Is There Evidence of Demographic Disparity in Provision and Outcomes of Artificial Nourishment?

Abstract

BackgroundPatients who are unable to eat normally can be provided with enteral nutrition (tube feeding) via several routes, including percutaneous endoscopic gastronomy (PEG)and nasogastric (NG) tubes. In this retrospective cohort study, we examined the relationship between PEG and NG tube placement and three outcomes: mortality, complication, and discharge location, controlling for demographic covariates that have been associated with health disparities. We also examined the relationship between demographic covariates and types of tube placement.MethodsWe use delectronic medical record data to identify patients (n=3416) receiving enteralnutrition (PEG n=2321; NG n=1095) during 2013–2015 in a large urban hospital. ICD‐9 codes were used to assess complications. Mortality and complication were binary variables and a binary variable for discharge location was calculated to represent patients who were discharged to an attended setting or an unattended(e.g., home) setting. Relationships between enteral access type and outcomes were analyzed using logistic regression. Demographic covariates were race, coded as non‐White and White; insurance, coded as Medicaid, Medicare, orprivate; age, and sex.ResultsTube type was not found to be significantly associated with mortality, with or without adjustment for covariates. Odds of having a complication in patients with PEG tubes was 59.6 times more than those with NG tubes, adjusted for covariates (p<0.0001). The odds of being sent to an unattended discharge location (ie, home) following hospital treatment was 92% less likely for patients with PEG than those with NG tubes, adjusted for covariates (OR=0.08; p<0.0001). Having a PEG placed was 1.19 times more likely in non‐Whites than in Whites (p=.0196), 2.2 times greater among patients with Medicare (p<0.0001), and 1.4 times greater among patients with Medicaid (p=.0019), compared to patients with private insurance. Sex was not significantly associated with tube type.ConclusionsMortality was not found to be associated with tube type. However, the complication ratio was heavily skewed toward those with PEG devices. Those with PEG devices were much more likely to be sent to an attended care facility compared to those with NG tubes. Non‐Whites and those with government insurance were also more likely to have a PEG device. This may beconfounded by patient comorbidities or factors relating to PEGs and the healthcare system like length of stay or costs. The risk of complication with PEG compared to NG tubes and the impact patient demographics have on these outcomes should be examined in future studies to clarify the role of health disparities in artificial nutrition.Support or Funding InformationThis study was funded by National Institutes of Health Training Grant #HL007343 Descriptivestatistics and sample outcome compositions Composition Covariates PEG (n=2321) NGT (n=1095) Sex Male 1192 (68.31%) 553 (31.69%) Female 1129 (67.56%) 542 (32.44%) Insurance (SES) Medicaid (low) 461 (63.94%) 260 (36.06%) Medicare (gov't) 1477 (73.52%) 532 (26.48%) Private 383 (55.83%) 303 (44.17%) Age Mean: 62.4 year Mean: 66.7 years Ethnicity Non‐White 1499 (69.37%) 662 (30.63%) White 822 (65.50%) 433 (34.50%) Outcomes Mortality Yes (expired) 223 (71.47%) 89 (28.53% No 2098 (67.59%) 1006 (32.41%) Complication Yes 120 (99.17%) 1 (0.83% No 2201 (66.80%) 1094 (33.20%) Discharge Location (Unattended vs Attended) Unattended 134 (22.64%) 458 (77.36%) Attended 1964 (78.18%) 548 (21.82%) Univariate predictors of tube type among potentialconfounders Composition Logistic Regression results PEG (n=2321) NGT (n=1095) β parameter Odds Ratio P‐value Sex Male 1192 (68.31%) 553 (31.69%) 0.0342 1.035 0.6408 Female 1129 (67.56%) 542 (32.44%) 0 (ref) ‐‐‐ ‐‐‐ Insurance (SES) Medicaid (low) 461 (63.94%) 260 (36.06%) 0.3385 1.403 0.0019 Medicare (gov't) 1477 (73.52%) 532 (26.48%) 0.7869 2.197 <.0001 Private 383 (55.83%) 303 (44.17%) 0 (ref) ‐‐‐ ‐‐‐ Age Mean: 62.4 year Mean: 66.7 years 0.0124 1.012 <.0001 Ethnicity Non‐White 1499 (69.37%) 662 (30.63%) 0.1763 1.193 0.0196 White 833 (65.50%) 433 (34.50%) 0 (ref) ‐‐‐ ‐‐‐ Power analysis >0.999 Multivariate logistic regression adjusted forpotential confounders sex, SES, age, and ethnicity Outcome Predictor ORcrude ORadjusted 95% CI Adj P‐value Adj Mortality PEG 1.201 1.154 0.889, 1.499 0.2829 NGT 1.0 1.0 ‐‐‐ ‐‐‐ Complication PEG 59.589 55.159 7.7, 395.124 <0.0001 NGT 1.0 1.0 ‐‐‐ ‐‐‐ Discharge Location (Unattended vs Attended ) PEG 0.082 0.080 0.064, 0.101 <0.0001 NGT 1.0 1.0 ‐‐‐ ‐‐‐