Current data suggest that gastroparesis is associated with an increased mortality, with reported rates ranging from 4% to nearly 40%. Considering this variability, the goal of this study was to determine mortality rates and risk factors for adverse outcomes in gastroparesis.
Using the diagnosis code for gastroparesis, admission rates, duration of hospitalizations, discharge status, and inpatient mortality were determined for emergency department encounters and admissions compiled in the Nationwide Emergency Department Sample and Nationwide Inpatient Sample of the Agency for Healthcare Research and Quality. Comorbid conditions, procedural evaluations, age cohort, and gender distribution were examined as potential risk factors.
More than 50% of the emergency encounters for gastroparesis resulted in admission with age, cardiovascular, renal, and infectious disorders, but not diabetes mellitus being associated with higher admission rates. Inpatient mortality was 1.2 ± 0.1%, was not negatively affected by diabetes mellitus as comorbidity, and increased with coexisting infections and with more aggressive therapy. Discharge status was similarly affected by comorbidities, treatment complications, and more aggressive therapy.
Conclusions & Inferences
These results demonstrate that gastroparesis does not come with a high mortality risk, with most deaths being due to comorbid conditions. Although gastrostomies and/or nutritional support were used in only a minority of admissions, the associated increase in morbidity and mortality highlights the need to carefully select the right candidates for such interventions and to discuss the common occurrence of adverse outcomes with patients.
Gastroparesis is a chronic disorder that significantly impairs the quality of life of affected individuals.[1, 2] Despite the often tremendous impact on the daily lives of patients, little is known about the prognosis of gastroparesis. Limited epidemiologic data suggest that most patients can be managed with dietary and medical therapy, with only about one quarter of the affected individuals receiving more aggressive interventions and/or requiring hospitalizations. Consistent with these results, we have recently reported that nearly 60% of patients seen in a tertiary referral center were not hospitalized during a follow-up period of 2 years. Information on mortality is similarly limited and confounded by often skewed populations seen in tertiary referral centers. Case series report highly variable findings, ranging from 4% in a mixed cohort of in- and outpatients followed for 2 years to 37% in patients with diabetic gastroparesis requiring nutritional support.[3-9] The only population-based study demonstrated an excess mortality compared with the expected survival, largely due to cardiovascular or other complications of diabetes; in contrast, idiopathic gastroparesis was not associated with increased mortality. Consistent with such a negative prognostic role of diabetes mellitus, one large multicenter study demonstrated differences between diabetics and patients with idiopathic forms of gastroparesis, with hospitalizations rates between 70% for patients with insulin-dependent diabetes compared to about 40% for idiopathic gastroparesis. Focusing on diabetic patients, prior studies concluded that the coexisting presence of impaired gastric emptying is associated with more frequent hospitalizations, showing the burden of gastroparesis in this subgroup.[11, 12] Overall, the limited evidence thus suggest that hospitalizations and/or the need for more complex or invasive therapy may affect a significant number of individuals with gastroparesis, with diabetes mellitus being a negative prognostic marker.
Considering the variability of data and the likely impact of tertiary referral bias in most published case series, admission and mortality rates were examined in patients with gastroparesis, using the data repository of the Agency of Healthcare Research and Quality. The underlying hypothesis was that admissions due to gastroparesis associated with diabetes mellitus (used as surrogate for diabetic gastroparesis) correlate with a higher morbidity and mortality. The specific aims were to determine: (i) hospitalization rates in patients seen in emergency departments for gastroparesis; (ii) inpatient mortality and morbidity (operationally defined by length of stay and discharge status as surrogate markers) for patients admitted with the diagnosis of gastroparesis; (iii) factors associated with increased inpatient mortality, prolonged hospitalization and/or need for ongoing care after discharge in patients admitted with gastroparesis.
