Number of surgeons (A) or number of hospitals (B) performing differing numbers of pyloromyotomies in 2000.
Number of pyloromyotomies performed by surgeons (A) and hospitals (B) at each volume in 2000.
Ly DP, Liao JG, Burd RS. Effect of Surgeon and Hospital Characteristics on Outcome After Pyloromyotomy. Arch Surg. 2005;140(12):1191-1197. doi:10.1001/archsurg.140.12.1191
Previous studies have suggested that the outcome after pyloromyotomy is improved with increased surgeon experience. Others have proposed that infants with pyloric stenosis are best treated by specialty-trained pediatric surgeons or at children’s hospitals.
Surgeon and hospital characteristics affect complications, length of stay, and hospital charges after pyloromyotomy.
Data for a nationally representative sample of infants (n = 1277) who underwent pyloromyotomy in 2000 in the United States were obtained from the Kids’ Inpatient Database. Surgeon and hospital volumes were stratified into quintiles. Multivariate analyses were performed to analyze the impact of surgeon and hospital volume on length of stay, charges, and major operative complications using models that accounted for the hierarchical structure of patient-, surgeon-, and hospital-level covariates.
No association between surgeon volume and either length of stay or charges was observed. Higher surgeon volume, however, was associated with fewer complications (P<.001). Surgeons with the highest volume had a 90% lower risk of complications than those with the lowest volume. Higher hospital volume was associated with shorter length of stay (P<.001). No association between hospital volume and either charges or risk of complications was observed.
Higher surgeon and hospital volumes are associated with better outcome among infants who are treated for pyloric stenosis. Identification of aspects of medical and surgical treatment that account for this finding may lead to improvement in the outcome of infants undergoing pyloromyotomy.
Pyloromyotomy is a common procedure performed for infants that is usually associated with a short hospitalization and few postoperative complications. Recently, management strategies have been proposed to reduce length of stay (LOS) and overall cost, including the adoption of structured preoperative and postoperative management.1,2 These strategies have generally been effective in reducing the LOS and cost, but their impact has been modest (LOS reduced by ~ 12 hours and hospital charges by ~ $1000). Several studies have suggested that provider characteristics may also have an impact on outcome after pyloromyotomy. These studies have suggested that infants undergoing pyloromyotomy may have a shorter LOS or decreased complications when operated on by surgeons who perform a larger number of these procedures,3- 6 surgeons with pediatric surgical training,7,8 or surgeons who operate at a specialty children’s center.9 Several of these studies, however, have used methods limiting the validity and generalizibility of the results by using data from only 1 hospital center,4- 7 using retrospective data for comparison,7 or using data obtained over an extended study period.5- 7 Most of these reports used only univariate analysis4- 9 and only 1 accounted for the contribution of surgeon and hospital effects on outcome with multilevel modeling.3 The purpose of this study was to provide a more comprehensive analysis of the effects of surgeon and hospital characteristics on LOS, hospital charges, and major surgical complications after pyloromyotomy using hierarchical modeling.
This study was approved by the institutional review board at the Robert Wood Johnson Medical School, University of Medicine and Dentistry, New Jersey, New Brunswick. Data was obtained from the Heathcare Cost and Utilization Project Kids’ Inpatient Database (KID) from 2000. The KID is a pediatric-specific database containing 2.5 million abstracted discharge records from 2784 short-term, general, and other specialty hospitals from 27 states. An 80% sample of discharges of patients younger than 21 years old from participating states are included. The database contains up to 15 diagnosis codes and up to 15 procedure codes as recorded at the time of discharge using the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM).
Infants younger than 6 months old (n = 4733) who were discharged with a diagnosis code for pyloric stenosis (ICD-9-CM 750.5) and a procedure code for pyloromyotomy (ICD-9-CM 43.3) were identified in the database. To accurately assess the number of pyloromyotomies performed by each surgeon, analysis was limited to infants (n = 1277) from 5 states (Florida, Maine, New York, Pennsylvania, and South Carolina) that reported surgeon identifiers unique to individual surgeons and that reported surgeon identifiers missing in less than 1% of records. The analysis was further limited to hospitalizations primarily for the treatment of pyloric stenosis, leading to the exclusion of 22 infants (1.7%). These included infants who were hospitalized since birth, who had a primary diagnosis of prematurity, or who underwent an additional major surgical procedure during the same hospitalization (including coarctation repair, tracheoesophageal fistula repair, Ladd procedure, gastrostomy tube placement, or abdominal wall or inguinal hernia repair).
