Diabetes-associated short-term mortality after first-onset acute myocardial infarction: age- and residence-specific patterns in an underserved region

Abstract

Background

Urban–rural disparities in acute myocardial infarction (AMI) outcomes are well documented, and may be related to delayed emergency access and uneven service distribution. However, the ways in which structural barriers may amplify vulnerabilities associated with age at AMI onset and diabetes in underserved settings remain understudied, despite their importance for equity-oriented emergency care planning.

Methods

We conducted a 16-year retrospective cohort study of 1,591 first-onset AMI patients who were identified through the sole AMI-capable medical centre in Eastern Taiwan. The primary endpoint was overall short-term mortality (30-day mortality), which was further categorized into pre-hospital and post-arrival 30-day mortality. The interaction between age and diabetes was first examined to assess differential short-term mortality risks. To contextualize these effects within structural healthcare disparities, patients were further stratified by urban or rural residence. Adjusted odds ratios were estimated across four strata defined by age group (younger vs. older) and residence (urban vs. rural) to assess how structural disparities relate to mortality patterns associated with age at AMI onset and diabetes.

Results

Short-term mortality rose stepwise across groups (young non-diabetes 1.9%, young diabetes 8.2%, old non-diabetes 10.7%, old diabetes 17.9%). Diabetes was associated with higher short-term mortality in both age groups, with a stronger association among younger patients (aOR: 3.83 vs. 1.60; p-for-interaction = 0.045). Among younger patients, diabetes was particularly associated with pre-hospital mortality (aOR: 6.32, 95% CI 1.21–32.88), suggesting a greater vulnerability during the early emergency phase. In residence-stratified analyses, the association between diabetes and short-term mortality was most pronounced among younger rural patients (aOR: 4.24, 95% CI 1.21–14.84). Exploratory subgroup analyses suggested longer onset-to-presentation intervals in some rural subgroups; however, these findings were based on limited event counts and should be interpreted cautiously.

Conclusions

Diabetes was associated with higher short-term mortality after first-onset AMI, particularly among younger patients and those living in rural areas. These findings identify a high-risk subgroup in an underserved region. Future studies incorporating direct systems-of-care measures are needed to determine whether geographic and emergency-care access factors contribute to this diabetes-related vulnerability.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12889-026-28437-3.

Introduction

Acute myocardial infarction (AMI), a severe manifestation of cardiovascular disease (CVD), is a time-critical emergency in which short-term outcomes are shaped by both biological vulnerability and the organization of emergency care systems [1]. Despite therapeutic advances [1], AMI continues to impose a considerable population burden, particularly in underserved regions. Younger adults typically exhibit a survival advantage following AMI; however, the increasing prevalence of diabetes mellitus (DM) at younger ages has been associated with high early mortality and atypical presentations [2, 3], potentially narrowing this advantage during the critical early phase of care. In underserved settings, these vulnerabilities may be further shaped by structural barriers to timely emergency care.

Beyond individual-level risk factors, structural determinants, especially the distribution and timeliness of emergency cardiac services, critically influence AMI outcomes. Urban–rural disparities in short-term mortality are well documented and may be related to delayed transport, diagnostic delays, and unequal access to reperfusion therapy. Clinical guidelines recommend a 30-minute door-to-needle target for thrombolysis, with pre-hospital administration further reducing delays and mortality [4]. However, these targets are often difficult to achieve in underserved regions due to healthcare system fragmentation. Additionally, while most studies focus on urban settings or in-hospital outcomes, how structural barriers interact with specific clinical vulnerability profiles, such as younger age at onset combined with diabetes, to shape early survival following AMI in structurally underserved settings remains unclear.

Eastern Taiwan (Hualien and Taitung) exemplifies an underserved setting where these constraints are salient. The region has a high diabetes burden and a lower life expectancy than the national average [5, 6]. Specialist capacity is limited, transport from remote communities often exceeds one hour, and mountainous terrain further complicates medical access. With only one AMI-capable medical centre located in urban Hualien City, urban–rural residence serves as a contextual proxy for geographic accessibility to emergency care [7]. These geographic and service-distribution constraints may contribute to attenuation of the survival advantage typically observed in younger adults with diabetes, particularly during the pre-hospital phase when system-level delays exert disproportionate effects on survival.

Age is a non-modifiable determinant of AMI outcomes and is central to Taiwan’s ongoing demographic shift, which combines rising early-onset diabetes with rapid population aging [8]. AMI incidence rises with age; individuals older than 60 years experience approximately a 2.5-fold higher incidence [1], and mortality risk is elevated in older adults [9]. Conversely, early-onset DM shortens life expectancy by 6–14 years among individuals diagnosed at 30–50 years [10]. Prolonged hyperglycaemia may drive earlier, more aggressive AMI in younger adults, while age-related physiological changes compound risk in older adults [11, 12]. Moreover, DM-induced endothelial dysfunction may synergize with limited access to secondary prevention in structurally disadvantaged populations, potentially amplifying early mortality risk among younger adults. Although diabetes has been linked to worse AMI outcomes in younger adults [13], little is known about how structural barriers modify these risks during the early, time-critical phase of care in underserved settings. Addressing this gap is critical for designing equity-oriented emergency planning and identifying vulnerable subgroups.

