Since the introduction of new antiplatelet and anticoagulant agents in the last decade, large-scale data studying gastrointestinal bleeding (GIB) in patients undergoing percutaneous coronary intervention (PCI) are lacking. Using the Nationwide Inpatient Sample, we identified all hospitalizations from 2006 to 2012 that required PCI. Temporal trends in the incidence and multivariate predictors of GIB associated with PCI were analyzed. A total of 4,376,950 patients underwent PCI in the United States during the study period. The incidence of GIB was 1.1%. Mortality rate in the GIB group was significantly higher (9.71% vs 1.1%, p <0.0001). Although the incidence of GIB remained stable during the study period (0.97% in 2006 to 1.19% in 2012), in-hospital mortality rate increased significantly from 7.9% in 2006 to 10.78% in 2012, with a peak of 12% in 2010. The GIB group had a longer median length of stay (5.80 vs 1.57 days) and an increased median cost of hospitalization ($26,564 vs $16,879). The predictors of GIB included cardiovascular co-morbidities such as acute myocardial infarction, cardiogenic shock, atrial fibrillation, congestive heart failure, valvular heart diseases, and a history of transient ischemic attack/stroke. Gastrointestinal co-morbidities including diverticulosis, esophageal cancer, stomach cancer, small intestine cancer, large intestine cancer, rectosigmoid cancer, gastrointestinal ulcer, and liver disease were predictors of GIB. Interestingly, a lower risk of GIB was associated with obese patients and patients with private insurance. A higher risk of GIB was noted in urgent versus elective admissions and weekend versus weekday admissions. In conclusion, the incidence of GIB in patients who underwent PCI remained stable from 2006 to 2012; however, the in-hospital mortality increased significantly. Identifying patients at higher risk for GIB is critically important to develop preventive strategies to reduce morbidity and mortality.
Percutaneous coronary intervention (PCI) plays an important role in the management of ischemic heart disease. This procedure reduces morbidity and mortality in patients with coronary artery disease when used in combination with evidence-based medical therapy. Although PCI has been shown to have high success rates and few ischemic complications, periprocedural bleeding complications have been shown to have an adverse impact on prognosis. Gastrointestinal bleeding (GIB) in patients undergoing PCI is a large health care burden. Studies have shown a significant increase in morbidity, mortality, and increased cost in these patients. Previous studies have shown that the incidence of GIB after PCI ranges from 0.6% to 2.3%. Recently introduced antiplatelet medications like prasugrel have led to changes in the management of ischemic heart disease and the complications associated with PCI. The increasing incidence of atrial fibrillation in the United States over the last decade has likely led to more PCI procedures being performed in this population. Anticoagulants such as dabigatran, rivaroxaban, and apixaban, which were initially irreversible, have been introduced, and their utilization has significantly increased in recent years, potentially impacting the outcomes of PCI procedures. Large-scale studies analyzing post-PCI GIB because these changes were implemented are lacking. In addition, large-scale data regarding the contemporary trends of post-PCI GIB are sparse. The main objectives of this study are to examine the frequency of in-hospital GIB associated with PCI and the likelihood of mortality, identify the predictors of GIB, examining temporal trends and evaluating the impact of bleeding on health care utilization and cost.
Methods
Data were obtained from the Nationwide Inpatient Sample (NIS), which is the largest all-payer inpatient database in the United States. The NIS is created through surveys of hospitalizations conducted by the Healthcare Cost and Utilization Project in association with participating states. It contains a stratified data sample that represents approximately 20% of all US community hospitals. The NIS consists of deidentified hospitalizations which include demographic information such has age, race, gender, primary and secondary procedures, insurance status, hospitalization outcomes, total cost, and length of stay (LOS) recorded by the discharging institution. In addition to this, the NIS also contains clinical and resource use information including safeguards to protect patient, physician, and hospital privacy. Data from the NIS correlate well with other discharge based hospitalization databases in the United States. The NIS has been used in previous studies that analyzed different diseases and procedures in terms of associated outcomes, temporal trends, health care use and access, and hospitalization rates. The results of these studies have been published in world-renowned journals.
