WHAT IS ALREADY KNOWN ON THIS TOPIC
External causes of morbidity and mortality, particularly falls and road traffic crashes, are major contributors to injury, disability and death worldwide.
However, large-scale epidemiological evidence from low- and middle-income countries on the patterns and severity of these causes has remained limited.
WHAT THIS STUDY ADDS
In a multicentre dataset (n=329 989), falls were the most frequent external cause, while road traffic crashes presented the strongest association with severe and fatal outcomes.
Mechanical forces and exposure to temperature, poison or nature contributed to the injury burden, with higher risk observed among adults and older males.
HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY
The findings highlight the need for targeted prevention strategies, including fall prevention, residential and workplace safety and traffic safety interventions, and support prioritising high-risk groups such as adult and older men, in policy and practice.
Introduction
Injuries remain a major global burden of disease, despite a 24.8% decline in disability-adjusted life-years (DALYs).1 Injuries are an important public health issue in Iran, with mortality of 4.66 per 1000 population in 2017. From 2011 to 2018, premature death accounted for a total of 8 62 542 years of life lost, with men predominating (499 300 years).2
External causes of morbidity and mortality (ECMaM) were among the top ten contributors to DALYs (2019), among individuals aged 10–49 years: road injuries (ranked first), self-harm (third) and interpersonal violence (fifth).1 Notably, they often result in major challenges for policymakers, health providers and affected families.3 Over 90% of trauma-related deaths occur in low- and middle-income countries (LMICs), driven by rapid urbanisation and motorisation.4 Disproportionate burden of injuries is placed on developing countries, as LMICs experience more than 90% of the DALYs lost.5 External causes now rank among the top three leading causes of mortality in Iran.2 These causes showed distinct demographic patterns, with road traffic crashes (RTCs) affecting young men in LMICs,6 falls among older adults in high-income countries (HICs).3 The contrast underscores the need for context-specific preventions and triage strategies, including investment in traffic safety policies and male-adult-targeted interventions in LMICs. Accordingly, understanding these nuances may improve resource allocation and emergency care triage systems.7
Injuries are likely under-reported in LMICs due to poor-quality data and limited surveillance systems, which hinder accurate estimation and underscore the need for better monitoring.8 This lack of all-inclusive injury data has been known as a critical gap and a research priority. In Iran, many studies rely on single-centre data, small-sample size and short-term designs, causing critical knowledge gaps.9 Longitudinal, large-scale data are needed to better understand local injury patterns3 and tailor interventions.
In this context, inclusive data derived from the Health Information System (HIS) and Tabriz integrated road traffic injury registry (TIRTIR) are crucial for understanding trauma-related outcomes.10 Although methodological discrepancies, inclusion criteria and case definitions hinder across-setting comparisons,3 leveraging multiple data sources can generate more comprehensive and informative injury statistics.11
Although HIS data may contain limitations regarding technical, user-related, data-quality issues and support deficiencies,12 its recent, contemporary, multicentre data and detailed outcome measures (mortality and length of stay (LOS)) provide valuable insights into ECMaM. Furthermore, using large multicentre, hospital-based datasets enhances understanding of patterns in Iran and supports targeted public health interventions to reduce incidence and improve outcomes.
To this end, this study addressed the gap on injury data in LMICs by using large population-based datasets from HIS (سیستم اطلاعات بیمارستانی) and TIRTIR to assess the epidemiology of ECMaM in the Northwest of Iran, based on the International Classification of Diseases, 10th version (ICD-10) coding.
Methodology
Study design and data collection
This population-based retrospective study was conducted using a dataset of 329 989 patients recorded in HIS and TIRTIR between March 2011 and January 2023. The primary outcome measure was the occurrence of injuries and ECMaM, as identified using ICD-10 codes. Secondary outcome measures included LOS, outcome and demographic and external causes-related predictors associated with the odds of clinical outcomes.
Data were collected from two northwestern referral trauma hospitals in Tabriz, Iran—Imam Reza and Shohada. Following institutional approval, access to the datasets was obtained on submission of an official signed request letter to the relevant authorities. Data cleaning and preprocessing were then performed using Microsoft Excel and Access 2013, including validation checks, duplicates removal and standardisation of missing value codes. Absent information was treated as missing data.
