The study cohort was derived from the National Inpatient Sample (NIS), Healthcare Cost and Utilization Project (HCUP), and Agency for Healthcare Research and Quality database (AHRQ). The NIS is the largest U.S. database of hospital stays, containing a de-identified sample of discharge records from hospitals nationwide; it’s one of several files in HCUP, a family of de-identified hospital databases and tools curated from state and hospital data systems; and AHRQ is the federal agency that sponsors HCUP and makes these datasets and methods available. HCUP NIS data are publicly available and deidentified, and thus, the study was exempt from institutional review board approval. The NIS database contains weighted, discharge‐level data approximating ∼35 million hospitalizations annually, derived from a 20% stratified sample of discharges from U.S. community hospitals, and is designed to yield national estimates representative of ∼98% of the U.S. population.

We queried the NIS for the 2016–2020 releases and constructed the cohort using both ICD-10-CM diagnoses and ICD-10-PCS procedure codes. Hospitalizations with PCI were identified by ICD-10-PCS codes 02703(4–7)Z, 02703(D–G)Z, 02703TZ, 02713(4–7)Z, 02713(D–G)Z, 02713TZ, 02723(4–7)Z, 02723(D–G)Z, 02723TZ, 02733(4–7)Z, 02733(D–G)Z, 02733TZ, 02H(0–3)3DZ, 02H(0–3)3YZ, 027(0–3)3ZZ, 02C(0–3)3Z7, 02C(0–3)3ZZ, and 02F(0–3)3ZZ. Cases of CTO were flagged using ICD-10-CM I25.82. Analyses were restricted to patients older than 30 years, and demographic variables extracted for description included age, sex, and race.

We evaluated in-hospital all-cause mortality and procedure-related complications identified by ICD-10-CM codes: postprocedural myocardial infarction (I97.89), contrast-induced nephropathy (N99.0), cardiac perforation (I97.51), procedural bleeding (I97.410, I97.411, I97.610, I97.611, I97.630, I97.631), cardiac tamponade (I31.4), acute postprocedural respiratory failure (J95.821), and postprocedural ischemic stroke (I97.821).

All analyses were conducted using SAS software (version 9.4; SAS Institute, Cary, NC), and STATA 19.5 (Stata Corporation, College Station, TX). Data were obtained from the National Inpatient Sample (NIS) for years 2016–2020 and analyzed in accordance with Healthcare Cost and Utilization Project (HCUP) recommendations to account for the complex survey design.

Baseline patient demographics, clinical characteristics, and hospital-level variables were summarized for patients undergoing CTO intervention and those without CTO intervention. Continuous variables were reported as means with standard deviations or medians with interquartile ranges, as appropriate, and compared using survey-weighted linear regression. Categorical variables were reported as weighted percentages and compared using survey-weighted Rao–Scott chi-square tests.

To reduce confounding due to baseline differences between CTO and non-CTO patients, propensity score matching was performed. Propensity scores representing the probability of undergoing CTO intervention were estimated using survey-weighted logistic regression, incorporating hospital clustering (HOSP_NIS), stratification (NIS_STRATUM), and discharge-level sampling weights (DISCWT). Covariates included age, gender race, chronic kidney disease, chronic obstructive pulmonary disease, three-vessel coronary artery disease, prior percutaneous coronary intervention, history of coronary artery bypass grafting, anemia, cardiomyopathy, smoking status (current and former), valvular heart disease, endocarditis, prior stroke, diabetes mellitus, hypertension, hyperlipidemia, alcohol use, and peripheral vascular disease (PVD).

Patients were matched 1:1 using greedy nearest-neighbor matching on the logit of the propensity score with a caliper width of 0.2 standard deviations of the logit of the propensity score. Only matched pairs were retained for downstream analyses. Covariate balance before and after matching was assessed using standardized mean differences, with values <0.1 indicating adequate balance.

Primary and secondary outcomes, including in-hospital mortality and postprocedural complications, were compared between matched CTO and non-CTO cohorts. Survey-weighted logistic regression models were used to estimate odds ratios (ORs) with 95% confidence intervals (CIs), incorporating propensity score matching weights to account for the matched design. A survey-weighted generalized linear model (GLM) with a negative binomial distribution and a log link to model Length of stay (LOS). We used incidence rate ratios (IRRs) to quantify the association between the predictor variables and LOS. All statistical tests were 2-sided, and a p-value <0.05 was considered statistically significant. All analyses were conducted following the implementation of population discharge weights.