset":false,"role":"supplement","size":737230,"visible":true,"origin":"","legend":"","description":"","filename":"Additionalfile5FigS3altrefearlydeath01.jpg","url":"https://assets-eu.researchsquare.com/files/rs-9352419/v1/43e008f9c627dcc7cfe04d25.jpg"},{"id":108183306,"identity":"3a992b45-3ce3-4f8d-a492-384325996ed0","added_by":"auto","created_at":"2026-04-30 09:00:37","extension":"jpg","order_by":6,"title":"","display":"","copyAsset":false,"role":"supplement","size":220137,"visible":true,"origin":"","legend":"","description":"","filename":"Additionalfile6FigS4stratified01.jpg","url":"https://assets-eu.researchsquare.com/files/rs-9352419/v1/bcd3d0346849849f15e926a2.jpg"},{"id":108183847,"identity":"de8e8ddd-f043-48ec-8890-a7803abb0570","added_by":"auto","created_at":"2026-04-30 09:02:57","extension":"jpg","order_by":7,"title":"","display":"","copyAsset":false,"role":"supplement","size":176098,"visible":true,"origin":"","legend":"","description":"","filename":"Additionalfile7FigS5purevscontains01.jpg","url":"https://assets-eu.researchsquare.com/files/rs-9352419/v1/b9d2cbb6ea34fab3ed0cee8c.jpg"},{"id":108183317,"identity":"bb54ab09-1e95-4be1-b8c8-605539418c98","added_by":"auto","created_at":"2026-04-30 09:00:40","extension":"jpg","order_by":8,"title":"","display":"","copyAsset":false,"role":"supplement","size":402536,"visible":true,"origin":"","legend":"","description":"","filename":"Additionalfile8FigS6leaveonecohort01.jpg","url":"https://assets-eu.researchsquare.com/files/rs-9352419/v1/52447b96a506953a3ceec886.jpg"},{"id":108183316,"identity":"8e035d27-9eb6-4c42-9e89-84f55d250a26","added_by":"auto","created_at":"2026-04-30 09:00:40","extension":"jpg","order_by":9,"title":"","display":"","copyAsset":false,"role":"supplement","size":530301,"visible":true,"origin":"","legend":"","description":"","filename":"Additionalfile9FigS7comboincrement01.jpg","url":"https://assets-eu.researchsquare.com/files/rs-9352419/v1/653851ff263c51b14c6a66f6.jpg"},{"id":108183338,"identity":"c9a5085f-705a-4c3f-81c8-28f0e5f50b81","added_by":"auto","created_at":"2026-04-30 09:00:48","extension":"jpg","order_by":10,"title":"","display":"","copyAsset":false,"role":"supplement","size":608949,"visible":true,"origin":"","legend":"","description":"","filename":"Additionalfile10FigS8locheatmap01.jpg","url":"https://assets-eu.researchsquare.com/files/rs-9352419/v1/61a8181c8d7489095a27330c.jpg"},{"id":108183486,"identity":"18bda680-5576-49c7-ae16-8031d5dc0950","added_by":"auto","created_at":"2026-04-30 09:01:36","extension":"jpg","order_by":11,"title":"","display":"","copyAsset":false,"role":"supplement","size":538234,"visible":true,"origin":"","legend":"","description":"","filename":"Additionalfile11FigS9withincountformal01.jpg","url":"https://assets-eu.researchsquare.com/files/rs-9352419/v1/c29d78b424221a06177c47b4.jpg"},{"id":108183772,"identity":"05fa91d4-b5ca-4826-bec2-846e31f232ea","added_by":"auto","created_at":"2026-04-30 09:02:43","extension":"jpg","order_by":12,"title":"","display":"","copyAsset":false,"role":"supplement","size":348890,"visible":true,"origin":"","legend":"","description":"","filename":"Additionalfile12FigS10organaxis01.jpg","url":"https://assets-eu.researchsquare.com/files/rs-9352419/v1/e2d3c286f889d3369d574a9a.jpg"},{"id":108183503,"identity":"1ee90be3-ec85-4629-a991-aa73fe931037","added_by":"auto","created_at":"2026-04-30 09:01:50","extension":"jpg","order_by":13,"title":"","display":"","copyAsset":false,"role":"supplement","size":452531,"visible":true,"origin":"","legend":"","description":"","filename":"Additionalfile13FigS11diseasecooccurrence01.jpg","url":"https://assets-eu.researchsquare.com/files/rs-9352419/v1/7630c16835101d3c768ca909.jpg"},{"id":108183786,"identity":"bbfb040f-02d9-4727-9c12-49e276984d65","added_by":"auto","created_at":"2026-04-30 09:02:47","extension":"jpg","order_by":14,"title":"","display":"","copyAsset":false,"role":"supplement","size":578500,"visible":true,"origin":"","legend":"","description":"","filename":"Additionalfile14FigS12stratifiedmeta01.jpg","url":"https://assets-eu.researchsquare.com/files/rs-9352419/v1/dbfede71ae5dc3f4151a1b5b.jpg"},{"id":108183326,"identity":"8100e6dd-fbc2-4ace-8a95-5008909eb9fe","added_by":"auto","created_at":"2026-04-30 09:00:42","extension":"jpg","order_by":15,"title":"","display":"","copyAsset":false,"role":"supplement","size":560412,"visible":true,"origin":"","legend":"","description":"","filename":"Additionalfile15FigS13coreincrement01.jpg","url":"https://assets-eu.researchsquare.com/files/rs-9352419/v1/360dadaf324ed7dcf1dc6bba.jpg"}],"financialInterests":"No competing interests reported.","formattedTitle":"Beyond Disease Count: Which Chronic Disease Combinations Drive Mortality Risk in Older Adults? Evidence from a Harmonised Analysis of Six Ageing Cohorts","fulltext":[{"header":"Background","content":"\u003cp\u003eMultimorbidity\u0026mdash;the co-occurrence of two or more chronic conditions in the same person\u0026mdash;has become a defining challenge for ageing health systems. Older adults carrying multiple conditions face higher mortality, faster functional decline, greater healthcare use, polypharmacy, and poorer quality of life. As survival after acute illness has improved and life expectancy risen, the number of people in this situation has grown steadily across high-income and middle-income settings alike. [\u003cspan additionalcitationids=\"CR2 CR3 CR4 CR5 CR6\" citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]\u003c/p\u003e \u003cp\u003eFor all its importance, multimorbidity is still most often operationalised as a simple disease count. Counts capture broad gradients of risk, but they rest on an implicit assumption that rarely holds: that combinations carrying the same number of conditions carry equivalent population-level mortality risk. The prognosis of hypertension plus arthritis bears little resemblance to that of heart disease plus chronic lung disease, and a cardiometabolic triad does not carry the same mortality burden as one that also involves cerebrovascular or respiratory disease. Treating these as interchangeable flattens real, population-level variation in mortality burden \u0026mdash; variation that matters for how we measure, monitor, and respond to multimorbidity at scale. [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]\u003c/p\u003e \u003cp\u003eThere is evidence that specific multimorbidity combinations carry distinct mortality risks, but several gaps remain. Much of the existing work has relied on latent classes or broad pattern groupings rather than direct comparisons between specific, named combinations. Many analyses also used \"contains\" exposure definitions\u0026mdash;classifying someone as exposed whenever they have at least the target conditions, regardless of additional comorbidities\u0026mdash;which obscures combination-specific estimates. And because most data come from a single cohort or country, it remains unclear how reproducible these patterns are across populations, health systems, and diagnostic environments. This gap matters: without cross-population evidence, combination-specific mortality risk cannot be used to inform health needs assessment or prioritise public health resources with confidence. [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e, \u003cspan additionalcitationids=\"CR11 CR12\" citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]\u003c/p\u003e \u003cp\u003eWhat is needed is a study that asks this question explicitly \u0026mdash; not in one country but across the full diversity of ageing populations \u0026mdash; and quantifies the answer in terms directly usable for population health planning. A multicohort design is well-suited to these questions. It enables direct comparison of cohort-specific estimates, helps separate genuine signal from setting-specific noise, and offers a more credible test of how generalisable combination-specific risk may be. We drew on harmonised data from six ageing cohorts spanning China, England, the United States, continental Europe, Korea, and Mexico to ask whether mortality risk depends not just on how many chronic conditions a person carries, but on which conditions co-occur.\u003c/p\u003e \u003cp\u003eThe study pursued four questions: how strongly all-cause mortality tracks disease count in this sample; whether combinations within the same count stratum differ meaningfully in risk; whether the most hazardous combinations share recognisable organ-system structures; and how consistently these patterns replicate across regions and health systems.\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eStudy design and data sources\u003c/h2\u003e \u003cp\u003eThe analysis drew on six longitudinal ageing studies: CHARLS (China), ELSA (England), HRS (United States), SHARE (continental Europe), KLoSA (Korea), and MHAS (Mexico)\u0026mdash;all of which provide comparable data on chronic conditions, relevant covariates, and mortality outcomes in older adults across diverse national settings. [\u003cspan additionalcitationids=\"CR15 CR16 CR17 CR18 CR19\" citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e]\u003c/p\u003e \u003cp\u003eDisease definitions, inclusion criteria, and the statistical pipeline were harmonised across all six cohorts. For each cohort the earliest available wave with complete ascertainment of all seven target conditions was selected as the analytic baseline. Cohort-specific analyses were run first, with effect estimates subsequently pooled using random-effects meta-analysis.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eParticipants\u003c/h3\u003e\n\u003cp\u003eParticipants aged 45 years or older at baseline were eligible, provided they had data on chronic disease status and survival. Those with missing information on all seven target conditions or with indeterminate follow-up status were excluded.\u003c/p\u003e \u003cp\u003eThe final harmonised analytic sample comprised 177,665 participants, with 36,622 deaths observed across follow-up periods ranging from 3 years in MHAS to 12 years in HRS.\u003c/p\u003e\n\u003ch3\u003eHarmonised chronic conditions\u003c/h3\u003e\n\u003cp\u003eSeven chronic conditions were selected based on their prevalence across all six cohorts, the comparability of their variable definitions, and their relevance to mortality in older adults: hypertension (HTN), diabetes mellitus (DM), heart disease (Heart), stroke (Stroke), chronic lung disease (Lung), cancer (Cancer), and arthritis (Arthritis). All were ascertained by self-report of a physician diagnosis; cohort-specific harmonisation decisions are described in Additional file 2: Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e.\u003c/p\u003e \u003cp\u003eA binary indicator was created for each condition, and total disease count was the sum of these seven indicators.\u003c/p\u003e\n\u003ch3\u003eDefinition of pure combination exposure groups\u003c/h3\u003e\n\u003cp\u003eThe primary analysis used strict pure combination definitions. A pure exposure group comprised participants whose total disease count equalled exactly the target number and who had exactly those target conditions\u0026mdash;no others from the seven-condition set. A pure dyad of hypertension plus diabetes, for instance, included only people with those two conditions and none of the other five.\u003c/p\u003e \u003cp\u003eThis definition keeps combination-specific comparisons clean by excluding additional conditions that could otherwise confound the estimate. Reassuringly, sensitivity analyses using contains definitions yielded broadly consistent risk rankings (Additional file 2: Figs. S2 and S5), suggesting that while pure definitions improve interpretability, the overall conclusions are robust to exposure coding choice. Participants with none of the seven conditions served as the common reference. Eligible exposures were 7 pure single conditions, 13 pure dyads, and 16 pure triads that met minimum sample size requirements (pooled N\u0026thinsp;\u0026ge;\u0026thinsp;50 across at least three contributing cohorts; see Additional file 2: Table \u003cspan refid=\"MOESM2\" class=\"InternalRef\"\u003eS2\u003c/span\u003e); the selection framework is described in the Additional file 2. One triad (heart disease, stroke, and arthritis) was available in only three of the six cohorts, yielding an unstable random-effects pooled estimate (τ\u0026sup2; \u0026gt; 100); this combination was retained in descriptive prevalence summaries but treated as exploratory and excluded from the within-count heterogeneity comparison. Fifteen triads were therefore included in the Layer 2 analysis.