Evaluating the “holiday season effect” of hospital care on the risk of mortality from pulmonary embolism: a nationwide analysis in Taiwan

We aimed to determine whether hospital admissions during an extended holiday period (Chinese New Year) and weekends were associated with increased mortality risk from pulmonary embolism (PE), compared to admissions on weekdays. We conducted a nationwide retrospective cohort study using Taiwan’s National Health Insurance Research Database. Data of newly diagnosed PEs during the months of January and February from 2001 to 2017 were obtained from patient records and classified into three admission groups: Chinese New Year (≥ 4 consecutive holiday days), weekends, and weekdays. The adjusted odds ratios (aORs) (95% confidence intervals [CIs]) for 7-day and in-hospital mortality were calculated using multivariable logistic regression models. The 7-day and in-hospital mortality risks were higher for patients admitted during the Chinese New Year holiday (10.6% and 18.7%) compared to those admitted on weekends (8.4% and 16.1%) and weekdays (6.6% and 13.8%). These higher mortality risks for holiday admissions compared to weekday admissions were confirmed by multivariable analysis (7-day mortality: aOR = 1.68, 95% CI 1.15–2.44, P = 0.007; in-hospital mortality: aOR = 1.41, 95% CI 1.05–1.90, P = 0.022), with no subgroup effects by sex or age. Hospital admission for PE over an extended holiday period, namely Chinese New Year, was associated with an increased risk of mortality.

www.nature.com/scientificreports/ PE is a critical condition that requires a careful and experienced clinical examination to make a timely diagnosis. Recent developments in the knowledge of PE have provided detailed guidelines for the evaluation and management of PE 6 . Early intervention and diagnosis of PE can be life-saving and cost-effective, as deterioration in respiratory function with PE can lead to right ventricular strain, the requirement for ventilation, and admission to the intensive care unit. Currently, however, prompt evaluation and management of PE is not as widely emphasized as for other conditions, such as stroke or myocardial infarction. Moreover, the negative effect of weekend admissions on the outcomes of various time-sensitive medical conditions, which has been widely described [7][8][9][10] , may further impact the outcomes of PE [11][12][13][14][15][16][17][18][19][20] . A study by Nanchal et al. 15 indicated that the time to intervention is a potential mediating factor for the observed weekend effect on the outcomes of PE, which may itself be associated with the reduced number of medical staff on weekends, who often have less practical experience.
In Asia, the Chinese New Year, which is a holiday of ≥ 4 days, can prolong the negative weekend effect on the diagnosis and management of PE. During this holiday period, hospitals in countries that celebrate this holiday usually have a substantial reduction in staffing, with a corresponding shut down of certain wards. However, evidence regarding the negative impact of the weekend effect on PE is limited overall, with varying conclusions presented [11][12][13][14][15][16][17][18][19][20] . Furthermore, whether the effect on time-sensitive medical conditions, such as PE, is more evident during consecutive holidays is still unknown. Therefore, the aim of this study was to compare the mortality risk among patients admitted for PE during the Chinese New Year holiday, weekends, and weekdays, in Asia, where this issue has not been sufficiently evaluated.

Methods
Data source. Our analysis was based on data from the Taiwan's National Health Insurance Research Database (NHIRD). This database contains the health data of approximately 25 million individuals enrolled in the National Health Insurance (NHI) program. The NHI program is a universal, single-payer health insurance program in Taiwan, which was founded in 1995 and includes > 99% of the Taiwanese population. Moreover, 97% of the 480 hospitals and 22,000 clinics in Taiwan have a contract with the NHI program. The database includes individual demographic and income data, and the following health data: diagnoses, treatments, and the cost of medical (in-and out-patient) care for hospitals of all levels and medical providers. All individual identifiable information in the NHIRD is removed before release for research purposes and is encrypted under an access policy.

Ethics declarations.
Our study was conducted in accordance with the World Medical Association Declaration of Helsinki: Ethical Principles for Medical Research Involving Human Subjects. The study protocol was approved by the Hualien Tzu Chi Hospital Research Ethics Committee (IRB107-152-C). The requirement for informed consent was waived owing to the retrospective nature of the study, and the use of anonymized NHIRD data.