To determine risk factors for hospitalization and adverse outcomes, the Nationwide Inpatient Sample (NIS) and the Nationwide Emergency Department Sample (NEDS) databases of the Agency for Healthcare Research and Quality were searched. The NIS contains a compilation of data from more than 1000 hospitals across the United States; the NEDS provides complementary information abstracted from about 30 million emergency room visits in nearly 1000 distinct emergency departments. Annual emergency encounters and admissions were extracted using International Classification of Diseases (ICD) 9 code 536.3 for gastroparesis as primary or secondary diagnosis. Aggregate data were abstracted using the publically available data sets. Based on the search algorithm, primary diagnosis refers to the diagnosis code responsible for the admission. Thus, only a single primary diagnosis code can be associated with any given encounter/admission. In contrast, several secondary diagnoses may be related to a single encounter/admission. The data provide information about age cohorts, gender, and insurance status. Associated procedures and diagnoses were retrieved based on the Clinical Classification Software that bundles codes into a limited number of clinically meaningful groups. Inpatient mortality was compared with other gastrointestinal disorders, using the ICD-9 codes for functional dyspepsia (563.6),functional diseases of the colon and anorectum (irritable bowel syndrome and constipation: 564.0 and 564.1), gastric ulcer with and without complications (531.0–531.9), cholelithiasis with and without complications (574.0–574.1), diverticular disease (562.1) Crohn's disease (555.0–555.9), ulcerative colitis (556.0–556.9), gastro-esophageal reflux disease (530.81–530.83), and chronic pancreatitis (577.1). Gastric electrical stimulation is used as a treatment in patients with refractory gastroparesis. As there is no ICD-9 code for gastric electrical stimulation, the procedural code 04.92 (insertion of a peripheral neurostimulator) was used to determine complications and associated diagnoses and procedures. To approximate the number of gastric electrical stimulator implantations, this information was correlated with diabetes mellitus and endoscopy as secondary codes. Nationwide Emergency Department Sample data were available and analyzed for the years 2006 through 2009, NIS data combining primary with secondary diagnoses or procedures were obtained for 2007 through 2010. During this time period, the number of states included into the NIS data bank increased from 40 to 45. To account for these differences, all results were normalized to the total number of encounters.
Annual data were averaged over the 4-year observation period. Although changes in the NIS data repository require the use of weighted data to determine time trends, the analysis did not focus on trends and was restricted to a period not requiring such adjustments based (http://hcup-us.ahrq.gov/db/nation/nis/nistrends.jsp). All data are given as mean with standard errors. The number of secondary diagnosis was limited to the 20 most common diagnoses and collated for the timer period of the study. Only comorbidities consistently listed during the entire period were included in the final tabulation of data. Differences were assessed using t-test and anova with the Holm–Sidak or Dunn's method to correct for multiple comparisons (SigmaStat 2.0; SPSS, Chicago, IL, USA).
Emergency department encounters
About 200 000 annual encounters listed gastroparesis as diagnosis, with it being the primary diagnosis in slightly less than 10%. The majority of emergency encounters resulted in hospitalization (Fig. 1A). Consistent with the known epidemiology of gastroparesis, women accounted for the majority of emergency encounters (65.7 ± 0.2%) with admission rates not differing based on gender (women: 75.7 ± 1.2% vs men: 78.1 ± 1.0). As shown in Fig. 1B, most patients were young and middle-aged adults. When normalized by the contribution of each cohort in the overall volume of emergency encounters, admission rates significantly increased with age (Fig. 1B).
This age dependence of hospitalizations points at the importance of comorbid conditions, which were examined next. The most common comorbidities could be grouped into separate categories, with diabetes mellitus being most frequently noted, followed by cardiovascular problems, disorders of electrolyte or fluid homeostasis or renal failure, other gastrointestinal illnesses, infections, and complications of medical treatment (Fig. 2). Interestingly, an associated diagnosis of diabetes mellitus did not increase the risk of admission, whereas heart or renal failure, gastrointestinal bleeding, coexisting infections, complications of prior treatment, and – in gastroparesis as primary diagnosis – mood disorders increased the likelihood of inpatient treatment (Table 1).