This study investigates whether the association between diabetes and short-term AMI mortality differs by onset age and whether these patterns vary across urban–rural structural contexts in Eastern Taiwan. In this study, structural disparities were operationalised using urban–rural residence as a proxy for geographic accessibility and access to time-critical AMI services. We hypothesised that the association between diabetes and short-term mortality would differ by age at AMI onset and may vary across urban–rural contexts, particularly during the pre-hospital phase, with the greatest excess risk in rural areas. By identifying equity-relevant high-risk subgroups, this study aims to provide evidence for designing equitable, system-level interventions to improve cardiovascular outcomes in structurally underserved regions.

Methods

Study design

This retrospective cohort study investigated how structural disparities modify short-term mortality patterns associated with age at AMI onset and diabetes. Specifically, we first examined the moderating effect of age at AMI onset on the relationship between DM and short-term mortality, and subsequently contextualized these effects within the structural urban–rural disparities characteristic of Eastern Taiwan. Data were obtained from medical records of first-onset AMI patients at Hualien Tzu Chi Hospital, which managing approximately 90% of AMI cases in eastern Taiwan. The study spanned 1 January 2000 to 31 December 2015, when standardised AMI diagnostic criteria and reperfusion strategies (e.g., percutaneous coronary intervention [PCI]) were widely adopted. AMI was diagnosed by cardiovascular specialists based on clinical assessment, electrocardiogram (ECG) changes, and elevated cardiac enzymes. Patients with incomplete data were excluded to define the complete-case analytic cohort, resulting in 1,591 eligible patients. A locator map (Fig. 1) illustrates geographic access disparities in Eastern Taiwan.

Fig. 1.

Locator map of Taiwan. Eastern Taiwan (Hualien & Taitung) is shown in green; other counties in grey. The red star marks the sole AMI-capable medical centre in urban Hualien City; black dots mark regional hospitals. Administrative boundaries: GADM v4.1, level-2 (WGS84), CC BY 4.0; processed with mapshaper

Study population

Patients were stratified by age at AMI onset into young (< 65 years) and old (≥ 65 years) groups, aligning with World Health Organization (WHO) criteria for old and European Cardiovascular Disease Statistics defining 65 years as the threshold for premature acute coronary syndrome [14, 15]. Each age group was further subdivided by DM status, defined by a prior physician-documented diagnosis and/or use of insulin or oral antihyperglycaemic agents, or by plasma glucose ≥ 126 mg/dL recorded in the medical record, according to US CDC guidelines [16]. Given that glucose measurements were obtained in the acute AMI setting, documented medication history was considered alongside glucose values to mitigate potential misclassification due to stress hyperglycaemia.

Study variables

Outcomes

The primary outcome was overall short-term mortality, defined as death within thirty days of AMI onset, regardless of whether death occurred before or after hospital arrival. To capture distinct phases of risk, mortality was categorized as pre-hospital and post-arrival 30-day mortality (excluding pre-hospital deaths from post-arrival analyses). Pre-hospital mortality was ascertained from routinely documented emergency department and referral records within the hospital information system. For this study, pre-hospital mortality included deaths documented before formal admission, whereas deaths after admission were classified as post-arrival 30-day mortality. Accordingly, this outcome represents hospital-documented pre-hospital mortality prior to admission, rather than a comprehensive capture of all out-of-hospital AMI fatalities in the community. Survival was defined as living beyond 30 days after AMI onset.

Covariates (risk factors)

Covariates were selected to adjust for demographic, metabolic, behavioural, and cardiovascular confounders potentially associated with short-term mortality:

• Demographic variables: sex (male or female); age (years); residential area (urban or rural; see Appendix 1, Appendix Fig. 1 and Appendix Fig. 2 in Supplementary Materials for detailed regional classification), was treated as a proxy for geographic accessibility and structural context for time-critical AMI emergency care; onset season (hot: April–October; cold: November–March).

• Metabolic risk factors: hypertension (ICD-9-CM 401–405) [17]; dyslipidaemia (ICD-9-CM 272.x) [18]; obesity (BMI ≥ 27 kg/m² according to Taiwan’s BMI classification) [19].

• Behavioural risk factors: cigarette use, betel nut chewing, and alcohol consumption (as recorded in medical records).

• Cardiovascular-related variables: ECG presentation (NSTEMI vs. STEMI); number of diseased vessels (0–1 vs. 2–3; ≥50% stenosis in the Left Main Coronary Artery (LMCA), Right Coronary Artery (RCA), or Posterior Descending Artery (PDA), indicating coronary disease severity); Time from symptom onset to hospital presentation (≥ 12 h vs. < 12 h), calculated using two timestamps recorded in the medical records: (1) symptom onset time, documented in emergency department physician notes or referral records (typically based on patient or family report), and (2) hospital presentation time, recorded as the emergency department triage time. This ≥ 12-hour cutoff was chosen because fibrinolytic therapy provides the greatest mortality reduction when administered within 12 h of symptom onset [20–22].

Statistical analysis

Patient characteristics were summarized as mean ± standard deviation (SD) for continuous variables and as counts (n) with percentages (%) for categorical variables. Standardized mean differences (SMD) were used to assess group differences, with values exceeding 0.1 indicating clinically significant differences [7, 23]. Short-term mortality rates were stratified by age and DM status, and compared using the chi-square test.