Using International Classification of Diseases, Ninth Revision, Clinical Modification codes 36.06 and 36.07, we identified all hospitalizations requiring PCI between years 2006 and 2012. All patients older than 18 years were included. Hospitalizations with missing age, gender, admission or discharge date, and in-hospital mortality status were excluded from the sample. In addition, entries with the same admission and discharge date were excluded to avoid events that do not represent acuity for hospital admission.
Our primary objective was to determine the incidence and mortality of GIB and its temporal trend in our study population between years 2006 and 2012. Our secondary objective was to investigate the independent predictors of post-PCI GIB. In our analysis, we included co-morbidities such as alcoholism, diverticulosis, peptic ulcer disease, atrial fibrillation, congestive heart failure, valvular heart disease, renal failure, anemia, previous significant bleeding, coagulation disorders, liver disease, and gastrointestinal malignancies including esophageal, gastric, small intestinal, large intestinal, and rectosigmoid cancers. All these are well-established conditions associated with GIB. In addition to this, the LOS and cost of each hospitalization were also investigated.
The NIS variables were used to identify patient age, gender, and race. Race was divided into white, black, Hispanic, and others. Age was divided into 5 groups: 18 to 34 years, 35 to 49 years, 50 to 64 years, 65 to 79 years, and 80 years or older. The severity of co-morbid conditions was defined using the Deyo modification of the Charlson Comorbidity Index, which contains 17 co-morbid conditions that have differential weights. The index has scores ranging from 0 to 33 with greater scores corresponding to a higher burden of co-morbid diseases and vice versa. We considered participating hospitals as teaching hospitals only if they had an Accreditation Council for Graduate Medical Education–accredited residency program, were a member of the Council of Teaching Hospitals, and/or had a full-time equivalent interns and resident to patient ratio of ≥0.25. We also took hospital location (rural/urban) and bed size into account for our study. We divided bed size cutoff points into small, medium, and large so that approximately 1/3 of the hospitals in a given combination of region, location, and teaching status fall within each bed size category. The LOS for each hospitalization was calculated after excluding those who died during their stay. The cost of each hospitalization was also determined after merging data with cost-to-charge ratio files available from the Healthcare Cost and Utilization Project website. The total cost of each hospitalization was determined by multiplying the cost-to-charge ratio with the total hospital charge. Inflation was accounted for by adjusting the cost of each year in reference to the US dollar value using Consumer Price Index data.
The weights provided by the NIS were used to generate national estimates. The chi-square test was used to compare categorical variables between patients with or without GIB. Continuous variables like LOS were compared using the Wilcoxon signed-rank test was used to compare continuous variables, as they did not show a normal distribution. Three hierarchical mixed-effects logistic regression models were created to identify the independent multivariate predictors of mortality, LOS, and cost. Hospital- and patient-level variables were included in all multivariate models using the unique hospital identification numbers incorporated as random effects within the models. These included region (Northeast, South, Midwest, and West), bed size, and teaching status for the hospitals, as well as age, gender, median household income, and primary payer information for the patients. To account for the hospital clustering effect, hospital identification was incorporated in the model as a random effect. Stata IC 11.0 was used for the analyses. A p value of less than 0.05 was considered statistically significant. For categorical variables, such as incidence of GIB and in-hospital mortality, the chi-square test of trend for proportions was used with the Cochran–Armitage test through the “ptrend” command in Stata. The nonparametric test for trend by Cuzick was used for continuous variables like cost of hospitalization, using the “nptrend” command in Stata.
