Variations Between Hospitals in Antireflux Procedures in Children FREE

Gastroesophageal reflux disease (GERD), the pathologic and symptomatic reflux of gastric content into the esophagus, affects about 7% of infants in the first year of life.^1- 2 The indication for referral for surgical therapy for GERD is most commonly a failure to respond to medical management, though some are also referred based on their medical comorbidities for prophylactic procedures, given the perceived natural history and risks associated with these underlying problems.³

Antireflux procedures (ARPs) are common in children and are being performed with increasing frequency.^3- 4 Many studies described excellent outcomes after ARP, with greater than 90% resolution of symptoms and improved quality of life.^5- 6 Given the prevalence of GERD and the reported success of ARP, it is not surprising that, historically, it has been described as one of the most common surgical procedures performed. A recent study examining national trends in the use of ARP, however, concluded that the indications for ARPs have “evolved during the laparoscopic era,” in that fewer operations are performed on neurologically impaired children or on children after infancy.⁷ A recent study used the Nationwide Inpatient Sample coupled with census data to identify national population-based trends with respect to ARP. Citing a study that demonstrated variation between individual hospitals and national trends, the authors also stated that they were not able to address how local variations in the use of ARP may affect infants and children.⁸

While on the macrolevel it is extremely helpful to identify the national trend in the rate of ARPs, it raises many questions. Among them is whether this trend is universal or if it is specific to a region, hospital, group practice, or even to an individual surgeon. While a moderate amount of variation in practice patterns between local groups or individuals is expected, large variations might suggest either similarly large variation in the indications for the operation or in the preoperative assessment and decision making. This is important, in that to identify valid measures of outcomes of an intervention like ARP beyond universal measures such as death or length of stay, we must first identify the indications. For example, if children have an ARP for aspiration pneumonia, it would be ideal to identify a cessation of events of aspiration pneumonia after the operation. A recent article demonstrated that when evaluating ARPs by more specific disease-related outcomes such as rate of hospitalizations for GERD-related diagnoses like aspiration pneumonia, some children were hospitalized more frequently after compared with before.⁹

No published articles to date describe clustered or regional prevalence of regurgitation or GERD and, similarly, no data suggest clustered or regional severity of disease. Given the lack of demonstrable clustering of the presence or severity of GERD across the United States, if the indications for ARP are standard, one would expect that ARPs would be performed equally often in pediatric hospitals nationwide. Our hypothesis, therefore, is that a critical evaluation of the incidence of ARPs should not exhibit regional or clustered patterns. The aim of this study is to examine differences in the annual rate of pediatric antireflux operations across individual pediatric hospitals over time.

We performed a retrospective examination of pediatric hospital discharge data relevant to antireflux operations. Issues explored included differences across hospitals and trends over time. Hospital-level denominator data were not feasible given the absence of clearly defined catchment areas for admissions to free-standing children's hospitals. As a result, we used comparisons with admissions for other relatively common gastrointestinal surgical procedures as a proxy, assuming that the referral patterns and incidence of these comparison operative procedures should remain relatively stable within institutions over time. For example, as the incidence of appendicitis in children would be expected to stay relatively constant over the study period, a decrease in the ratio of antireflux operations to appendectomy procedures within a hospital would suggest a drop in the frequency of antireflux operations.

We used the Pediatric Health Information System (PHIS) database developed by the Child Health Corporation of America that includes demographic, diagnostic, and charge data from 36 freestanding, noncompeting children's hospitals.¹⁰ These hospitals include private as well as University pediatric tertiary referral centers. The PHIS includes diagnosis and procedure codes using the International Classification of Diseases, Ninth Revision, Clinical Modification(ICD-9-CM) format. In a secondary post hoc analysis, we also examined the incidence of antireflux operations in the Healthcare Cost and Utilization Project Kid's Inpatient Database (HCUP KID) for the year 2000. The sole purpose of this post hoc analysis was to estimate the overall proportion of pediatric antireflux operations performed in the United States that are included in the PHIS database.

Our study included pediatric patients younger than 18 years with discharge dates between January 1, 2001, and March 31, 2006. The primary sample included all patients with procedure codes for ARP (ICD-9 procedure codes 44.66-44.67). Comparison samples were drawn for patients meeting the same age and date criteria, with procedure codes for appendectomy, pyloromyotomy, and gastrostomy; individual patients may be represented in more than 1 category, although this was rare. Unique patient identifiers were used, however, to ensure that an individual patient only counted toward a category for the first admission, even if subsequent admissions occurred with the same procedure code.

Other variables examined included demographic characteristics (age, sex, Medicaid status), hospital stay characteristics (year of discharge, length of stay, intensive care unit use, laparoscopic procedures), and relevant comorbid conditions.

The distribution of collected variables was described across each of the operative categories. The ratio of antireflux operations to each of appendectomy, pyloromyotomy, and gastrostomy procedures were then calculated by year at the hospital level. As only one-fourth of the data were available for 2006, ratios were only calculated for the years 2001 through 2005. The distribution of ratios across hospitals was then examined, calculating the median, range, and interquartile range.

