Gastrostomy Placement and Mortality Among Hospitalized Medicare Beneficiaries FREE

APPROPRIATE NUTRITION is fundamental to curative medical care. Poor outcomes accompany malnutrition in a variety of clinical settings.^1- 2 Consequently, physicians and other caregivers may consider enteral feeding appropriate not only for patients who are unable to swallow, but also when oral intake is inadequate, sometimes irrespective of the ability to eat. Whereas permanent feeding tube placement previously necessitated major surgery, the percutaneous endoscopic gastrostomy³ introduced in the early 1980s requires only sedation for insertion. Among older persons, a possible consequence of both attitudes toward nutritional support and the introduction of the percutaneous endoscopic gastrostomy has been a substantial increase in gastrostomy placement among hospitalized patients. For example, between 1988 and 1995, the number of gastrostomies placed in hospitalized patients aged 65 years or older in the United States increased from approximately 61000⁴ to 121000.⁵

Whether enteral feeding leads to improved quantity or quality of life, particularly among frail older persons, is unclear. Moreover, the use of feeding tubes among older individuals provokes considerable debate.^6- 8 Yet, despite controversy and increasing use of permanent feeding tubes among older persons, few population-based descriptions of gastrostomy use and outcomes have been published.^9- 10 Thus, the purpose of this study was to describe hospitalized older Medicare beneficiaries having gastrostomies placed and their outcomes using linked claims and vital status data.

Hospitalized Medicare beneficiaries aged 65 years or older discharged from the hospital in 1991 in whom gastrostomies were placed (International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] codes 43.1, 43.11, 43.19) were eligible for inclusion in the cohort. Patients enrolled in health maintenance organizations (HMOs) were excluded because of incomplete ascertainment in Medicare claims (6% of beneficiaries were enrolled in HMOs in 1991).^11- 12 Of 82650 patients discharged from the hospital with the recorded procedure codes, 1545 were repeated admissions; in these cases, only the first admission was included in the analyses. Hospital claims and Social Security Administration files matched to claims by the Health Care Financing Administration provided vital status information for 80108 (98.8%) of the 81105 older individuals with a recorded gastrostomy placement in 1991.

The cohort was described according to age, sex, race (black, white, or other), and primary and secondary diagnoses. Diagnoses and procedures were categorized according to summary classification schemes¹³ developed for use with health care claims and for medical effectiveness research. A diagnostic category was added for swallowing disorders (ICD-9-CM 787.2 [dysphagia], 783.3 [feeding difficulties], V41.6 [problems with swallowing and mastication]), and carcinomas in situ were excluded from the neoplasm category. Twenty-nine patients had ICD-9-CM code 43.1 (gastrostomy) recorded and were included as having percutaneous endoscopic gastrostomies (ICD-9-CM code 43.11). The discharge locations of individuals who were alive at hospital discharge were determined from claims records. Gastrostomy placement rates were calculated according to age, sex, and race (black and white) categories per 1000 eligible Medicare beneficiaries not enrolled in HMOs (Health Care Financing Administration, unpublished data, 1993).

Survival and mortality were calculated from the day of gastrostomy placement with crude 30-day, 1-year, and 3-year mortality rates estimated using the Kaplan-Meier method.¹⁴ Three years of follow-up were analyzed. To depict mortality rates for primary and secondary diagnoses, sex-specific age-adjusted and race-adjusted (for black and white classifications only) Cox models¹⁴ were fitted for each diagnostic category. Comparable age-adjusted and race-adjusted 30-day, 1-year, and 3-year mortality rates were calculated from the fitted models using the mean sex-specific covariate values for black and white patients.^15- 16

In 1991, a total of 81105 older Medicare beneficiaries were discharged from hospitals with gastrostomies (59969 percutaneous endoscopic gastrostomies and 21136 operatively placed). Likely reflecting demographic patterns, most patients were women (Table 1). About three quarters were aged 75 years or older. Slightly fewer than one third were discharged to home and approximately one half were transferred to long-term care facilities (ie, skilled nursing and intermediate care facilities). The in-hospital mortality rate was 15.3%.

Cerebrovascular disease was the most common primary diagnosis (Table 2). A primary or secondary cerebrovascular disease diagnosis was recorded for slightly fewer than one third of the sample, and in 29.3% and 27.7% of black and white patients, respectively. Diagnoses of neoplasms were commonly recorded, as were diagnoses of fluid and electrolyte disorders and pneumonia (with or without aspiration). Malnutrition was the listed primary diagnosis for 4.4% and was either a primary or secondary diagnosis in just over one quarter of patients. In addition, although evidence of benefit may be lacking, enteral feeding is sometimes considered for those with decubitus ulcers¹⁷ (a primary or secondary diagnosis in 10.5% of patients). In addition to gastrostomy placement, major gastrointestinal, cardiac, and respiratory procedures were performed in 15.2% of patients.