Patient-, surgeon-, and hospital-level variables potentially associated with outcome were obtained from the database. Age, prematurity, fluid and electrolyte disturbances, weight loss, and upper gastrointestinal tract hemorrhage data were abstracted from the database using discharge diagnoses codes (Table 1). These patient-level variables were previously found to be associated with an increased preoperative or postoperative LOS among infants undergoing pyloromyotomy.10- 13 Additional comorbidities with a potential impact on LOS and hospital charges were also abstracted, including metabolic disorders, chromosomal disorders, congenital cardiac anomalies, acute respiratory tract infection or pneumonia, gastrointestinal tract infection, and urinary tract infection. Other patient-level covariates, including weekend admission, payer source, and patient income estimate, were available in the KID. Primary payer sources were grouped into 4 categories: Medicaid and Medicare, private insurance, self-pay, and other or no charge (Table 2).
Surgeon volume was defined as the total number of pyloromyotomies performed by each surgeon in 2000. This variable accounted for procedures performed on all patients. Because of highly skewed distribution, surgeon volume was divided into quintiles. Hospital-level variables included volume of pyloromyotomies, location (urban vs rural), teaching status (nonteaching vs teaching), and volume and percentage of pediatric discharges in 2000. Hospital volume was divided into quintiles similar to those for surgeon volume. Hospital teaching status was based on definitions used by the American Hospital Association (Chicago, Ill), and children’s hospital designation (not identified as a children’s hospital vs a children’s unit in a general hospital vs a children’s general hospital) was based on definitions used by the National Association of Children’s Hospitals and Related Institutions (Alexandria, Va). The presence of a pediatric surgery residency (fellowship) program was determined using data obtained from the Association of American Medical Colleges (Washington, DC) (Table 2).14 Outcome variables included postoperative LOS, total LOS, hospital charges, and the occurrence of a major operative complication. Total LOS and hospital charges were obtained from the KID. A major operative complication was defined as the occurrence of an intestinal perforation (gastric or duodenal) or wound dehiscence (Table 1).
The impact of the surgeon and hospital characteristics on the continuous outcome variables, LOS and hospital charges, was analyzed using a linear mixed model (PROC MIXED, SAS 8.2, SAS Institute, Cary, NC) to account for the nested structure of the data (patients nested within surgeons and surgeons nested within hospitals). Values of LOS and hospital charges underwent logarithmic transformation to more closely approximate a normal distribution. A backward elimination strategy was used to select predictors for inclusion in the final model. Estimates of the adjusted regression coefficients and corresponding 95% confidence intervals were evaluated. The impact of surgeon and hospital characteristics on the binary outcome, the occurrence of a major complication, was modeled using a generalized estimating equation approach (PROC GENMOD, SAS 8.2, SAS Institute). Differences between hospitals treating pyloric stenosis and those not treating pyloric stenosis were analyzed using the χ2 test for the categorical characteristics and a 2-sample t test for continuous characteristics.
An overview of the infants analyzed in this study is outlined in Table 2. The average ± SD LOS was 2.8 ± 1.6 days (range, 1-27 days), and the average ± SD hospital charge was $7743 ± $4869 (range, $1639-$81 472). Surgeons performed an average ± SD of 4.2 ± 4.6 pyloromyotomies (range, 1-32), with most surgeons (56%) performing this procedure on fewer than 3 patients (Figure 1A). Most infants (56%) were operated on by a surgeon who performed fewer than 9 pyloromyotomies, and 10% of infants were operated on by a surgeon who performed only 1 pyloromyotomy in 2000 (Figure 2A). At least 1 pyloromyotomy was performed in 169 (25%) of 674 hospitals from the states included in the study. An average ± SD of 7.7 ± 11.0 procedures (range, 1-58) were performed at each hospital, with most hospitals (56%) treating fewer than 4 patients (Figure 1B). Most infants (53%) were operated on at hospitals that treated fewer than 21 patients, and 4% of infants were treated at hospitals at which only 1 pyloromyotomy was performed in 2000 (Figure 2B). Compared with hospitals at which no pyloromyotomies were performed, hospitals treating at least 1 infant were more likely to be urban or teaching hospitals and more likely to have experience or specialized expertise in treating children (children’s hospital designation, pediatric surgery residency program, or larger number or percentage of pediatric discharges; Table 3). Major surgical complications occurred in 13 patients (1.1%), including 11 patients who had a duodenal or gastric perforation and 2 patients who underwent reoperation for wound dehiscence. There were no deaths.