\u003c/p\u003e \u003cp\u003eSensitivity analyses using \"contains\" definitions\u0026mdash;broader exposure coding that does not exclude additional comorbidities\u0026mdash;were run to check whether the more inclusive approach changed direction or magnitude materially.\u003c/p\u003e\n\u003ch3\u003eOutcome\u003c/h3\u003e\n \u003cp\u003eThe primary outcome was all-cause mortality, ascertained by follow-up interview, official death records, or proxy report according to cohort design. Follow-up time was calculated from baseline interview to death, last known follow-up, or the end of the study period, whichever occurred first.\u003c/p\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003eCovariates\u003c/h2\u003e \u003cp\u003eAll main models adjusted for age, sex, and educational attainment\u0026mdash;covariates that were harmonisable across all six cohorts. Smoking entered the main models for the same reason; alcohol was deferred to sensitivity analyses given less consistent ascertainment. Age- and sex-only models provided a secondary sensitivity check.\u003c/p\u003e \u003cp\u003eRestricting main-model adjustment to universally available covariates was a deliberate choice to keep pooled estimates comparable across cohorts; more extensive adjustment was explored where data permitted, in sensitivity analyses.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eStatistical analysis: primary analysis\u003c/h3\u003e\n\u003cp\u003eWithin each cohort, Cox proportional hazards models were fitted for each pure exposure group against the disease-free reference. Log hazard ratios and standard errors were pooled using DerSimonian\u0026ndash;Laird random-effects meta-analysis. [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e] We report pooled HRs, 95% CIs, and 95% prediction intervals (PIs), with I\u0026sup2; as a descriptive measure of between-cohort heterogeneity. [\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e] PIs were included because they convey what a comparable new cohort might expect to observe, rather than the uncertainty around a single average. [\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e, \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e] The 95% PI was calculated as exp(βpooled\u0026thinsp;\u0026plusmn;\u0026thinsp;1.96\u0026times;\u0026radic;(τ\u0026sup2; + SE\u0026sup2;pooled)), where τ\u0026sup2; is the DerSimonian\u0026ndash;Laird between-study variance and SEpooled is the standard error of the pooled log-HR.\u003c/p\u003e \u003cp\u003eThe first analytic layer quantified the count gradient by comparing pure 1-disease, 2-disease, and 3-disease groups against disease-free participants.\u003c/p\u003e\n\u003ch3\u003eStatistical analysis: structure beyond count\u003c/h3\u003e\n\u003cp\u003eThe second layer asked whether specific combinations carried risk beyond what their count alone would predict. For each dyad and triad, we summarised the spread of pooled HRs within count class and calculated deviation ratios\u0026mdash;each combination's pooled HR relative to the count-class gradient HR (the pooled HR for all combinations at that disease count, as estimated in Layer 1). Sensitivity analyses also used the within-class median HR as an alternative background metric.\u003c/p\u003e \u003cp\u003eKaplan\u0026ndash;Meier absolute risk differences (ARDs) at 5 and 10 years were estimated against the disease-free reference within each cohort, then pooled using random-effects meta-analysis with standard errors derived from the Greenwood formula. These ARDs are descriptive absolute-risk summaries, not fully adjusted causal estimates.\u003c/p\u003e \u003cp\u003eCore high-risk combinations were identified across four dimensions: risk magnitude, deviation from count-class background, absolute risk burden, and cross-cohort stability. The thresholds were data-adaptive, selected to capture the upper portion of the risk distribution while retaining a sufficient number of combinations for clinical interpretation: (1) pooled HR in the top 50% of its count class; (2) deviation ratio\u0026thinsp;\u0026ge;\u0026thinsp;0.85 (combination HR divided by count-class gradient HR (the pooled HR for all combinations at that disease count, as estimated in Layer 1)); (3) 10-year ARD in the top 55% of its count class; and (4) I\u0026sup2; \u0026lt;88% with directionally stable leave-one-cohort estimates. Combinations meeting at least three of the four were designated core high-risk; full scoring details are in the Additional file 2. The intent was to flag combinations that were not merely statistically elevated but clinically meaningful and reasonably stable across settings. Because the framework required meeting at least three of the four criteria, a combination whose 95% prediction interval crossed the null could still qualify if its pooled HR, absolute risk burden, and deviation ratio were sufficiently high; the wide PI in such cases reflects between-cohort variability in effect size rather than absence of risk elevation. Given the number of combinations tested, p-values for formal deviation tests were adjusted for multiple comparisons using the Benjamini\u0026ndash;Hochberg false discovery rate (FDR) procedure (Additional file 2: Table \u003cspan refid=\"MOESM5\" class=\"InternalRef\"\u003eS5\u003c/span\u003e). The core prioritisation framework was applied to dyads and triads only; single-condition groups were reported separately as reference comparators.\u003c/p\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003eStatistical analysis: cross-regional reproducibility\u003c/h2\u003e \u003cp\u003eCross-regional reproducibility was assessed through directional consistency across cohorts, prediction intervals, and leave-one-cohort analyses. Combinations were classified descriptively as consistently high-risk, directionally consistent but heterogeneous in magnitude, or unstable and mixed.\u003c/p\u003e \u003cp\u003eAn exploratory stratified meta-analysis contrasting European and non-European cohorts was also run to probe whether broad regional context shifted the magnitude of associations. These stratified analyses were hypothesis-generating. The direction and magnitude of EU-to-non-EU differences were not consistent across combinations\u0026mdash;some combinations showed larger magnitudes in European cohorts and others in non-European cohorts\u0026mdash;precluding any simple conclusion about a regional risk gradient.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003ePost hoc structural axes\u003c/h2\u003e \u003cp\u003eOnce core high-risk combinations were identified, we examined their disease composition to see whether any structural patterns kept recurring. Organ-system axes were not defined in advance; they emerged from the data as a post hoc, descriptive summary of repeatedly appearing disease structures.\u003c/p\u003e \u003cp\u003eSeven descriptive axes were identified post hoc: (1) cardiometabolic combinations compounded by respiratory disease (CM\u0026thinsp;+\u0026thinsp;Respiratory); (2) cardiometabolic combinations compounded by cerebrovascular disease (CM\u0026thinsp;+\u0026thinsp;Cerebrovascular); (3) cancer-containing cardiometabolic combinations (Cancer\u0026thinsp;+\u0026thinsp;CM); (4) pure cardiometabolic structures (Pure CM); (5) musculoskeletal-dominant combinations, typically involving arthritis and vascular or pulmonary disease (MS-dominant); (6) multi-axis combinations spanning more than two organ systems; and (7) cancer-dominated combinations where cancer is the sole or predominant condition. Axes are used for descriptive visualisation only and do not constitute predefined exposure categories or latent classes.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003eSensitivity analyses\u003c/h2\u003e \u003cp\u003ePrespecified sensitivity analyses examined alternative exposure definitions (contains rather than pure), alternative reference groups, exclusion of deaths occurring within the first year of follow-up and exclusion of deaths within the first two years (two cutoffs examined), reduced-adjustment models, subgroup analyses by age and sex, and leave-one-cohort robustness. Proportional hazards assumptions were evaluated using Schoenfeld residuals. Exploratory increment analyses were undertaken for selected core combinations to contextualise structural risk escalation relative to simpler combinations and component conditions; these analyses were not interpreted as mechanistic interaction tests.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003eSoftware and statistical thresholds\u003c/h2\u003e \u003cp\u003eAll analyses were run in Python 3.10 using pandas, numpy, scipy, and lifelines. Tests were two-sided. Given the descriptive and exploratory nature of several analyses, the emphasis was placed on effect sizes, intervals, and reproducibility rather than dichotomous significance cutoffs.\u003c/p\u003e \u003cp\u003e This study is reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines (Additional file 1).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003eUse of AI-assisted technology\u003c/h2\u003e \u003cp\u003eDuring the preparation of this study, the authors used Claude (Anthropic) to assist with reviewing and refining statistical analysis code. All analytical code and outputs were reviewed and verified by the authors, who take full responsibility for the integrity of the data analysis and the content of the manuscript.\u003c/p\u003e \u003c/div\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec17\" class=\"Section2\"\u003e\n \u003ch2\u003eCohort characteristics and prevalence structure\u003c/h2\u003e\n \u003cp\u003eThe six cohorts contributed 177,665 participants in total, with 36,622 deaths during follow-up\u0026mdash;an overall mortality of 20.6%. That figure varied substantially across cohorts, driven by differences in follow-up length, age distribution, and national context. Hypertension was the most prevalent condition in all cohorts; cancer and chronic lung disease were generally less common. The extent of cross-cohort variation in disease prevalence underscored why reproducibility needed to be tested rather than assumed (Fig. \u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e; Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e\n \u003cdiv class=\"gridtable\"\u003e\u0026nbsp;\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eOverview of the six ageing cohorts included in the analysis\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003ccolgroup cols=\"7\"\u003e\u003c/colgroup\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eCohort\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003eRegion\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colname=\"c3\"\u003e\n \u003cp\u003eTotal N\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003eDeaths\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colname=\"c5\"\u003e\n \u003cp\u003eCombinations (n)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colname=\"c6\"\u003e\n \u003cp\u003eHR (1 vs 0 disease)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colname=\"c7\"\u003e\n \u003cp\u003eNotes\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eCHARLS (China)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003eEast Asia\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\n \u003cp\u003e17,517\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003e\u0026ndash;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\n \u003cp\u003e34\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c6\"\u003e\n \u003cp\u003e\u0026ndash;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c7\"\u003e\n \u003cp\u003eCommunity-dwelling adults aged\u0026thinsp;\u0026ge;\u0026thinsp;45\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eELSA (England)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003eEurope\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\n \u003cp\u003e18,492\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003e\u0026ndash;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\n \u003cp\u003e36\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c6\"\u003e\n \u003cp\u003e\u0026ndash;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c7\"\u003e\n \u003cp\u003eCommunity-dwelling adults aged\u0026thinsp;\u0026ge;\u0026thinsp;45\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eHRS (USA)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003eNorth America\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\n \u003cp\u003e33,838\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003e\u0026ndash;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\n \u003cp\u003e36\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c6\"\u003e\n \u003cp\u003e\u0026ndash;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c7\"\u003e\n \u003cp\u003eCommunity-dwelling adults aged\u0026thinsp;\u0026ge;\u0026thinsp;45\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eSHARE (Europe)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003eEurope\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\n \u003cp\u003e89,290\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003e\u0026ndash;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\n \u003cp\u003e36\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c6\"\u003e\n \u003cp\u003e\u0026ndash;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c7\"\u003e\n \u003cp\u003eCommunity-dwelling adults aged\u0026thinsp;\u0026ge;\u0026thinsp;45\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eKLoSA (Korea)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003eEast Asia\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\n \u003cp\u003e8,967\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003e\u0026ndash;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\n \u003cp\u003e32\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c6\"\u003e\n \u003cp\u003e\u0026ndash;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c7\"\u003e\n \u003cp\u003eCommunity-dwelling adults aged\u0026thinsp;\u0026ge;\u0026thinsp;45\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eMHAS (Mexico)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003eLatin America\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\n \u003cp\u003e9,561\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003e\u0026ndash;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\n \u003cp\u003e30\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c6\"\u003e\n \u003cp\u003e\u0026ndash;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c7\"\u003e\n \u003cp\u003eCommunity-dwelling adults aged\u0026thinsp;\u0026ge;\u0026thinsp;45\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n \u003cp\u003e\u003cem\u003eARD\u0026thinsp;=\u0026thinsp;absolute risk difference; HR\u0026thinsp;=\u0026thinsp;hazard ratio; CI\u0026thinsp;=\u0026thinsp;confidence interval. Total N\u0026thinsp;=\u0026thinsp;total analytic sample per cohort (including disease-free participants). Cohort-specific deaths and single-condition HRs are reported in\u003c/em\u003e Fig. \u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e \u003cem\u003eand Additional file 2: Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e. Dashes (\u0026ndash;) indicate values presented in\u003c/em\u003e Fig. \u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e \u003cem\u003eto avoid misinterpretation of unweighted totals across cohorts with different follow-up durations.\u003c/em\u003e\u003c/p\u003e\n \u003cdiv class=\"gridtable\"\u003e\u0026nbsp;\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003ePooled prevalence and mortality risk by disease-count class and single conditions\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003ccolgroup cols=\"6\"\u003e\u003c/colgroup\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eCategory\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003eN exposed (pooled)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colname=\"c3\"\u003e\n \u003cp\u003ePooled HR (95% CI)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003eI\u0026sup2; (%)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colname=\"c5\"\u003e\n \u003cp\u003eCohorts (k)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colname=\"c6\"\u003e\n \u003cp\u003e10-yr ARD, % (95% CI)\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003ePure 1 disease (any)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\n \u003cp\u003e54,033\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\n \u003cp\u003e1.28 (1.18\u0026ndash;1.39)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003e78%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\n \u003cp\u003e6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c6\"\u003e\n \u003cp\u003e\u0026ndash;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003ePure 2 diseases (any)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\n \u003cp\u003e30,684\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\n \u003cp\u003e1.72 (1.51\u0026ndash;1.95)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003e90%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\n \u003cp\u003e6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c6\"\u003e\n \u003cp\u003e\u0026ndash;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003ePure 3 diseases (any)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\n \u003cp\u003e13,008\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\n \u003cp\u003e2.29 (1.81\u0026ndash;2.88)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003e95%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\n \u003cp\u003e6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c6\"\u003e\n \u003cp\u003e\u0026ndash;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eLung disease (single)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\n \u003cp\u003e2,327\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\n \u003cp\u003e1.89 (1.49\u0026ndash;2.40)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003e79%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\n \u003cp\u003e6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c6\"\u003e\n \u003cp\u003e+\u0026thinsp;12.7 (6.0\u0026ndash;19.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eCancer (single)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\n \u003cp\u003e2,887\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\n \u003cp\u003e1.83 (1.52\u0026ndash;2.20)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003e65%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\n \u003cp\u003e6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c6\"\u003e\n \u003cp\u003e+\u0026thinsp;12.9 (7.7\u0026ndash;18.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eStroke (single)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\n \u003cp\u003e977\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\n \u003cp\u003e1.80 (1.61\u0026ndash;2.02)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003e0%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\n \u003cp\u003e6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c6\"\u003e\n \u003cp\u003e+\u0026thinsp;27.6 (22.9\u0026ndash;32.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eDiabetes (single)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\n \u003cp\u003e4,092\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\n \u003cp\u003e1.56 (1.27\u0026ndash;1.90)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003e77%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\n \u003cp\u003e6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c6\"\u003e\n \u003cp\u003e+\u0026thinsp;7.8 (1.5\u0026ndash;14.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eHeart disease (single)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\n \u003cp\u003e4,122\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\n \u003cp\u003e1.43 (1.24\u0026ndash;1.65)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003e58%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\n \u003cp\u003e6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c6\"\u003e\n \u003cp\u003e+\u0026thinsp;12.8 (6.1\u0026ndash;19.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eHypertension (single)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\n \u003cp\u003e23,992\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\n \u003cp\u003e1.21 (1.07\u0026ndash;1.37)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003e83%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\n \u003cp\u003e6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c6\"\u003e\n \u003cp\u003e+\u0026thinsp;5.5 (2.7\u0026ndash;8.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eArthritis (single)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\n \u003cp\u003e15,636\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\n \u003cp\u003e1.09 (1.02\u0026ndash;1.16)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003e29%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\n \u003cp\u003e6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c6\"\u003e\n \u003cp\u003e+\u0026thinsp;4.0 (0.8\u0026ndash;7.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n \u003cp\u003e\u003cem\u003eAll HRs are adjusted for age, sex, educational attainment, and smoking, and are relative to the disease-free group (0 of 7 conditions). ARD\u0026thinsp;=\u0026thinsp;10-year Kaplan\u0026ndash;Meier absolute risk difference versus disease-free reference, pooled across cohorts. I\u0026sup2; = between-cohort heterogeneity statistic. HR\u0026thinsp;=\u0026thinsp;hazard ratio; CI\u0026thinsp;=\u0026thinsp;confidence interval; k\u0026thinsp;=\u0026thinsp;number of cohorts contributing.\u003c/em\u003e\u003c/p\u003e\n \u003cp\u003eFigure \u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e and Table \u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e summarise cohort structure, follow-up, mortality, and the prevalence of the seven harmonised chronic conditions. The prevalence of core dyads and triads is shown in Table \u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec18\" class=\"Section2\"\u003e\n \u003ch2\u003eLayer 1: disease-count gradient in mortality risk\u003c/h2\u003e\n \u003cp\u003eA clear count gradient in all-cause mortality was observed. Relative to disease-free participants, pooled hazard ratios were 1.28 (95% CI 1.18\u0026ndash;1.39) for pure single-condition groups, 1.72 (1.51\u0026ndash;1.95) for pure dyads, and 2.29 (1.81\u0026ndash;2.88) for pure triads. Ten-year absolute risk differences also increased stepwise across count classes (Fig. \u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e; Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e).