Nationwide analysis of cardiac arrest risk in patients transported from seacoast by ambulances in Japan

Introduction

Background

Seacoasts are generally considered to be dangerous locations with drownings being a typical geographical threat (1,2). Some reports have focused on the epidemiology of drowning at seacoasts (3,4). However, these reports covered a limited area, whereas others were restricted to specific age groups, particularly children (5,6). Some reports incidence of drowning based on the nationwide registry studies (7,8). Notably, drowning, other types of injuries, or medical emergencies are health threats that occur in coastal settings (9).

Japan’s emergency response approach was once primarily “scoop and run”, but has increasingly incorporated a “stay and play” approach. This study serves as foundational knowledge to consider such approach policies. Such findings may also provide useful information for the current policy of medical control systems by physicians prevalent in Japan.

Rationale and knowledge gap

Considering these situations, a summative verification based on a population-based registry, not confined to a particular region, age group, or cause of injury or illness, is needed. If the ratio of cardiac arrests among patients transported from the seacoast is greater than that of other patients, exploring this cause may broaden our understanding of the underlying factors contributing to the acute health risk associated with the seacoast and be used to promote safety. Alternatively, such acute health risks may also be linked to environmental factors, including recent global warming.

Objective

Therefore, we conducted an epidemiological survey of patients transported from seacoasts to clarify the epidemiological features of acute health risks associated with seacoasts. Using nationwide inclusive data concerning patients transported by ambulances in the emergency medical system in Japan, we described the characteristics of patients transported from seacoasts compared with those transported from other locations, with a particular focus on the rate of cardiac arrest upon arrival at hospitals. Therefore, we investigated the influence of related factors that characterize seacoast patients on the rate of cardiac arrest. This study is expected to contribute not only to Japan’s emergency medical response system but also to provide fundamental knowledge on acute health risks of the patients from the seacoast in general. We present this article in accordance with the STROBE reporting checklist (available at https://jphe.amegroups.com/article/view/10.21037/jphe-25-34/rc).

Methods

Study design and data source

This nationwide retrospective observational study was based on data provided by the Fire and Disaster Management Agency of Japan. The data were obtained from a population-based registry of patients transported to the hospital by ambulance. In this study, we enrolled all patients who were transported by ambulance between January 1, 2020, and December 31, 2023.

This study was approved by the Research Ethics Committee of Kyoto Tachibana University (No. 24-14) and was exempted from the requirement for informed consent. All research was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Study setting

The study area covered Japan, spanning 377,976 km² as of 2020 (10). Its population was 126,146,099 (61,349,581 men and 64,796,518 women) as registered in the national census in 2020 (11).

Data collection

In Japan, there were 726 fire stations with dispatch centers as of April 1, 2020. All ambulances included in this study were operated by these fire departments. The present analysis focused exclusively on patients transported by ground ambulances. In most cases, each ground ambulance is staffed by a crew of three emergency medical providers, including at least one emergency life-saving technician qualified to provide advanced life support. Among 5,270 ambulance teams nationwide, 5,241 were capable of providing advanced life support and advanced prehospital care in 2020 (12).

Patient data were categorized by the location of injury or illness occurrence and compared, with a focus on particularly seacoast versus non-seacoast areas. Regarding the location, the ambulance crew reported where the event occurred and classified it according to the master code for location (Table S1). Events were categorized as “seacoast” when they occurred outdoors in an area near the sea. Thus, the classification of “seacoast” was made essentially by exclusion of other categories. For example, a restaurant located on the beach was categorized as a public access facility rather than as seacoast. Data on patient sex and age were collected as baseline demographic data. In addition to this data, response time was collected as a time factor. This was defined as the period from the emergency call to the contact of the ambulance crew with the patient. The cause of injury or illness was also collected as one of the variables related to the outcome. Season and day of the week were collected as information on when the incidents occurred. The types of injuries or illnesses were categorized according to the 10th revision of the International Classification of Disease (13), and whether the case involved drowning was recorded.