Results
We identified a total of 4,376,950 patients who underwent PCI procedures for coronary artery disease between years 2006 and 2012 ( Table 1 ). The overall incidence of GIB in these patients was 1.1%. This incidence remained relatively stable during our study period (0.97% in 2006 and 1.19% in 2012, Figure 1 ). In contrast, the in-hospital mortality rate of patients with post-PCI GIB increased significantly from 7.9% in 2006 to 10.78% in 2012 ( Table 2 , Figure 2 ). Mortality related to GIB was significantly increased throughout our study period, being highest in 2010.
| Gastrointestinal bleeding | Overall | P value | ||
|---|---|---|---|---|
| No | Yes | |||
| Overall study population | 98.9% | 1.1% | 4376950 | |
| Patient level variables | ||||
| Age in years | <.0001 | |||
| 18-34 | 0.56% | 0.47% | 0.56% | |
| 35-49 | 11.6% | 6.13% | 11.54% | |
| 50-64 | 37.87% | 26.61% | 37.74% | |
| 65-79 | 37.5% | 43.92% | 37.57% | |
| >=80 | 12.47% | 22.87% | 12.59% | |
| Gender | <.0001 | |||
| Male | 66.47% | 59.11% | 66.39% | |
| Female | 33.53% | 40.89% | 33.61% | |
| Race | <.0001 | |||
| White | 63.91% | 62.03% | 63.88% | |
| Black | 6.59% | 8.05% | 6.61% | |
| Hispanic | 5.48% | 5.96% | 5.48% | |
| Others | 5.43% | 5.14% | 5.42% | |
| Missing | 18.6% | 18.82% | 18.6% | |
| Comorbidities | ||||
| Myocardial infarction | 44.36% | 71.61% | 44.66% | <.0001 |
| Obesity | 12.47% | 9.99% | 12.44% | <.0001 |
| Hypertension | 70.61% | 63.89% | 70.53% | <.0001 |
| Diabetes | 33.38% | 34.37% | 33.39% | <.0001 |
| Alcoholism | 1.95% | 4.33% | 1.98% | <.0001 |
| Diverticulosis | 0.86% | 5.53% | 0.91% | <.0001 |
| Any gastrointestinal cancer | 0.98% | 2.91% | 1% | <.0001 |
| Peptic ulcer disease | 0.02% | 0.16% | 0.02% | <.0001 |
| Atrial fibrillation | 9.25% | 18.29% | 9.35% | <.0001 |
| Congestive heart failure | 14.11% | 34.18% | 14.33% | <.0001 |
| Valvular disorder | 0.26% | 1.48% | 0.28% | <.0001 |
| Renal failure | 10.11% | 21.95% | 10.24% | <.0001 |
| Anemia | 8.12% | 25.59% | 8.31% | <.0001 |
| Significant bleed | 0.36% | 10.63% | 0.47% | <.0001 |
| Coagulation disorder | 2.02% | 8.71% | 2.1% | <.0001 |
| Median household income | <.0001 | |||
| 1. 0–25th percentile | 25.75% | 28.6% | 25.78% | |
| 2. 26–50th percentile | 26.15% | 25.75% | 26.14% | |
| 3. 51–75th percentile | 24.21% | 24.26% | 24.21% | |
| 4. 76–100th percentile | 21.63% | 19.05% | 21.61% | |
| Primary Payer | <.0001 | |||
| Medicare | 50.45% | 65.77% | 50.62% | |
| Medicaid | 5.58% | 6.17% | 5.59% | |
| Private | 35.28% | 20.92% | 35.12% | |
| Self-pay/no pay/others | 8.53% | 6.97% | 8.51% | |
| Hospital bed size | <.0001 | |||
| Small | 7.49% | 6.82% | 7.48% | |
| Medium | 20.2% | 20.17% | 20.2% | |
| Large | 71.78% | 72.35% | 71.79% | |
| Hospital location & Teaching status | 0.0166 | |||
| Rural | 5.62% | 5.77% | 5.62% | |
| Urban – Non Teaching | 39.92% | 40.34% | 39.93% | |
| Urban – Teaching | 53.93% | 53.24% | 53.92% | |
| Hospital region | <.0001 | |||
| Northeast | 19.21% | 16.62% | 19.18% | |
| Midwest | 25.19% | 26.18% | 25.2% | |
| South | 38.99% | 39.5% | 38.99% | |
| West | 16.2% | 16.93% | 16.21% | |
| Admission days | <.0001 | |||
| Weekdays | 84.33% | 77.21% | 84.25% | |
| Weekend | 15.67% | 22.79% | 15.75% | |
