Effects of the COVID‐19 pandemic on heart failure hospitalizations in Japan: interrupted time series analysis

Introduction

The Coronavirus Disease 2019 (COVID-19) pandemic has had unprecedented effects on health care systems and patient care delivery worldwide.1 In addition to the direct involvement of COVID-19 with cardiovascular organs, the pandemic disrupted health care systems for acute and chronic conditions, such as acute myocardial infarction and heart failure (HF), at the patient and health care provider levels.2-5 Recent studies from the EU6-9 and the USA10, 11 have revealed a substantial decrease in hospitalization rates for acute myocardial infarction with a sustained decrease in the proportion of urgent percutaneous coronary interventions and increased risk-adjusted mortality rates.

Although decreased hospitalizations for acute coronary syndrome have been reported during the COVID-19 pandemic, the implications for hospitalizations due to the more severe HF are not well understood. The EU,12, 13 the USA,14-17 and other regions18, 19 have reported reduced hospitalizations due to HF. There is, however, a paucity of data on the trends and management of HF cases during the pandemic from Japan and East Asia, the region that was affected earlier during the COVID-19 pandemic.

In addition, concerns have been raised regarding a potential increase in susceptibility to COVID-19 from using angiotensin-converting enzyme inhibitors (ACEIs) or angiotensin receptor blockers (ARBs) and subsequent up-regulation of angiotensin-converting enzyme 2.20-22 On the other hand, the abrupt withdrawal of ACEIs and ARBs from patients with HF may result in decompensated HF and adverse outcomes.23-25

Therefore, we sought to define changes in the volume of urgent HF hospitalizations from the COVID-19 state of emergency in Japan via an interrupted time series analysis. We also investigated changes in the prescription of ACEIs and ARBs, and in-hospital mortality during the COVID-19 pandemic.

Methods Data source and study participants

This retrospective cohort study analysed data extracted from the Diagnosis Procedure Combination (DPC) database collected from April 2018 to September 2020. The DPC/per-diem payment system (PDPS) is a Japanese prospective payment system applied to acute care hospitals. There were 1730 hospitals adopting the DPC/PDPS in 2018, which accounted for 54% (482 618 out of 891 872) of all general beds in Japanese hospitals in 2018. The DPC data consist of claims and discharge summaries, including the International Classification of Diseases, Tenth Revision (ICD-10) codes for classifying the primary diagnosis, the cause of admission, the most and second-most medical-resource-intensive diagnoses, up to 10 comorbidities, and 10 complications. The DPC data also contain codes of all services and medications provided during each hospitalization as well as PDPS information. In the present study, administrative data from 583 hospitals were available during the COVID-19 pandemic period. For evaluation, we included all hospital inpatients that provided DPC data continuously during the study period.

The study period started on 1 April 2018 (1st week) and ended on 4 July 2020 (118th week); all inpatients whose admission date was within this period were included. The ICD-10 classifications were used for defining comorbidities (Supporting Information, Table S1). This study included patients aged 18 years or older with a diagnosis of HF. We defined HF cases as ICD-10, I50.x with unscheduled admissions. We determined unscheduled admissions, other than planned or scheduled hospitalizations, using a specific DPC variable. The recalibrated Elixhauser Comorbidity Index was calculated based on a previously described coding algorithm.26-28 This study was conducted in accordance with the principles of the Declaration of Helsinki, and the study was approved by the Ethics Committee of Kyoto University Graduate School and Faculty of Medicine (approval number: R0135) with a waiver of informed consent prior to data extraction from the database.

The primary outcome measure was the change in the weekly number of HF hospitalizations before and after Japan's COVID-19 state of emergency. Secondary outcome measures included the prescription rates ACEIs and ARBs, and in-hospital mortality among HF patients.