\u003c/p\u003e\n \u003cdiv class=\"gridtable\"\u003e\u0026nbsp;\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eDisease-count gradient and single-condition mortality risk\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003ccolgroup cols=\"8\"\u003e\u003c/colgroup\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eSection\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003eCombination\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colname=\"c3\"\u003e\n \u003cp\u003eN exposed\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003eHR (95% CI)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colname=\"c5\"\u003e\n \u003cp\u003eI\u0026sup2; (%)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colname=\"c6\"\u003e\n \u003cp\u003e10-yr ARD, %\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colname=\"c7\"\u003e\n \u003cp\u003ePrediction interval\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colname=\"c8\"\u003e\n \u003cp\u003eConsistency type\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eCount gradient\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003e1 disease (any), pure\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\n \u003cp\u003e54,033\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\n \u003cp\u003e1.28 (1.18\u0026ndash;1.39)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c5\"\u003e\n \u003cp\u003e78%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c6\"\u003e\n \u003cp\u003e\u0026ndash;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c7\"\u003e\n \u003cp\u003e\u0026ndash;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c8\"\u003e\n \u003cp\u003e\u0026ndash;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eCount gradient\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003e2 diseases (any), pure\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\n \u003cp\u003e30,684\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\n \u003cp\u003e1.72 (1.51\u0026ndash;1.95)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c5\"\u003e\n \u003cp\u003e90%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c6\"\u003e\n \u003cp\u003e\u0026ndash;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c7\"\u003e\n \u003cp\u003e\u0026ndash;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c8\"\u003e\n \u003cp\u003e\u0026ndash;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eCount gradient\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003e3 diseases (any), pure\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\n \u003cp\u003e13,008\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\n \u003cp\u003e2.29 (1.81\u0026ndash;2.88)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c5\"\u003e\n \u003cp\u003e95%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c6\"\u003e\n \u003cp\u003e\u0026ndash;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c7\"\u003e\n \u003cp\u003e\u0026ndash;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c8\"\u003e\n \u003cp\u003e\u0026ndash;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eSingle conditions\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003eLung disease\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\n \u003cp\u003e2,327\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\n \u003cp\u003e1.89 (1.49\u0026ndash;2.40)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c5\"\u003e\n \u003cp\u003e79%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c6\"\u003e\n \u003cp\u003e+\u0026thinsp;12.7 (6.0\u0026ndash;19.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c7\"\u003e\n \u003cp\u003e1.11\u0026ndash;3.21\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c8\"\u003e\n \u003cp\u003eT2: High-risk heterogeneous\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eSingle conditions\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003eCancer\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\n \u003cp\u003e2,887\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\n \u003cp\u003e1.83 (1.52\u0026ndash;2.20)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c5\"\u003e\n \u003cp\u003e65%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c6\"\u003e\n \u003cp\u003e+\u0026thinsp;12.9 (7.7\u0026ndash;18.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c7\"\u003e\n \u003cp\u003e1.27\u0026ndash;2.64\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c8\"\u003e\n \u003cp\u003eT2: High-risk heterogeneous\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eSingle conditions\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003eStroke\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\n \u003cp\u003e977\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\n \u003cp\u003e1.80 (1.61\u0026ndash;2.02)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c5\"\u003e\n \u003cp\u003e0%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c6\"\u003e\n \u003cp\u003e+\u0026thinsp;27.6 (22.9\u0026ndash;32.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c7\"\u003e\n \u003cp\u003e1.61\u0026ndash;2.02\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c8\"\u003e\n \u003cp\u003eT1: Consistent high-risk\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eSingle conditions\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003eDiabetes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\n \u003cp\u003e4,092\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\n \u003cp\u003e1.56 (1.27\u0026ndash;1.90)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c5\"\u003e\n \u003cp\u003e77%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c6\"\u003e\n \u003cp\u003e+\u0026thinsp;7.8 (1.5\u0026ndash;14.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c7\"\u003e\n \u003cp\u003e1.00\u0026ndash;2.42\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c8\"\u003e\n \u003cp\u003eT2: High-risk heterogeneous\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eSingle conditions\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003eHeart disease\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\n \u003cp\u003e4,122\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\n \u003cp\u003e1.43 (1.24\u0026ndash;1.65)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c5\"\u003e\n \u003cp\u003e58%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c6\"\u003e\n \u003cp\u003e+\u0026thinsp;12.8 (6.1\u0026ndash;19.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c7\"\u003e\n \u003cp\u003e1.09\u0026ndash;1.88\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c8\"\u003e\n \u003cp\u003eT2b: Moderate stable\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eSingle conditions\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003eHypertension\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\n \u003cp\u003e23,992\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\n \u003cp\u003e1.21 (1.07\u0026ndash;1.37)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c5\"\u003e\n \u003cp\u003e83%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c6\"\u003e\n \u003cp\u003e+\u0026thinsp;5.5 (2.7\u0026ndash;8.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c7\"\u003e\n \u003cp\u003e0.92\u0026ndash;1.60\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c8\"\u003e\n \u003cp\u003eT3: Mixed/unstable\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eSingle conditions\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003eArthritis\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\n \u003cp\u003e15,636\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\n \u003cp\u003e1.09 (1.02\u0026ndash;1.16)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c5\"\u003e\n \u003cp\u003e29%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c6\"\u003e\n \u003cp\u003e+\u0026thinsp;4.0 (0.8\u0026ndash;7.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c7\"\u003e\n \u003cp\u003e0.98\u0026ndash;1.20\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c8\"\u003e\n \u003cp\u003eT2b: Moderate stable\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n \u003cp\u003e\u003cem\u003eHRs are adjusted for age, sex, educational attainment, and smoking versus the disease-free group. ARD\u0026thinsp;=\u0026thinsp;10-year absolute risk difference (pooled). Prediction interval\u0026thinsp;=\u0026thinsp;95% PI for the pooled HR. Consistency type: T1\u0026thinsp;=\u0026thinsp;consistently high-risk in all cohorts; T2\u0026thinsp;=\u0026thinsp;directionally consistent but heterogeneous; T3\u0026thinsp;=\u0026thinsp;unstable or mixed direction. All dyad and triad estimates are shown in Additional file 2: Table \u003cspan refid=\"MOESM2\" class=\"InternalRef\"\u003eS2\u003c/span\u003e.\u003c/em\u003e\u003c/p\u003e\n \u003cp\u003eEven so, I\u0026sup2; values of 78\u0026ndash;95% at the disease-count level signalled that count alone was not fully capturing the prognostic landscape\u0026mdash;substantial variation remained within each count stratum. This count-level pattern served as the background against which combination-specific heterogeneity was evaluated. Among individual single conditions, stroke showed the most stable risk elevation (I\u0026sup2; = 0%, Type 1 consistency), whereas hypertension\u0026mdash;the most prevalent single condition\u0026mdash;showed the highest between-cohort heterogeneity (I\u0026sup2; = 83%) with a prediction interval that crossed the null (0.92\u0026ndash;1.60, Type 3), likely reflecting cross-cohort variation in hypertension management and ascertainment practices.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec19\" class=\"Section2\"\u003e\n \u003ch2\u003eLayer 2: within-count heterogeneity among dyads and triads\u003c/h2\u003e\n \u003cp\u003eWithin count strata, the spread in mortality risk was striking. Among the 13 dyads, pooled hazard ratios spanned from 1.26 (hypertension plus arthritis) to 5.10 (heart disease plus lung disease) \u0026mdash; a more than four-fold difference in relative mortality risk within the same disease-count class. The 15 analysed triads showed a narrower but still substantial spread, from 1.62 to 3.22 (approximately 1.98-fold) (Fig. \u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e; Table \u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e; Additional file 2: Fig. \u003cspan refid=\"MOESM9\" class=\"InternalRef\"\u003eS9\u003c/span\u003e).\u003c/p\u003e\n \u003cp\u003eThe population-level scale of this difference was substantial. Heart disease plus lung disease\u0026mdash;the highest-risk dyad\u0026mdash;carried a 10-year pooled ARD of +\u0026thinsp;41.8 percentage points relative to disease-free participants. Hypertension plus arthritis, the most common dyad, had an excess 10-year risk of +\u0026thinsp;8.9 percentage points. Combinations with the same disease count differed not modestly but by over 30 percentage points in absolute mortality burden.\u003c/p\u003e\n \u003cp\u003eFormal within-count summaries confirmed that dyads showed the widest relative spread, and that only a minority of combinations deviated markedly upward from the count-class background (Additional file 2: Fig. \u003cspan refid=\"MOESM9\" class=\"InternalRef\"\u003eS9\u003c/span\u003e and Table \u003cspan refid=\"MOESM3\" class=\"InternalRef\"\u003eS3\u003c/span\u003e). Knowing the number of conditions is not enough\u0026mdash;which conditions co-occur matters too.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec20\" class=\"Section2\"\u003e\n \u003ch2\u003eLayer 3: core high-risk combinations and priority structure\u003c/h2\u003e\n \u003cp\u003eUsing the prespecified four-dimensional prioritisation framework, 17 combinations were identified as core high-risk combinations: 7 dyads and 10 triads. Among dyads, the most hazardous combinations included heart disease plus lung disease (pooled HR 5.10), diabetes plus heart disease (2.50), hypertension plus lung disease (2.03), hypertension plus cancer (2.03), and hypertension plus stroke (1.97). Among triads, the highest-risk combinations included hypertension plus diabetes plus lung disease (3.22), hypertension plus diabetes plus stroke (3.10), hypertension plus heart disease plus lung disease (2.83), and hypertension plus heart disease plus stroke (2.67) (Fig. \u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e; Table 4).\u003c/p\u003e\n \u003cp\u003eThis figure displays pooled hazard ratios with 95% confidence intervals and prediction intervals for the core single conditions, dyads, and triads retained in the final main analysis. The figure shows seven single conditions (Panel A), the six highest-risk dyads (Panel B), and the five highest-risk triads (Panel C), selected from the 17 core combinations by pooled HR ranking. Complete estimates for all 17 core combinations are in Table\u0026nbsp;4.