Survival status upon arrival at the hospital was collected as outcome information. Survival status and the types of injuries or illnesses were diagnosed by the physicians in charge of initial care upon arrival at the hospital. Concerning survival status, we collected data on whether the patients were in cardiac arrest upon arrival at the hospitals. Cardiac arrest upon arrival at the hospital was defined as the absence of signs of circulation at the time of ambulance arrival at the hospital. These cases are nothing more than out-of-hospital cardiac arrest, except for patients who were successfully resuscitated until arrival at the hospital. In Japan, when ambulance crews attempt resuscitation, they always transport the patient to a hospital. If there is obvious evidence of irreversible death, such as decapitation, incineration, decomposition, rigor mortis, or dependent cyanosis, the crew does not initiate transport.

The cases with one more missing item were excluded from this study.

Statistical analyses

To elucidate the health risks associated with seacoasts, we compared the rates of cardiac arrest upon arrival in patients transported from seacoasts with those of patients from other locations. We also described other characteristics, such as sex, age, and response time. The cause of injury or illness and when these incidents occurred, such as season and day of the week, were also compared between the two groups. The cause of injury or illness was focused on whether patients were drowning or not. Regarding the seasons, we compared summer with the other seasons. Summer is defined as the period from June to August in Japan. We also compared the weekend with weekdays. The weekend was defined as Saturday and Sunday. Comparison between the two groups were evaluated using the Wilcoxon rank-sum test for continuous variables and the Pearson chi-square test for categorical variables. We first conducted logistic regression analysis among seacoast patients to identify variables associated with cardiac arrest, then conducted a logistic regression analysis to examine the associations between incident location (seacoast vs. non-seacoast) and factors such as sex, age, response time, cause of injury or illness, season, and day of the week. These factors were set as explanatory variables, and the location of the incident (seacoast vs. non-seacoast) was set as the outcome variable. Using logistic regression analysis, propensity scores were calculated for each patient based on these explanatory variables to estimate the probability that the patient was transported from the seacoast. Patients from the seacoast were matched with those from other locations based on propensity scores. One-to-one greedy nearest neighbor matching without replacement was performed (14-16) with a caliper width set to 0.2 standard deviations of the propensity score (17). The balance between the two groups was evaluated using the Wilcoxon rank-sum test for continuous variables and the Pearson chi-square test for categorical variables, following the matching. Propensity score matching was conducted via the MatchIt package (18).

To evaluate the discriminative ability of the estimated propensity scores in distinguishing between the seacoast and non-seacoast groups, we constructed a receiver operating characteristic (ROC) curve. We assessed the performance using the area under the curve (AUC). Sensitivity analyses were performed to assess the degree of unmeasured confounders on the basis of Rosenbaum’s sensitivity analysis (19). For this analysis, we used the rbounds package (20).

Propensity score matching and sensitivity analyses were conducted in R version 4.2.0 (R Foundation for Statistical Computing). All other statistical analyses were performed via IBM SPSS version 29. A two-tailed P value less than 0.05 was considered statistically significant. These analyses were also conducted as subgroup analyses restricted to patients with external causes.

Results

Patients

During the 4-year study period, 21,982,175 patients were transported by ambulance to a hospital in Japan. Among the patients, 14,142 cases were from the seacoast and 21,586,384 were from other locations. In the remaining 381,649 cases, the locations where incidents occurred were unknown. Patients with missing values, representing 0.9% of patients from the seacoast and 3.5% from other locations, were excluded from logistic regression analysis and propensity score matching (Table 1). There were no missing values in the season and on the day of the week because there was no missing information about the day when the event occurred. We also found no missing values in the cause of injury or illness. The cases in those causes were not determined and categorized as “unidentified”.

