- Email: [email protected]
Hospital-Level Factors Associated with Nonoperative Management in Common Pediatric Surgical Procedures
YJPSU-59433; No of Pages 6 Journal of Pediatric Surgery xxx (xxxx) xxx
Contents lists available at ScienceDirect
Journal of Pediatric Surgery journal homepage: www.elsevier.com/locate/jpedsurg
Original Investigation
Hospital-Level Factors Associated with Nonoperative Management in Common Pediatric Surgical Procedures☆,☆☆ Martha Ingram a, Heather L. Short b, Chethan Sathya a, Helene Fevrier b, Mehul V. Raval a,⁎ a b
Division of Pediatric Surgery, Department of Surgery, Northwestern University Feinberg School of Medicine, Ann & Robert H. Lurie Children's Hospital of Chicago, Chicago, IL, USA Division of Pediatric Surgery, Department of Surgery, Emory University School of Medicine, Children's Healthcare of Atlanta, Atlanta, GA, USA
a r t i c l e
i n f o
Article history: Received 27 February 2019 Received in revised form 18 September 2019 Accepted 19 October 2019 Available online xxxx Key words: Pediatric surgery Nonoperative management Variation in care Healthcare Utilization
a b s t r a c t Purpose: Our purpose was to examine patient- and hospital-level factors associated with nonoperative management in common pediatric surgical diagnoses. Methods: Using the 2012 Kid's Inpatient Database (KID), we identified patients b 20 years old diagnosed with cholecystitis (CHOL), bowel obstruction (BO), perforated appendicitis (PA), or spontaneous pneumothorax (SPTX). Logistic regression models were used to identify factors associated with nonoperative management. Results: Of 36,026 admissions for the diagnoses of interest, 7472 (20.7%) were managed nonoperatively. SPTX had the highest incidence of NONOP (55.9%; n = 394), while PA had the lowest incidence (9.2%; n = 1641). Utilization of operative management varied significantly between hospitals. Patients diagnosed with BO (OR 0.41; 95% CI 0.30–0.56) and SPTX (OR 0.28; 95% CI 0.14–0.56) had decreased odds of operative management when treated at an urban, teaching hospital compared to a rural hospital. Patients with PA had increased odds of operative management when treated at an urban, teaching hospital (OR 2.42; 95% CI 1.78–3.30). Hospitallevel factors associated with decreased odds of nonoperative management included urban, nonteaching status (OR 0.54; 95% CI 0.31–0.91) and location in the South (OR 0.53; 95% CI 0.34–83) and West (OR 0.47; 95% CI 0.30–0.74). Conclusions: Despite representing more than 20% of pediatric surgical care for several conditions, nonoperative management is an understudied aspect of care with significant variation that warrants further research. Level of evidence: III © 2019 Published by Elsevier Inc.
There are operative and nonoperative treatment options for many pediatric surgical conditions. The decision to operate is multifactorial and reflects patient and surgeon preferences but also the institutional and/or practice familiarity with use of nonoperative surgical management. Clinical decision making, namely knowing when to operate and when not to operate, is an essential skill required of surgeons [1–3]. The right answer is not always readily apparent, however, and both options often carry risks and benefits, which need to be carefully weighed by the surgeon [4]. Previous studies have demonstrated wide variation in the reliability and reproducibility of clinical surgical
Abbreviations: KID, Kid's Inpatient Database; CHOL, Cholecystitis; BO, Bowel obstruction; PA, Perforated appendicitis; SPTX, Spontaneous pneumothorax; HCUP, Healthcare Cost and Utilization Project. ☆ Disclosures: None ☆☆ IRB approval: Exempt ⁎ Corresponding author at: Division of Pediatric Surgery, Department of Surgery, Faculty, Surgical Outcomes and Quality Improvement Center (SOQIC), Center for Healthcare Studies, Institute for Public Health And Medicine (IPHAM), Northwestern University Feinberg School of Medicine, Surgical Quality Improvement System Lead, Ann & Robert H. Lurie Children's Hospital of Chicago, 633 N. St. Clair St., 20th Floor, Chicago, IL 60611, USA E-mail address: [email protected] (M.V. Raval).
judgment when surgeons are surveyed regarding how they would manage several clinical vignettes [5–7]. The apparent discrepancies in when to use operative versus nonoperative treatment modalities may, in part, explain previously described regional variation in surgical utilization for some procedures [8–10]. Surgeons have been shown to be more likely to recommend operative management if they perceived the risks of operating to be low and the benefits of operating to be high and to be more likely to recommend nonoperative management if they perceived the risks of nonoperative management to be low and the benefits to be high [4]. These perceptions are dependent upon an assessment of the patient's operative risk factors [11,12]. Additionally, surgeon characteristics have been shown to have a small influence on the decision to operate or not [1,13]. This previous research has focused on overall surgical decision-making. Despite persistent controversy over the use of nonoperative versus operative management for certain diagnoses (i.e. small bowel obstruction), there are no studies examining factors associated with operative versus nonoperative management for specific pediatric diagnoses of interest. The purpose of this study was to examine patient- and hospital-level factors associated with use of operative versus nonoperative management in four common pediatric surgical diagnoses.
https://doi.org/10.1016/j.jpedsurg.2019.10.006 0022-3468/© 2019 Published by Elsevier Inc.