\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eTable\u0026nbsp;4. Core high-risk multimorbidity combinations: dyads and triads\u003c/strong\u003e\u003c/p\u003e\n \u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"602\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eCombination\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eN exposed\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eHR (95% CI)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eI\u0026sup2; (%)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003e10-yr ARD, %\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003ePrediction interval\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eDirection consistency\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eConsistency type\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eOrgan axis\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eHeart disease + Lung disease\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e380\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e5.10 (2.55\u0026ndash;10.22)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e92%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e+41.8 (32.5\u0026ndash;51.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1.09\u0026ndash;23.79\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e100%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eType 2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eCM + Resp\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eDiabetes + Heart disease\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e559\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e2.50 (1.57\u0026ndash;3.99)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e81%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e+26.9 (17.2\u0026ndash;36.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.90\u0026ndash;6.95\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e83%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eType 2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003ePure CM\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eHypertension + Cancer\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1,330\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e2.03 (1.56\u0026ndash;2.64)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e68%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e+18.3 (10.1\u0026ndash;26.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1.21\u0026ndash;3.41\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e100%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eType 2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eCancer + CM\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eHypertension + Lung disease\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1,072\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e2.03 (1.40\u0026ndash;2.93)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e83%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e+18.5 (8.3\u0026ndash;28.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.88\u0026ndash;4.70\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e100%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eType 2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eCM + Resp\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eHypertension + Stroke\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1,145\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1.97 (1.58\u0026ndash;2.46)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e63%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e+24.9 (15.9\u0026ndash;33.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1.27\u0026ndash;3.07\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e100%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eType 2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eCM + CV\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eStroke + Arthritis\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e415\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1.93 (1.58\u0026ndash;2.35)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e21%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e+25.9 (10.6\u0026ndash;41.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1.45\u0026ndash;2.56\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e100%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eType 1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eMS-dominant\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eLung disease + Arthritis\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1,281\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1.92 (1.53\u0026ndash;2.42)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e67%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e+20.2 (15.0\u0026ndash;25.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1.21\u0026ndash;3.05\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e100%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eType 2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eMS-dominant\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eHTN + Diabetes + Lung disease\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e264\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e3.22 (1.66\u0026ndash;6.23)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e79%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e+25.7 (17.1\u0026ndash;34.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.76\u0026ndash;13.67\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e83%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eType 2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eCM + Resp\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eHTN + Diabetes + Stroke\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e383\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e3.10 (2.30\u0026ndash;4.19)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e52%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e+44.7 (38.0\u0026ndash;51.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1.78\u0026ndash;5.41\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e83%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eType 2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eCM + CV\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eHTN + Heart disease + Lung disease\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e374\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e2.83 (1.94\u0026ndash;4.12)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e66%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e+42.0 (24.9\u0026ndash;59.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1.34\u0026ndash;5.99\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e83%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eType 2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eCM + Resp\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eHTN + Heart disease + Stroke\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e466\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e2.67 (1.71\u0026ndash;4.16)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e79%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e+35.8 (23.1\u0026ndash;48.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1.06\u0026ndash;6.71\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e83%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eType 2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eCM + CV\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eHTN + Cancer + Heart disease\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e281\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e2.56 (1.70\u0026ndash;3.86)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e59%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e+43.1 (35.1\u0026ndash;51.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1.24\u0026ndash;5.30\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e80%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eType 2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eCancer + CM\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eHTN + Diabetes + Cancer\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e326\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e2.56 (2.13\u0026ndash;3.08)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e+30.8 (17.4\u0026ndash;44.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e2.12\u0026ndash;3.08\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e100%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eType 1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eCancer + CM\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eHTN + Diabetes + Heart disease\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1,169\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e2.36 (1.57\u0026ndash;3.56)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e87%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e+23.8 (10.7\u0026ndash;36.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.95\u0026ndash;5.85\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e83%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eType 2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003ePure CM\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eDiabetes + Lung disease + Arthritis\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e105\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e2.27 (1.09\u0026ndash;4.76)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e73%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e+29.8 (13.1\u0026ndash;46.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.58\u0026ndash;8.89\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e100%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eType 2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eCM + Resp\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eHeart disease + Lung disease + Arthritis\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e315\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e2.25 (1.60\u0026ndash;3.16)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e67%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e+31.3 (12.2\u0026ndash;50.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1.21\u0026ndash;4.19\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e100%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eType 2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eCM + Resp\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eHTN + Stroke + Arthritis\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e652\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1.96 (1.55\u0026ndash;2.48)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e49%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e+35.4 (29.3\u0026ndash;41.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1.28\u0026ndash;2.99\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e83%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eType 2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eCM + CV\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003cp\u003e\u003cem\u003eAbbreviations: HR, hazard ratio; CI, confidence interval; PI, prediction interval; ARD, absolute risk difference; CM, cardiometabolic; CV, cerebrovascular; Resp, respiratory. Axis labels in this table are simplified from the full seven-category post-hoc classification; see Additional file 2: Fig. S10 and Table S16 for the complete classification scheme.\u003c/em\u003e\u003c/p\u003e\n \u003cp\u003e\u003cem\u003eCore high-risk combinations meeting \u0026ge;3 of 4 prioritisation criteria. Dyads (top 7 rows) and triads (bottom 10 rows) are listed in descending order of pooled HR within each group. HTN = hypertension; CM = cardiometabolic; Resp = respiratory; CV = cerebrovascular; MS = musculoskeletal. HRs adjusted for age, sex, educational attainment, and smoking versus disease-free reference. ARD = 10-year absolute risk difference (pooled). Direction consistency = proportion of cohorts showing HR \u0026gt; 1. Type 1 = consistently high-risk; Type 2 = directionally consistent but heterogeneous in magnitude.