Characteristics of patients from the seacoast

The cardiac arrest rate upon hospital arrival clearly differed between the patients from the seacoast and the patients from other locations. Among patients from the seacoast, 8.7% were in cardiac arrest upon arrival at the hospital, whereas only 1.5% of patients transported from other locations were in cardiac arrest (Table 2). Patients from the seacoast also clearly differed from the other patients in terms of sex and age. Among patients transported from the sea, the male-to-female ratio was 77.2% male and 22.8% female, showing a significant disparity. In contrast, among patients transported from other locations, the ratio was nearly equal at 50.9% male and 49.1% female. The median age of patients from the coast was 49 years [interquartile range (IQR): 28–66 years]. This number was markedly lower than that of patients from non-seacoast areas, where the median age was 73 years (IQR: 48–84 years). Among the factors related to this different survival state, a longer response time was observed in patients from the seacoast than in those from other locations. The median response time for seacoast patients was 14.0 minutes (IQR: 10–20 minutes), which was longer than for patients from other locations (median: 10.0 minutes; IQR: 8–13 minutes) (Table 3). This may reflect that reaching seacoast patients often requires more time. However, the physical distance is not necessarily greater, indicating lower prehospital medical accessibility in terms of time and potentially placing these patients at a disadvantage. The patients from the seacoast were characterized by season and day of the week. Compared to other seasons, there are disproportionately more patients who were transported from the seacoast in summer. They were also more frequently transported on weekends compared to weekdays among patients from the seacoast (Table 3).

The causes of incidents also differed between seacoast and non-seacoast patients. Drowning accounted for 14.6% (2,063/14,142) of cases among seacoast patients, whereas it accounted for only 0.2% (33,640/21,586,384) among patients from other locations (Table 4). However, the proportion of cardiac arrests within drowning cases was comparable between the two groups, representing 705/2,063 (34.2%) for seacoast patients and 10,360/33,640 (30.8%) for non-seacoast patients [standardized mean difference (SMD): 0.07], although this difference was statistically significant (P<0.01) (Table 5).

The logistic regression analysis identified age, response time, season (summer vs. other seasons), day of the week (weekend vs. weekdays), and cause (drowning vs. other causes) as significant predictors of cardiac arrest among seacoast patients (Table S2).

Associations of the factors with the locations of incidents

The results of the logistic regression analysis revealed that sex, age, response time, season (summer vs. other seasons), day of the week (weekend vs. weekdays), and cause (drowning vs. not drowning) were clearly associated with the location of the incident that occurred (seacoast vs. non-seacoast; Table 6). As both univariate and multivariate analyses demonstrated, males were more likely to experience incidents at seacoasts than females were. The adjusted odds ratio (OR) for males was 3.0 (vs. females). Both the crude and adjusted ORs for age revealed that younger age was closely associated with the seacoast. There was a significant association between seacoast cases and longer emergency medical system response times. Patients from the seacoast are more likely to suffer from drowning. The adjusted OR of drowning for patients from the seacoast was 122.6 in multivariate analysis (Table 6). Regarding the season and day of the week, patients from the seacoast tended to have incidents in the summer and on weekends, respectively, compared with non-seacoast patients. The AUC in this binary logistic regression analysis was 0.80 (Figure 1).

Figure 1 ROC curve for the location of occurrence of incidents (seacoast vs. non-seacoast). ROC, receiver operating characteristic.

Cardiac arrest rate after propensity score matching

Matching based on the propensity scores calculated via logistic regression analysis revealed that 13,883 pairs were formed between patients from the seacoast and patients from other locations (Figure 2). For example, with this process, 1,880 and 1,691 drowning patients were extracted by matching for the seacoast patients and non-seacoast patients, respectively (Table 7). Those were 2,063 and 33,640 before the matching (Table 4). The variables of sex, age, response time, season, day of the week, and whether the patient suffered from drowning between the two groups were balanced via this procedure, and no significant differences were detected between patients from the seacoast location and those from other locations (Table 7).

Figure 2 Flowchart of patient selection and formation of matched pairs (seacoast vs. non-seacoast).