Study population.
We included all adult patients aged ≥ 20 years who had been admitted to hospital in January or February from 2001 through 2017, and had been diagnosed with new-onset PE (as per the International Classification of Diseases [ICD]-9 code, 415.1 or the ICD-10 code, I26). We excluded patients with missing common demographic information (such as sex and date of birth), those with a previous diagnosis of PE (prior to the index hospitalization), and those with a duration of hospitalization of > 6 months. As our study aimed to evaluate the adult population, we excluded patients aged < 20 years. In Taiwan, a person is legally considered an adult at the age of 20, according to Civil Law. The patient selection process and a brief overview of the demographic information of the excluded patients are summarized in Supplementary Fig. S1 online.
Definition of independent variables. The independent variable in the analysis was the day of admission, namely during the Chinese New Year holiday (consecutive days of holiday based on the lunar calendar), on a non-holiday weekend (Saturday or Sunday), or on a non-holiday weekday (Monday through Friday). Of note, the exact dates of the Chinese New Year differ each year due to the discrepancy between the solar and lunar calendars; however, it does fall in the months of January and February, as per the Gregorian calendar. For our study, we obtained the annual dates of the Chinese New Year from 2001 to 2017 from Taiwan's national holiday lists. In order to minimize the variation of seasonal diseases resulting from seasonal variation in mortality, we only extracted hospitalization records for the months of January and February, annually.
Outcomes. The primary outcomes were the 7-day and in-hospital mortality, defined after admission for PE. The date of death was confirmed by linking patient files in the NHIRD to the National Register of Deaths in Taiwan. The mortality risk was compared between the three admission periods (Chinese New Year holiday, weekends, and weekdays), using the weekday admission as the reference group.
Covariates and potential confounders. Information on patients' pre-existing comorbidities was obtained by retrieving individuals' ICD-9-CM diagnostic codes from in-and out-patient records. Baseline comorbidities were those diagnosed (on an in-or out-patient basis) within the 12-month period preceding the index hospitalization. Charlson Comorbidity Index scores were used to quantify the burden of comorbidities and provide the overall health status, providing a valid index of a patient's mortality risk 21 . Hospitals in Taiwan are classified into three levels (medical center, regional hospital, and district hospital), as determined by the Joint Commission of Taiwan, based on the quality of care provided, completeness of specialty services, and the number of beds. The level of a hospital and its location determine the services available and resources, both of which Statistical analyses. Baseline characteristics were compared between the three admission groups using a chi-squared test for categorical variables and analysis of variance (ANOVA) for continuous variables. A multivariable logistic regression analysis was used to calculate the odds ratios (ORs) for the in-hospital and 7-day mortality, adjusted for all baseline characteristics (Table 1). Sub-analyses stratified by sex and age were also performed. We conducted a test for interaction, using multivariable logistic regression models with the interaction terms "admission type*age group" and "admission type*sex group, " to evaluate whether our study estimates (the effect of admission type on the risk of mortality) differ significantly according to age and sex, respectively 22,23 .

Results
Characteristics of the study group. The characteristics of our study group are summarized in Table 1.
Our study group included 5637 patients admitted to participating hospitals in January or February from 2001 to 2017 with a primary diagnosis of PE. The mean age of the study population was 67 years, with women accounting for 52%. Except for diabetes mellitus, there were no significant differences in baseline characteristics across the three admission groups (Table 1).
In-hospital and 7-day mortality. The 7-day mortality rates for PE were 10.6%, 8.4%, and 6.6% for the Chinese New Year holiday, weekend, and weekday admission groups, respectively. The in-hospital mortality rates were 18.7%, 16.1%, and 13.8% for the three admission groups, respectively (Table 2). In both univariable and multivariable logistic regression models, the 7-day and in-hospital mortality risks were higher for patients admitted during the Chinese New Year holiday compared to those admitted on weekdays. After adjusting for confounders, admission during the Chinese New Year holiday was associated with a 68% higher 7-day and a 41% higher in-hospital mortality risk, with significant adjusted ORs (aORs): 7-day mortality (aOR = 1.68, 95% confidence interval [CI] 1.15-2.44, P = 0.007) and in-hospital mortality (aOR = 1.41, 95% CI 1.05-1.90, P = 0.022). The 7-day mortality (aOR = 1.29, 95% CI 1.01-1.66, P = 0.044) and in-hospital mortality (aOR = 1.21, 95% CI 1.00-1.46, P = 0.052) risks were also higher for patients admitted on weekends compared to those admitted on weekdays, although the difference was only borderline significant ( Table 2).

Sex-and age-stratified analyses and interaction tests.
The sex-and age-stratified analyses identified a consistent trend of higher mortality for patients admitted during the Chinese New Year holiday compared to weekday admissions, regardless of sex or age (Table 3), although not all comparisons were significant. The interaction tests revealed that the effect of Chinese New Year holiday admission on the risk of mortality was not significantly modified by sex or age. Similarly, there was no evidence of sex-and age-specific effects on the risk of mortality for patients admitted on weekends compared to those admitted on weekdays.