Inclusion/exclusion criteria
Inclusion criteria for the study encompassed all patients admitted to the emergency department or hospitalised in two referral trauma hospitals, from 2011 to 2023. We included all patients whose primary diagnosis codes were provided, namely S00–T88 (chapter XIX) and V00–Y99 (chapter XX), in line with international standards for trauma surveillance. Their demographic information and ICD-10 codes were documented and available through the HIS or the TIRTIR. Patients with missing data on key primary variables were excluded.
A large proportion of data recorded in the HIS was used, based on the ICD-10 classification system Chapters XIX (Injury, Poisoning and Certain Other Consequences of External Causes) and XX (External causes of morbidity and mortality). Chapter XIX comprises injuries to specific body regions (S00–S99), multiple body regions (T-codes), and head injuries (S00–S09). Chapter XX includes the codes V01–Y98.13
To address missing discharge dates for 2597 patients—resulting in LOS values exceeding 90 days—two analytic approaches were applied: (1) excluding records with LOS >90 days and (2) recoding all LOS values above 90 to 90 days. Both methods produced similar patterns of statistical significance. Therefore, the first approach was chosen (excluding LOS >90 days) to maintain data quality and minimise the influence of data entry errors or unrecorded discharges.
Patient and public involvement
No patient and public involvement.
Statistical analyses
Descriptive statistics included medians and IQR for non-normally distributed continuous variables. The skewness and kurtosis tests for normality, using the D’Agostino-Pearson omnibus test (sktest command in Stata (StataMP, version 17)),14 were statistically significant (p<0.05) for age and LOS. Therefore, non-parametric methods were used for their analysis. Categorical variables were summarised by frequencies and percentages.
A χ² test was used to compare sex distribution between outcome categories (discharged, severe, deceased). Hereafter, ‘severe’ refers to ‘patients discharged in a severe condition and in need of continued care’, and ‘deceased’ denotes in-hospital mortality (IHM). Then, pairwise χ² tests with Bonferroni correction were applied as a post hoc test for χ² with three categories. According to the adjusted alpha (α=0.05/3=0.0167), a p<0.0167 indicates statistical significance.15
Moreover, Standardised residuals were examined to determine which categories contributed most to the overall χ² result. The Mann-Whitney U test compared the distribution of age between groups with and without a specific injury type. The Kruskal-Wallis test examined differences in age and LOS across three outcome categories (discharged, severe and deceased). To further explore these differences, the Dunn’s post hoc test performed pairwise comparisons between the outcome groups and identified which specific groups differ significantly in age and LOS.
An ordered logistic regression model was applied to assess the association between an ordinal dependent variable of clinical outcomes (discharge, severe and death) and independent variables of external causes, adjusting for age and LOS groups. However, since the effect of some variables affecting outcomes differed with the level of outcomes, and the assumption of proportional odds and parallel lines was violated (χ2(8)=1168.02, Prob>χ²<0.001), a generalised ordinal logistic regression was applied.16 The model was estimated using the Stata package of gologit2 command with the autofit option, which tests and relaxes the proportional odds assumption only for variables that violate it. Coefficients (SE) with 95% CIs were reported for each outcome level. The model had two coefficient sets (discharged vs severe and death; discharged and severe vs death). Thus, the analysis identified all differences due to the impact of each independent variable on the outcome.16
Results
In total, there were 329 989 hospitalised patients, of which 194 504 (58.94%) were male. The total mean age (SD) was 36.21 (20.57). The mean age (SD; range) by sex was 37.09 (21.49; 0–99) for women and 35.60 (19.89; 0–99) for men.
The outcome was recorded for a total of 2 64 402 patients, with 258 887 (97.91%) recorded as discharged, 2653 (1.00%) as severe and 2862 (1.08%) as deceased. There was a difference in sex distribution across outcome categories (χ²(2)=276.662, p<0.001). This was confirmed by a Bonferroni corrected post hoc test (p<0.001), which revealed that male patients experienced a significantly higher rate of being recorded as severe (1.27%) compared with female patients (0.66%) (χ²(1)=231.21, p<0.001). Sex difference was significantly confirmed in IHM, with higher mortality among males (1.21%) than females (0.93%) (χ²(1)=47.48, p<0.001).