The primary endpoint was overall short-term mortality, further categorized into pre-hospital and post-arrival 30-day mortality. To examine whether the association between DM and short-term mortality differed by age, binary logistic regression models were fitted, including an interaction term between age group (< 65 vs. ≥65 years) and DM status. The interaction between onset age group and DM was prespecified based on the study hypothesis; all other interaction tests and subgroup analyses were considered exploratory.

Multinomial logistic regression was then applied to assess whether these associations varied across different types of short-term mortality (overall, pre-hospital and post-arrival 30-day mortality), with survivors as the reference group. The multinomial models were adjusted for DM status, age, sex, residential area, and onset season, hypertension, dyslipidaemia, obesity, smoking, betel nut use, alcohol consumption, STEMI, number of diseased vessels, and time from symptom onset to hospital presentation.

Additional exploratory subgroup analyses were conducted to determine if the association between DM and short-term mortality, as well as the moderating effect of AMI onset age, differed by the presence or absence of other risk factors.

Finally, to examine whether these patterns varied by geographical context, analyses were stratified by age and residence (Young–Urban, Young–Rural, Old–Urban and Old–Rural), and adjusted odds ratios (aORs) with 95% confidence intervals (CIs) were estimated within each stratum.

To evaluate the potential impact of missingness, we conducted sensitivity analyses including summaries of variable-level missingness patterns overall and by outcome status, comparisons of included versus excluded records and evaluation of the DM effect between the available and excluded missing cases, and multiple imputation analyses; detailed imputation procedures are provided in the supplementary materials. To address low event counts and potential (quasi-)separation in the exploratory subgroup analyses, we refitted all models using Firth’s penalized logistic regression as a sensitivity analysis to evaluate the robustness of the estimates. Analyses were performed using IBM SPSS Statistics Version 26.0 (IBM Corp., Armonk, NY, USA). Two-sided p < 0.05 was considered statistically significant.

Reporting guideline adherence

This manuscript follows STROBE guidelines [24] and checklist [25], used during drafting and editing, as detailed in the supplementary materials.

Results

Patterns of clinical vulnerability and structural healthcare disparities were jointly associated with short-term AMI outcomes in this cohort, indicating that both biological and structural factors contributed to inequitable early mortality patterns.

Cohort selection

A total of 3,005 first-time AMI patients were identified (Fig. 2). After excluding those with incomplete data on mortality (n = 35), DM (n = 176), and other variables (n = 1,203), 1,591 patients remained. Patients were grouped into young (< 65 years, n = 702) and old (≥ 65 years, n = 889) groups. In the young group, 281 had DM and 421 did not; in the old group, 420 had DM and 469 did not.

Fig. 2.

Participants flow chart

Baseline characteristics

The cohort’s mean age was 65.5 ± 13.53 years; 71.0% were male and 44.1% had DM. The prevalence of diabetes was higher in old than young patients (59.9% vs. 40.1%). Table 1 summarises baseline characteristics by DM status and age group. Clinically significant differences (SMD ≥ 0.1) were evident across several variables. In both age strata, the proportion of males was markedly lower in the DM group (young: 75.4% vs. 84.3%; old: 55.2% vs. 70.4%). Young diabetic patients were older on average (55.3 vs. 51.3 years), whereas old diabetic patients had a slightly lower mean age (75.0 vs. 75.7 years). Diabetics had higher prevalences of hypertension (young: 74.0% vs. 49.9%; old: 82.9% vs. 68.7%) and dyslipidaemia (young: 56.6% vs. 51.8%; old: 43.1% vs. 33.9%). Behavioural factors also differed: diabetics reported less smoking (young: 51.2% vs. 63.4%; old: 23.8% vs. 29.9%) and alcohol consumption (young: 39.9% vs. 50.1%; old: 15.7% vs. 23.5%). Betel nut use was more frequent among non-diabetics, with a significant difference in young patients (26.8% vs. 21.0%). STEMI was more common in non-diabetics (young: 59.1% vs. 48.4%; old: 48.8% vs. 38.1%), whereas diabetics more often had multi-vessel disease (young: 45.2% vs. 20.9%; old: 50.0% vs. 31.6%). Young diabetics had a significantly longer delay from symptom onset to presentation (≥ 12 h: 44.1% vs. 38.7%). These patterns suggest that younger diabetic patients face overlapping clinical and structural vulnerabilities.

Table 1.

