International Journal of Information Management
Spatial analysis and data mining techniques for identifying risk factors of Out-of-Hospital Cardiac Arrest
Introduction
Out-of-Hospital Cardiac Arrest (OHCA) is a critical issue of emergency medical service (EMS). In the United States, annually approximately 420,000 OHCA cases occurred (Go, Mozaffarian, & Roger, 2014). Researches have revealed that the survival rate of OHCA patients is related to the location of the OHCA events (Nichol et al., 2008; Sasson, Keirns et al., 2010). One important factor related to survivability after OHCA is the presence of witnesses and timely administration of cardiopulmonary resuscitation (CPR). However, not all OHCA patients receive CPR or have a witness at the time of cardiac arrest (McNally et al., 2011). The guidelines of CPR chain on survivability from the American Heart Association stress the importance of timely administration of CPR. However, approximately 80% of OHCA and 55% of Intra-Hospital Cardiac Arrest (IHCA) adult patients still did not regain ROSC (Return of Spontaneous Circulation) after receiving CPR (Manuel, 2013). The age makeup of the OHCA patients varied from young to old. Older patients commonly also have various chronic diseases like cardiovascular disease (CVD), high blood pressure, and diabetes (Lee, 2010). A major concern with a CVD patient is delayed medical attention during the critical moment after the onset of a heart attack. Studies have shown that the survival rate of a cardiac arrest patient is significantly improved if the patient receives basic CPR within 4 min, or advanced CPR within 8 min (Valenzuela, Roe, Cretin, Spaite, & Larsen, 1997).
In Taiwan, the four major types of EMS cases are (1) motor vehicle accidents, (2) acute illness, (3) mental illness, and (4) roadside collapse. Many cases of acute illness and roadside collapse involve acute CVD, which occupies the second position among the 10 major causes of death in Taiwan. On average, one life was lost every 31 min and 50 s due to CVD in 2015, according to Ministry of Health and Welfare, Republic of China. The EMS systems of the governments at different levels have various EMS resources, such as Basic Life Support (BLS) units consisting of an ambulance with only Emergency Medical Technicians (EMT), and Advanced Life Support (ALS) units consisting of an ambulance with EMT-P (Paramedic). Some ambulances may equip with remote vital sign monitor, which can send vital signs to hospital by advanced ICT (Information and Communication Technology). There are also Patrols equipped with portable first-aid devices. They ride motorcycles to fast pass through traffic jams to the site of OHCA cases and then give a first aid before the arrival of EMT. The EMS systems can organize training and education programs of CPR and cardiac arrest (Sasson et al., 2013). They can preplace AED (Automated External Defibrillator) in suitable places, for example convenient stores, which are of a very high density in Taiwan. However, the first-aid resources are not limitless. It heavily challenges the resource allocation and management policy for EMS of a metropolis. The EMS systems have to decide how many first-aid resources are allocated in different districts. It had been proposed to organize CPR education and training programs in high-risk areas by spatial analysis (Nassel et al., 2014).
Ubiquitous computing technologies can enhance EMS very much. For example, the EMS can locate the OHCA event by the help of calling smartphones equipped with GPS sensor. Geography Information System (GIS) can provide much valuable spatial information, such as the location of the most nearby EMT-P team, suitable hospitals, and traffic condition. Spatial analysis and data mining technologies can help the administrators identify the areas of high risks and important factors to the survivability of OHCA. Thus, the administrators can allocate EMS resources to effectively improve the survival rate of OHCA patients.
Although there are geographical variations in OHCA survival rates in different cities, there also exist variations in populations (Sasson, Rogers, Dahl, & Kellermann, 2010). The frequency of bystander CPR also appears to aggregate within cities (Root et al., 2013). Districts of high risks can be defined as those of a higher prevalence of OHCA and a lower prevalence of the bystander CPR. In this study, we adopted spatial cluster analysis with a finer granularity to identify the districts of high OHCA risks. Furthermore, we use data mining technologies to find important factors to the survival rate of OHCA patients. Consequently, the EMS systems can maximize public health resources to the communities most in need.
There are various spatial analysis methods, e.g., spatial scan statistic (Kulldorff, 1997) and kernel density (Waller and Gotway, 2004). However, there is currently no consensus on how to identify districts of high risks. The main objective of this study is to propose an integrated method to identify districts of high OHCA risks and important factors to the 2-h survivability after OHCA through spatial analysis and data mining technologies. Given OHCA hotspots and information of an OHCA patient’s medical history, public health institutions may adopt suitable ICT to increase the probability of received CPR and prioritize the dispatch of emergency aid resources.
Section snippets
Data acquisition and preprocessing
This study analyzed the OHCA data registered in the Fire Department of the New Taipei City Government from January 1 to December 31, 2011.