Table 1. Comorbid conditions and admission rates in patients with gastroparesis seen in emergency departments, normalized as percentage of overall encounters. Blank fields indicate that these comorbidities were not among the 20 most commonly listed conditions. anova testing demonstrated significant differences between groups corrected for multiple comparisons
*Denotes differences from the entire cohort; †denotes difference between gastroparesis as primary and secondary diagnosis (only assessed for overall admission rates).
Mortality of patients admitted for gastroparesis was at 1.26 ± 0.13% with significant differences between admissions for gastroparesis as primary and secondary diagnosis (0.44 ± 0.03% vs 1.26 ± 0.13%; P < 0.01). Compared with other gastrointestinal disorders, inpatient deaths were more common in gastroparesis than in gastro-esophageal reflux and Crohn's disease (Fig. 3A).The coexistence of diabetes mellitus was not associated with an increase in mortality, whereas severe infectious complications (sepsis) correlated with a higher likelihood of inpatient death (Fig. 3B).
Complications of devices, medical or surgical therapy increased hospitalization rates for emergency encounters (Table 1) and accounted for 5.0 ± 0.2% of the primary diagnoses in patients admitted for gastroparesis. Therefore, the association between procedures and mortality was assessed next (Table 2). Endoscopic procedures were the most commonly performed interventions with about one third of admissions with gastroparesis as the primary and about one fifth with gastroparesis as the secondary diagnosis being associated with an upper endoscopy. Other procedures associated with the underlying diagnosis were endoscopic or surgical insertion of a feeding tube, performed in about 2–4% of the admissions, and initiation of nutritional support, implemented in 3–5% of the hospitalizations. Many interventions targeted comorbid or confounding problems. About one in six admissions led to a vascular procedure other than coronary angiography, whereas cardiac catheterization was performed in about 2.5% of the admissions. Patients underwent hemodialysis in about 13% of the hospitalizations. Surgical interventions ranging from exploratory laparotomies with adhesiolysis to cholecystectomy or gastrectomy were performed in about 1–2% of the admissions. A group-wise comparison revealed significant differences in mortality with aggressive supportive therapy with intubation and/or mechanical ventilation, the insertion of a gastrostomy tube and the use of nutritional support being associated with higher rates of inpatient death (Table 2).
Table 2. Use of diagnostic tests and procedures in patients with gastroparesis as primary or secondary diagnosis expressed as percentage of admissions; mortality is displayed for patients with secondary diagnosis only and also given in percentages (normalized by admissions). Blank fields indicate that these interventions were not among the 20 most commonly listed procedures. Gastrectomy rates were only available for 1 year (††). anova demonstrated that mortality significantly differed between groups (P < 0.001); after correcting for multiple comparisons, only mechanical ventilation, insertion of a gastrostomy, and nutritional support were associated with increased mortality
Primary diagnosis (%)
Secondary diagnosis (%)
*Denotes a difference from the entire cohort; †denotes difference between gastroparesis as primary and secondary diagnosis.
Length of stay and discharge status may function as surrogate markers of severity and/or complexity of illness and its treatment. Men trended to have a slightly shorter stay compared with women (P = 0.06; Fig. 4A), but discharge status did not differ based on gender. Age significantly affected length of stay and discharge status (Fig. 4B). After correcting for multiple comparisons, patients with 18 and 45 years of age were discharged earlier than geriatric patients and together with pediatric patients had higher rates of regular discharge (P < 0.01). Conversely, geriatric patients were significantly more likely to be transferred to nursing homes or require homecare than young adults (P < 0.01). As shown in Table 3, the length of stay was about 1 day longer in admissions with gastroparesis as secondary compared gastroparesis as primary diagnosis. The coexistence of diabetes mellitus or complicating cardiac or renal problems did not prolong hospitalizations, while infectious processes and complication of medical therapy carried a significant burden with higher rates of nursing home transfers, while diabetes mellitus, hypertension or the diagnosis of gastritis or duodenitis were associated with lower likelihood of such transfers.