The difference in cardiac arrest rates upon hospital arrival decreased after matching. However, even after matching, the cardiac arrest rates remained significantly different, occurring in 8.5% of seacoast patients, whereas in non-seacoast patients, it was 3.0% (P<0.001, SMD =0.24). The result of Rosenbaum’s sensitivity analysis revealed that the upper bound of the P value exceeded 0.05 when the gamma value surpassed 2.62. In the subgroup of patients with external causes, cardiac arrest occurred in 7.5% of seacoast patients and 5.0% of non-seacoast patients (P<0.001, SMD =0.10). Logistic regression results (Table S3) and the results of propensity-score–matching (Table S4) were similar to those observed in all patients.

Discussion

Key findings

One of the major findings of our research was the notably high proportion of cardiac arrests among patients from the seacoast upon hospital arrival. The rate of cardiac arrest upon hospital arrival was 8.7% among patients from the coast, which contrasts with the 1.5% rate reported in patients from other locations (Table 2). Therefore, clarifying the background characteristics of patients from the seacoast is essential for understanding the health risks associated with the seacoast.

Furthermore, considering the good discrimination between patients from the seacoast and those from other locations in the logistic regression analysis, which was demonstrated by a sufficiently high area under the ROC curve, we expected the survival status of these patients to be equal after propensity matching. As expected, the difference in the outcomes between the two groups narrowed after propensity score matching. However, significant differences were still observed in the outcomes. Accordingly, the differences remaining after matching should be discussed in more detail.

Therefore, we conclude that patients transported from seacoast areas exhibited a clearly greater rate of cardiac arrest upon hospital arrival compared to those from other locations. Although longer response times and a higher proportion of drowning incidents among seacoast patients partially explain this increased risk, the disparity persisted even after adjusting for these factors through propensity score matching.

Strengths and limitations

The most important limitation of this study is the accuracy and quality of the recorded data. In the Japanese ambulance system, certain variables, such as the time of the emergency call, are automatically recorded. However, many variables, including other time factors, must be recorded manually by ambulance crews. Regarding the location, the classification of incident locations was determined by ambulance crews using the master code system, which is applied uniformly across all fire stations in Japan. However, some ambiguity may remain in distinguishing seacoast from adjacent inland wilderness. Missing data is inevitable in this large-scale, comprehensive survey. Our missing table shows that if one of the categories is missing in a case, this case contains two or more missing data. For example, if the location of injury occurrence or illness was missing, the percentage of missing sex data was nearly half (Table 1). The number of cases without missing data from the seacoast and other locations reached 99.5% and 98.9%, respectively. Consequently, to obtain unbiased results in our dataset, we excluded cases containing missing data from logistic regression analysis and propensity score matching.

In addition, another important limitation of this study is that the physicians’ decisions were recorded by the ambulance crew only at initial hospital care for transported patients. Determining the cause of injury or illness can be challenging in some cases, especially when internal causes are involved. However, we believe that drowning as the cause of injury was accurately decided. Survival status (cardiac arrest vs. noncardiac arrest) was also accurately determined.

Moreover, our study was based on nationwide registry data from Japan, and the research methods and analyses are highly reproducible. Whether near the sea or not, the level and protocols of emergency services are considered uniform across Japan, but individual cases have not been verified in this study. For example, ambulance equipment and crew expertise are standardized nationwide, and within each fire department headquarters, crew duty stations are rotated regularly. However, potential differences may still exist between seacoast and non-seacoast areas with respect to factors such as patient occupation or distinctions between tourists and local residents. We also could not obtain the information concerning with distance from scene to hospital. However, due to differences in healthcare systems and geographical environments among countries, caution should be exercised when generalizing these results to other regions. Furthermore, since our nationwide emergency medical data only recorded the initial diagnoses made by physicians at hospital arrival, final confirmed diagnoses might not have been fully captured. Our dataset lacked information on in-hospital mortality, neurological outcomes limiting our ability to assess final outcome by follow-up study.

Comparison with similar research

To date, the number of prior studies in the nationwide epidemiological survey of emergency cases from seacoasts has been limited. We believe that this study provides valuable data for promoting seacoast safety and contributes to the existing body of knowledge. Although these characteristics have been reported in other regions (4), it may be worthwhile to describe and confirm these findings using large-scale epidemiological data. In addition, traumatic risk among seacoasts was indicated in a previous report (21).

