Data sources
The data analyzed in the present study were secondary data. Data for the period from 2010 to 2018 were retrieved from the National Health Insurance Research Database for analysis. We collected data files including outpatients, emergency visits, inpatients, catastrophic illnesses, and death cause statistics. These data were compiled, and those that corresponded to the relevant variables were selected for analysis. Since the patient identifications in the National Health Insurance Research Database have been scrambled random identification numbers for insured patients by the Taiwan government for academic research use, the informed consent was waived by the Research Ethics Committee of the Taichung Jen-Ai Hospital. The research was conducted in accordance with the 1964 Declaration of Helsinki and amendments and was approved by the institutional review board (IRB) of the Taichung Jen-Ai Hospital (IRB NO. 111–20), Taiwan.
Study participants
This study included patients with AMI who visited emergency departments between 2012 and 2018. Patients were excluded if they were aged < 18 years, had been diagnosed as having MI, had committed suicide or died in an accident within 30 days after visiting the emergency department, had no record of admission after visiting the emergency department, or had not undergone PCI. The emergency department visits were defined as medical behaviors that corresponded to the billing codes 00201B, 00202B, 00203B, 00204B, and 00225B. A diagnosis of AMI was defined as a diagnosis made with the following codes: 410.0X–410.6X and 410.8X of ICD-9-CM or I21.01, I21.02, I21.09, I21.11, I21.19, I21.21, I21.29, I22.0, I22.1, I22.8, and I22.9 of ICD-10-CM. Figure 1 depicts the flowchart for patient selection.
Definition and description of variables
The definition of the variables considered in the present study are as follows. We defined 24-h PCI service as a treatment performed in a hospital offering PCI service with 24 h per day. In Taiwan, the evaluation of emergency capacity grading is conducted by the Joint Commission of Taiwan. According to the emergency medical capabilities of each hospital, they are categorized into three levels. In this study, hospitals providing 24-h PCI service were defined as Level I hospitals according to the Emergency Capability Grading. In Taiwan’s hospital accreditation system, the guidelines state that ‘Level I hospitals are required to provide 24-h cardiac catheterization services and achieve a D2B time of less than 90 min for emergency treatment of STEMI patients in over 75% of cases.’ A patient with STEMI was defined as a patient who was diagnosed with STEMI in accordance with the following the primary discharge diagnosis codes: 410.0X–410.6X and 410.8X of ICD-9-CM or I21.01, I21.02, I21.09, I21.11, I21.19, I21.21, I21.29, I22.0, I22.1, I22.8, and I22.9 of ICD-10-CM. The sex of the patients was either male or female. The patients were divided into the following age groups: < 45, 45 to 54, 55 to 64, 65 to 74 years, 75 to 84 years, and ≥ 85 years. For monthly salary, the patients were divided into the following groups: ≤ NT$17 280, NT$17 281 to NT$22 080, NT$22 081 to NT$36 300, and ≥ NT$36 301. The urbanization levels of the patients’ insured areas were divided into Levels 1 to 7; Levels 1 and 7 indicated the highest and lowest levels of urbanization, respectively36.
The hospitals investigated in the present study were classified as medical centers regional hospitals, and district hospitals. On the basis of ownership, the hospitals were classified as public and nonpublic hospitals. The comorbidity severity was estimated using Deyo’s Charlson comorbidity index (CCI). Deyo CCI is a modified version of the CCI, and it classifies comorbidities into 17 types. Patients’ primary and secondary diagnostic codes within 1 year before their visit to the emergency department were converted into weighted scores and then summed to calculate the corresponding Deyo CCI scores37. The comorbidity severity was rated on a scale from 0 to 4 points. Classification of catastrophic illnesses comprised the presence (yes) and absence (no) of any major illness. The applied triage categories were Level 1 (Code 00201B), Level 2 (Code 00202B), Level 3 (Code 00203B), Level 4 (Code 00204B), and Level 5 (Code 00225B); Level 1 patients were those who most urgently required medical attention. Administration of PCI referred to patients who underwent PCI (codes 33076B, 33077B, and 33078B) during emergency visits or hospitalization.