\u003c/em\u003e\u003c/p\u003e\n \u003cp\u003eThe highest-risk combinations were not the most prevalent ones. Combinations carrying the greatest absolute mortality burden were often uncommon, while many of the most frequently occurring combinations were only moderately hazardous. This systematic mismatch between prevalence and mortality burden challenges frameworks that prioritise multimorbidity management by condition frequency alone \u0026mdash; and points toward a population health approach that distinguishes common combinations from genuinely dangerous ones (Fig. 5; Table 4).\u003c/p\u003e\n \u003cp\u003eRanking combinations by frequency alone would systematically miss those carrying the highest mortality burden.\u003c/p\u003e\n \u003ch2\u003eLayer 4: post hoc structural axes and cross-regional reproducibility\u003c/h2\u003e\n \u003cp\u003eLooking at the composition of the 17 core combinations, a clear structural pattern emerged: high-risk combinations were not scattered randomly. They congregated along two main axes\u0026mdash;cardiometabolic disease compounded by respiratory disease, and cardiometabolic disease compounded by cerebrovascular disease. Pure cardiometabolic combinations also appeared in the core set but generally carried lower average risk than these more complex extensions. Cancer-containing combinations contributed in selected instances but were less consistently recurrent (Fig. 5; Additional file 2: Figs. S10 and S11).\u003c/p\u003e\n \u003cp\u003eAcross the 36 analysed exposures, 19 showed directionally consistent risk elevation in all six cohorts, and 25 had prediction intervals that excluded the null. Under the descriptive consistency typology, 3 combinations were consistently high-risk, 28 directionally consistent but variable in magnitude, and 5 unstable or mixed. These two tallies reflect different standards. Nineteen combinations showed HR \u0026gt; 1 in every individual cohort (strict directional consistency). Using the three-type consistency typology\u0026mdash;which additionally considers prediction intervals and leave-one-cohort stability\u0026mdash;3 were classified as consistently high-risk, 28 as directionally consistent but heterogeneous in magnitude, and 5 as unstable or mixed; the first two categories sum to 31. Directional reproducibility was far more common than strict quantitative homogeneity (Fig. 6; Table 4).\u003c/p\u003e\n \u003cp\u003eHeart disease plus lung disease illustrates the trade-off between magnitude and stability\u0026mdash;it was the highest-risk dyad, yet also the most heterogeneous, with cohort-specific estimates spanning a wide range. Several cardiometabolic combinations sat lower on the risk scale but maintained their direction more reliably across settings. Sensitivity analyses using contains definitions, alternative references, early-death exclusions, and reduced adjustment did not materially change the overall picture (Additional file 2: Figs. S2, S3, S5, S6, S8, and S12).\u003c/p\u003e\n\u003c/div\u003e"},{"header":"Discussion","content":"\u003ch2\u003ePrincipal findings\u003c/h2\u003e\n\u003cp\u003eDisease count and disease structure each shape mortality risk, but in different ways. Across 177,665 older adults in six countries, count was associated with a clear stepwise gradient\u0026mdash;yet combinations at the same count level differed markedly in both relative and absolute risk. The highest-risk dyads and triads fell along recognisable structural axes rather than being scattered at random, and those axes showed directional reproducibility across regions even when effect sizes varied.\u003c/p\u003e\n\u003cp\u003eDisease count is, in short, useful but insufficient. Among older adults with the same number of conditions, prognosis depends substantially on which conditions co-occur\u0026mdash;a distinction that matters particularly given evidence that existing mortality prediction models relying on standard morbidity summaries often perform only modestly across external validation settings. [9,25]\u003c/p\u003e\n\u003ch2\u003eComparison with previous studies\u003c/h2\u003e\n\u003cp\u003eOur findings extend the disease-count literature by demonstrating what count-based frameworks leave unanswered: whether all dyads or all triads of the same size are prognostically equivalent. They are not.\u003c/p\u003e\n\u003cp\u003eOur results sit alongside, rather than against, prior work on multimorbidity patterns. Latent class and cluster-based approaches have established that multimorbidity is structured, but they produce groupings that are difficult to compare across datasets and do not yield combination-specific effect estimates. Working with strict pure combinations allowed us to compare named dyads and triads directly \u0026mdash; a more tractable framing for both clinical risk assessment and population-level epidemiological surveillance. [10-13]\u003c/p\u003e\n\u003cp\u003eThere is single-cohort evidence pointing toward particularly dangerous dyads\u0026mdash;combinations of heart failure and COPD being the most studied. Our analysis situates those observations within a broader framework: such combinations are not isolated findings but recurring structural patterns in the epidemiology of multimorbidity mortality across diverse ageing populations worldwide \u0026mdash; a finding with implications for how international health surveillance systems characterise this burden. [8,26,27]\u003c/p\u003e\n\u003cp\u003eWhat sets this study apart from prior pattern-based work is the direct quantification of same-count heterogeneity. Rather than inferring that structure matters, we show it explicitly: within identical disease-count strata, risk varies substantially, and that variation is structured.\u003c/p\u003e\n\u003ch2\u003eInterpretation and implications\u003c/h2\u003e\n\u003cp\u003eThe gap between prevalence and danger has practical consequences. Risk stratification tools anchored to condition count or disease frequency may systematically under-prioritise less common but genuinely hazardous combinations\u0026mdash;heart disease with chronic lung disease being the clearest example. Common combinations such as hypertension plus arthritis may drive the bulk of service demand while contributing far less to mortality. [28,29]\u003c/p\u003e\n\u003cp\u003eThe cardiometabolic\u0026ndash;respiratory axis that emerged from this analysis is mechanistically plausible. Cardiovascular and chronic respiratory diseases share pathways through systemic inflammation, hypoxia, pulmonary vascular changes, reduced exercise tolerance, and exacerbation cycles. The cardiometabolic\u0026ndash;cerebrovascular axis is grounded in shared vascular risk, impaired post-stroke recovery, frailty, and downstream complications. Our observational design cannot establish mechanism, but the concentration of high-risk combinations along biologically coherent axes\u0026mdash;rather than scattered at random\u0026mdash;supports the structural interpretation. [26,27,30]\u003c/p\u003e\n\u003cp\u003eThese findings suggest that multimorbidity profiling in older adults could be sharpened by asking not just how many conditions an individual carries, but which ones co-occur. Condition count alone misses the structure. [29,31]\u003c/p\u003e\n\u003cp\u003eFrom a population health standpoint, these findings point toward three concrete implications. First, surveillance systems that aggregate multimorbidity burden by disease count alone may systematically underestimate the mortality concentration in specific combination subgroups. A person with heart disease plus lung disease faces more than four times the hazard of one with hypertension plus arthritis \u0026mdash; yet count-based metrics assign them equal weight. Second, health needs assessments \u0026mdash; which guide resource allocation across health systems \u0026mdash; could be refined by profiling the prevalence\u0026ndash;danger landscape documented here: the most prevalent combinations (hypertension plus arthritis, 14% of pure dyad participants) are not the most lethal, while the most lethal (heart disease plus lung disease, +41.8 percentage points in ten-year absolute risk) are comparatively rare. Prioritising by frequency will systematically underinvest in the highest-risk groups. Third, integrated care programmes for multimorbidity are expanding across health systems; the cardiometabolic\u0026ndash;respiratory and cardiometabolic\u0026ndash;cerebrovascular axes identified here offer empirically grounded structural targets for coordinated care design \u0026mdash; a framework that goes beyond managing individual conditions in isolation.\u003c/p\u003e\n\u003ch2\u003eHeterogeneity and external validity\u003c/h2\u003e\n\u003cp\u003eThe substantial between-cohort heterogeneity was expected\u0026mdash;baseline mortality rates, disease ascertainment practices, healthcare access, treatment patterns, and follow-up duration all differ across six countries. High I\u0026sup2; does not invalidate pooled estimates, but it does change how they should be read: as average signals in heterogeneous populations, not portable, fixed-point risk multipliers. Prediction intervals and leave-one-cohort analyses communicate this uncertainty more honestly than confidence intervals alone.\u003c/p\u003e\n\u003cp\u003eDirectional consistency was more common than quantitative homogeneity, and we regard it as the more appropriate standard. Risk elevation in the same direction across settings spanning China, Europe, and Mexico\u0026mdash;even with wide prediction intervals\u0026mdash;is consistent with genuine structural risk that regional context modulates in magnitude.\u003c/p\u003e\n\u003ch2\u003eStrengths and limitations\u003c/h2\u003e\n\u003cp\u003eThe study\u0026apos;s main strengths are its scale and analytical design. Six geographically diverse cohorts with harmonised analyses offer a breadth of evidence not achievable in single-country work. Strict pure-combination definitions reduce contamination from additional conditions. Reporting both hazard ratios and absolute risk differences makes findings interpretable at both the individual and population level \u0026mdash; the former supporting prognostic reasoning, the latter directly relevant to public health burden estimation. Heterogeneity is addressed directly through prediction intervals, leave-one-cohort analyses, and descriptive consistency classes rather than being treated as a nuisance.\u003c/p\u003e\n\u003cp\u003eSeveral limitations warrant acknowledgement. All six cohorts recruited community-dwelling participants, potentially underrepresenting institutionalised or very frail older adults in whom multimorbidity burden may be higher and combination-specific risks may differ. All seven conditions relied on self-reported physician diagnosis, leaving recall bias, under-diagnosis, and cross-country differences in healthcare access as plausible sources of misclassification. [32] In particular, \u0026ldquo;heart disease\u0026rdquo; encompasses a heterogeneous category whose scope may vary across cohorts and respondents\u0026mdash;potentially including coronary heart disease, heart failure, valvular disease, and arrhythmias\u0026mdash;and cross-country differences in lay interpretation of this term may contribute to between-cohort heterogeneity for heart-disease-containing combinations. Restricting the analysis to conditions available across all six cohorts understates the real complexity of multimorbidity\u0026mdash;chronic kidney disease, dementia, and depression are conspicuously absent. Strict pure-combination definitions, while analytically cleaner, reduce sample sizes for rarer combinations; triads with fewer than four contributing cohorts should be interpreted with particular caution. The DerSimonian\u0026ndash;Laird random-effects estimator used here is known to underestimate between-study variance when the number of studies is small (k = 6) [21], which may yield confidence intervals that are too narrow; alternative estimators such as REML or the Hartung\u0026ndash;Knapp adjustment could be considered in future work. Residual confounding\u0026mdash;from disease severity, medication use, and unmeasured lifestyle factors\u0026mdash;cannot be excluded. ARDs from cohort-specific Kaplan\u0026ndash;Meier estimates are descriptive, not causal. And the structural axes identified are post hoc patterns in this data, not predefined subtypes; they need replication in independent cohorts before they can be treated as validated risk categories.