Fig. 1 shows the Taipei metropolitan area, the largest metropolis situated in the Taipei Basin in the northern Taiwan Island. It consists of Taipei City, New Taipei City (previously Taipei County), and Keelung City. In general, Taipei City in the flat center of the basin has a higher standard of living. Keelung City is a major harbor city located in the
Population study
A simply statistic showed that most OHCA events occurred during spring and winter; and most incidents occurred during daytime hours. Most cases occurred between 7 am to 10 am, followed by 15 pm to 18 pm. OHCA occurred mainly in household residences. Most cases were males and the subjects were mostly 18–65 years old, with the mean value 64.6 and standard deviation 20.4. OHCA occurred mostly with internal medical illness; the emergency category was predominantly acute illnesses. The most
Discussion and conclusion
In this study, important OHCA first-aid factors are identified according to the chi-square analysis. These factors are types of OHCA, EMT-P dispatched, intubation, drug administration, onsite ROSC, AED usage, bystander witness, AED initial cardiac rhythm, cardiac rhythm recovered before admission, and past histories of diabetes and renal diseases. Based on the finding of this study, the quality of service of the EMS can be improved. For example, when the condition of a patient was critical, the
Acknowledgment
This study was supported by the internal research project “Multidisciplinary Health Cloud Research Program: Technology Development and Application of Big Health Data” funded by Academia Sinica. It was approved by the Institutional Review Board (IRB) of Academia Sinica (IRB#: AS-IRB01-15011).
References (26)
- et al.
Dynamic ambulance reallocation for the reduction of ambulance response times using system status management
The American Journal of Emergency Medicine
(2015) - et al.
Multiple cluster analysis for the identification of high-risk census tracts for out-of-hospital cardiac arrest (OHCA) in Denver, Colorado
Resuscitation
(2014) Cognitive informatics in image semantics description: identification and automatic pattern understanding
Neurocomputing
(2013)- et al.
Cognitive systems for intelligent business information management in cognitive economy
International Journal of Information Management
(2014) - et al.
Geographic factors are associated with increased risk for out-of hospital cardiac arrests and provision of bystander cardio-pulmonary resuscitation in Singapore
Resuscitation
(2014) - et al.
A tale of two cities: the role of neighborhood socioeconomic status in spatial clustering of bystander CPR in Austin and Houston
Resuscitation
(2013) Local indicators of spatial association—LISA
Geographical Analysis
(1995)- et al.
Relationship between renal dysfunction and outcomes in emergency department patients with potential acute coronary syndromes
Emergency Medicine Journal
(2013) - et al.
Admission glycaemia and its association with acute coronary syndrome in Emergency Department patients with chest pain
Emergency Medicine Journal
(2014) - et al.
Data mining: a tutorial based primer
(2011)
The analysis of spatial association by use of distance statistics
Geographical Analysis
Heart Disease and Stroke statistics–2014 update: a report from the American Heart Association
Circulation
Out-of-hospital cardiac arrest in Cork, Ireland
Emergency Medicine Journal
Cited by (21)
A hierarchical clustering based panel data approach: A case study of regional incentives
2022, International Journal of Information Management Data InsightsCitation Excerpt :In addition to these data-driven studies, data mining based approaches are extensively applied to several fields such as healthcare (Obenshain, 2004; Tăranu, 2015), higher education (Kumar & Chadha, 2011), bioinformatics (Raza, 2012), agriculture (Milovic & Radojevic, 2015; Patel & Patel, 2014), and weather prediction and climate change studies (Olaiya & Adeyemo, 2012). The details on applications of data mining can be found in Bole et al., Papa and Jaklič (2015), Huang et al., Liu and Chang (2012), Huang et al., Wu and Chou (2013), Mostafa and El-Masry (2013), Yan et al., Wang and Hu (2020), and Kao et al. (2017). Furthermore, review studies on data mining can be found in Al-Hashedi and Magalingam (2021), Dogan and Birant (2021), Espadinha-Cruz et al., Godina and Rodrigues (2021), and Ewieda et al., Roushdy and Shaaban (2021), Shin and Shim (2021), Xu et al., Yue and Qu (2021), Ngai et al. (2009), Otari and Kulkarni (2012) and Chen and Liu (2005).
Use of artificial intelligence in emergency medicine
2020, Artificial Intelligence in Precision Health: From Concept to ApplicationsArtificial intelligence for decision making in the era of Big Data – evolution, challenges and research agenda
2019, International Journal of Information ManagementCitation Excerpt :Nearly all of the AI systems mentioned up to the year 2000 use rule-based inference, and this element of AI techniques remains with us today; three of the specific AI systems in articles from 2017 and 2018 (Araujo & Pestana, 2017; Kao et al., 2017; Rekik, Kallel, Casillas, & Alimi, 2018) are also rule-based. The main change over time is that originally the rules were usually elicited from human experts by a human knowledge engineer, whereas now they are more likely to have been developed using an automated method such as CART (Classification and Regression Trees) (Kao et al., 2017) or association rule mining (Rekik et al., 2018). Almost as pervasive are the various methods of semantic linguistic analysis.
Spatial autocorrelation analysis of cargo trucks on highway crashes in Chile
2018, Accident Analysis and PreventionCitation Excerpt :Other studies have applied spatial statistical methods to analyze patterns of highway crashes that involve this type of vehicles (Chien et al., 2002; Eckley and Curtin, 2013; Harkey, 1999; Pulugurtha and Pasupuleti, 2013; Teoh et al., 2017; Yalcin and Duzgun, 2015). Among different existing spatial statistical methods, spatial autocorrelation analysis has been employed to characterize the degree of spatial associations in different fields such as economy (Kondo, 2015), health (Kao et al., 2017), geology (Sarp and Duzgun, 2015), and environment (Meng, 2016). This type of statistical analysis has also been widely used to identify spatial clusters in traffic safety research.