Characteristics of first-onset AMI patients by diabetes status and age group

Characteristics	Young (n = 702)			Old (n = 889)
	With DM (n = 281)	Without DM (n = 421)	Standardized Mean Differences (SMD)	With DM (n = 420)	Without DM (n = 469)	Standardized Mean Differences (SMD)
Demographic variables
Male, n (%)	212(75.4)	355(84.3)	-0.22	232(55.2)	330(70.4)	-0.32
Age (years), mean ± SD	55.3 ± 7.05	51.3 ± 9.02	0.49	75.0 ± 6.55	75.7 ± 7.10	-0.10
Urban area, n (%)	155(55.2)	243(57.7)	-0.05	220(52.4)	236(50.3)	0.04
Onset during cold season, n (%)	132(47.0)	179(42.5)	0.09	198(47.1)	209(44.6)	0.05
Metabolic risk factors
Hypertension, n (%)	208(74.0)	210(49.9)	0.51	348(82.9)	322(68.7)	0.34
Dyslipidaemia, n (%)	159(56.6)	218(51.8)	0.10	181(43.1)	159(33.9)	0.19
Obesity, n (%)	118(42.0)	165(39.2)	0.06	105(25.0)	105(22.4)	0.06
Behavioural risk factors
Smoking, n (%)	144(51.2)	267(63.4)	-0.25	100(23.8)	140(29.9)	-0.14
Betel nut chewing, n (%)	59(21.0)	113(26.8)	-0.14	42(10.0)	54(11.5)	-0.05
Alcohol consumption, n (%)	112(39.9)	211(50.1)	-0.21	66(15.7)	110(23.5)	-0.20
Cardiovascular-related variables
STEMI, n (%)	136(48.4)	249(59.1)	-0.22	160(38.1)	229(48.8)	-0.22
The number of diseased blood vessels, n (%)			-0.53			-0.38
0–1	154(54.8)	333(79.1)		210(50.0)	321(68.4)
2–3	127(45.2)	88(20.9)		210(50.0)	148(31.6)
Time from symptom onset to hospital presentation, n (%)			-0.11			-0.09
<12 h	157(55.9)	258(61.3)		204(48.6)	248(52.9)
≥12 h	124(44.1)	163(38.7)		216(51.4)	221(47.1)

Note: Standardized Mean Differences (SMD) compares AMI patients with DM versus without DM within each onset age group (young: <65 years; old: ≥65 years); SMD ≥0.1 indicates clinically significant differences

DM Diabetes mellitus, STEMI ST-Elevation Myocardial Infarction

Interaction between age and diabetes on short-term mortality

Overall, 156 patients (9.8%) died within 30-day of first AMI, 98 of these deaths (62.8%) occurred in patients with DM. Short-term mortality rose stepwise across groups: 1.9% in young non-DM, 8.2% in young DM, 10.7% in old non-DM, and 17.9% in old DM (Supplementary Figure S1). In both age strata, mortality was higher among diabetic patients (young: p < 0.001; old: p = 0.002). Decomposition of short-term mortality revealed higher 30-day than pre-hospital mortality in both age groups (young non-DM 1.4% vs. 0.5%; young DM 5.0% vs. 3.2%; old non-DM 8.1% vs. 2.6%; old DM 13.3% vs. 4.5%) (Supplementary Figure S2). Pre-hospital deaths accounted for 26.9% (42/156) of all short-term deaths overall and 39.1% (9/23) among younger patients with diabetes. Phase-specific deaths within 30 days (pre-hospital vs. post-arrival) are summarized in Supplementary Figure S3. In multivariable models, DM was associated with higher overall short-term mortality in both age strata, with a stronger effect in young patients (aOR: 3.83, 95% CI: 1.583–9.29, p = 0.003) than old patients (aOR:1.60, 95% CI: 1.02–2.41, p = 0.027). The DM × age-group interaction for overall short-term mortality was significant (p-for-interaction = 0.045), indicating effect modification by age at AMI onset (Table 2).

Table 2.

Effect of diabetes on pre-hospital and post-arrival 30-day mortality by age group

	Pre-hospital mortality					Post-arrival 30-day mortality
	Number of patients died					Number of patients died
Onset age group	With DM n(%)	Without DM n(%)	aOR(95% CI)#	p-value	P-value (interact)†	With DM n(%)	Without DM n(%)	aOR(95% CI)#	p-value	P-value (interact)†
Young (n = 702)	9(3.2)	2(0.5)	6.32(1.21,32.88)	0.029*	0.179	14(5.0)	6(1.4)	3.14(1.10,8.95)	0.033*	0.243
Old (n = 889)	19(4.5)	12(2.6)	1.75(0.81,3.80)	0.155		56(13.3)	38(8.1)	1.55(0.97,2.47)	0.067

Note: Model adjusted for age, sex, residential area, onset season, hypertension, dyslipidaemia, obesity, smoking, betel nut use, alcohol consumption, STEMI, number of diseased vessels, and time from symptom onset to hospital presentation. Patients who died pre-hospital were excluded from the post-arrival 30-day mortality analysis. The chi-square test indicated a significant difference in pre-hospital and post-arrival 30-day mortality between patients with and without DM in both age groups: p < 0.001 for the young group and p = 0.009 for the old group

P-for-interaction (†) represents the interaction effect between onset age group and DM on pre-hospital and post-arrival 30-day mortality

DM Diabetes Mellitus, aOR adjusted Odds Ratio, CI Confidence Interval

*p < 0.05

#without DM was selected as the reference group for calculating odds ratio

When overall mortality was partitioned, DM remained associated with higher risk among younger patients for both pre-hospital (aOR: 6.32, 95% CI: 1.21–32.88, P = 0.029) and post-arrival 30-day mortality (aOR: 3.14, 95% CI: 1.10–8.95, P = 0.033) (Table 2). In contrast, risks in the old group were not statistically significant (pre-hospital mortality: aOR: 1.75, 95% CI: 0.81–3.80, P = 0.155; post-arrival 30-day mortality: aOR: 1.55, 95% CI: 0.97–2.47, P = 0.067). Interactions between age group and DM were not significant for pre-hospital and post-arrival 30-day mortality (p-for-interaction: 0.179 and 0.243, respectively), likely due to limited events (thirty-one deaths in the young group). Nevertheless, the consistent direction of estimates highlights possible early-phase vulnerability among younger diabetic patients.