Statistical analysis

Continuous variables were expressed as the median and the interquartile range. Comparisons between the two groups (before and after the state of emergency) for continuous variables were assessed by Mann–Whitney U tests. Categorical data were compared using χ2 or Kruskal–Wallis tests, as appropriate. Statistical significance was set at P < 0.05 (two-tailed). Statistical analyses were performed with R Version 4.0.2 (R Foundation for Statistical Computing, Vienna, Austria) using R packages ‘forecast’, ‘tsModel’, ‘tseries’, ‘foreign’, ‘lmtest’, ‘Epi’, ‘splines’, ‘vcd’, ‘ggplot2’, ‘nlme’, ‘car’, and ‘MASS’.

Interrupted time series analyses

Interrupted time series analysis was used to compare HF hospitalizations before and after the COVID-19 state of emergency. Weekly numbers of HF hospitalizations are presented as dots in Figure 1. Segmented regression analysis was used to verify changes in the trend and level over time.29 A 1 week time unit was selected in order to provide optimal precision to the model. Because the Japanese government announced the state of emergency on 7 April 2020, the first week of April 2020 was used as the identified breakpoint. Accordingly, HF cases were divided into one of two periods for analysis, either before (the first week of April 2018 to the first week of April 2020) or after (the second week of April 2020 to the first week of July 2020) the state of emergency was declared. The number of HF admissions in the post-state of emergency period was compared with the expected estimates had the state of emergency not occurred to determine the state of emergency's effect on hospitalizations. Segmented regression analysis was performed to identify the level and trend changes in weekly HF hospitalizations during the pandemic, with consideration given to time dependence in the model. Seasonal variation was taken into account by including harmonic terms (sines and cosines).

image

Interrupted time series analysis for HF cases per week. The number of HF cases per week is plotted on the y-axis and the date (year and month) on the x-axis. The red dots show the number of HF cases. The solid lines show the predicted trends based on the seasonally adjusted regression model before (blue) and after (red) April 2020. The grey dashed line shows the 95% confidence interval. The red dashed line shows the 95% prediction interval. The horizontal lines represent the linear predicted values through the study period (dashed, blue) and after the state of emergency was declared (solid, red). The vertical dotted line denotes the declaration of the state of emergency in Japan. HF, heart failure.

The changes in weekly proportions of in-hospital mortality and prescription of ACEIs and/or ARBs over time were also assessed by interrupted time series analysis with identical conditions (time unit, breakpoint, grouping periods, and segmented regression model) as described above. For the evaluation of in-hospital deaths and drug prescriptions, HF cases per week were included as an offset term to adjust for any potential changes in HF inpatient cases over time. The validity of the model was assessed with correlograms (autocorrelation and partial autocorrelation functions) and residual plots.

Secondary and sensitivity analysis

Secondary analysis was performed between 2018 and 2019 by the interrupted time series analysis with a breakpoint of April 2019 to ascertain that April 2020 have a different impact due to the COVID-19 pandemic. Another secondary analysis for the evaluation of HF cases per month, the Japanese population per month was included as an offset term to adjust for any potential changes in the Japanese population over time. Sensitivity analysis was done by changing the break point and evaluating the segmented regression model for each break point. Candidate break points were as follows: (i) 11 March 2020, the World Health Organization has declared the COVID-19 pandemic. (ii) End-March, 2020, in Japan, the number of COVID-19 infections substantially increased in this period. (iii) 7 April 2020, the Japanese government declared the state of emergency.

There are two outlier points in the data. These points in time correspond to the public holiday at the end of the year in Japan. It is known that hospitalization for cardiovascular diseases increases during this period every year. Thus, we conducted a sensitivity analysis controlling for these wild data points. The point can be controlled for in the model by entering an indicator variable that has value 1 in that period and 0 in all others.

Results Overall study population

There were 213 784 HF cases requiring hospitalization during the study period [age; 83.0 [75.0, 89.0] years; sex (male); 110 317 cases (51.6%)], among which 90.9% of cases were admitted before Japan's state of emergency was declared, with only 19 363 cases (10.1%) admitted thereafter (Table 1). Although patient demographics and drug prescriptions were consistent across the periods, patients with HF hospitalized during the COVID-19 period (after the state of emergency was declared) were more frequently males, had dyslipidaemia as a comorbidity, and were prescribed ACEIs and/or ARBs, beta-blockers, and anti-aldosterone blockers. The duration of hospitalization for HF was similar throughout the entire study period, with a median of 17.0 days. In-hospital mortality was 10.5% (22 547 cases) across the total cohort, with 20 555 deaths (10.6%) before the state of emergency and 1992 deaths (10.3%) thereafter (Table 1).