Explanations of findings

Demographic features, such as sex and age, of patients from the coast demonstrated unique characteristics. They had a higher proportion of males, approximately half of the patients from other locations, and they were clearly younger. However, these demographic characteristics did not explain the high percentage of cardiac arrests in patients transported from seacoasts.

Time is an essential factor in emergencies. In general, densely populated areas where emergency dispatch centers are located are distant from the coast. Therefore, response time should be highlighted, and it is essential to note that response times are often extended for patients from the coast. Patients at seacoasts might experience longer delays from the occurrence of the incident to emergency reporting, potentially due to fewer eyewitnesses and limited communication facilities at these locations, which can further exacerbate the severity of their condition upon hospital arrival. Other time factors such as scene time and transport time are also potentially critical contributors to cardiac arrest upon hospital arrival. However, the gamma value of 2.62 in the Rosenbaum sensitivity bound did not increase when these factors were added as explanatory variables. The sensitivity parameter gamma represents the maximum allowable imbalance in the odds of being transported from a seacoast location between two matched patients who are identical in observed covariates. In our matched sample, the upper bound of the significance level for the difference in cardiac arrest rates exceeded 0.05 only when gamma was greater than 2.62. This indicates that an unmeasured confounder would need to increase the odds of being in the seacoast group by more than 2.62-fold to render the association statistically non-significant. We believe that response time appears to be an appropriate variable to represent time-related factors in this study (22,23). In addition, the close relationship between drowning and patients from the seacoast accentuates the characteristics of these patients. We speculated that these two factors play a major role in explaining the high percentage of cardiac arrests upon hospital arrival.

Nonetheless, the most plausible answer is the existence of hidden variables. The cause of injury or illness factor should be reconsidered, particularly since the only factor included was whether the injury or illness involved drowning. Since the percentage of cardiac arrests due to drowning was similar between seacoast and non-seacoast patients, we adjusted for this variable by focusing on drowning cases.

Specifically, two possible points might be considered to explain the remaining difference in the percentage of cardiac arrest. However, in our results, although we had more traumatic cases from the seacoast (Table 4), the number of cases of cardiac arrests was limited (Table 5). Instead, we noted the high percentage of case occurrences and cardiac arrests in the category of “other external cause”. Although we could not clarify the detailed patient information in this category, consideration of the influence of environmental factors, such as heat stroke, might be an issue for further investigation. Another possible explanation is the high percentage of cardiac arrests due to cardiovascular diseases among the patients from the seacoast. Although patients with cardiovascular diseases transported by ambulances tend to be seriously ill in general, patients from the coast were particularly critical, with one-third experiencing cardiac arrest (Table 5). Thus, speculation regarding causes other than drowning may be needed through a more detailed epidemiological survey.

The category “other external causes” might include conditions such as heat stroke due to high environmental temperatures or physical stress caused by intense activity, both of which potentially increase the risk of cardiac arrest. Additionally, cardiovascular events occurring at the seacoast could be associated with environmental stressors, physical exhaustion, or temperature differences, warranting further investigation.

Taken together, this finding indicates that additional injuries, illnesses, or unmeasured factors may also contribute to the elevated rate of cardiac arrests in this population.

Implications and actions needed

Accordingly, Future international comparative studies would help confirm the universal applicability of our findings. Furthermore, the acute health risks we examined may also be influenced by broader environmental factors such as heavy metal air pollution (24) or climatic influences (25,26), including those related to global warming. Further epidemiological research integrating these environmental dimensions would enhance our understanding of acute health risks in diverse settings.

Since the epidemiological features of drowning patients were reported at a younger age and in males (27), an epidemiological study of patients with other causes of illness, including cardiovascular disease, may be expected in future studies. Future studies should link individual patient identification with hospital follow-up diagnostic data to enhance the accuracy of outcome measures. Operationally, given the observed higher incidence of seacoast emergency cases during the summer and on weekends, targeted resource allocation and emergency preparedness strategies are recommended for these peak periods. Efforts should also include improving public awareness of the cardiovascular risks associated with seacoast activities, ultimately aiming to reduce the incidence of cardiac arrest upon hospital arrival among patients on the seacoast. In addition, raising public awareness of the need to install automated external defibrillators (AEDs) in densely populated locations, including large bathing facilities, is essential.