| Type of admission | <.0001 | |||
| Non-Elective | 72.34% | 85.68% | 72.49% | |
| Elective | 27.37% | 14.12% | 27.23% | |
| Length of stay | <.0001 | |||
| Median (interquartile range) | 1.57 days (1-2.96) | 5.80 days (3.23-10.46) | 1.59 (1-3) | |
| Cost | <.0001 | |||
| Median (interquartile range) | $16879 ($12904-$22544) | $26564 ($19105-$40083) | $16937 ($12935-$22672) | |
| Disposition | <.0001 | |||
| Routine | 90% | 58.3% | 89.65% | |
| Transfer to facility | 8.6% | 31.55% | 8.85% | |
| AMA | 0.29% | 0.4% | 0.29% | |
| Died | 1.1% | 9.71% | 1.19% | |
| Mortality | 1.1% | 9.71% | 1.19% | <.0001 |
| 2006 | 2007 | 2008 | 2009 | 2010 | 2011 | 2012 | Overall | |
|---|---|---|---|---|---|---|---|---|
| Odds Ratio (95% CI) | 2.55 (1.99-3.26) | 2.368 (1.80-3.11) | 2.928 (2.31-3.71) | 2.975 (2.34-3.79) | 3.242 (2.51-4.19) | 1.781 (1.41-2.26) | 2.599 (2.01-3.36) | 2.631 (2.39-2.90) |
| P value | <.0001 | <.0001 | <.0001 | <.0001 | <.0001 | <.0001 | <.0001 | <.0001 |
| Incidence of GI Bleed | 0.97% | 1.04% | 1.11% | 1.11% | 1.24% | 1.16% | 1.19% | 1.1% |
| In-hospital mortality without GI Bleed | 0.84% | 0.93% | 1.03% | 1.12% | 1.2% | 1.36% | 1.47% | 1.1% |
| In-hospital mortality with GI Bleed | 7.9% | 8.17% | 9.92% | 10.4% | 12% | 9.38% | 10.78% | 9.71% |
Baseline patient characteristics are listed in Table 1 . The relative proportion of GIB cases was relatively low in younger patients but was significantly greater in patients over the age of 65 years. Women made up a significantly larger relative proportion of GIB cases than men. Blacks and Hispanics had a greater relative proportion of GIB cases than whites. A greater relative proportion of patients who developed GIB had Medicare or Medicaid as a primary source of insurance in comparison to patients who did not develop GIB. In patients who did not develop GIB, a greater proportion had private insurance or self-pay/no pay/other in comparison to those who developed GIB. Obesity, hypertension, and diabetes mellitus were associated with a lower incidence of GIB. The history of significant bleeding, coagulation disorder, alcoholism, diverticulosis, GI cancer, peptic ulcer disease, atrial fibrillation, congestive heart failure, valvular heart disorder, renal failure, or anemia was associated with an increased risk of GIB. Patients developing post-PCI GIB had an increase in LOS by 4.23 days and cost of hospitalization by $9,685. Mortality rate was significantly higher in patients who developed GIB in comparison to those who did not. Patients with GIB were more likely to be placed in a facility upon discharge.
Hospital characteristics are identified in Table 1 . Hospital location (urban vs rural) or teaching status did not have an association with post-PCI GIB. A greater proportion of patients who developed GIB were admitted under urgent or emergent settings compared with elective admissions. Weekend admissions were more likely to be associated with GIB than weekday admissions. Hospital bed size also was not associated with GIB; however, there was a trend toward more GIB in large hospitals and less in small hospitals. Hospitals located in the West and Midwest made up a greater proportion of GIB cases than hospitals located in the South and Northeast.