Geographic disparities and structural context

Structural disparities further modified the relationship between diabetes and short-term mortality (Fig. 3). Across all age–geography strata, patients with diabetes exhibited higher aOR of short-term mortality compared with their non-diabetic counterparts. The highest aOR was observed in the Young–Rural group (aOR 4.24, 95% CI 1.21–14.84), followed by Young–Urban (aOR 2.74, 95% CI 0.70–10.75), Old–Rural (aOR 2.14, 95% CI 1.22–3.76), and Old–Urban (aOR 1.53, 95% CI 0.78–3.00). This gradient suggests diabetes disproportionately increased early mortality among younger patients, particularly those residing in rural areas.

Fig. 3.

Forest plot of diabetes and overall 30-day mortality after AMI by age–geography groups. Forest plot showing adjusted odds ratios (aORs) with 95% confidence intervals (CIs) for the association between diabetes mellitus (DM) and short-term mortality after acute myocardial infarction (AMI), stratified by age–geography groups. Sample sizes and death counts are indicated for each stratum

Subgroup analyses

As shown in Table 3, exploratory subgroup analysis suggested broadly consistent directions of association between DM and short-term mortality across several subgroups. However, some estimates, particularly among younger patients and sparse strata such as those without smoking, were based on small cell counts and wide confidence intervals; therefore, these results should be interpreted as exploratory and hypothesis-generating rather than confirmatory, and should be used primarily to inform future studies with larger event counts. The impact of DM was more pronounced in young patients, especially in those without hypertension (aOR:6.21 vs. 0.84, p-for-interaction = 0.019), without smoking (aOR:21.68 vs. 1.63, p-for-interaction = 0.015), without alcohol consumption (aOR:11.63 vs. 1.73, p-for-interaction = 0.015), or multi-vessel disease (aOR:5.76 vs. 1.15, p-for-interaction = 0.048). Elevated risks were observed in both young urban (aOR: 5.47) and rural (aOR: 3.75) residents (both p < 0.05).

Table 3.

Effect of diabetes on short-term mortality by age group: overall and subgroup analyses

Characteristics		Young (n = 702)				Old (n = 889)
		Number of patients died				Number of patients died				P-value(interact)†
		With DMn(%)	Without DMn(%)	aOR(95% CI)	p-value	With DMn(%)	Without DMn(%)	aOR(95% CI)	p-value
All subjects		23(8.2)	8(1.9)	3.83(1.58,9.29)	0.003*	75(17.9)	50(10.7)	1.60(1.02,2.41)	0.027*	0.045*
Demographic variables
Sex	Male	19(9.0)	8(2.3)	4.27(1.84,9.94)	0.001*	41(17.7)	36(10.9)	1.75(1.08,2.84)	0.022*	0.073
	Female	4(5.8)	0(0.0)	-	-	34(18.1)	14(10.1)	1.97(1.01,3.84)	0.043*	0.997
Residential area	Urban	13(8.4)	4(1.6)	5.47(1.75,17.10)	0.001*	49(22.3)	28(11.9)	2.13(1.28,3.53)	0.003*	0.138
	Rural	10(7.9)	4(2.2)	3.75(1.15,12.24)	0.020*	26(13.0)	22(9.4)	1.43(0.79,2.62)	0.240	0.156
Onset season	Cold	11(8.3)	3(1.7)	5.33(1.46,19.52)	0.005*	32(16.2)	21(10.0)	1.73(0.96,3.11)	0.067	0.121
	Hot	12(8.1)	5(2.1)	4.15(1.43,12.04)	0.005*	43(19.4)	29(11.2)	1.91(1.15,3.19)	0.012*	0.198
Metabolic risk factors
Hypertension	Yes	17(8.2)	5(2.4)	3.65(1.32,10.08)	0.008*	67(19.3)	31(9.6)	2.24(1.42,3.53)	0.001*	0.390
	No	6(8.2)	3(1.4)	6.21(1.51,25.51)	0.004*	8(11.1)	19(12.9)	0.84(0.35,2.03)	0.701	0.019*
Dyslipidaemia	Yes	12(7.5)	5(2.3)	3.48(1.20,10.08)	0.015*	31(17.1)	16(10.1)	1.85(0.97,3.52)	0.060	0.319
	No	11(9.0)	3(1.5)	6.61(1.81,24.18)	0.001*	44(18.4)	34(11.0)	1.83(1.13,2.97)	0.013*	0.069
Obesity	Yes	13(11.0)	5(3.0)	3.96(1.37,11.44)	0.007*	15(14.3)	8(7.6)	2.02(0.82,4.99)	0.122	0.344
	No	10(6.1)	3(1.2)	5.51(1.49,20.34)	0.004*	60(19.0)	42(11.5)	1.80(1.18,2.77)	0.006*	0.111
Behavioural risk factors
Smoking	Yes	6(4.2)	7(2.6)	1.62(0.53,4.90)	0.393	16(16.0)	10(7.1)	2.48(1.07,5.72)	0.030*	0.547
	No	17(12.4)	1(0.6)	21.68(2.84,165.18)	0.001*	59(18.4)	40(12.1)	1.63(1.06,2.52)	0.026*	0.015*
Betel nut chewing	Yes	3(5.1)	1(0.9)	6.00(0.61,59.00)	0.083	6(14.3)	1(1.9)	8.83(1.02,76.51)	0.020*	0.809
	No	20(9.0)	7(2.3)	4.26(1.77,10.25)	0.001*	69(18.3)	49(11.8)	1.67(1.12,2.48)	0.011*	0.057
Alcohol consumption	Yes	6(5.4)	6(2.8)	1.93(0.61,6.14)	0.256	6(9.1)	6(5.5)	1.73(0.54,5.62)	0.354	0.896
	No	17(10.1)	2(1.0)	11.63(2.65,51.10)	0.001*	69(19.5)	44(12.3)	1.73(1.15,2.61)	0.008*	0.015*
Cardiovascular-related variables
Presentation on electrocardiography	STEMI	12(8.8)	7(2.8)	3.35(1.29,8.71)	0.009*	33(20.6)	27(11.8)	1.94(1.12,3.39)	0.018*	0.336
	NSTEMI	11(7.6)	1(0.6)	14.04(1.79,110.09)	0.001*	42(16.2)	23(9.6)	1.82(1.06,3.13)	0.029*	0.060
The number of diseased blood vessels	0–1	8(5.2)	6(1.8)	2.99(1.02,8.76)	0.037*	32(15.2)	23(7.2)	2.33(1.32,4.11)	0.003*	0.689
	2–3	15(11.8)	2(2.3)	5.76(1.28,25.86)	0.011*	43(20.5)	27(18.2)	1.15(0.68,1.97)	0.600	0.048*
Time from symptom onset to hospital presentation	< 12 h	11(7.0)	6(2.3)	3.16(1.15,8.74)	0.020*	30(14.7)	24(9.7)	1.61(0.91,2.85)	0.101	0.255
	≥ 12 h	12(9.7)	2(1.2)	8.63(1.89,39.29)	0.001*	45(20.8)	26(11.8)	1.97(1.17,3.34)	0.010*	0.072