Higher hospital volume was associated with shorter postoperative and total LOS (Table 4). No difference was found in LOS between children treated in a children’s unit in a general hospital and at a children’s general hospital, but postoperative and total LOS were shorter in hospitals without a children’s hospital designation (Table 4). No effect of surgeon volume, hospital location or teaching status, the presence of a pediatric surgery residency program, or the number or percentage of pediatric discharges on LOS was observed. Patients with a complication had a significantly longer LOS than those without. Total LOS had the largest impact on hospital charges. While surgeon volume and other hospital characteristics had no effect, charges were less in rural compared with urban hospitals (Table 4 and Table 5).
Higher surgeon volume was associated with fewer complications (odds ratio, 0.56 per volume quintile; 95% confidence interval, 0.40 to 0.78; P<.001), with surgeons with the highest volume having a 90% lower risk than those with the lowest volume. No association between any hospital characteristic and the risk of complications was observed.
The current study shows that infants undergoing pyloromyotomy are often treated by low-volume surgeons and at low-volume hospitals. Most infants were operated on by a surgeon performing fewer than 3 pyloromyotomies, and most underwent surgery at a hospital admitting fewer than 4 patients with pyloric stenosis in 2000. The finding that the largest number of patients are treated by low-volume providers has been previously observed for other surgical procedures.15 It is not possible to identify whether surgeons performing these procedures had pediatric surgical training or to determine the experience of individual surgeons with this procedure because these variables are not available in the KID. We did observe, however, that infants undergoing pyloromyotomy were more likely to be treated at hospitals with more pediatric or pediatric surgical experience, suggesting that many infants may already have been referred to centers with appropriate expertise.
We first analyzed the impact of surgeon and hospital volume on LOS. Because the impact of surgeon management practices on LOS is more likely to be observed after surgery, postoperative LOS was studied in addition to total LOS. While no effect of surgeon volume was observed, higher hospital volume was strongly associated with shorter postoperative and total LOS. These observations differ from findings by Langer and To,3 who observed that higher hospital volume was associated with shorter postoperative but not total LOS after pyloromyotomy. In their study, specialty training in pediatric surgery was found to be a potential explanation for shorter postoperative LOS at higher-volume hospitals because hospital volume was eliminated as a covariate when specialty training was added to their model. Because specialty training cannot be assessed in the KID, we could not test this finding. One possible explanation for shorter LOS in higher-volume hospitals may be the use of clinical pathways or postoperative feeding protocols. Although these measures have been shown to reduce LOS among infants undergoing pyloromyotomy,1,2 it is less likely that hospitals treating few cases of pyloric stenosis each year would have these available.
In addition to hospital volume, the type of children’s hospital designation had a significant effect on postoperative and total LOS. Interestingly, infants treated at hospitals without a children’s hospital designation had a shorter LOS than those treated at hospitals with at least some level of children’s hospital designation. The reasons for this observation are not apparent from the current study. More frequent presence of comorbidities among infants treated at referral children’s hospitals or the hospital teaching status do not explain this finding because these factors were also considered in the multivariate model.
While hospital volume was associated with LOS, neither surgeon nor hospital volume was found to affect hospital charges. The only hospital characteristic associated with lower charges was rural hospital status. The lower cost of health care in rural hospitals has been previously observed among children treated for asthma and adults treated for congestive heart failure.16,17 Lower cost in rural hospitals has been attributed to cost savings that may arise from having fewer hospital services or a lower cost of living in a rural setting. Not surprisingly, longer LOS was associated with higher hospital charges. Hospital volume remained an independent predictor of charges even when removing LOS from the model (data not shown).