Conclusions

This nationwide population-based analysis demonstrated that patients transported by ambulance from seacoast locations faced a substantially higher likelihood of cardiac arrest on hospital arrival than those originating from non-seacoast areas. Among the characteristics of patients from the seacoast, longer response times and a higher percentage of drowning cases contributed to a high percentage of cardiac arrest on hospital arrival. However, even after adjusting these factors through propensity score matching, the rate of cardiac arrest remained high among the patients from the seacoast. Thus, other injuries or illnesses may also contribute to a higher percentage of cardiac arrests in these patients.

Acknowledgments

We thank the Ambulance Service Planning Office of the Fire and Disaster Management Agency for providing the database. We would like to thank Editage (www.editage.com) for its English language editing services.

Footnote

Reporting Checklist: The authors have completed the STROBE reporting checklist. Available at https://jphe.amegroups.com/article/view/10.21037/jphe-25-34/rc

Data Sharing Statement: Available at https://jphe.amegroups.com/article/view/10.21037/jphe-25-34/dss

Peer Review File: Available at https://jphe.amegroups.com/article/view/10.21037/jphe-25-34/prf

Funding: None.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://jphe.amegroups.com/article/view/10.21037/jphe-25-34/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. This study was approved by the Research Ethics Committee of Kyoto Tachibana University (No. 24-14). The ethics committee waived the need for informed consent for this study.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Koon W, Peden A, Lawes JC, et al. Coastal drowning: A scoping review of burden, risk factors, and prevention strategies. PLoS One 2021;16:e0246034. [Crossref] [PubMed]
2. Abelairas-Gómez C, Tipton MJ, González-Salvado V, et al. Drowning: epidemiology, prevention, pathophysiology, resuscitation, and hospital treatment. Emergencias 2019;31:270-80.
3. Bessereau J, Fournier N, Mokhtari T, et al. Epidemiology of unintentional drowning in a metropolis of the French Mediterranean coast: a retrospective analysis (2000-2011). Int J Inj Contr Saf Promot 2016;23:317-22. [Crossref] [PubMed]
4. da F. Klein AH. Analysis of Hazards Associated with Sea Bathing: Results of Five Years Work in Oceanic Beaches of Santa Catarina State, Southern Brazil. J Coast Res 2003;107-16.
5. Webb AC, Wheeler A, Ricci A, et al. Descriptive Epidemiology of Pediatric Drowning Patients Presenting to a Large Southern US Children's Hospital. South Med J 2021;114:266-70. [Crossref] [PubMed]
6. Cohen N, Scolnik D, Rimon A, et al. Childhood Drowning: Review of Patients Presenting to the Emergency Departments of 2 Large Tertiary Care Pediatric Hospitals Near and Distant From the Sea Coast. Pediatr Emerg Care 2020;36:e258-62. [Crossref] [PubMed]
7. Claesson A, Krig A, Jonsson M, et al. Incidence and characteristics of drowning in Sweden during a 15-year period. Resuscitation 2021;162:11-9. [Crossref] [PubMed]
8. Lunetta P, Haikonen K. Hospital admissions for non-fatal drowning in Finland, 2002-2023: a nationwide population-based register study. Inj Epidemiol 2025;12:37. [Crossref] [PubMed]
9. Kelly S, Daw S, Lawes JC. Beyond drowning: Characteristics, trends, the impact of exposure on unintentional non-drowning coastal fatalities between 2012 and 22. Aust N Z J Public Health 2024;48:100113. [Crossref] [PubMed]