Note: Subgroup analyses are exploratory and intended to describe potential heterogeneity; estimates may be imprecise due to limited events. Model adjusted for age, sex, residential area, onset season, hypertension, dyslipidaemia, obesity, smoking, betel nut use, alcohol consumption, STEMI, number of diseased vessels, and time from symptom onset to hospital presentation. The chi-square test indicated a significant difference in overall short-term mortality between patients with and without DM in both age groups: p < 0.001 for the young group and p = 0.002 for the old group

P-for-interaction (†) represents the interaction effect between onset age group and DM on overall short-term mortality

DM Diabetes Mellitus, aOR adjusted Odds Ratio, CI Confidence Interval, STEMI ST-Elevation Myocardial Infarction

—: Not reported because zero cell counts resulted in complete separation in logistic regression. Sensitivity analyses using Firth’s penalized logistic regression are provided in Supplementary Table S6; estimates were generally directionally consistent, although significance differed in some sparse-data subgroups

*p < 0.05

In old patients, stronger DM-mortality associations appeared in selected subgroups; those with hypertension (aOR: 2.24, 95% CI: 1.42–3.53, p = 0.001), without dyslipidaemia (aOR: 1.83, 95% CI: 1.13–2.97, p = 0.013), without obesity (aOR: 1.80, 95% CI: 1.18–2.77, p = 0.006), without alcohol consumption (aOR: 1.73, 95% CI: 1.15–2.61, p = 0.008). DM also conferred higher risk in patients with 2–3 diseased vessels (aOR: 2.33, 95% CI: 1.32–4.11, p = 0.003), with presentation ≥ 12 h after symptom onset (aOR: 1.97, 95% CI: 1.17–3.34, p = 0.010), or during hot season (aOR: 1.91, 95% CI: 1.15–3.19, p = 0.012). Significant DM-related risk was observed among old urban residents (aOR: 2.13, 95% CI: 1.28–3.53, p = 0.003), with broadly similar patterns by sex (aOR: 1.75 for male and aOR:1.97 for female, both p < 0.05), smoking, betel nut chewing, and ECG presentation.

Sensitivity analyses

This exclusion rate (47.1%) primarily reflected missing non-clinical variables such as behavioural factors, typical in emergency data collection contexts, and was addressed through sensitivity analyses. Missingness patterns (overall and by outcome status) are summarised in Supplementary Table S1; missingness was predominantly observed in behavioural and angiographic variables, with minimal missingness in core demographics and the outcome. Differences between included and excluded records are quantified in Supplementary Table S2. Excluded records had a higher observed short-term mortality and were older, consistent with potential non-random missingness in emergency-care data, whereas diabetes status was broadly comparable between groups. Baseline characteristics stratified by diabetes status and onset age group of included versus excluded cases were similar (Table 1 and Supplementary Table S3). In unadjusted comparisons limited to variables available in excluded cases, the DM–mortality associations were directionally consistent with the main analyses for overall short-term, pre-hospital, and post-arrival 30-day mortality (Supplementary Table S4). Results from multiple imputation sensitivity analyses were also consistent with the complete-case estimates (Supplementary Table S5). In exploratory subgroup analysis with sparse events or (quasi-)separation, Firth penalized logistic regression (Supplementary Table S6) yielded effect directions consistent with the corresponding conventional subgroup results (Table 3), with differences mainly in statistical significance.

