This study used inpatient data obtained from the National Inpatient Sample Database (NIS) database between 2008 and 2017. The NIS is the largest publicly available all-payer data base established by the Agency for Healthcare Research and Quality under the Healthcare Cost and Utilization Project. More details regarding the NIS are available at www.hcup-us.ahrq.gov. Our study was exempt from formal Institutional Review Board approval owing to the use of publicly available database.

Of 74,730,240 hospitalizations recorded between 2008 and 2017 in the NIS database, 256,239 hospitalization records with patients aged ≥18 years old with a prostate cancer (PC) diagnosis were included. International Classification of Disease, Ninth Revision (ICD-9) and Tenth Revision (ICD-10) codes were used to identify PC hospitalizations (ICD-9-CM codes including 233.4 or 185 and ICD-10-CM codes D07.5 or C61) (11). Records that meet the following criteria were excluded: (1) age <18 years; (2) transferred to another hospital or institution on the day of admission; (3) length of stay missing or ≤1 day; (4) hospitalization expenditure missing or was 0 and (5) missing in-hospital death status. The study cohort was divided into two groups based on the presence or absence of atrial fibrillation (AF) (ICD-9-CM code 427.3 and ICD-10-CM code I48). Figure 1 illustrates the steps used to locate the target population.

Information regarding patient- and hospital-level characteristics for each hospitalization is included in the NIS. The patient-level variables were age, sex, race, median household income of residents in the patient' s ZIP Code, insurance type, admission day and admission season. Hospital-level characteristics included the region, teaching status, bed size, and hospital ownership. We also collected information on comorbidities, as suggested in previous studies, such as alcohol abuse, drug misuse, coronary heart disease, lipoid metabolism disorders, atherosclerosis, nephrotic syndrome, thyrotoxicosis, and chronic kidney disease, according to ICD-9/10 diagnostic codes and procedure codes (12). All the variables are shown in the first column of Table 1. Corresponding ICD codes of these variables are listed in Supplementary Table S1.

The main endpoints included the temporal trend of AF prevalence in PC inpatients and in-hospital mortality, total cost, length of hospital stay (LOS), complications including congestive heart failure, pulmonary circulation disorders, renal failure, fluid and electrolyte disorders and cardiogenic shock. The total cost was derived from the total charges in the database using the cost-to-charge ratio and Consumer Price Index.

Weighting and stratification methods were used to obtain the total national estimates because the NIS database is based on a sampling design. We then summarized the baseline characteristics of PC patients in the two groups (with or without AF) to demonstrate the descriptive analyses results. We used the independent samples t-test or rank sum test and chi-square test to compare the different characteristics between PC patients with and without AF. Mean and standard deviation (SD) were calculated for continuous variables, and frequencies and percentages were calculated for categorical variables.

The Cochran-Armitage trend test was conducted to analyze the temporal trend in AF prevalence, and subgroup analyses were conducted to analyze the temporal trend in AF prevalence based on patient- and hospital-level characteristics. In addition, multivariable-adjusted logistic analyses were conducted to evaluate the association between AF and adverse outcomes (in-hospital mortality, congestive heart failure, pulmonary circulation disorders, renal failure, fluid and electrolyte disorders and cardiogenic shock), controlling for the potential confounders of patient-level, hospital-level, and comorbid factors. The trend in in-hospital mortality was compared between PC inpatients with and without AF. As the total cost and LOS indicated a right-skewed distribution, a logarithmic transformation for these two outcomes was performed before constructing the multivariable linear models.

Sensitivity analyses were performed using the propensity score weighting (PSW) method, which combines logistic regression and overlap weighting (logistic-OW) to balance the variables and assess the outcomes. Absolute standardized mean differences (ASMD) calculated for each covariate between the two groups were used for balance assessment, indicating adequate covariate balance with ASMD < 0.1 (13).

The R software (Version 4.0.3) and SAS (Version 9.4) were used to conduct all the statistical analyses. All tests were two-tailed, and P < .05 was considered significant.

The missing rate for most variables was less than 3%, except for race (9.60%), admission season (4.32%) and median household income of residents in the patient' s ZIP Code (3.15%). We imputed missing data with the dominant category for categorical variables and the median for continuous variables (14). Missing race data were imputed into missing subgroups.