Please cite this article as: M. Ingram, H.L. Short, C. Sathya, et al., Hospital-Level Factors Associated with Nonoperative Management in Common Pediatric Surgical Procedures, Journal of Pediatric Surgery, https://doi.org/10.1016/j.jpedsurg.2019.10.006
2
M. Ingram et al. / Journal of Pediatric Surgery xxx (xxxx) xxx
1. Methods
were designated operative hospitals. This process was performed for all 4 diagnoses of interest.
1.1. Data source 1.3. Statistical analysis A cross-sectional analysis of the 2012 release of the Agency for Healthcare Research and Quality (AHRQ) sponsored Healthcare Cost and Utilization Project (HCUP) Kids' Inpatient Database (KID) was performed to identify children b20 years old with a diagnosis of cholecystitis (CHOL), bowel obstruction (BO), perforated appendicitis (PA), or spontaneous pneumothorax (STPX). The KID is a national administrative dataset containing hospital discharge-level data for patients aged less than 21 years. The dataset contains data from 44 states and more than 4100 US hospitals including information for more than 3 million pediatric discharges per year sampled from all community, nonrehabilitation hospitals of states participating in HCUP. National estimates of discharges were generated using discharge weights provided by HCUP that are based on the poststratification of hospitals according to characteristics including ownership, bed size, teaching status, rural/ urban location, and region in the United States. Records from each hospital are randomly sampled to include 10% of uncomplicated births and 80% of complicated births because less variation in outcomes among uncomplicated births is expected. More than 100 clinical and nonclinical data elements are available for each discharge record including patient demographics, hospital charges, and diagnosis and procedure codes [14]. Using the International Classification of Diseases, Ninth Revision, Clinical Manifestation (ICD-9CM), all patients (age b 21 years old) with 1 of the 4 diagnoses of interest were identified. CHOL cases included ICD-9-CM codes 574.0 (calculus of gallbladder with acute cholecystitis), 574.1 (calculus of gallbladder with other cholecystitis), 574.3 (calculus of bile duct with acute cholecystitis), 574.4 (calculus of bile duct with other cholecystitis), 575.0 (acute cholecystitis), 575.10 (cholecystitis, unspecified), and 575.11 (chronic cholecystitis). BO cases included ICD-9-CM code 560 (intestinal obstruction without mention of hernia). PA cases included ICD-9-CM codes 540.0 (acute appendicitis with generalized peritonitis) and 540.1 (acute appendicitis with peritoneal abscess). SPTX cases included ICD-9-CM codes 512.0 (spontaneous tension pneumothorax) and 512.8 (other spontaneous pneumothorax). Using the KID major operating room procedure indicator (ORPROC), individual cases were classified as operative if an associated major operating room procedure was performed and nonoperative if there was no major operating room procedure performed. Based on KID documentation, the OPROPC variable is present for all cases. Patient-level demographics obtained from the database included age, sex, race, expected primary payer, and median household income. Hospital-related information obtained from the database included bed size, teaching hospital status, and census region (Northeast, South, Midwest, West). Cases with invalid values for sex, death, insurance, or race were excluded. Additionally, any case that was transferred in or out of a hospital facility and those that carried more than 1 of the diagnoses of interest were excluded. 1.2. Hospital-level analysis Hospitals were characterized as favoring operative or nonoperative management based on utilization of nonoperative management for each diagnosis of interest after applying disease specific minimum case thresholds. Only hospitals that saw greater than 5 cases of any diagnosis of interest were included in the hospital-level analysis. It was assumed that, if any hospital reported fewer than 5 cases of any diagnosis of interest, the hospital experienced rare to never occurrence of decision-making around nonoperative management for the diagnosis of interest, or too much information would be missing from records to include in analysis. The median percentage of time that a major OR procedure was performed for each diagnosis was determined for all hospitals. Hospitals performing less than the median were designated nonoperative hospitals, and hospitals performing above the median