Discussion
Using a nationwide database, we found a "holiday season effect" on PE, with patients admitted with PE during the Chinese New Year holiday having a 68% higher risk of 7-day mortality, and a 41% higher risk of in-hospital mortality, compared to those admitted with PE on weekdays. We also identified a similar trend towards a higher mortality risk for patients with PE who were admitted on weekends compared to weekdays, although this "weekend effect" was only borderline significant. To our knowledge, this is the first study to specifically explore the effect of hospital admission during consecutive national holidays on the mortality risk of PE. Prior studies have reported worse outcomes for patients who were admitted for time-sensitive conditions on weekends, which constitutes the so-called "weekend effect". Nanchal et al. 16 reported an increased risk of mortality from PE for patients admitted on weekends, compared to those admitted on weekdays (OR = 1.17, 95% CI 1.11-1.22). They further explored the time to thrombolytic therapy or inferior vena cava filter placement between weekend and weekday admission groups, with longer time to treatment reported on weekends, which may explain the worse patient-related outcomes. Longer holidays may be associated with a greater impact through a similar pathway. A previous study demonstrated that consecutive days of holiday have a significant impact on the risk of mortality for patients with stroke 24 . Our present study further demonstrated the effect of an extended holiday period on the risk of mortality for patients with PE.
We also performed a subgroup analysis of patients admitted to hospital within 5 days prior to the Chinese New Year holiday (termed the pre-holiday group), as shown in Supplementary Table S1 online. This group may also be influenced by the "holiday season effect", if PE occurred during hospitalization and very close to the Chinese New Year's Eve. However, we found that the pre-holiday group did not have a higher risk of mortality than the weekend or weekday groups. Nevertheless, the Chinese New Year holiday group still had a higher risk www.nature.com/scientificreports/ of mortality than the pre-holiday group (see Supplementary Table S1 online). As noted above, PE is a critical condition that requires timely management. It was possible that most "holiday season effect" determinants, such as the timing of PE events, PE diagnosis, and time to treatment, occurred in the pre-holiday period (before the Chinese New Year), when medical resources, staffing levels, and patient load were more similar to those on normal working days. These findings further support the hypothesis that patients admitted to hospital on consecutive holiday days (namely the Chinese New Year holiday) tend to have higher risk of mortality, compared to those admitted on weekdays, weekends, or during the pre-holiday period. In addition, we observed a decrease in PE mortality over the years in Taiwan, which may be attributable to better management of PE in recent years (see Supplementary Fig. S2 online). Ideally, hospitals should have the capacity to provide the same level of care, regardless of the day of the week, hour of the day, or holidays. However, in the real world, holidays are associated with a reduction in medical and non-medical staff 25 . Consequently, physicians working during holidays are responsible for a larger patient load, compared to coverage during non-holidays and weekdays. As such, service providers are less familiar with hospitalized patients during these periods of high coverage. Moreover, with a decrease in subspecialty coverage during weekends and holidays, timely diagnosis and advanced diagnostic approaches are limited. All of these factors may contribute to the unavailability of timely intervention, which can have a negative impact on the patient outcomes of PE.
Although our study could not infer a causal relationship of the longer "holiday season effect" directly on the worse outcome of PE, we did show that the previously reported "weekend effect" on patient mortality risk was enhanced during the prolonged holiday period for the Chinese New Year in Taiwan. Knowledge of this "holiday season effect" is important to increase awareness among medical personnel to pursue advanced diagnostic modalities for patients with suspected PE, such as echocardiography and computed tomography pulmonary angiography, to lower the risk of mortality. Our findings also have implications for policy makers, highlighting the need to develop and implement strategies to maintain the quality of hospital medical care during extended holiday periods. Table 3. Sex-and age-stratified analyses of the risk of mortality from pulmonary embolism between the three admissions groups: Chinese New Year holiday, weekend, and weekday. aOR adjusted odds ratio, CI confidence interval, n number, Ref. reference. a Multivariable logistic regression model with adjustments for all covariates listed in Table 1. b The interaction test was performed to determine whether the association between the Chinese New Year holiday admission (vs. weekday admission) and the risk of mortality differed significantly according to age and sex. c The interaction test was performed to determine whether the association between weekend admission (vs. weekday admission) and the risk of mortality differed significantly according to age and sex. www.nature.com/scientificreports/ The limitations of our study need to be acknowledged and addressed in future studies. First, our analysis did not include detailed clinical data, which may influence the outcomes of PE, including body mass index, lifestyle, smoking, and alcoholism, and did not include imaging reports and laboratory tests. Further, we could not determine the severity of PE. Therefore, unknown or unmeasured confounders may have existed, leading to bias in this observational study. Second, as we used claims-based data, process indicators, such as the time to anti-coagulation therapy initiation, time to achieve target levels, and time from symptom onset to diagnosis and treatment, could not be evaluated. Information on time-specific hospital workforce, availability of interventions, and diagnostic and therapeutic procedure capacity during a period of extended holiday are also warranted to establish real causality between the "holiday season effect" and worse outcomes of time-sensitive medical conditions, such as PE.
In conclusion, this nationwide, retrospective cohort study found that hospital admission for PE during the Chinese New Year holiday was associated with a 68% higher risk of 7-day mortality, and a 41% higher risk of in-hospital mortality, compared to weekday admissions. Our findings also highlight the need to develop and implement strategies to maintain the quality of hospital medical care during extended holiday periods. Further studies are needed to explore the exact underlying mechanism and determine causality.