Table 3. Impact of comorbid conditions on the duration of hospitalization (data given in days) and discharge status (data given as percentage of all admissions). anova testing demonstrated significant differences between groups
Gastroparesis primary diagnosis
Gastroparesis secondary diagnosis
*Denotes differences from the entire cohort after correcting for multiple comparisons; †denotes difference between gastroparesis as primary and secondary diagnosis (only assessed for overall length of stay and discharge status).
The picture that emerged when relating procedures to duration of inpatient stays and discharge status mirrored the results seen when assessing mortality. Initiation of nutritional support and/or placement of a gastrostomy or other feeding tube more than doubled the duration of hospitalizations and changed the discharge status with most admissions resulting in need for ongoing care through nursing homes or homecare (Table 4). The need for other supportive therapy, such as blood transfusion, hemodialysis or ventilator therapy, was similarly associated with lower rates of routine discharge, but did not significantly affect length of stay compared with the group as a whole after correcting for multiple comparisons.
Table 4. Impact of diagnostic or therapeutic procedures on the duration of hospitalization (data given in days) and discharge status (data given as percentage of all admissions). anova testing demonstrated significant differences between groups
Gastroparesis primary diagnosis
Gastroparesis secondary diagnosis
*Denotes differences from the entire cohort after correction for multiple comparisons.
Annually, 1733 ± 154 admissions listed implantation of a peripheral neurostimulator as the primary diagnosis with epilepsy and Parkinson's disease being the associated diagnoses in 30.8 ± 1.8% and 10.4 ± 1.7%, respectively. Diabetes mellitus with complications accounted for 10.6 ± 1.8% of the associated diagnoses, which correlated reasonably well with the reported rate of endoscopies of 9.2 ± 2.6% as related procedure. Device-related complications were the primary reason for admission in 10.2 ± 0.9%.
Despite the increasing recognition of gastroparesis as a chronic, mostly idiopathic disorder, relatively little is known about its prognosis with mortality data ranging from 4% to nearly 40%.[3-9] This analysis of emergency department and hospitalization data adds important information, suggesting a more benign course with a low likelihood of death during hospitalizations and also a relatively low fraction of admissions involving more aggressive interventions, such as placement of feeding tubes or initiation of nutritional support.
Emergency department encounters
The first goal of this study was to determine admission rates in patients with gastroparesis seen in emergency departments, which ranges around 80%. As expected, comorbid conditions and/or potential complications of gastroparesis affect the likelihood of hospitalization. Coexisting cardiovascular, renal or infectious problems will lead to admissions in nearly all instances as will the presence of gastrointestinal bleeding or anemia. The importance of infections has previously been described in a case series of patients admitted for treatment of gastroparesis. Surprisingly, mood disorders fell into the same category, being associated with a high likelihood of admissions from the emergency departments of about 85% without correlating with worsened outcome as defined by mortality or discharge status. Although a compilation of related diagnosis codes cannot answer how mood disorders influence hospitalizations, a correlation between affect and subjectively perceived or physician-rated severity of gastroparesis had previously been reported in two case series and may thus explain this finding.[2, 14]
Inpatient mortality in gastroparesis
Considering the tremendous variability in mortality described in case series on gastroparesis, the second goal was to determine mortality rates using this large national data bank. As NIS data compile information from rural and urban areas, small and large centers independent of their role as teaching or university hospitals, tertiary referral bias that often skews published case series is excluded. Interestingly, the results fall well below the previously reported findings that range from about 4% in a sample in and outpatients to nearly 40% obtained in groups of hospitalized patients undergoing implantation of jejunostomy catheters for nutritional support.[4, 8] Considering the fact that most patients with gastroparesis can be managed as outpatients and will not require hospitalizations,[3, 4] an approach based on hospitalization records by definition excludes this large fraction of the potential patient base and may still skew data. Despite this caveat, most of deaths in the United States are still occurring in hospitals, thus providing us with at least with an approximation of the mortality associated with gastroparesis.