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eIn 177,665 older adults across six countries, disease count predicted mortality in a stepwise gradient \u0026mdash; but it concealed far more than it revealed about which combinations actually drive population-level mortality burden.\u003c/p\u003e\n\u003cp\u003eAmong people with the same number of conditions, mortality risk differed substantially depending on which conditions co-occurred. The highest risks clustered in cardiometabolic combinations compounded by respiratory or cerebrovascular disease, and these structural patterns held directionally across diverse settings even when effect sizes varied.\u003c/p\u003e\n\u003cp\u003eThe evidence supports a shift toward combination-specific epidemiological characterisation of multimorbidity \u0026mdash; one that identifies not just how many conditions co-occur but which structural patterns carry the greatest population-level mortality burden. Such an approach can strengthen multimorbidity surveillance, inform health needs assessment, and provide an empirical basis for prioritising prevention and integrated care for the highest-risk combination structures across diverse health systems.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cdiv class=\"DefinitionList\"\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eARD\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eAbsolute risk difference\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eCHARLS\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eChina Health and Retirement Longitudinal Study\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eCI\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eConfidence interval\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eDM\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eDiabetes mellitus\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eELSA\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eEnglish Longitudinal Study of Ageing\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eHR\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eHazard ratio\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eHRS\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eHealth and Retirement Study\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eHTN\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eHypertension\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eKLoSA\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eKorean Longitudinal Study of Ageing\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eKM\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eKaplan\u0026ndash;Meier\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eLOC\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eLeave-one-cohort\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eMHAS\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eMexican Health and Aging Study\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003ePH\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eProportional hazards\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003ePI\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003ePrediction interval\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eSHARE\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eSurvey of Health, Ageing and Retirement in Europe\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003c/div\u003e"},{"header":"Declarations","content":"\u003ch2\u003eEthics approval and consent to participate\u003c/h2\u003e\n\u003cp\u003eThis study used de-identified secondary data from six established longitudinal ageing cohorts. Each original cohort study obtained ethics approval from the relevant institutional review board or ethics committee, and all participants provided informed consent in accordance with local study procedures. The current work is a secondary analysis of harmonised cohort data and did not require new ethics approval. Ethics approvals for the parent cohort studies are described in their cohort profile publications. [14-19]\u003c/p\u003e\n\u003ch2\u003eConsent for publication\u003c/h2\u003e\n\u003cp\u003eNot applicable. No individually identifiable data are included in this manuscript.\u003c/p\u003e\n\u003ch2\u003eAvailability of data and materials\u003c/h2\u003e\n\u003cp\u003eThe datasets analysed in this study are derived from established longitudinal ageing studies. Data access is available through each study\u0026rsquo;s data portal: CHARLS (http://charls.pku.edu.cn), ELSA (https://www.elsa-project.ac.uk), HRS (https://hrs.isr.umich.edu), SHARE (https://share-eric.eu), KLoSA (https://survey.keis.or.kr), MHAS (http://www.mhasweb.org). Derived summary statistics and supplementary outputs are provided with this manuscript.\u003c/p\u003e\n\u003ch2\u003eCompeting interests\u003c/h2\u003e\n\u003cp\u003eThe authors declare that they have no competing interests.\u003c/p\u003e\n\u003ch2\u003eFunding\u003c/h2\u003e\n\u003cp\u003eThis work was supported by the Jiangsu Province Frontier Technology R\u0026amp;D Program (BF2025613) and the Jiangsu Provincial Key Research and Development Program (BE2023818). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.\u003c/p\u003e\n\u003ch2\u003eAuthors\u0026rsquo; contributions\u003c/h2\u003e\n\u003cp\u003eTZ and YX contributed equally to this work. TZ, LL, and JW conceived and designed the study. TZ and YX performed data harmonisation and statistical analysis. GJ, RH, YF, and XM contributed to data processing and quality control. XT and XZ assisted with interpretation of results. TZ drafted the manuscript. LL and JW supervised the study and critically revised the manuscript. All authors read and approved the final manuscript.\u003c/p\u003e\n\u003ch2\u003eAcknowledgements\u003c/h2\u003e\n\u003cp\u003eThe authors thank the participants and research teams of CHARLS, ELSA, HRS, SHARE, KLoSA, and MHAS for making these data available for research.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eNicholson K, Liu W, Fitzpatrick D, Hardacre KA, Roberts S, Salerno J, et al. Prevalence of multimorbidity and polypharmacy among adults and older adults: a systematic review. Lancet Healthy Longev. 2024;5(4):e287\u0026ndash;96. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/S2666-7568(24)00007-2\u003c/span\u003e\u003cspan address=\"10.1016/S2666-7568(24)00007-2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChowdhury SR, Chandra Das D, Sunna TC, Beyene J, Hossain A. Global and regional prevalence of multimorbidity in the adult population in community settings: a systematic review and meta-analysis. EClinicalMedicine. 2023;57:101860. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.eclinm.2023.101860\u003c/span\u003e\u003cspan address=\"10.1016/j.eclinm.2023.101860\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFerris JK, Fiedeldey LK, Kim B, Clemens F, Irvine MA, Haji Hosseini S, et al. A systematic review and meta-analysis of disease clusters in multimorbidity. Nat Commun. 2026;17(1):634. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41467-025-67372-6\u003c/span\u003e\u003cspan address=\"10.1038/s41467-025-67372-6\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHan S, Chen W, Shen M, Shao R, Yang W, Wang C, et al. Multimorbidity progression and the heterogeneous impact of healthy ageing risk factors: a multicohort study. BMJ Public Health. 2025;3(2):e002474. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1136/bmjph-2024-002474\u003c/span\u003e\u003cspan address=\"10.1136/bmjph-2024-002474\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eThe Academy of Medical Sciences. Multimorbidity: a priority for global health research. London: Academy of Medical Sciences. 2018. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://acmedsci.ac.uk/policy/policy-projects/multimorbidity\u003c/span\u003e\u003cspan address=\"https://acmedsci.ac.uk/policy/policy-projects/multimorbidity\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Accessed 15 Mar 2026.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBarnett K, Mercer SW, Norbury M, Watt G, Wyke S, Guthrie B. Epidemiology of multimorbidity and implications for health care, research, and medical education: a cross-sectional study. Lancet. 2012;380(9836):37\u0026ndash;43. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/S0140-6736(12)60240-2\u003c/span\u003e\u003cspan address=\"10.1016/S0140-6736(12)60240-2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMakovski TT, Schmitz S, Zeegers MP, Herzig S, Besi C, Schneider S, et al. Multimorbidity and quality of life: systematic literature review and meta-analysis. Ageing Res Rev. 2019;53:100903. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.arr.2019.04.005\u003c/span\u003e\u003cspan address=\"10.1016/j.arr.2019.04.005\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDamiano C, Dam V, Marengoni A, Ferrara N, Rizzuto D, Grande G, et al. Multimorbidity patterns and mortality in older adults: a two-cohort pooled analysis. Aging Clin Exp Res. 2025;37(1):250. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s40520-025-03150-0\u003c/span\u003e\u003cspan address=\"10.1007/s40520-025-03150-0\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNunes BP, Flores TR, Mielke GI, Thum\u0026eacute; E, Facchini LA. Multimorbidity and mortality in older adults: a systematic review and meta-analysis. Arch Gerontol Geriatr. 2016;67:130\u0026ndash;8. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.archger.2016.07.008\u003c/span\u003e\u003cspan address=\"10.1016/j.archger.2016.07.008\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOgaz-Gonz\u0026aacute;lez R, Zou Q, Maroto-Rodriguez J, Guti\u0026eacute;rrez-Robledo LM, Escamilla-Santiago R, L\u0026oacute;pez-Cervantes M, et al. Mortality among older adults across multimorbidity categories and lifestyle patterns. Aging Dis. 2026. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.14336/AD.2025.1012\u003c/span\u003e\u003cspan address=\"10.14336/AD.2025.1012\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Online ahead of print.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJoseph JJ, Bennett A, Ojo O, Ogunmoroti O, Melissa AE, Durkalski-Mauldin V, et al. Associations of cardiometabolic multimorbidity with all-cause and coronary heart disease mortality among Black adults in the Jackson Heart Study. JAMA Netw Open. 2022;5(10):e2237552. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1001/jamanetworkopen.2022.37552\u003c/span\u003e\u003cspan address=\"10.1001/jamanetworkopen.2022.37552\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePrados-Torres A, Calder\u0026oacute;n-Larra\u0026ntilde;aga A, Hancco-Saavedra J, van den Poblador-Plou B. Multimorbidity patterns: a systematic review. J Clin Epidemiol. 2014;67(3):254\u0026ndash;66. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.jclinepi.2013.09.021\u003c/span\u003e\u003cspan address=\"10.1016/j.jclinepi.2013.09.021\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGarin N, Koyanagi A, Chatterji S, Tyrovolas S, Olaya B, Leonardi M, et al. Global multimorbidity patterns: a cross-sectional, population-based, multi-country study. J Gerontol Biol Sci Med Sci. 2016;71(2):205\u0026ndash;14. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1093/gerona/glv128\u003c/span\u003e\u003cspan address=\"10.1093/gerona/glv128\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhao Y, Hu Y, Smith JP, Strauss J, Yang G. Cohort profile: the China Health and Retirement Longitudinal Study (CHARLS). Int J Epidemiol. 