To test for changes in the procedures over time within the hospitals, a series of linear regression analyses were performed with hospital as a fixed effect. The unit of analysis for these regressions was a given year of data within an individual hospital, and the coefficient for the discharge year was examined. Again, only data for the years 2001 through 2005 were included for these analyses, as the outcome of interest was the total number of antireflux operations for the year. Four regression models were tested, controlling for (1) the number of appendectomy procedures during the year for each hospital, (2) the same for pyloromyotomy and (3) gastrostomy, and (4) including the annual figures for all 3 procedures. These analyses allowed us to estimate the annual change in the number of antireflux procedures across hospitals.

In a secondary post hoc analysis, we used the KID data with population-based sample weights to estimate the total proportion of pediatric admissions for each procedure that were captured within the PHIS database, with the aim of adding context to the overall findings. Unique patient identifiers were not available within the public-use KID data files, however, so repeat admissions of the same child within a category could not be excluded.

This study was approved by the institutional review board at the Children's Hospital and Regional Medical Center in Seattle, Washington.

During our study period, 13 691 ARPs, 41 441 nonincidental appendectomies, 14 895 pyloromyotomies, and 23 527 gastrostomy tube placements were identified. More than half of the patients undergoing each operation were male (Table 1). Of the patients undergoing an ARP, 48% had at least 1 hospital day in the intensive care unit during their hospitalization. Thirty-nine percent of the ARP population had a diagnostic code consistent with a neurologic condition or developmental delay, 33% had a diagnosis of aspiration pneumonia, and more than half had failure to thrive. Barrett esophagus was associated with 0.1% of patients. Of note, though only 11% of patients were identified as having had an operation performed laparoscopically, an ICD-9 code specific to this approach was only introduced in late 2004. In the years following for which we have data in PHIS, 44% of cases were performed laparoscopically.

Between 2001 and 2005, the average number of ARPs per appendectomy, pyloromyotomy, and gastrostomy tube placement declined annually across free-standing pediatric institutions (Table 2). For example, the 2001 ratio of ARP to appendectomy was 0.36, declining to 0.26 in 2005. This change is indicative of either a decrease in the number of ARPs, an increase in the number of appendectomies, or a combination of both. Similarly, the ratio for pyloromytomy declined from 0.96 to 0.76, and that of gastrostomy tube placement from 0.67 to 0.47. The repetition of the decline across ratio categories suggests that a true decrease in ARP rates was taking place. When we visually examine the frequency of ARP per appendectomy, pyloromyotomy, and gastrostomy placement by individual hospitals, we identify considerable interhospital variability in these ratios (Figure 1). Each dot on these graphs represents an individual hospital within PHIS. We do not observe a strong correlation between the numbers of ARPs performed per year per hospital relative to the other control procedures.

The regression analyses demonstrated that the frequency with which ARPs are being performed in freestanding pediatric hospitals is significantly decreasing over time relative to appendectomies, pyloromyotomies, and gastrostomy placement (Table 3). Within each hospital, an average of 2.9 fewer ARPs are being performed each year relative to the number of appendectomies (P < .01), and more than 4 fewer relative to the number of pyloromyotomies and gastrostomy tube placements each (P < .001). When these annual changes are examined within each hospital individually, however, it appears that such changes are not occurring equally everywhere (Figure 2). At one extreme, one of the hospitals is performing over 40 fewer ARPs per year relative to the number of appendectomies, pyloromyotomies, and gastrostomy tube placements per year, whereas at the other extreme, one of the hospitals is performing over 10 more ARPs relative to these common procedures.

In a post hoc analysis, we used the KID data to estimate the proportion of pediatric admissions for these procedures nationwide that were captured by the PHIS database. As the KID database is not available for every year, the closest comparison was between the 2000 KID data and the 2001 PHIS data. Assuming no significant change over the single year, we can estimate that the PHIS database captured 37% of all Nissens, 8% of all appendectomies, 13% of laparoscopic appendectomies, 24% of pyloromyotomies, and 36% of gastrostomies placed in patients younger than 18 years in the United States.

The number of ARPs being performed in 36 free-standing children's hospitals is decreasing each year relative to several commonly performed operations performed at these same institutions. Despite this overall annual decrease, there is tremendous variation between individual hospitals in how frequently ARPs are being performed relative to these procedures. These findings compel us to question our a priori assumptions that either (1) the prevalence and severity of GERD or (2) the indications for ARP are consistent across populations and between hospitals.

Prior studies have addressed the prevalence of childhood regurgitation and GERD across populations.^11- 13 Though no environmental factors have been clearly identified as responsible for the development of GERD, there may indeed be cultural, environmental, or ethnic factors that influence the natural history and geographic distribution of this disease.^14- 15 Most of the hospitals contributing to the PHIS database, however, are large referral centers that draw from diverse ethnic and socioeconomic populations, and it is therefore unlikely that the variability observed in our data are entirely due to significant differences in each hospital's population. In fact, we looked at variables that might be associated with hospital-level characteristics that might explain the observed variation such as census region, population size of each hospital's metropolitan area, and bed size of the hospital. Neither region nor population were correlated with the outcome, and the association with bed size was not statistically significant (P = .08, with hospitals with >350 beds more likely to exhibit a decrease). We think it is more likely that the indications for ARP vary significantly between pediatric hospitals. As ARPs are major operations that can be associated with significant benefits as well as significant complications, these are important findings in that inconsistent indications may translate into 2 possible adverse scenarios—one in which children with unremitting GERD that would benefit from this intervention are not receiving it, and another in which children undergo these operations despite having the potential benefits outweighed by the risks.