For each procedure of interest, logistic regression models were used to determine patient- and hospital-level characteristics associated with operative versus nonoperative management. At the individual patientlevel, a univariate analysis was performed by individual diagnosis with use of a major OR procedure as the outcome and individual demographic variables and hospital characteristics (i.e. bed size, teaching status, and census region) as the exposure. If an observation was missing information for a particular variable, it was excluded from that univariate analysis. The observation was reincorporated into univariate analysis for the patient factor information that it could provide. In order to determine hospital-level factors associated with the use of nonoperative management, a separate analysis was performed for each of the hospital with at least 5 cases for a diagnosis with nonoperative management as the outcome and hospital characteristics as the exposure. Statistical significance was set to two-tailed p b 0.05. All statistical analyses were performed using SAS 9.4 (Cary, NC). 2. Results Demographics for the overall cohort and by individual diagnoses are included in Table 1. In 2012, there were 36,026 admissions for the diagnoses of interest with nearly a quarter representing nonoperative management (20.7%, n = 7472). The highest incidence of nonoperative management was seen with spontaneous pneumothorax (55.9%, n = 394), while the lowest incidence was seen with perforated appendicitis (9.2%, n = 1641). The proportion of cases undergoing a major OR procedure was similar for all diagnoses based on census region and hospital bed size and for PA and CHOL based on hospital teaching status. However, there was more variation in the use of a major OR procedure for BO and SPTX based on teaching status with urban, teaching hospitals having a higher proportion of cases with a major OR procedure than rural or urban, nonteaching hospitals (Fig. 1). Among hospitals of any type, the utilization of a major OR procedure varied significantly between hospitals for all diagnoses ranging from a 10-fold difference for BO and a 5-fold difference for all other diagnoses (Fig. 2). Table 2 details the results of the univariate analysis and odds ratios (ORs) represent patient-level factors associated with the use of a major OR procedure as treatment for each of the 4 diagnoses of interest. Patients with a diagnosis of CHOL had increased odds of operative management if they were Black or Asian and decreased odds of operative management if they were female, had a median household income in the second through fourth quartile, or treated in an urban, nonteaching hospital. Additionally, CHOL patients treated in the West had the lowest odds of operative management compared to other census regions. BO patients had increased odds of undergoing a major OR procedure if they were female, Hispanic, self-pay or treated in the West. Odds of undergoing a major OR procedure were decreased if BO patients were Black, or treated at an urban hospital (teaching or nonteaching) compared to a rural hospital. Patients with PA had increased odds of operative management if they were female, Black, in the fourth quartile for median household income, or treated at an urban, teaching hospital. Lastly, decreased odds of operative management were observed for SPTX patients who were in the highest median household income quartile, and those treated at urban (teaching and nonteaching) hospitals compared to rural hospitals, while SPTX patients treated in the Midwest had increased odds of operative management compared to other census regions. Hospitals were designated as favoring operative or nonoperative management based on the median percentage of time that a major OR procedure was performed for each diagnosis as described in the Methods section. Of the 4179 hospitals included in the database, 1102 (26%) met the criteria of seeing N 5 cases of at least one of the
Please cite this article as: M. Ingram, H.L. Short, C. Sathya, et al., Hospital-Level Factors Associated with Nonoperative Management in Common Pediatric Surgical Procedures, Journal of Pediatric Surgery, https://doi.org/10.1016/j.jpedsurg.2019.10.006
M. Ingram et al. / Journal of Pediatric Surgery xxx (xxxx) xxx
3
Table 1 Demographics for all procedures and by individual diagnoses of interest. Freq = frequency. SD = standard deviation * = cells reporting b10 patients cannot be reported as per instruction from Healthcare Cost and Utilization Project (HCUP) Kids' Inpatient Database (KID).