Discussion

This study explores how age at AMI onset influences the link between DM and short-term mortality in Eastern Taiwan, a geographically underserved region with centralized healthcare resources and complex topographic barriers [26]. By linking clinical vulnerability profiles to structural healthcare disparities, we found that the usual short-term survival advantage of younger AMI patients with diabetes was markedly diminished, with the strongest associations observed during the early phase of care and among rural residents. These findings may inform efforts to integrate cardiovascular emergency planning with broader rural health equity strategies.

The prespecified analysis showed that DM significantly increases short-term mortality following AMI, with a more pronounced effect in young patients, as evidenced by a significant age interaction. By contrast, phase-specific differences in pre-hospital and post-arrival 30-day mortality were directionally similar but based on fewer events, and the corresponding interaction tests were not statistically significant. These phase-specific findings should therefore be interpreted cautiously. Young patients with DM had higher pre-hospital and post-arrival 30-day mortality, a pattern not observed in old patients. Although AMI is less common in young populations, their expected survival advantage is attenuated in the presence of diabetes [27–29], likely due to atypical presentations, such as dyspnoea or reduced pain sensitivity, which may relate to diagnostic delays, and larger infarct sizes associated with hyperglycemia [30]. In contrast, higher short-term mortality in old adults likely reflects age-related physiological decline and comorbidities, which increase overall risk but may attenuate the independent effect of DM. This pattern is biologically and clinically plausible and may reflect additional diabetes-related vulnerabilities. Taken together, these vulnerabilities may help explain why the attenuation of the usual survival advantage appears more evident in the pre-hospital phase, when delays in symptom recognition and access to timely care may disproportionately affect early survival.

The stratified analyses suggest a possible role for structural healthcare disparities in shaping these patterns. Among the four age and geography strata, younger rural patients with diabetes experienced the greatest excess mortality risk, followed by younger urban, older rural, and older urban groups. This gradient should be interpreted in the context of Eastern Taiwan’s geography, where mountainous terrain and centralized specialized cardiac services may contribute to prolonged symptom-to-treatment times. However, the present dataset could not distinguish transport, referral, Emergency Medical Services (EMS), or in-hospital treatment delays. Such systemic barriers disproportionately affect younger diabetic patients [31], potentially attenuating the early survival advantage typically associated with younger age. Structural disparities may also influence current urban–rural heterogeneity through differential access to newer cardioprotective diabetes therapies (e.g., SGLT2 inhibitors or GLP-1 receptor agonists), which have been shown to improve cardiovascular outcomes in patients with diabetes [32]. Such access differences may remain relevant to contemporary public health planning in underserved regions.

These findings are consistent with prior Asian studies, such as Song et al. (2022) [13] which examined age specific differences in long term outcomes among AMI patients with diabetes. Unlike those studies, which primarily focused on mid to long term mortality, our analysis concentrates on the early post event period and distinguishes pre-hospital and post-arrival 30-day mortality as distinct clinical phases. This temporal lens is particularly important because geographic factors may have a greater impact before hospital arrival, especially in underserved settings where challenges in emergency access have been documented. This pre-hospital interval may reflect both delay in help-seeking and delay in reaching definitive care; these components could not be distinguished in the present dataset. Furthermore, our study incorporates Asia-specific risk factors [33], including betel nut chewing, a known diabetogenic risk with high prevalence in Eastern Taiwan, urban–rural disparities, and symptom-to-hospital delay. Such contextual factors are often underrepresented in Western cohorts, and their inclusion provides a more nuanced understanding of diabetes related vulnerability in AMI. The implications extend beyond Taiwan to other rural and under-resourced areas in Asia, where limited access to timely cardiac care and similar geographic variations may contribute to comparable disparities [34].

Exploratory subgroup analysis suggested possible heterogeneity in the association between DM–mortality across clinical subgroups. However, several subgroup estimates, particularly among younger patients, were based on sparse events and wide confidence intervals. These findings should therefore be interpreted cautiously as hypothesis-generating signals rather than evidence of independent subgroup-specific effects. DM’s effect was particularly pronounced in young patients without traditional risk factors, such as those without hypertension or smoking, suggesting an independent role of DM in early-onset AMI mortality. Conversely, in old patients, DM’s effect was often complicated by additional risk factors, including those with hypertension or longer delays from symptom onset to hospital presentation. These findings suggest that the role of environmental factors, where rural patients may experience potential transport delays associated with geographic distances [31], which could contribute to higher pre-hospital mortality in young DM patients. Urban patients, despite better access, showed elevated risk, possibly due to socioeconomic and lifestyle factors like sedentary behavior [35].

From a public health perspective, these findings illustrate how structural and clinical vulnerabilities intersect to produce inequitable outcomes in time sensitive cardiovascular emergencies. Eastern Taiwan provides a relevant setting in which geographic barriers and centralized healthcare infrastructure may intersect with age- and disease-related vulnerability. Addressing these inequities may be considered through targeted structural interventions at multiple levels. First, enhancing public awareness of atypical symptoms among younger diabetic adults may improve timely help seeking and support earlier presentation [36, 37]. Second, expanding pre-hospital emergency capacity in rural areas, including community-based triage and the use of telemedicine, may reduce delay [38, 39]. Third, integrating cardiovascular and diabetes care into primary health services, such as universal screening in rural areas could improve early risk identification and reduce system level delays during emergencies [40].