Table 1. Patient characteristics All HF (N = 213 784) Before COVID-19 (n = 194 421) After COVID-19 (n = 19 363) P-value Demographics Age, yearsa 83.0 [75.0, 89.0] 83.0 [75.0, 89.0] 83.0 [75.0, 89.0] 0.35 Sex (male), n (%) 110 317 (51.6) 100 122 (51.5) 10 195 (52.7) <0.01 Comorbidities Hypertension, n (%) 113 363 (53.0) 103 008 (53.0) 10 355 (53.5) 0.19 DM, n (%) 63 729 (29.8) 58 009 (29.8) 5720 (29.5) 0.40 Dyslipidaemia, n (%) 49 289 (23.1) 44 666 (23.0) 4623 (23.9) <0.01 Smoking history, n (%) 45 207 (41.3) 30 319 (41.3) 14 888 (41.2) 0.71 Prior MI, n (%) 17 119 (8.0) 15 624 (8.0) 1495 (7.7) 0.13 Stroke, n (%) 3797 (1.8) 3466 (1.8) 331 (1.7) 0.48 PAD, n (%) 501 (0.2) 450 (0.2) 51 (0.3) 0.39 Previous heart failure, n (%) 19 212 (9.0) 17 443 (9.0) 1769 (9.1) 0.45 Elixhauser Comorbidity Index 5 [0, 7] 5 [0, 7] 5 [0, 7] 0.25 Medications ACEI/ARB, n (%) 111 563 (52.2) 101 293 (52.1) 10 270 (53.0) 0.01 Beta-blocker, n (%) 132 658 (62.1) 120 225 (61.8) 12 433 (64.2) <0.01 Aldosterone receptor antagonist, n (%) 98 404 (46.0) 89 167 (45.9) 9237 (47.7) <0.01 Duration of hospitalization, days 17.0 [11.0, 28.0] 17.0 [11.0, 28.0] 17.0 [11.0, 27.0] <0.01 In-hospital mortality, n (%) 22 547 (10.5) 20 555 (10.6) 1992 (10.3) 0.23 ACEI, angiotensin-converting enzyme inhibitor; ARB, angiotensin receptor blocker; COVID-19, Coronavirus Disease 2019; DM, diabetes mellitus; HF, heart failure; MI, myocardial infarction; PAD, peripheral arterial disease. Interrupted time series analyses for changes in heart failure admissions

Figure 1 demonstrates the weekly numbers of HF cases with the interrupted time series analysis. The COVID-19 pandemic led to a significant and immediate decrease in HF admissions per week [−3.6%, 95% confidence interval (CI): −0.3% to −6.7%, P = 0.03], which corresponded with the absolute reduction of 77.5 HF cases per week between the two periods. Thereafter, HF hospitalizations remained at a lower level without a significant change in the trend of HF admissions per week during the pandemic (slope change: −0.04%; 95% CI: −0.4% to 0.5%, P = 0.86). The trend in HF hospitalizations from January through June 2020 and the corresponding period in 2019 are displayed in Supporting Information, Figure S1. After the state of emergency was declared, the number of HF admissions continued to decline between April and June 2020.