Age-specific, race-specific, and sex-specific gastrostomy placement rates for eligible Medicare beneficiaries are depicted in Figure 1. Placement rates increased considerably with increasing age and were markedly higher among black patients than white patients. For example, 1 in 131 white and 1 in 58 black non-HMO Medicare beneficiaries aged 85 years or older were discharged alive or deceased from a hospital in 1991 following gastrostomy placement.

The 30-day, 1-year, and 3-year mortality rates were substantial (Table 3). One year after gastrostomy insertion, 63.0%(95% confidence interval [CI], 62.7%-63.4%) of patients had died, with mortality rates reaching 74.5% (95% CI, 74.2%-74.8%) and 81.3% (95% CI, 81.0%-81.5%) at 2 and 3 years, respectively. The median survival for women was 28.9 weeks, whereas one half of men had died by 17.6 weeks. Mortality rates were higher among men and older patients; black patients had slightly lower 30-day mortality rates but higher 1-year and 3-year mortality rates. Despite their higher mortality rates, among those with available vital status information, men were younger than women. The mean (SD) age for men was 78.4 (7.6) years vs 81.8 (8.1) years for women (P<.001, t=59.9, df=80106). The average age of black patients was, however, similar to white patients (80.5 [8.4] years vs 80.6 [8.0] years, respectively; P=.46; t=.74; df=77081).

Thirty-day, 1-year, and 3-year mortality rates varied among the common primary and secondary diagnoses (Figure 2 and Figure 3). At 30 days, primary diagnoses of malnutrition and fluid and electrolyte disorders and secondary diagnoses of swallowing disorders, dementia, or cerebrovascular disease were accompanied by the lowest mortality rates. Thirty-day mortality rates were highest among those with primary diagnoses of nonaspiration pneumonia or influenza, secondary diagnoses of congestive heart failure, or any neoplasm. Among men and women, the patterns of mortality rates were somewhat similar at 1 year, but by 3 years they were inconsistent. Finally, excluding patients having major gastrointestinal, cardiac, or respiratory procedures yielded minimal absolute differences in mortality rates at 1 and 3 years (less than 3%) in all diagnostic categories except neoplasms in men and women and obstructive pulmonary disease in men, with higher mortality rates following the exclusion.

Few population-based studies have described gastrostomy use and patient outcomes in the growing older population.^9- 10 The current study indicates that among hospitalized older Medicare beneficiaries gastrostomies are placed frequently and mortality in the years following placement is high. The diagnoses and high mortality rates reflect the frailty of many older individuals in whom gastrostomies are placed. Whether the high mortality is caused by nutritional depletion, underlying diseases, or gastrostomy placement cannot be determined from these data; however, some observations are striking.

First, the frequency of gastrostomy placement is substantial (0.8% of white and 1.7% of black Medicare beneficiaries aged 85 years or older). Second, the difference between black and white patients noted here, although consistent with data from the National Hospital Cost and Utilization Project,¹⁸ is difficult to explain. Although the disparity might be attributed to different risks of experiencing a stroke,¹⁹ the proportions of black and white patients diagnosed as having cerebrovascular disease were similar. However, determining whether stroke occurrence can account for the ethnic disparity requires data not obtained in the current study. The difference could also be related to attitudes toward life-sustaining therapies,²⁰ with black patients possibly more inclined to desire enteral feeding.

Although these results may be intriguing, there are limitations to the use of administrative claims data. Although some diagnoses listed in Medicare claims records have been found valid,²¹ inaccuracies remain.^21- 22 It is also possible that for some patients, a temporary feeding tube preceding the gastrostomy led to a poor outcome. For some patients enteral feeding may have been started too late, with earlier nutritional intervention possibly being more effective in prolonging life²³ (gastrostomies were placed a median of 8 days following hospital admission). In addition, claims data include information regarding gastrostomy placement, but do not include information on how they are used for enteral feeding. Claims data also lack sufficient clinical information to examine factors such as weight loss and biochemical markers of malnutrition. Finally, confounding by indication is likely inherent in any observational study of enteral feeding; however, a randomized controlled trial examining the life-prolonging effect of enteral feeding is ethically inconceivable.

Limitations notwithstanding, this study should have implications for older patients, their families, and physicians. Particularly when an older individual fails to eat, these results could help inform the decision to place a gastrostomy. The substantial mortality rates may be reason to consider that some enterally fed patients who do not have swallowing disorders are not dying because of lack of nutrition, but rather, lack the need to eat because they are dying. Enteral feeding is at times a lifesaving procedure (for example, in older individuals with dysphagic stroke or who are not eating because of severe depression). Yet for others, gastrostomy placement appears to portend spending the final weeks or months of life artificially fed. The mortality and placement rates among these Medicare beneficiaries emphasize the need to critically examine potential benefits and risks of enteral feeding by gastrostomy in older individuals, particularly those nearing the end of life.