As a post hoc test, the standardised residuals indicated males had significantly more records of severe outcomes (residual=9.727) and deaths (residual=4.325), while females had fewer severe outcomes (residual=−11.499) and fewer deaths (residual=−5.114).
There was a statistically significant difference in age distribution across outcome groups (χ²=799.721, p<0.001). The Dunn’s post hoc test indicated significant differences between ‘Deceased’ (rank sum/number of observations=mean rank:171.96) versus ‘Discharged’ (mean rank:131.72) and ‘Deceased’ versus ‘severe outcome’ (mean rank: 127.83) (p<0.001), suggesting that deceased patients were significantly older than the two other groups (table 1).
Table 1
Frequency distribution of multiple injured organs based on International Classification of Diseases-version 10, among trauma patients presenting to referral hospitals in Tabriz, Iran, 2011–2023 (n=329 989)
Most patients with multiple injured organs suffered mainly from wrist/hand (24.82%), head (18.08%) and forearm/elbow (15.09%) injuries (table 1). In other words, the value for wrist/hand injury (79 528) represents the number of patients sustained wrist/hand injury in combination with injuries to one or more additional body organs, rather than isolated wrist/hand injuries.
Age distribution across outcome groups differed significantly by injured organ, including injuries to head, neck, thorax, abdomen, upper arm, forearm, wrist/hand, hip/thigh, knee/leg and ankle/foot (p<0.001). Patients with head, upper arm/shoulder, forearm/elbow and wrist/hand injuries were significantly younger than those without these injuries. In contrast, patients with neck, thorax, abdomen, hip/thigh, knee/leg and ankle/foot injuries were significantly older than their counterparts (table 1, online supplemental table).
Across all injury categories, deceased patients were significantly younger than those discharged. Severe cases were younger than discharged patients in neck, upper arm/shoulder, forearm/elbow and knee/lower leg injuries, but older in abdomen and hip/thigh injuries. Deceased patients were also significantly younger than severe cases in all injury types except neck injuries, where the difference was not significant (table 1).
There were statistically significant differences in median LOS across outcome groups for all injury types, including injuries to head, neck, thorax, abdomen, upper arm, forearm, wrist/hand, hip/thigh, knee/leg and ankle/foot (p<0.001). Using Dunn’s post hoc test, patients with severe head, neck, thorax and abdomen injuries had longer stays than discharged patients. Conversely, for upper and lower limb injuries, both severe and deceased patients typically had shorter stays (table 1).
The preliminary analysis of patients’ ICD-10 codes indicated that the majority of hospitalised patients with a single injured organ (n=6903), coded as S0–S9, suffered from head injury (59%), followed by wrist/hand (10%) (figure 1). Among male patients with single injured organs (n=4661), the most commonly affected sites were head, wrist/hand, thorax and abdomen/lower back, in descending order. In female counterparts (n=2242), the most frequent injuries involved the head, wrist/hand, abdomen/lower back and shoulder/upper arm, respectively (figure 2).
Figure 1
Distribution of single injured organs, S0–S9, based on International Classification of Diseases-version 10, among trauma patients presenting to referral hospitals in Tabriz, Iran, 2011–2023 (n=6903).
Figure 2
Distribution of single injured organs, S0–S9, by sex (male=4661; female=2242), based on International Classification of Diseases-version 10, among trauma patients presenting to referral hospitals in Tabriz, Iran, 2011–2023.
External causes of morbidity and mortality
For total injuries, the leading external causes were falling (30.22%), mechanical forces (27.79%), exposure to temperature/poison/nature (ExTPN) (18.62%) and RTCs (17.36%), in descending order. The sex, age and LOS differences were statistically significant in cases involving RTCs, falls, mechanical forces, intentional self-harm, assault, events of undetermined intent, medical and surgical complications and sequelae of external causes (table 2). The discrepancy between the total number of patients (329 989) and total number of injuries with defined external causes (282 183) reflects cases with missing documentation, unidentifiable external cause codes or those not classified under the ICD-10 external cause categories.
Table 2
Distribution of external causes of morbidity and mortality based on International Classification of Diseases-version 10, among trauma patients presenting to referral hospitals in Tabriz, Iran, 2011–2023 (n=329 989)
Table 3 presents the adjusted ORs across two outcome thresholds: (1) discharged versus more severe outcomes (severe or death) and (2) discharged or severe versus death.
