Data source
In South Korea, registration with the national health insurance system is compulsory for all citizens, and hence, its coverage rate is more than 99% of the country’s population. Currently, the National Health Insurance Service (NHIS) manages all databases of Korea’s health service utilization, including outpatient care, inpatient care, emergency medicine, dental care, and all other medically necessary services. The NHIS released a National Sample Cohort database (NHIS-NSC) for research purposes. It consists of 1,025,340 Koreans (about 2.2% of the total population of South Korea) as an initial 2002 cohort and has followed the subjects for 12 years (2002–2013). The cohort data were sampled systematically within 1,476 strata defined by combinations of the age, gender, eligibility status, and household income. All information on the clinical visits, hospitalization, medical treatment, and prescribed drugs are included in the NHIS-NSC. The information of the subject’s mortality, obtained from the Korean National Statistical Office, was also included. Disease diagnoses for all individuals were classified according to the International Classification of Diseases 10th (ICD-10) codes. The details of the NHIS-NSC are described in previous reports [9,10,11 ].
Informed consent was not specifically obtained individually because this study was based on the data from the NHIS-NSC. The data were fully anonymized and de-identified for the analysis. This study was approved by the institutional review board of Yonsei University Health System. The first author vouches for the integrity of the data and the accuracy of the results.
Data collection and study population
In the NHIS-NSC, we identified the patients with SCA using ICD-10 code I46.x (cardiac arrest) and I49.0 (ventricular fibrillation). The patients who were discharged from an emergency department or inpatient clinic with diagnosis code I46.x or I49.0 were included. The dead-on-arrival subjects with an underlying cause of death reported as I46.x or I49.0 were also included. We did not use the 2002 cohort data because the initial 2002 cohort started with subjects who did not die in 2002. A total of 5,675 patients with SCA were identified between 2003 and 2013. The study population included both in-hospital and out-of-hospital SCA. The patients’ age, gender, household income, and urbanization level of the region were identified. The household income levels were stratified into 10 quantiles from grade 1 to 10, and grade 0 meant medical aid beneficiaries (low household income: grade 0–3, middle household income: grade 4–7, and high household income: grade 8–10). The urbanization level of the region was classified into populated urban regions (regions with a population of more than a million) and other regions. The outcome of SCA was evaluated by the rate of survival to hospital discharge.
Definition and validation of SCA
We defined the SCA by ICD-10 codes I46.x (cardiac arrest) and I49.0 (ventricular fibrillation) after excluding non-cardiac arrest. Patients diagnosed with sudden arrest accompanied by the following diagnosis were considered non-cardiac arrest; respiratory arrest (R09.0, R09.2), gastrointestinal bleeding (I85.0, K25.0, K25.4, K26.0, K26.4, K27.0, K27.4, K92.0-K92.2), brain hemorrhage (I60.x–I62.x, S06.4–S06.6), septic shock (A41.9, R57.2), pregnancy and delivery (O00-O99), diabetic ketoacidosis (E14.1), anaphylaxis (T78.2), and accidents including asphyxiation, drowning, poisoning, traffic accident, fall, and suicide (T71, T75.1, T36–T65, V01–V99, W00–99, X60–X84).
To evaluate the accuracy of our definition of SCA, we conducted a validation study with medical records of two independent tertiary hospitals from 2009 to 2013. We found 731 patients with code I46.x or I49.0 after excluding those with diagnosis codes for non-cardiac causes, as mentioned above. Their medical records were then reviewed by five physicians, and we ascertained the patients with true SCA. The positive predictive value was 80.2% (586 of 731) by using our criteria of SCA, suggesting the good diagnostic accuracy of our definition. False-positive cases were respiratory arrest (7.0%), history of SCA (4.2%), arrest due to cancer progression (1.9%), accidents (1.8%), bleeding (1.8%), metabolic acidosis (1.0%), septic shock (0.8%), stroke (0.5%), and others (0.9%).
Statistical analysis
To evaluate the changes in the variables by the calendar year, we used the Cochran-Armitage trend test for categorical variables and linear regression for continuous variables. The age and gender standardized incidence rates per 100,000 person-years for the study population were calculated to compare the trends by years and strata with the year 2003 as the reference (direct standardization method).
To assess whether survival to hospital discharge had changed over time, multivariable Poisson regression models were constructed. The adjusted variables in the model were the age, gender, household income, and urbanization level of the region. The Poisson distribution was used to directly estimate the rate ratios instead of the odd ratios to avoid any potential exaggeration [12]. Our independent variable, the calendar year, was included as a categorical variable, with 2003 as the reference year. We multiplied the adjusted rate ratio for each year (2004 through 2013) by the observed survival rate for the reference year to obtain the yearly risk-adjusted survival rates for the study period. Those rates represented the estimated survival for each year if the patient case mix was identical to that in the reference year [5]. We also evaluated the calendar year as a continuous variable to obtain the adjusted rate ratios for the year-to-year survival trends.
All statistical analyses were conducted with the use of SPSS 20, R 3.2.1, or GraphPad Prism 6 software. All hypothesis tests were two-sided, with a significance level of 0.05.