Angiotensin-converting enzyme inhibitors and angiotensin receptor blockers prescriptions during COVID-19 pandemic

Prescriptions of ACEIs and ARBs for HF inpatients before and after Japan's state of emergency are shown in Figure 2. The proportions of ACEIs and ARBs prescriptions were not significantly different between the study periods (change: 4.2%; 95% CI: −0.3% to 8.9%, P = 0.07), although there was a decrease in the absolute number of patients with worsening HF. There were no significant changes in prescription trends for ACEIs and ARBs with interrupted time series analysis (slope change: −0.1%; 95% CI: −0.2% to 0.1%, P = 0.38).

image

Interrupted time series analysis for ACEI and ARB prescriptions per month. The proportion of ACEI and ARB prescriptions per week is plotted on the y-axis and the date (year and month) is on the x-axis. The curved green line represents a locally estimated smoothing spline fitted through the weekly proportions. The shaded green area shows the 95% confidence interval. The grey dashed line shows the 95% prediction interval. The horizontal green dashed line shows the linear predicted values through the entire study period, while the horizontal blue dashed line shows the linear predicted values after April 2020. The blue shaded area shows the period after the declaration of the state of emergency in Japan. ACEI, angiotensin-converting enzyme inhibitor; ARB, angiotensin receptor blocker.

In-hospital mortality

Among HF inpatients, interrupted time series analysis revealed that there was no significant change in the proportion of in-hospital deaths associated with the COVID-19 pandemic (change: 5.3%; 95% CI: −4.9% to 16.6%, P = 0.32; slope change: −1.7%; 95% CI: −3.1% to −0.4%, P = 0.01; Figure 3).

image

Interrupted time series analysis for in-hospital deaths per week. The proportion of in-hospital deaths per week is plotted on the y-axis and the date (year and month) is on the x-axis. The red curved line shows a locally estimated smoothing spline fitted through the weekly proportions. The shaded red area shows the 95% confidence interval. The grey dashed line shows the 95% prediction interval. The horizontal red dashed line represents the linear predicted values, and the horizontal blue dashed line represents the linear predicted values after April 2020. The blue shaded area shows the time period after the declaration of the state of emergency in Japan.

Secondary and sensitivity analysis

To determine whether our findings were specific to COVID-19 pandemic, we conducted post hoc, additional secondary and sensitivity analyses. A secondary analysis showed that HF admission cases with the breakpoint of April 2019 did not significantly decrease as seen with the breakpoint of April 2020 during COVID-19 pandemic but rather increased (change: 5.4%; 95% CI: 3.4% to 7.6%, P < 0.001).

We refit the model substituting the breakpoint to candidate breakpoints; (i) 11 March 2020, the World Health Organization has declared the COVID-19 pandemic. (ii) End-March, 2020, in Japan, the number of COVID-19 infections substantially increased in this period. The COVID-19 pandemic was accompanied by significant and immediate decreases in HF admissions cases for both breakpoints (change: −3.6%, 95% CI: −0.3% to −6.7%, P = 0.03 and change: −3.6%, 95% CI: −0.3% to −6.7%, P = 0.03, respectively). Another secondary analysis for the evaluation of HF cases per month, the number of HF admission led to a significant and immediate decrease in HF admissions per month (−33.3%, 95% CI: −35.9% to −30.6%, P < 0.001) between the study periods.

We added the indicator variable for wild data points to test the robustness of our findings based on the model controlling for wild data points. Sensitivity analysis verified that our findings were qualitatively unaffected by controlling for these wild data points (change: −4.9%, 95% CI: −8.0% to −1.7%, P = 0.003).

Discussion

The main findings of our study suggest that (i) urgent hospitalizations due to HF significantly decreased immediately after the state of emergency was declared in Japan; (ii) prescriptions for ACEIs and ARBs were similar before and after the state of emergency; and (iii) there was no significant deterioration of in-hospital mortality before and after the declaration of the state of emergency in Japan.