Table 3
Generalised ordered logistic regression analysis of clinical outcomes by external causes of morbidity and mortality among trauma patients presenting to referral hospitals in Tabriz, Iran, 2011–2023 (n=329 989)
Female patients were significantly more likely to be discharged or to avoid death compared with males. Compared with patients 17 years and below, other age groups, particularly the older adults, were more likely to be in worse categories. They were especially more likely to die rather than being discharged in good or severe conditions. Patients with longer LOS had a higher likelihood of being in worse categories, compared with those staying <1 day. Worse outcomes increased slightly compared with <1 day. Patients with LOS ≥10 days were specifically more unlikely to have better outcomes. Falls with positive coefficients showed the likelihood of being in worse outcomes. The highest coefficients of RTCs and other land crashes indicated a much higher likelihood to lead to severe or fatal outcomes compared with other external causes. Assaults and ExTPN also led to worse outcomes but were especially less likely to result in death.
The Wald test of the parallel lines assumption indicated that the assumption was partially relaxed using the gologit2 command with the autofit option, which improved model fit (n=264 402; χ²(8)=7.85, p=0.165; Pseudo R2=0.22).
Discussion
The current hospital-based study analysed 329 989 patients’ data recorded in HIS and TIRTIR database from two northwestern Iranian referral trauma centres during 2011–2023. It contributed to identifying consistent patterns between patient characteristics, ECMaM, and clinical outcomes.
Females showed a better clinical outcome at discharge compared with males, similar to a recent study presenting males’ higher IHM.17 In line with our finding of better survival with longer hospital stays in women,18 two Finnish studies recognised male dominance in terms of overall deaths and the absolute number of IHM cases.19 20 It underscores the importance of sex-sensitive prevention strategies. Another study revealed male sex as an independent predictor of morbidity.21 However, analysing poor outcomes after excluding mortality discovered no significant difference in outcomes between sexes.22 The better clinical outcome for female compared with male patients, possibly reflects gender-based biological or healthcare access disparities, or severer injuries for males. Moreover, study location is also a determinant, since around 54.8% of DALY losses were attributed to males in Europe and Central Asia, and 49.9% in South Asia.23 Therefore, the sex differences need further investigation through stratified analyses by key factors such as age and injured organs.
Ageing as a biological process leads to reduced tissue function and a higher risk of diseases, especially after age 60.24 The study reported that the risk of worse outcomes rises further with age and is even higher in older adults. Similarly, about 80% of deaths in the USA occur among older adults, who follow different trajectories of dying, according to the research.25 This pattern suggests that adults at productive age26 and elder adults with comorbidities may be disproportionately exposed to risk factors for worse clinical outcomes.
The analysis indicated that the risk of worse outcomes rises further with longer LOS, compared with patients with LOS <1 day. However, a quasi-experimental study revealed that despite increases in LOS, significant effects on 90-day mortality were not reported.27 Longer hospitalisation did not alter the risk of IHM since it was not for clinical reasons or medical needs.27
RTCs and other land crashes, assaults, ExTPN, and falls were well-established as having much higher risks of mortality and severity compared with other causes. Consistently, unintentional injuries are among the leading causes of mortality and severity in the USA. Unintentional injuries can be prevented through awareness of risks and lifestyle changes that reduce related factors.28 This consistency across geographic and health system contexts highlights the urgent need for targeted prevention strategies—particularly in road safety enforcement, considering that RTCs will become the third main cause of mortality and disability by 2030 because of urbanisation and increased number of motor vehicles.26 High-risk environment and jobs can lead to occupational hazards including exposure to extreme temperature, poisoning substances, and assaults.29 30
External causes of morbidity and mortality
The leading external causes in our study population were falls, exposure to mechanical forces, ExTPN and RTCs. This pattern is broadly consistent with previous Iranian studies reporting similar distributions of injury mechanisms.31 The same major external causes of falls and ExTPN were identified as well.32 Additionally, Australian Hospital Statistics (1999–2000) show a similar pattern, where falls and mechanical forces dominate as ECMaM and account for a large share of injury-induced hospitalisations.33
Consistent with international studies, our findings suggest that injuries of falls, mechanical trauma, ExTPN and RTCs are affected by socioeconomic and geographic conditions, reinforcing the broader public health significance of our results.31 34 35 Similarly, a study in Canada concluded that falls were the leading cause of injury among adults and seniors.35 This consistency across diverse populations highlights the global importance of targeted prevention strategies.