2014;43(1):61\u0026ndash;8. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1093/ije/dys203\u003c/span\u003e\u003cspan address=\"10.1093/ije/dys203\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSteptoe A, Breeze E, Banks J, Nazroo J. Cohort profile: the English Longitudinal Study of Ageing. Int J Epidemiol. 2013;42(6):1640\u0026ndash;8. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1093/ije/dys168\u003c/span\u003e\u003cspan address=\"10.1093/ije/dys168\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSonnega A, Faul JD, Ofstedal MB, Langa KM, Phillips JWR, Weir DR. Cohort profile: the Health and Retirement Study (HRS). Int J Epidemiol. 2014;43(2):576\u0026ndash;85. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1093/ije/dyu067\u003c/span\u003e\u003cspan address=\"10.1093/ije/dyu067\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eB\u0026ouml;rsch-Supan A, Brandt M, Hunkler C, Kneip T, Korbmacher J, Malter F, et al. Data resource profile: the Survey of Health, Ageing and Retirement in Europe (SHARE). Int J Epidemiol. 2013;42(4):992\u0026ndash;1001. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1093/ije/dyt088\u003c/span\u003e\u003cspan address=\"10.1093/ije/dyt088\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLee J. KLoSA-Korean Longitudinal Study of Aging. Korean J Fam Med. 2020;41(1):1\u0026ndash;2. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.4082/kjfm.19.0160\u003c/span\u003e\u003cspan address=\"10.4082/kjfm.19.0160\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWong R, Michaels-Obreg\u0026oacute;n A, Palloni A. Cohort profile: the Mexican Health and Aging Study (MHAS). Int J Epidemiol. 2017;46(2):e2. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1093/ije/dyu263\u003c/span\u003e\u003cspan address=\"10.1093/ije/dyu263\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLee J, Ailshire J, Crimmins E, et al. J Gerontol B Psychol Sci Soc Sci. 2021;76(Suppl 1):S5\u0026ndash;16. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1093/geronb/gbab050\u003c/span\u003e\u003cspan address=\"10.1093/geronb/gbab050\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Gateway to Global Aging Data: resources for cross-national comparisons of aging.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials. 1986;7(3):177\u0026ndash;88. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/0197-2456(86)90046-2\u003c/span\u003e\u003cspan address=\"10.1016/0197-2456(86)90046-2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHiggins JPT, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency in meta-analyses. BMJ. 2003;327(7414):557\u0026ndash;60. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1136/bmj.327.7414.557\u003c/span\u003e\u003cspan address=\"10.1136/bmj.327.7414.557\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIntHout J, Ioannidis JPA, Rovers MM, Goeman JJ. Plea for routinely presenting prediction intervals in meta-analysis. BMJ Open. 2016;6(7):e010247. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1136/bmjopen-2015-010247\u003c/span\u003e\u003cspan address=\"10.1136/bmjopen-2015-010247\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRiley RD, Higgins JPT, Deeks JJ. Interpretation of random effects meta-analyses. BMJ. 2011;342:d549. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1136/bmj.d549\u003c/span\u003e\u003cspan address=\"10.1136/bmj.d549\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHo L, Pugh C, Seth S, Arakelyan S, Lone NI, Lyall MJ, et al. Performance of models for predicting 1-year to 3-year mortality in older adults: a systematic review of externally validated models. Lancet Healthy Longev. 2024;5(3):e227\u0026ndash;35. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/S2666-7568(23)00264-7\u003c/span\u003e\u003cspan address=\"10.1016/S2666-7568(23)00264-7\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAxson EL, Sundaram V, Bloom CI, Bottle A, Cowie MR, Quint JK. Hospitalisation and mortality in patients with comorbid COPD and heart failure: a systematic review and meta-analysis. Respir Res. 2020;21(1):54. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1186/s12931-020-1312-y\u003c/span\u003e\u003cspan address=\"10.1186/s12931-020-1312-y\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen W, Thomas J, Sadatsafavi M, FitzGerald JM. Risk of cardiovascular comorbidity in patients with chronic obstructive pulmonary disease: a systematic review and meta-analysis. Lancet Respir Med. 2015;3(8):631\u0026ndash;9. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/S2213-2600(15)00241-6\u003c/span\u003e\u003cspan address=\"10.1016/S2213-2600(15)00241-6\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eN\u0026uuml;esch E, Dieppe P, Reichenbach S, Williams S, Iff S, J\u0026uuml;ni P. All cause and disease specific mortality in patients with knee or hip osteoarthritis: population based cohort study. BMJ. 2011;342:d1165. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1136/bmj.d1165\u003c/span\u003e\u003cspan address=\"10.1136/bmj.d1165\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNational Institute for Health and Care Excellence. Multimorbidity: clinical assessment and management. NICE guideline [NG56]. London: NICE. 2016. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.nice.org.uk/guidance/ng56\u003c/span\u003e\u003cspan address=\"https://www.nice.org.uk/guidance/ng56\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Accessed 15 Mar 2026.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDi Angelantonio E, Kaptoge S, Wormser D, Willeit P, Butterworth AS, Banber N, et al. Association of cardiometabolic multimorbidity with mortality. JAMA. 2015;314(1):52\u0026ndash;60. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1001/jama.2015.7008\u003c/span\u003e\u003cspan address=\"10.1001/jama.2015.7008\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDumbreck S, Flynn A, Nairn M, Wilson M, Treweek S, Mercer SW, et al. Drug-disease and drug-drug interactions: systematic examination of recommendations in 12 UK national clinical guidelines. BMJ. 2015;350:h949. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1136/bmj.h949\u003c/span\u003e\u003cspan address=\"10.1136/bmj.h949\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOkura Y, Urban LH, Mahoney DW, Jacobsen SJ, Rodeheffer RJ. Agreement between self-report questionnaires and medical record data was substantial for diabetes, hypertension, myocardial infarction and stroke but not for heart failure. J Clin Epidemiol. 2004;57(10):1096\u0026ndash;103. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.jclinepi.2004.04.005\u003c/span\u003e\u003cspan address=\"10.1016/j.jclinepi.2004.04.005\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"bmc-public-health","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"pubh","sideBox":"Learn more about [BMC Public Health](http://bmcpublichealth.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/pubh/default.aspx","title":"BMC Public Health","twitterHandle":"@BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"multimorbidity, chronic disease combinations, all-cause mortality, population health, ageing cohorts, multicohort study, public health surveillance, cardiometabolic disease","lastPublishedDoi":"10.21203/rs.3.rs-9352419/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9352419/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e \u003cp\u003eMultimorbidity affects the majority of older adults worldwide and is a leading driver of mortality in ageing populations. Yet frameworks for measuring multimorbidity almost universally reduce this complexity to a disease count, treating all combinations of the same size as equivalent \u0026mdash; an assumption rarely tested across diverse countries and health systems.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eWe conducted a harmonised multicohort analysis of six longitudinal ageing studies from China, England, the United States, continental Europe, Korea, and Mexico, including 177,665 adults and 36,622 deaths. Seven chronic conditions were harmonised across cohorts: hypertension, diabetes, heart disease, stroke, lung disease, cancer, and arthritis. Pure exposures were defined as individuals with exactly the specified conditions and no others among the seven. Eligible exposures comprised 7 single conditions, 13 dyads, and 16 triads. Cohort-specific Cox models were fitted and pooled with random-effects meta-analysis. We quantified count-based gradients, within-count heterogeneity, 10-year Kaplan\u0026ndash;Meier absolute risk differences, and cross-region reproducibility using prediction intervals, leave-one-cohort analyses, and descriptive consistency classes.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eMortality increased stepwise with disease count (pure 1 disease: pooled hazard ratio [HR] 1.28, 95% CI 1.18\u0026ndash;1.39; pure 2 diseases: 1.72, 1.51\u0026ndash;1.95; pure 3 diseases: 2.29, 1.81\u0026ndash;2.88; each versus disease-free participants), but high between-cohort heterogeneity (I\u0026sup2; 78\u0026ndash;95%) indicated substantial within-count variation. Among dyads, pooled HRs ranged from 1.26 (hypertension plus arthritis) to 5.10 (heart disease plus lung disease), a 4.05-fold range within the same count class. Ten-year absolute risk differences ranged from +\u0026thinsp;8.9 (hypertension plus arthritis) to +\u0026thinsp;41.8 percentage points (heart disease plus lung disease). High-risk combinations clustered along cardiometabolic structures compounded by respiratory or cerebrovascular disease. Nineteen of 36 combinations showed HR\u0026thinsp;\u0026gt;\u0026thinsp;1 in all six cohorts; under a stricter consistency typology incorporating prediction intervals, 3 were consistently high-risk, 28 directionally consistent, and 5 unstable.\u003c/p\u003e\u003ch2\u003eConclusions\u003c/h2\u003e \u003cp\u003eDisease count captures a mortality gradient but obscures substantial variation in which combinations drive population-level burden. Among older adults with identical disease counts, which conditions co-occur is at least as important as how many. These findings support combination-specific epidemiological profiling \u0026mdash; rather than count-based summaries alone \u0026mdash; to characterise multimorbidity mortality risk across populations and health systems.\u003c/p\u003e","manuscriptTitle":"Beyond Disease Count: Which Chronic Disease Combinations Drive Mortality Risk in Older Adults? Evidence from a Harmonised Analysis of Six Ageing Cohorts","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-04-29 06:11:56","doi":"10.21203/rs.3.rs-9352419/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"editorInvitedReview","content":"","date":"2026-04-24T13:40:03+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"211481072426167080851314094461143118853","date":"2026-04-23T22:00:04+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-04-20T09:02:22+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2026-04-10T09:19:40+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-04-09T01:32:39+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-04-09T01:32:18+00:00","index":"","fulltext":""},{"type":"submitted","content":"BMC Public Health","date":"2026-04-08T06:06:03+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"bmc-public-health","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"pubh","sideBox":"Learn more about [BMC Public Health](http://bmcpublichealth.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/pubh/default.aspx","title":"BMC Public Health","twitterHandle":"@BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"5df17120-dc5a-47f7-a944-b3c5569ebc41","owner":[],"postedDate":"April 29th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-04-29T06:11:57+00:00","versionOfRecord":[],"versionCreatedAt":"2026-04-29 06:11:56","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9352419","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9352419","identity":"rs-9352419","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}