While LOS and hospital charges are important outcomes that measure health care utilization and cost, complication rate is an outcome that measures morbidity and potential mortality. Among provider factors studied, only surgeon volume was associated with the occurrence of a complication. Although the complication rate of pyloromyotomy is low, surgeons in the highest volume quintile were observed to have a 90% lower risk of a major operative complication than those in the lowest quintile. Similar results were reported by Langer and To,3 who found that the lower complication rate observed among pediatric surgeons was attributable to the number of pyloromyotomies performed rather than specialty training. Our findings are consistent with other studies that have supported the safety of pyloromyotomy when performed by surgeons experienced with pyloromyotomy regardless of specialty training or clinical setting.3- 6,18,19
The current study has several limitations. Administrative data was used for case identification and risk stratification. While this type of data has limitations when used for these purposes,20 administrative data facilitate accrual of a larger number of patients in a shorter time interval than would be possible in even a well-designed multicenter study. Because the KID captures hospital admissions regardless of hospital type in each state, these data are more representative of current practice than previous studies that have been based at referral centers. The use of administrative data did not permit us to consider other outcomes such as readmission for additional medical or surgical care, the need for additional outpatient visits, or other morbidities that may have resulted from surgical treatment. Because of these limitations, we focused only on outcomes that were readily available through discharge abstracts—LOS, hospital charges, and operative complications. Although the use of data from a single year did not permit the analysis of the longitudinal effects of surgeon or hospital experience, it is less likely that our observations were biased by the evolution of medical and surgical care that may occur in studies using data over a longer period. Our database also did not allow us to distinguish the specialty training or previous operative experience of individual surgeons. As shown by Langer and To3 using data from Canadian hospitals, these additional surgeon characteristics may affect outcome, particularly postoperative LOS.
The current finding of a volume-outcome relationship among infants with pyloric stenosis confirms the findings of several studies that have addressed this issue using only univariate analysis and a Canadian study that used more appropriate multilevel modeling methods.3- 6 Establishment of a volume-outcome relationship has several important implications for the management of infants with pyloric stenosis. It is premature to conclude from these findings that infants should simply be referred from low- to high-volume providers. Although recommended for high-risk adult operations, regionalization of care may not be as beneficial for pyloromyotomy. Transportation from a low- to a high-volume hospital may actually increase overall LOS and cost, offsetting the benefits achieved by care at the high-volume hospital. In the absence of a proven effect of regionalization on either LOS or charges, the main benefit of transport would be a reduction in surgical complications. The current study, however, shows that many children with pyloric stenosis may already be treated at centers with significant experience with pediatric care, suggesting that regionalization may already be occurring in some settings.
While the current study shows that surgeon volume influences complication rate, pyloromyotomy is a procedure associated with an excellent outcome in most cases and is rarely associated with either mortality or significant postoperative morbidity. Although duodenal perforation or other operative complications are not desirable outcomes, these events rarely are associated with prolongation of hospitalization for more than a few days, major long-term morbidity, or death.21 For these reasons, transport to hospitals with high-volume surgeons may have a small overall impact on morbidity and mortality. Because of costs associated with transport, the cost-effectiveness of this strategy should be carefully considered. While pediatric specialty training and experience with pyloromyotomy may be convenient for identifying appropriate surgeons for performing this procedure, internal tracking of the complication rates of individual surgeons may be more effective than regional referral to high-volume surgeons, particularly because this process will identify the low-volume but low-complication surgeon.
While large academic children’s hospitals have taken the lead in publishing preoperative and postoperative management protocols that may reduce LOS and cost, it is not known how widespread these practice recommendations have been adopted or even whether their benefits can be transported to other settings. Having established a volume-outcome relationship for pyloromyotomy, the next step is to identify the medical strategies and surgical techniques that account for this relationship. These data might then be used to improve outcomes achieved by both low- and high-volume hospitals and surgeons. Additional study is needed to determine whether transport of management strategies or patient transport is the most effective method for improving outcome after pyloromyotomy.
Correspondence: Randall S. Burd, MD, PhD, Department of Surgery, Division of Pediatric Surgery, Robert Wood Johnson Medical School, One Robert Wood Johnson Place, PO Box 19, New Brunswick, NJ 08903 (email@example.com).
Accepted for Publication: February 21, 2005.
This article was corrected on 12/28/06, prior to publication of the correction in print.