Notably, our findings differentiate between the frequency of injury mechanisms and their clinical severity. Although falls were the most frequent external cause, RTCs indicated the strongest association with severe and fatal outcomes. Therefore, prevention strategies should reflect both the burden of injuries and the possibility of severe outcomes.
The predominance of falls, exposure to mechanical forces, and ExTPN as well as their higher occurrence among men may be influenced by factors such as unsafe living environments, occupational hazards and activity-related exposures.31 36 They suggest that injury prevention should seriously include not only road traffic safety but also fall-prevention programmes, residential and occupational safety. These factors are applicable to Iran, where men are often at more hazard such as more driving, workplace threats and violence, because of the Iranian socio-cultural context.36 37 A better understanding of the contributing factors could support the development of more suitable and targeted prevention programmes. Moreover, the unsafe road infrastructure and risky traffic behaviours among road users could still explain the disparity too.38
Several factors such as age, cause of injury, severity and body region are associated with hospitalisation. Older patients with more severe injuries have higher LOS than younger counterparts.39 In a LOS study, patients with severe injuries—including multiple traumas, spine, head, chest and abdomen injuries had longer LOS.40 Our findings are consistent with this pattern across similar injury regions, with the exception of upper and lower limb injuries. Extended LOS among patients with more severe injuries is expected, as these injuries often require intensive monitoring, surgical interventions, prolonged rehabilitation and carry a higher risk of complications such as infection. This explains why patients with severe injuries are particularly vulnerable to extended hospitalisations.
Strengths and limitations
This study used an adequate and representative population from trauma referral centres to address the research questions. Moreover, a large proportion of trauma data from HIS and TIRTIR databases were used which makes the study more credible and its large sample size strengthens the validity of its findings. However, the limitations of this study were the cross-sectional nature of the study design and the lack of data on alcohol consumption and smoking status. The retrospective nature of this study, with missing outcome data may introduce selection bias, especially if outcome data were more likely to be recorded for severe or deceased patients. Moreover, we only studied those injury patients who were admitted to the study hospitals. Those patients who died before hospital admission were not included in this study. Compared with other studies, the substantially larger sample size in our study may partly explain the differences in findings, as it increases the statistical power to detect even small differences that might not be identified in studies with smaller sample sizes. ICD-10-based studies are limited, leading to infrequent reporting of external causes such as ExTPN and mechanical forces.
Conclusions
Through this investigation, valuable insights into the complex interplay between external causes and injury types were gained, providing a basis for the development of evidence-based prevention and intervention strategies mitigating the impact of injuries on victims and societies. Our findings indicate falls as the most frequent external cause in the study population, while RTCs as the strongest cause associated with severe and fatal outcomes. The findings underscore the importance of comprehensive injury prevention strategies for adults and older adults, especially for men. Prevention efforts should include interventions in fall prevention, improvements in residential and workplace safety, and public awareness campaigns to address injuries related to falls and exposure to mechanical forces and ExTPN. Simultaneously, traffic safety enforcement and improvements in infrastructure and road-user behaviours remain necessary.
Data availability statement
Data are available on reasonable request. The data could be available by a request by sending email to the author HS-B.
Ethics statements
Patient consent for publication
Ethics approval
This study appreciated the ethical principles of the Declaration of Helsinki. The road traffic injury research centre ethics committee and regional ethics committee, Tabriz University of Medical Sciences, approved the research project (PERSIAN Traffic Cohort Study ethics code: IR.TBZMED.REC.1398.543; PhD thesis ethics code: ir.tbzmed.rec.1399.971, in 25 January 2021). Data with anonymous information were preserved, so informed consent was not required in this study.
Acknowledgments
The authors acknowledge the Hanse-Wissenschaftskolleg (Institute for Advanced Studies), Delmenhorst, Germany, for the fellowship grant during the preparation of this manuscript. The authors would like to thank the patients in the study.