When compared with other gastrointestinal disorders, inpatient mortality in patients with gastroparesis exceeds the rates seen in gastro-esophageal reflux and Crohn's disease, with the latter being associated with higher likelihood of death compared with reference populations.[16, 17] While the different approach based on admissions rather than a defined number of individuals limits the ability to compare results, the data are consistent with the those from the only population-based study, which demonstrated that gastroparesis is indeed associated with shorter life expectancy.
Surrogate outcome measures of disease severity
The NIS data also provide the opportunity to assess indirect measures of disease severity by examining length of stay and the need for ongoing treatment in nursing home or via homecare delivery. More than two thirds of the hospitalizations resulted in routine discharges with the remaining patients requiring some form of continuing care, mostly in nursing homes, rehabilitation centers or in the form of homecare support. The need for such ongoing treatment is even lower in patients under 65 years of age, which account for more about 80% of the admissions for gastroparesis.
Factors associated with adverse outcomes
The next goal of this study was to identify factors associated with adverse outcomes, in the form of either higher mortality rates or the predefined surrogate measures of disease burden. Focusing first on diabetes mellitus, which is not only the most commonly associated comorbidity, but likely caused gastroparesis in many of these patients, there was no association with worse outcomes. These results stand in contrast with prior reports from tertiary referral centers showing higher hospitalization rates,[10, 13] and also differ from the only published population-based study, which demonstrated excess mortality in patients with diabetic, but not idiopathic gastroparesis. The NIS focuses on admissions rather than patients and does by definition not differentiate multiple admissions of single individuals from single admissions of multiple individuals. Therefore, the similar mortality rates in admissions due to diabetes mellitus and gastroparesis could potentially be significantly higher when assessed per patient if a substantial percentage of patients are hospitalized repeatedly within the time frame of a year. While more frequent hospitalizations for diabetic patients are common in patients seen in tertiary referral centers,[10, 13, 18] the relative frequency of diabetes mellitus as listed comorbidity matches the prevalence of diabetic gastroparesis in most larger case series, arguing against repeat hospitalizations as a major confounder.[4, 5, 14, 19] The results extracted from the large data repository of the NIS are also consistent with a previously published smaller study of gastroparesis-related hospitalization in North Carolina, which concluded that diabetic patients may account for a significant proportion of admissions due to gastroparesis, but that these patients rarely face adverse outcomes. Longitudinal studies of patients with diabetic gastroparesis followed for 10 years similarly argue against gastroparesis as a negative prognostic indicator.[9, 21] Taken together, the results suggest that the diagnosis of diabetic gastroparesis does not come with a very negative outlook for most of the affected patients.
Infectious disorders, cardiac, and renal diseases surfaced as additional and important confounders, associated with increased admission rates, lower rates of regular discharge and increased mortality. These results are consistent with previously reported factors contributing to the excess mortality in gastroparesis. One of the most striking findings of this study was the role of device and treatment complications, which were listed as the primary diagnosis in more emergency encounters and hospitalizations than cardiac or renal problems and which were associated with more frequent referrals to nursing homes or homecare agencies. The burden of hospital-acquired infections or adverse effects of therapies has been demonstrated repeatedly and contributes significantly to morbidity and mortality of hospitalized patients.[22-24] The NIS does not provide sufficient details to systematically investigate different treatments and their potential impact on infections or other adverse outcomes. However, the data repository includes information about comorbid conditions, procedural evaluations, and therapies, which at least allows some inferences. Sepsis, pneumonia, procedures to gain vascular access, and the need for intubation and/or mechanical ventilation were among the 20 most common comorbid conditions and all associated with poor outcomes.