There is much indirect evidence to support this concern. Recent studies have questioned the excellent outcomes of ARP in infants and children.^9,16- 20 Gastroesophageal reflux disease is poorly understood in the pediatric population, and often a diagnosis made clinically by reporting of subjective symptoms rather than objective evidence.¹ Additionally, the symptoms leading to the diagnosis can vary significantly with age and underlying medical conditions.²¹ Though several articles have linked objective studies with the reported symptoms, use of these studies preoperatively is neither universal nor standardized, as it is in adults.²² In fact, a publication reviewing more than 7000 ARPs performed in 7 major hospitals over 20 years demonstrates that the most frequent objective measure of reflux used preoperatively was upper gastrointestinal series, and was performed in 68% of the population. The next most commonly used study was a pH probe, used in 54% of patients.⁵ In terms of the clarity of the link between subjective symptoms and objective measures of GERD in children, another recent study demonstrated that only 52% of children evaluated for symptoms suggestive of GERD had a positive pH test, which has been considered the diagnostic gold standard.²³

Though many articles have been written describing the excellent outcomes after these operations, most are based on subjective measures of outcome, and few use objective measures to evaluate their efficacy.^{5- 6,24- 27} This becomes especially problematic in light of the recent publication by Lee et al.²⁰ The authors found that not only are patients hospitalized with equal frequency before and after an ARP but, more importantly, many patients who were previously without specific symptoms ended up hospitalized with that symptom postoperatively. For example, these authors found that only 3 of the 24 patients who were hospitalized with aspiration pneumonia before ARP were hospitalized after ARP with this same diagnosis. They found, however, that an additional 20 patients that had never been hospitalized for aspiration pneumonia were hospitalized for this diagnosis after their ARP. The next obvious questions in light of the marginal association between symptoms and GERD and objective studies and GERD is, what are the indications for these procedures and what measures of outcome should we use to evaluate their efficacy? Should we aim to relieve the specific GERD-related symptom for which an individual was referred, should we aim to relieve any GERD-related symptom including the one for which the patient was referred, should we focus on normalizing objective measures of GERD such as pH probe results, or should it be a combination of these things?

There are several limitations to this study. First, we use an administrative database that does not capture all of the possible clinical data. As a result, we are using this information to identify disparity in clinical practices across hospitals, to suggest the need for a more standardized approach, and to use these associations to design better prospective studies. Second, the hospitals in this database are large referral centers, and we cannot account for possible changes in referral patterns over the years of this study. We attempted to account for this possibility by using patients who had 3 different procedures (appendectomy, pylorolyotomy, and gastrostomy tube placement) as reference populations.

Despite these limitations, this article has some important implications. The significant disparity in the use of ARPs across free-standing children's hospitals in this country relative to other common procedures raises the question of whether too many or too few operations are being performed. There is little doubt that infants and children benefit from both medical and surgical interventions for GERD. It is a disease that is driven clinically by symptoms that are variable and diverse, such as pain, aspiration, acute life-threatening events, apnea, and failure to thrive, each of which is often poorly specific to GERD, and each of which is not clearly associated with a positive objective study. If the variability between hospitals is due to variability in the indications, then our next task is to clarify the indications for ARPs. We need to clarify the association between symptoms and the disease. We need to identify which objective studies will confirm the presence of GERD, and the relationship between specific symptoms and each of these objective studies. We need to develop disease-specific and population-specific measures of not only the quality of life of patients, but also measures of the impact of GERD on the patient and family (S. C. Acierno, et al, unpublished data, December 2008). Only then will we be able to identify measures of outcomes appropriate and specific to this population. This should be done by identifying a cohort of patients referred for GERD-related symptoms, and prospectively following up this group through a diagnostic and therapeutic algorithm that will carefully record underlying medical conditions, presenting symptoms, objective results, and the interactions between these variables that might predict the optimal interventions.

Correspondence: Adam Goldin, MD, MPH, Pediatric General and Thoracic Surgery, Children's Hospital and Regional Medical Center, M/S W-7729, PO Box 5371, Seattle, WA 98105-0371 (adam.goldin@seattlechildrens.org).

Accepted for Publication: November 3, 2008.

Author Contributions:Study concept and design: Goldin, Garrison, and Christakis. Acquisition of data: Goldin, Garrison, and Christakis. Analysis and interpretation of data: Goldin, Garrison, and Christakis. Drafting of the manuscript: Goldin. Critical revision of the manuscript for important intellectual content: Goldin, Garrison, and Christakis. Statistical analysis: Goldin, Garrison, and Christakis. Obtained funding: Goldin and Christakis. Administrative, technical, and material support: Goldin and Garrison. Study supervision: Garrison and Christakis.

Financial Disclosure: None reported.