Hospital Variation in Gastrostomy Tube Use among the Critically Ill

To the Editor:

In the last 20 years, the number of critically ill patients receiving gastrostomy tubes has increased dramatically, such that critically ill patients are now the most common recipients of gastrostomy tubes among hospitalized patients (1). However, the decision to place a gastrostomy tube during critical illness is not guided by quality evidence of long-term benefits or optimal patient selection. Practices with little supporting evidence may be subject to large variation, and observations of large variation can highlight areas needing further evidence to guide practice (2). Using a nationally representative population-based database, we sought to quantify variation in gastrostomy use in critically ill patients among U.S. hospitals.

Methods

We used the Agency for Healthcare Research and Quality’s Healthcare Cost and Utilization Project’s National Inpatient Sample, 2014 (3). We identified hospitalized adults (≥18 yr), with International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes consistent with critical illness, who received gastrostomy tubes (4–8). Hospitals with fewer than 25 admissions of critically ill patients in 2014 were excluded (9). Using survey-weighted methods (3), we derived national estimates. We calculated risk-adjusted gastrostomy rates at the hospital level using hierarchical multivariable logistic regression models with hospital-level random intercepts, adjusting for patient (age, sex, race, primary insurance payer, median zip code income, Elixhauser comorbidities and mortality index [10], surgical status [11], type of critical illness [4], type of acute organ failure [5]) and hospital characteristics (bed number, location/teaching status, U.S. region). Intraclass correlation coefficients (ICCs) (the proportion of variance attributable to between-hospital variance) were used to quantify the level of variation in gastrostomy tube placement between hospitals; proportional changes in variance (the contribution of patient and hospital characteristics to variance, respectively) were calculated by comparing model ICCs with and without patient and hospital factors. We compared patient and hospital characteristics of high–gastrostomy rate and low–gastrostomy rate hospitals (top and bottom decile of adjusted rate, respectively) using chi-square tests. Statistical testing was two-tailed, with α = 0.05, using SAS 9.4 (SAS Institute). The Beth Israel Deaconess Medical Center Institutional Review Board deemed the study exempt from review.

Results

In 2014, 91,990 (2.1%) patients (18,398 unweighted) of 4,293,521 (858,704 unweighted) critically ill patients received a gastrostomy tube. Of 2,895 hospitals, 672 (23.2%) hospitals placed zero gastrostomy tubes, and 576 (19.9%) hospitals placed more than 10 gastrostomy tubes. Risk-adjusted rates of gastrostomy tube for each hospital, from lowest to highest, are shown in Figure 1; the median was 2.0 (interquartile range, 1.7–2.6) per 100 critically ill patients. Hospitals in the top decile of gastrostomy tube placement rates (3.5 or more gastrostomy tubes per 100 critically ill patients) accounted for 29% of all gastrostomy tubes placed in 2014. Compared with hospitals in the bottom decile (fewer than 1.5 gastrostomy tubes per 100 critically ill patients), high-gastrostomy hospitals were more likely to be urban teaching hospitals (61.4% vs. 14.9%, P < 0.01) and larger (43.1% vs. 21.8%, P < 0.01); there was no significant difference in measured patient characteristics.

Figure 1.

Risk-adjusted rates of gastrostomy tube placement were ranked from lowest to highest and plotted (blue dots). Error bars indicate 95% confidence interval for each point estimate.

We observed a crude ICC (percentage of total variation attributable to between-hospital variance) of 11% (95% confidence interval [CI], 10–12%), which decreased to 9% (95% CI, 8–10%) after adjusting for patient characteristics, and 8% (95% CI, 7–9%) after also adjusting for hospital characteristics. Thus, the proportional change in variance (the degree to which between-hospital variance could be explained by characteristics) was 17% for patient factors and 9% for hospital factors.

Discussion

We examined variation in rates of gastrostomy use across U.S. hospitals. Despite the paucity of evidence or guidelines regarding gastrostomy tube placement in the critically ill, there was surprisingly little variation in the rates of gastrostomy placement, with an ICC of only 8% (for comparison, variation between hospitals in intensive care unit use is ∼15%) (12). Most hospitals in the highest 10% of gastrostomy rates were large, urban teaching hospitals, which placed nearly 30% of gastrostomy tubes during critical illness.

Despite the few outlier hospitals, the generally consistent and low rate of gastrostomy placement during critical illness may occur for multiple reasons. First, prior surveys of patient preferences showed that 55.6% of patients viewed a reliance on a feeding tube to be a health state equivalent to or worse than death (13). Second, physicians may extrapolate evidence against placing feeding tubes for patients with dementia to patients with critical illness. Our findings of high gastrostomy placement rate within large, urban teaching hospitals provide further data to support prior studies showing higher rates of interventions and fewer treatment limitations within these types of hospitals (14). Importantly, although high between-hospital variation would highlight the need for clarification of best practices, the absence of variation does not exclude this need, as it remains uncertain if the current ad hoc selection of patients is optimally aligned with patient prognosis and patient preferences. Further studies are necessary to determine drivers of gastrostomy placement during critical illness.

Our study has several limitations. First, we relied on the use of ICD-9-CM coding to define our critically ill patient cohort and gastrostomy placement. The ICD-9-CM codes for gastrostomy tubes have been previously validated (1, 15). The algorithm used to define critical illness identifies severe diagnoses that are likely to be among top-billed diagnoses at discharge and has been used in prior literature to define critical illness as an alternative to simple admission to an intensive care unit (4, 12, 16, 17), which can be influenced by local practice variation and intensive care unit bed availability. Second, unmeasured differences in case-mix at hospitals may further reduce the limited between-hospital variation. Third, we could not assess drivers of the relative agreement between hospitals in gastrostomy rates; further research is needed to determine if these decisions are driven by patient values, low physician enthusiasm for recommending gastrostomy tubes, or external factors such as procedure reimbursement.

In summary, despite a lack of data or guidelines on gastrostomy placement in the critically ill, we found little between-hospital variation in rates of gastrostomy tube placement. As the total number of patients receiving gastrostomy tubes is high because of a large critically ill population, further studies delineating the benefits, morbidity, and long-term mortality are critical to ensure alignment of gastrostomy tube placement with patient prognosis and patient preferences in this growing population.