Another striking finding was the association between endoscopic placement of gastrostomies, surgical enterostomies, and initiation of nutritional support, all more aggressive treatments for gastroparesis, are associated with increased mortality or other adverse outcomes. Although patient selection potentially confounds these results, published case series consistently demonstrated high complication rates of such interventions, suggesting that procedure-related complications likely contribute to these results.[8, 25-29] Only a minority of patients underwent these procedures. Nonetheless, these findings should caution healthcare providers and patients when deciding on these more aggressive therapies for gastroparesis.
Gastric electrical stimulation has been proposed as a therapeutic option for patients with refractory gastroparesis,[6, 30-34] even though results of randomized trials with periods of sham intervention show only marginal or no benefit during the blinded phase of the trial.[30, 35, 36] The lack of a separate procedural code for the use of gastric electrical stimulation did not allow a reliable identification of this procedure and related outcomes. However, the correlation between diabetes mellitus with complications and endoscopy as associated diagnosis or procedure, respectively, likely approximates the number of patients with gastroparesis and thus suggests that only a small number of admission are related to this procedure. With the small number, no meaningful outcome data could be extracted. However, device-related complications were among the most commonly listed diagnoses related to this procedural code and fall within the range of previously reported case series.[18, 30, 37, 38] While at best an approximation, that data fit the pattern of more aggressive interventions correlating with worsening outcomes. Although associations do not establish causality or directionality of interactions, these results still point at potential adverse effects that need to be considered and viewed in the context of marginal results in appropriately designed trials when discussing this therapy with patients.
Data extracted from the NIS and NEDS provide detailed information of about emergency encounters and admissions independent of insurance coverage, age or gender, covering about 25% of all hospitalizations from all geographically distinct areas within the United States. Such information thus avoids skewing data due to tertiary referral or selection bias based on age, insurance coverage or economic problems. However, most patients with gastroparesis do not require hospitalizations, which form the basis of this large data repository. While we are left with a less negative outlook based on the NIS data, the focus on inpatient treatment still skews data. When looking at the data, we have to recognize some additional important limitations. As is true for all studies based on diagnostic codes, the validity of diagnostic coding is unclear. The relatively high frequency of diagnostic tests and therapeutic interventions targeting other organs or systems certainly demonstrates the importance of potential confounders and comorbid conditions, which can in turn influence morbidity and mortality. Many of these comorbidities may indeed be interrelated or could even be caused by gastroparesis. For example, repeated vomiting and inability to maintain oral intake may lead to fluid and electrolyte disorders. Similarly, renal failure, hypertension, and other cardiovascular disorders will often coexist and may all be related to underlying problems, such as diabetes mellitus. However, the age and gender distribution of NIS and NEDS data matches results of published case series of patients with confirmed gastroparesis.[4, 5, 14, 19, 41, 42] The data banks are based on encounters rather than individual patients, thus potentially confounding the results due to repeat admissions. While the current approach highlights several important variables associated with adverse outcomes, only more complex statistical modeling will provide more detailed insight into independent predictors.
Despite their limitations, these results highlight several important points. First, the diagnosis gastroparesis does not come with a high mortality risk, with most death being due to comorbid conditions, a conclusion that also holds for patients with diabetic gastroparesis. This information should be reassuring for patients and is important for healthcare providers, as we need to weigh the risks and benefits of interventions in the context of likely outcomes. The second key finding relates to such risk-benefit considerations, as the use of more aggressive therapies, be it in the form of gastrostomies, enterostomies or nutritional support, is associated with significant morbidity and mortality. Beyond selecting the right candidates for such interventions, we need to discuss these common adverse outcomes with patients, acknowledging that such therapies may correct nutritional or other problems, but potentially lead to new symptoms or complications, which often necessitate further interventions and/or hospitalizations and may even have a negative impact on overall survival.