These observations align with the United Nations Sustainable Development Goals (SDGs), particularly SDG 3 (Good Health and Well-Being), which calls for reducing premature mortality from non-communicable diseases and achieving universal health coverage (UHC); and SDG 10 (Reduced Inequalities), which emphasizes eliminating geographic and structural disparities in access to essential health services. By addressing system-level inequities in emergency cardiovascular care, this study may contribute evidence relevant to SDG-oriented health system reforms, not only in Taiwan but also across other Asian regions facing similar geographic and structural healthcare challenges.

Strengths and limitations

This study is the first to examine the moderating effect of AMI onset age on DM in short-term mortality, distinguishing pre-hospital and post-arrival 30-day mortality as distinct phases, using 16 years of data (2000–2015) from eastern Taiwan’s only AMI treatment centre. This study period also provides a historical benchmark for understanding structural healthcare inequality in an underserved region. It uniquely includes preventive parameters like betel nut chewing and urban-rural disparities, offering insights relevant to Taiwan and similar Asian settings.

However, several limitations exist. First, the absence of DM severity, HbA1c levels at admission, cardiovascular medication data, diabetes duration, and diabetes subtype (Type 1 vs. Type 2) may confound and underestimate DM’s effect; diabetes misclassification due to stress hyperglycaemia in the acute AMI setting is also possible. Similarly, key clinical severity markers and more granular systems-of-care metrics—including Killip class, renal function, heart failure status, comorbidity burden, reperfusion strategy, EMS response time, and door-to-balloon/needle times—were not consistently available and could not be adjusted for. These data constraints may have resulted in residual confounding, particularly in residence-stratified comparisons, and prevent the definitive attribution of observed patterns to specific mechanisms such as transport barriers or in-hospital care differences. Second, the small number of pre-hospital deaths limited statistical power for interaction analyses. Third, missing data due to AMI’s emergency nature (e.g., body weight) was mitigated by sensitivity analysis, which showed consistent trends; nevertheless, selection bias due to non-random missingness cannot be fully excluded. Fourth, patient- or family-reported, chart-documented symptom-onset times are prone to measurement error and may be differentially misclassified by residential context. Moreover, time from symptom onset to hospital presentation represents an aggregate pre-hospital interval and cannot distinguish symptom-recognition delay, help-seeking delay, EMS activation or response time, transport delay, referral delay, or in-hospital reperfusion delay. Therefore, associations involving onset-to-presentation delay and rurality should be interpreted cautiously. Finally, as a hospital-identified cohort, deaths occurring entirely outside the hospital documentation pathway were not captured. This may have led to an underestimation of rural pre-hospital mortality and could have biased urban–rural comparisons. Accordingly, these findings should be interpreted as associations observed within the documented hospital/referral pathway rather than as population-level estimates. Urban–rural residence therefore served as a contextual proxy for geographic accessibility and regional service distribution, rather than as a direct measure of access-to-care mechanisms.

In Eastern Taiwan, a geographically and medically underserved region with a high diabetes burden, younger AMI patients with diabetes showed a diminished usual short-term survival advantage, especially in the pre-hospital phase. These findings suggest that structural and geographic barriers may be relevant to this diabetes-related vulnerability, possibly through transport-related barriers and longer symptom onset-to-hospital presentation intervals. Future studies incorporating more granular systems-of-care measures are needed to clarify these potential pathways, and equity-oriented, community-based public health strategies may be needed to reduce geographic inequities and improve cardiovascular outcomes in resource-limited settings.

Abbreviations

AMI

Acute myocardial infarction

aOR

Adjusted odds ratio

BMI

Body mass index

CDC

Centers for Disease Control and Prevention

CI

Confidence interval

CVD

Cardiovascular disease

DM

Diabetes mellitus

ECG

Electrocardiogram

ICD-9-CM

International Classification of Diseases, Ninth Revision, Clinical Modification

IRB

Institutional Review Board

LMCA

Left main coronary artery

NSTEMI

Non-ST elevation myocardial infarction

PCI

Percutaneous coronary intervention

PDA

Posterior descending artery

RCA

Right coronary artery

SD

Standard deviation

SMD

Standardized mean difference

STEMI

ST elevation myocardial infarction

WHO

World Health Organization

Authors’ contributions

D.J.Y.T. serves as the guarantor for this study. All authors contributed to the conception and design of the study. D.J.Y.T., T.-C.H., J.-C.H., and C.-W.L. were responsible for the acquisition, analysis, and interpretation of data. D.J.Y.T., T.-C.H., and C.-W.L. drafted the manuscript and were involved in its critical revision. All authors reviewed and approved the final manuscript.

Funding

This research was funded by Hualien Tzu Chi Hospital, Buddhist Tzu Chi Medical Foundation, grant number TCMMP 106-01. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Data availability

The data that support the findings of this study are derived from medical records of Hualien Tzu Chi Hospital and are not publicly available due to institutional regulations and patient privacy considerations. Data access may be granted to qualified researchers upon reasonable request to the corresponding author.

Declarations

Ethics approval and consent to participate

This study was conducted in accordance with the principles of the Declaration of Helsinki and was approved by the Research Ethics Committee at Buddhist Tzu Chi General Hospital (IRB106-118-B). The requirement for informed consent was waived by the ethics committee because this was a retrospective study using medical record data.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.