10. Geospatial Information Authority of Japan. The Report of Statistical reports on the land area by prefectures and municipalities in Japan in 2020. Ministry of Land, Infrastructure and Transport of Japan, Japan. 2020. Available online: https://www.gsi.go.jp/KOKUJYOHO/MENCHO/backnumber/GSI-menseki20201001.pdf. Accessed 8 June 2025.
11. Statistics Bureau of Japan. 2020 Population Census. Ministry of Internal Affairs and Communications of Japan. 2020. Available online: https://www.e-stat.go.jp/en/stat-search/files?page=1&layout=datalist&toukei=00200521&tstat=000001136464&cycle=0&year=20200&month=24101210&tclass1=000001136466. Accessed 8 June 2025.
12. Fire and Disaster Management Department. Emergency Medical Service System. In: White Paper on Fire Service. Fire and Disaster Management Department, Tokyo. 2021. Available online: www.fdma.go.jp/publication/hakusho/r2/items/r2_all.pdf#page=230. Accessed 5 Oct 2025.
13. World Health Organization. ICD-10: international statistical classification of diseases and related health problems: tenth revision. World Health Organization, Geneva. 2004. Available online: https://iris.who.int/handle/10665/42980. Accessed 8 June 2025.
14. Rosenbaum PR, Rubin DB. Constructing a Control Group Using Multivariate Matched Sampling Methods That Incorporate the Propensity Score. Am Stat 1985;39:33-8.
15. Austin PC. A comparison of 12 algorithms for matching on the propensity score. Stat Med 2014;33:1057-69. [Crossref] [PubMed]
16. Cenzer I, Boscardin WJ, Berger K. Performance of matching methods in studies of rare diseases: a simulation study. Intractable Rare Dis Res 2020;9:79-88. [Crossref] [PubMed]
17. Austin PC. Optimal caliper widths for propensity-score matching when estimating differences in means and differences in proportions in observational studies. Pharm Stat 2011;10:150-61. [Crossref] [PubMed]
18. Ho D, Imai K, King G, et al. MatchIt: Nonparametric Preprocessing for Parametric Causal Inference. J Stat Softw 2011;42:1-28.
19. Rosenbaum PR. Sensitivity analysis for certain permutation inferences in matched observational studies. Biometrika 1987;74:13-26.
20. Keele LJ. Perform Rosenbaum Bounds Sensitivity Tests for Matched and Unmatched Data. 2022. Available online: https://cran.r-project.org/web/packages/rbounds/rbounds.pdf. Accessed 5 Oct 2025.
21. Thom O, Roberts K, Leggat PA, et al. Cervical spine injuries occurring at the beach: epidemiology, mechanism of injury and risk factors. BMC Public Health 2022;22:1404. [Crossref] [PubMed]
22. Baiocchi M, Woo YJ, Chiu P, et al. The role and significance of sensitivity analyses in enhancing the statistical validity of clinical studies. J Thorac Cardiovasc Surg 2022;163:749-53. [Crossref] [PubMed]
23. Becker SO, Caliendo M. Sensitivity Analysis for Average Treatment Effects. The Stata Journal 2007;7:71-83.
24. Kaziród-Wolski K, Sielski J, Jóźwiak M, et al. Does PM 2.5 and PM 10-associated heavy metals affect short-term and long-term survival after out-of-hospital cardiac arrest? Four-year study based on regional registry. Minerva Med 2024;115:14-22. [Crossref] [PubMed]
25. Ho AFW, Ho J, Ong M, et al. Influence of ambient temperature and absolute humidity on sudden cardiac arrest in Singapore: a nationwide time-series study. J Epidemiol Community Health 2025;79:954-9. [Crossref] [PubMed]
26. Burger M, Ristau P, Bohn A, et al. Air temperature and humidity impact out-of-hospital-cardiac-arrests in Germany: A 10-year cohort study from the German Resuscitation Registry. Resusc Plus 2024;20:100750. [Crossref] [PubMed]
27. Erasmus E, Robertson C, van Hoving DJ. The epidemiology of operations performed by the National Sea Rescue Institute of South Africa over a 5-year period. Int Marit Health 2018;69:1-7. [Crossref] [PubMed]