In a multivariate model, after adjusting for various confounders, older age and female gender were noted to be associated with a greater risk of GIB. The various co-morbidities associated with an increased risk of GIB are listed in Table 3 . Cardiovascular conditions such as acute myocardial infarction, cardiogenic shock, atrial fibrillation, congestive heart failure, or a history of transient ischemic attack/stroke was associated with a higher propensity of developing GIB. In addition to cardiovascular co-morbidities, there was a dramatic increase in the incidence of GIB in patients with existing gastrointestinal co-morbidities. Patients with a history of diverticulosis, stomach cancer, small intestine cancer, large intestine cancer, rectosigmoid cancer, gastrointestinal ulcer, and liver disease, were more likely to have GIB. In contrast, obese patients had lower incidence of GIB. Furthermore, alcohol use, anemia, and coagulation disorders were also associated with an increased incidence of GIB.
| Variables | Odds ratio (95% CI) | P value |
|---|---|---|
| Age | 1.017 (1.01-1.02) | <.0001 |
| Female | 1.065 (1.02-1.12) | 0.01 |
| Acute myocardial infarction | 2.217 (2.09-2.36) | <.0001 |
| Cardiogenic Shock | 3.303 (3.06-3.56) | <.0001 |
| Drug eluting stent | 0.653 (0.62-0.69) | <.0001 |
| Diverticulosis | 4.506 (4.03-5.04) | <.0001 |
| Esophageal cancer | 1.711 (0.67-4.35) | 0.2586 |
| Stomach cancer | 6.104 (3.04-12.27) | <.0001 |
| Large intestine cancer | 6.137 (4.55-8.28) | <.0001 |
| Small intestine cancer | 4.906 (1.19-20.20) | 0.0276 |
| Rectosigmoid cancer | 6.495 (3.80-11.01) | <.0001 |
| Gastrointestinal ulcer | 4.172 (2.10-3.02) | <.0001 |
| Liver diseae | 2.588 (2.22-3.02) | <.0001 |
| Atrial fibrillation | 1.201 (1.13-1.28) | <.0001 |
| Congestive heart failure | 1.437 (1.36-1.52) | <.0001 |
| Valvular disorder | 1.732 (1.39-2.16) | <.0001 |
| Chronic renal failure | 1.142 (1.07-1.22) | <.0001 |
| TIA/Stroke | 1.45 (1.21-1.73) | <.0001 |
| Anemia | 3.304 (3.08-3.54) | <.0001 |
| Coagulation disorder | 1.691 (1.54-1.86) | <.0001 |
| Obesity | 0.837 (0.78-0.91) | <.0001 |
| Alcohol use | 1.803 (1.60-2.03) | <.0001 |
| Peripheral vascular disease | 1.127 (1.05-1.21) | 0.0006 |
| Median household income | ||
| 1st quartile | Reference | |
| 2nd quartile | 0.93 (0.87-0.98) | 0.014 |
| 3rd quartile | 0.93 (0.87-1.00) | 0.0392 |
| 4th quartile | 0.85 (0.79-0.92) | 0.0001 |
| Primary payer | ||
| Medicare | Reference | |
| Medicaid | 1.109 (0.99-1.24) | 0.0651 |
| Private including HMO | 0.783 (0.72-0.85) | <.0001 |
| Self-pay/No-pay/Others | 0.827 (0.74-0.92) | 0.0007 |
| Hospital bed size | ||
| Small | Reference | |
| Medium | 1.014 (0.91-1.13) | 0.7986 |
| Large | 1.006 (0.91-1.11) | 0.8959 |
| Hospital location & Teaching status | ||
| Rural | Reference | |
| Urban Non-Teaching | 1.016 (0.89-1.16) | 0.8202 |
| Urban Teaching | 1.028 (0.90-1.18) | 0.6882 |
| Hospital region | ||
| Northeast | Reference | |
| Midwest | 1.033 (0.97-1.15) | 0.5586 |
| South | 1.012 (0.91-1.12) | 0.824 |
| West | 0.957 (0.84-1.10) | 0.4909 |
| Elective vs. Non-Elective | 1.359 (1.26-1.47) | <.0001 |
| Weekend vs. weekdays admission | 1.13 (1.07-1.20) | <.0001 |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