Among the 256,239 hospitalizations (1,275,345 weighted admissions) aged ≥18 years who were diagnosed with PC from 2008 to 2017 and met the inclusion criteria, 1,069,372 weighted hospitalizations (83.8%) did not have AF, and 205,973 (16.2%) had AF. Comparison of the demographics, hospital characteristics, complications, treatments and clinical outcomes between patients with and without AF is presented in Table 1. For patient-level features, the AF group was characterized by an older age (68.8% vs. 37.0%, P < .001), higher proportion of white individuals (74.3% vs. 61.2%, P < .001), and greater prevalence of Medicare insurance coverage (84.4% vs. 62.1%, P < .001) (Table 1). Regarding comorbidities, the AF group had higher rates of collagen vascular diseases, deficiency anemias, chronic pulmonary disease, coagulopathy, diabetes, hypertension, hypothyroidism, neurological disorders, paralysis, obesity, peripheral vascular disorder, valvular disease, and weight loss. However, these patients were less likely to have lymphatic metastasis or hepatic metastases (Table 1).

AF prevalence among PC inpatients significantly increased from 14.0% in 2008 to 20.1% in 2017 (P for trend <.001). Further subgroup analysis also revealed a consistent rise in AF prevalence across all groups, with all trends being statistically significant (P for trend <.001) (Figure 2, Supplementary Figure S1). The prevalence of AF positively correlated with age, with the highest rate observed among patients aged ≥75 years in 2017 (32.8%). Besides, there was an increase in AF prevalence, by 1.9 times, among patients aged 45–64. Furthermore, AF was more prevalent among hospitalized White patients than other racial groups, and the highest prevalence was observed in patients covered by Medicare. Additionally, higher median household income of residents in the patient's ZIP Code and comorbid hypertension were associated with a higher AF prevalence. Supplementary Figure S1 provides further evidence of the consistent increase in AF prevalence across different hospital regions, bed sizes, teaching statuses, and metastatic status reinforcing the findings of this study.

This study revealed a significant increase in in-hospital mortality among PC inpatients. In patients without AF, the mortality rate increased from 3.1% in 2008 to 3.9% in 2017 (P for trend <.001), whereas in patients with AF, the rate increased from 5.1% in 2008 to 8.1% in 2017 (P for trend <.001). Throughout the entire study period, patients with AF had a higher in-hospital mortality rate than those without AF (Figure 3).

Multivariable-adjusted logistic model results showed that AF was associated with incremental in-hospital mortality (odds ration [OR], 1.49; 95% confidence interval [CI], 1.42–1.56), P < .001). AF was also associated with other adverse outcomes, including congestive heart failure (OR, 1.76; 95% CI, 1.69–1.83), P < .001), pulmonary circulation disorders (OR, 1.50; 95% CI, 1.40–1.62), P < .001), renal failure (OR, 1.30; 95% CI, 1.21–1.39), P < .001), fluid and electrolyte disorders (OR, 1.09; 95% CI, 1.06–1.12), P < .001) and cardiogenic shock (OR, 2.07; 95% CI, 1.74–2.46), P < .001; Figure 4).

The subgroup analysis results of patients admitted for radical prostatectomy, PC patients with or without metastatic, and those with or without comorbid hypertension were consistent with the findings of the entire cohort (Supplementary Tables S2,S3).

The relationship between insurance type and AF prevalence, and the association between insurance type and in-hospital mortality were further investigated. Compared with other insurance, even after adjusting covariates, Medicare is still significantly associated with increment of AF prevalence. However, compared with other insurance, patients with Medicare do not undergo an increased risk of death (Supplementary Table S4).

We also analyzed the median in-hospital LOS and total costs for patients with and without AF.

The median length of stay for patients with AF and without AF were 4.00 days [interquartile range (IQR): 3.00–7.00] and 3.00 days (IQR: 2.00–6.00), respectively. The median total cost for patients with AF was 10,563.17 dollars (IQR: 6,335.66–18,596.05), whereas that for those without AF was 11,045.36 dollars (IQR: 6,784.84–17,179.64) (Table 1). After performing a logarithmic transformation and multivariate regression, we found that the AF group had a 9% higher total cost and 9% longer LOS. Furthermore, the LOS and total costs for inpatients with and without AF decreased slightly over the entire study period (Table 2).

Most of the baseline variables were unbalanced between the AF and without AF groups. After PSW model, all the ASMD were <.001, suggesting a good balance between these comparisons (Supplementary Figure S2). In the PSW, we obtained similar results, showing that AF group was associated with incremental in-hospital mortality (OR, 1.40; 95% CI, 1.34–1.46, P < .001), congestive heart failure (OR, 1.57; 95% CI, 1.52–1.62, P < .001), pulmonary circulation disorders (OR, 1.47; 95% CI, 1.37–1.57, P < .001), renal failure (OR, 1.04; 95% CI, 1.02–1.07, P < .001), fluid and electrolyte disorders (OR, 1.08; 95% CI, 1.06–1.11, P < .001) and cardiogenic shock (OR, 1.81; 95% CI, 1.54–2.14, P < .001) (Supplementary Table S5). In addition, AF group had a higher total cost and longer LOS than the without AF group (Supplementary Table S5).