The annual volume of PCI used for each hospital was within one of the quartile ranges (≤ Q1, Q1 to Q2, Q2 to Q3, and > Q3) established based on the PCI reimbursement claims by hospitals across the country. MI severity was evaluated on the basis of the number of blood vessels operated during PCI; considering this, patients were stratified into the following 3 groups: 1 blood vessel (code, 33076B), 2 blood vessels (code, 33077B), and ≥ 3 blood vessels (code, 33078B). On the basis of stent use, patients were stratified into the with-stent and without-stent groups. In the relevant studies, the volume of PCI was divided into high, medium, and low levels5,7. In another study, the quartile method was used to precisely analyze the effects of the workload of admitting physicians8. The average volume of patients seen per day (PSPD) by emergency physicians was calculated by dividing the volume of reimbursement claims for emergency treatment on the day of patient admission by the number of emergency physicians on that day. Each hospital’s PSPD by emergency physicians fell into one of the quartile ranges (≤ Q1, Q1 to Q2, Q2 to Q3, and > Q3) established based on the average volume of PSPD by emergency physicians across the country on the corresponding day. The cumulative volume of reimbursement claims for PCI performed by each PCI surgeon was evaluated for the period between 2010 and the day of undergoing PCI. The volume fell into one of the quartile ranges (≤ Q1, Q1 to Q2, Q2 to Q3, and > Q3) established based on the cumulative volume of reimbursement claims for PCI performed by all PCI surgeons in Taiwan in the corresponding year. The volume of reimbursement claims for PCI performed by each PCI surgeon in the preceding year was evaluated using data corresponding to the 1-year period preceding the day of undergoing PCI. The volume would fall into one of the quartile ranges (≤ Q1, Q1 to Q2, Q2 to Q3, and > Q3) established based on the preceding year’s annual volume of PCIs performed by all PCI surgeons across the country. The annual hospital volume of PCIs performed in each hospital in the preceding year was evaluated using data corresponding to the 1-year period preceding the day of PCI and fell into one of the quartile ranges (≤ Q1, Q1 to Q2, Q2 to Q3, and > Q3) established based on the preceding year’s annual volume of PCIs performed in all hospitals across the country. In this study, important relevant variables were selected based on previous research, and potential influencing factors were included in the model to control for their effects. In terms of the patients’ conditions, we incorporated age, gender, CCI score, and other catastrophic illnesses. For socioeconomic status, we controlled for monthly salary and urbanization level. Regarding disease severity, we controlled for factors such as the number of blocked coronary vessels, stent placement, and emergency triage level. At the hospital characteristics, we controlled for whether 24-h cardiac catheterization services were available, hospital level, hospital ownership, and the previous year’s PCI institutional volume. Additionally, for the workload and experience of the cardiac catheterization operators, we controlled for accumulated PCI operator volume and the previous year’s PCI operator volume. After controlling for the above significant variables in the study, we discussed the association between emergency physicians’ workload and AMI outcomes.
Assessment of major outcomes
We primarily explored whether the average daily volume of emergency physicians affected the risks of mortality within 30 days after visiting the emergency department, revisit within 3 days after discharge from the hospital, and readmission within 14 days after discharge from the hospital in patients with AMI undergoing PCI.
Statistical analysis
The present study adopted a retrospective cohort design. SAS 9.4 (SAS Institute, Cary, NC, USA) was employed for secondary data processing and statistical analysis. The following descriptive data are presented in terms of number and percentage values: age, sex, monthly salary, urbanization level, CCI score, catastrophic illnesses, triage level, coronary stent implantation, MI severity, mortality within 30 days after visiting the emergency department visit, revisit within 3 days after discharge from the hospital, readmission within 14 days after discharge from the hospital, hospitals’ ability to offer 24-h PCI service, hospital level, hospital ownership.
Inferential statistics were used to investigate the primary outcomes. First, a chi-square test was performed to analyze the differences in the primary outcomes between subgroups with different average daily volumes of emergency physicians. Subsequently, a logistic regression model employing generalized estimating equations (GEEs) was used to compare the subgroups in terms of the primary outcomes after adjusting for covariates (eg. age, sex, monthly salary, urbanization level, CCI score, catastrophic illnesses, hospital level, hospital ownership, and triage level). The logistic regression with GEE model was also used to reduce hospital cluster effects.