Analysis of the National Inpatient Sample database for safety of primarily outpatient cardiovascular procedures is biased due to selective hospitalization of patients who develop complications, falsely inflating complication rates.
Key Findings
Results
Patients who underwent atrial fibrillation catheter ablation on day 1 of admission or before (group 1) had significantly higher complication rates than those ablated on day 2 or later (group 2), despite being younger and healthier.
Total analysis included 47,925 weighted hospitalizations of patients who underwent atrial fibrillation catheter ablation between 2016 and 2022
Complication rate was 9.5% in group 1 versus 6.6% in group 2 (P<0.001)
Multivariate analysis showed group 1 was an independent predictor of complications (OR 1.54 [95% CI, 1.30–1.83]; P<0.001)
This paradoxical finding — sicker patients in group 2 yet lower complication rates — is attributed to selection bias in the NIS dataset
Results
Patients in group 1 (ablation on day of admission or before) were younger and had significantly fewer comorbidities compared to group 2.
Group 1 had fewer octogenarians than group 2 (6.9% versus 10.4%, P<0.001)
Group 1 had lower rates of renal disease (12.6% versus 20.0%, P<0.001)
Group 1 had lower rates of heart failure (28.3% versus 45.8%, P<0.001)
Group 1 had lower rates of ischemic heart disease (16.0% versus 20.5%, P<0.001)
Discussion
The NIS database captures a biased sample of outpatient cardiovascular procedures because patients admitted to the hospital are disproportionately those who develop complications during otherwise outpatient procedures.
AF catheter ablation is primarily performed as an outpatient procedure, so inpatient NIS records represent a non-random subset
Patients undergoing outpatient procedures who are at higher risk or develop complications are admitted as inpatients, thus falsely inflating the complication rate observed in the NIS dataset
The counterintuitive finding that healthier patients (group 1) had higher complication rates than sicker patients (group 2) is explained by this selective hospitalization mechanism
This bias challenges the validity of using NIS data to assess the safety of procedures that are primarily outpatient in nature
Methods
The study design used day of ablation relative to admission date as a proxy to distinguish planned inpatient procedures from procedures complicated by adverse events requiring hospitalization.
Group 1 was defined as ablation performed on day 1 of admission or before, representing likely planned or same-day procedures
Group 2 was defined as ablation performed on day 2 of admission or later, representing patients admitted for other reasons or complications prior to ablation
This retrospective cross-sectional analysis used NIS data from 2016 to 2022
Patient baseline characteristics and procedural complication rates were compared between the two groups
What This Means
This research examined a widely used hospital database called the National Inpatient Sample (NIS) to understand whether it can reliably be used to study the safety of heart procedures that are usually done on an outpatient basis — meaning patients go home the same day. Using atrial fibrillation catheter ablation (a common heart rhythm procedure) as a test case, the researchers found a puzzling pattern: patients who had their procedure on the day they were admitted to the hospital (who were younger and had fewer health problems) actually had more complications than patients who were admitted earlier and were older and sicker.
The explanation for this seemingly backward finding is a statistical phenomenon called selection bias. Because catheter ablation is normally an outpatient procedure, most patients who have it done never appear in the hospital database at all. The patients who do end up in the hospital database are disproportionately those who had problems — either complications that forced them to be admitted, or pre-existing conditions serious enough to warrant planned inpatient care. This means the NIS database does not represent a typical cross-section of everyone getting the procedure; it over-represents the sickest and most complicated cases, making the procedure look more dangerous than it really is.
This research suggests that researchers and clinicians should be cautious when interpreting studies that use the NIS or similar hospital databases to evaluate the safety of procedures that are primarily performed in outpatient settings. The findings highlight the importance of understanding where data comes from and what biases may be built into a dataset before drawing conclusions about procedural risks. Studies using these databases may systematically overestimate complication rates for outpatient cardiovascular procedures.
Oakes K, Port J, Furer A, Ehrhardt-Humbert L, John K, Chee J, et al.. (2026). Proper Use of the National Inpatient Sample Database to Analyze Safety of Cardiovascular Procedures.. Journal of the American Heart Association. https://doi.org/10.1161/JAHA.125.047611