Firstly, unprecedented reductions in HF hospitalizations were documented in our Japanese cohort following the state of emergency. A range of plausible explanations can be proposed for this observation.12-16, 18, 19, 30 The declaration of the state of emergency and fear of being infected with COVID-19 in the hospital setting may have kept HF patients away from hospitals. Also, society-wide measures introduced to reduce transmission, such as social distancing, hand hygiene, and wearing masks, may have effectively reduced the incidence of other common respiratory viral infections. Respiratory infections are an acute precipitant exacerbating HF.31 As such, reducing respiratory infections may consequently limit HF hospitalizations. In Japan, there was a marked reduction in the number of inpatients with non-COVID-19 acute respiratory infections during the COVID-19 pandemic.32, 33 Therefore, this factor could be a contributor to reducing infection-related hospital admissions of HF patients. Furthermore, because medical resources were transferred to prioritize COVID-19 patients, there may have been delays in the treatment of less urgent HF patients. Of these plausible explanations, the most worrisome is that HF patients may have been reluctant to seek urgent care for fear of contracting COVID-19.2, 4, 34

Secondly, concern exists regarding the interaction between severe acute respiratory syndrome coronavirus-2 and angiotensin-converting enzyme 2 potentially increasing susceptibility to COVID-19. This concern led to a behavioural change in which both health care providers and patients suspended the prescription and use of ACEIs or ARBs in the previous studies.20-22 However, the proportion of ACEIs and ARBs prescriptions did not differ before or after the pandemic in the present study. Similarly, recent investigations support the maintenance of ACEIs or ARBs prescriptions during the COVID-19 pandemic.23-25 In this study, there was no significant increase in susceptibility to COVID-19 among HF patients receiving ACEIs or ARBs, and as such, there was no clear evidence for ceasing ACEIs and ARBs prescriptions prophylactically. Thus, the maintenance of ACEIs or ARBs prescriptions may have contributed to avoiding a rise in mortality rates.

Thirdly, in-hospital mortality due to HF did not worsen during the COVID-19 pandemic. Previous studies have shown that in-hospital mortality was worse among patients admitted during the COVID-19 pandemic, and there were higher rates of class III or IV symptoms (New York Heart Association) when compared with the same period in 2019.18, 19 Our results, which contradict these previous studies, may reflect the extent to which the pandemic affected the health care system overall. For example, disruptions in the health care system in northern Italy in the early stages of the pandemic6, 35 resulted in a limited supply of medical resources. Consequently, the demands for care exceeded the available resources, which limited the quality of medical care for other diseases. In Japan, we presume that resources and health care providers in cardiovascular care units were adequately resourced and still able to ensure appropriate acute care for all HF patients requiring urgent attention.36

This study had several limitations. Firstly, only hospitalized HF patients were included, which does not provide information regarding HF cases in the outpatient setting. Survival bias should also be considered in patients hospitalized with HF. Our study also lacked information about several factors associated with in-hospital mortality, such as ejection fraction, renal function, and natriuretic peptide levels. Segmented regression analysis of interrupted time series data allows us to assess how much a well-defined intervention changed an outcome of interest. Interrupted time series (ITS) design is also increasingly being used to evaluate the impact of interventions ranging from national public health policy implementation to health impact of natural disaster, political, and sociocultural events. In case of natural disaster, like COVID-19 pandemic, it may be difficult to define when the intervention began. A sufficient number of time points before and after the intervention are needed to conduct segmented regression analysis. We had deal with relatively short time series data with 30 monthly measures. Thus, further time series analyses are warranted to verify current findings controlling for temporal and seasonal trends with use of long-term longitudinal data.

In conclusion, we demonstrated a decline in HF cases in acute cardiovascular care settings during the COVID-19 pandemic. Furthermore, there was no clear evidence of a negative effect of the pandemic on the prescription of ACEIs and ARBs, nor in-hospital mortality.

Acknowledgements

We gratefully acknowledge the participating hospitals in the DPC and their staff.

Funding

This study was supported by the JSPS KAKENHI (Grant Number JP19H01075) from the Japan Society for the Promotion of Science, by a Health Labour Sciences Research Grant from the Ministry of Health, Labour and Welfare, Japan (20HA2003), and by the GAP Fund Program of Kyoto University type B (recipient: Yuichi Imanaka). This study was also supported by a Health, Labour and Welfare Policy Research Grants for Research on Policy Planning and Evaluation (Grant Number: 20AA2005) (recipient: Kiyohide Fushimi). The funders played no role in the study design, data collection and analysis, decision to publish, or manuscript preparation.

留言 (0)

沒有登入
gif