Patient Characteristic Age Male Race Missing White Black Hispanic Asian/Pacific Islander Native American Other Length of Stay (days) Mortality Major OR Procedure (Y) Primary Payer Missing Medicare Medicaid Private insurance Self-pay No charge Other Quartile Median Household Income Missing 1 ($1–$38,999) 2 ($39,000–$47,999) 3 ($48,000–$62,999) 4 ($63,000+)
All Diagnoses N = 36,026
Cholecystitis N = 12,760 (35.42%)
Bowel Obstruction N = 4655 (12.92%)
Perforated Appendix N = 17,906 (49.70%)
Spontaneous Pneumothorax N = 705 (1.96%)
Mean / Freq (SD/%)
Mean / Freq (SD/%)
Mean / Freq (SD/%)
Mean / Freq (SD/%)
Mean / Freq (SD/%)
13.38 (5.70) 16,790 (46.61)
17.19 (3.36) 2269 (17.78)
10.92 (7.62) 3000 (64.45)
11.27 (4.90) 10,934 (61.06)
14.42 (6.68) 587 (83.26)
2179 (6.05) 15,745 (43.70) 3649 (10.13) 11,569 (32.11) 807 (2.24) 383 (1.06) 2884 (4.70) 4.46 (5.40) 42 (0.12) 28,554 (79.26)
602 (4.72) 5146 (40.33) 1778 (13.93) 4397 (34.46) 132 (1.03) 170 (1.33) 535 (4.19) 2.97 (2.80) b10 * 9842 (77.13)
381 (8.18) 2403 (51.62) 699 (15.02) 774 (16.63) 145 (3.11) 18 (0.39) 235 (5.05) 5.42 (8.64) b10 * 2136 (45.89)
1153 (6.44) 7776 (43.43) 1108 (6.19) 6306 (35.22) 486 (2.71) 188 (1.05) 889 (4.96) 5.05 (3.61) b10* 16,265 (90.84)
43 (6.10) 420 (59.57) 64 (9.08) 92 (13.05) 44 (6.24) b10* 35 (4.96) 10.23 (20.92) 32 (4.54) 311 (44.11)
75 (0.21) 111 (0.31) 16,152 (44.83) 15,656 (43.46) 2507 (6.96) 119 (0.33) 1406 (3.90)
37 (0.29) 52 (0.41) 6388 (50.06) 4518 (35.41) 1206 (9.45) 63 (0.49) 496 (3.89)
b10* 24 (0.52) 1766 (37.94) 2402 (51.60) 217 (4.66) 15 (0.32) 226 (4.85)
33 (0.18) 29 (0.16) 7825 (43.70) 8271 (46.19) 1055 (5.89) 36 (0.20) 657 (3.67)
0 (0.00) b10* 173 (24.54) 465 (65.96) 29 (4.11) b10* 27 (3.83)
705 (1.96) 10,920 (30.31) 8731 (24.24) 8494 (23.58) 7176 (19.92)
242 (1.90) 4506 (35.31) 3234 (25.34) 2857 (22.39) 1921 (15.05)
94 (2.02) 1176 (25.26) 999 (21.46) 1122 (24.10) 1264 (27.15)
350 (1.95) 5079 (28.36) 4353 (24.31) 4353 (24.31) 3771 (21.06)
19 (2.70) 159 (22.55) 145 (20.57) 162 (22.98) 220 (31.21)
Fig. 1. Proportion of cases with a major operating room (OR) procedure by diagnosis. (A) Overall cases managed operatively versus nonoperatively by diagnosis. (B) Cases with a major OR procedure by census region. (C) Cases with a major OR procedure by bed size. (D) Cases with a major OR procedure by teaching status. OP=operative; NONOP=nonoperative
Please cite this article as: M. Ingram, H.L. Short, C. Sathya, et al., Hospital-Level Factors Associated with Nonoperative Management in Common Pediatric Surgical Procedures, Journal of Pediatric Surgery, https://doi.org/10.1016/j.jpedsurg.2019.10.006
4
M. Ingram et al. / Journal of Pediatric Surgery xxx (xxxx) xxx
Fig. 2. Variation in operative management by hospital by diagnosis of interest. Only hospitals with 5 or more cases of a diagnosis are included.
diagnoses of interest. More than 80% (n = 894) of the included hospitals were designated as operative hospitals. Hospital-level factors associated with decreased odds of nonoperative management included urban, nonteaching status and location in the South and West. Hospital bed size was not associated with the use of nonoperative management for any diagnosis (Table 3).
3. Discussion To our knowledge, this is the first paper to assess patient and hospital level factors associated with operative versus nonoperative management in children or to assess variability in nonoperative management across centers. We found that 21% of cases of four common pediatric surgical diagnoses were managed nonoperatively and there was considerable variation in the utilization of operative management among individual hospitals for all diagnoses. A prior study evaluated children undergoing appendectomy using the KID and found significant variation in different processes of care and outcomes among children treated at children vs. nonchildren hospitals [18], however this study did not evaluate nonoperative management. Another study evaluated children with intussusception using the PHIS database and found significant variation in practice patterns (operative management) and resource utilization across freestanding children's hospitals [19]. The majority of studies evaluating variability in nonoperative vs. operative management in children have focused on trauma, particularly solid organ injury and pancreatic injury management [20]. Hence, data with respect to
variation in nonoperative management of the common pediatric surgical conditions that we evaluated are lacking. Our study demonstrated that nonoperative management is utilized for 1 out of every 5 children with a common surgical diagnosis, representing a significant portion of the surgical population for which there are currently no quality indicators. This finding highlights that there is significant variation in application of nonoperative versus operative management for several common pediatric surgical conditions. We also identified particular populations of children and hospital types that were associated with nonoperative management of these pediatric conditions. As a result, quality improvement initiatives can be targeted to these populations and hospitals types. In adults, a significant number of patients with surgical diagnoses are managed nonoperatively and demonstrate wide variations in associated morbidities, length of stay, and mortality outcomes [21]. Given the wide variation in surgical practice of nonoperative management, major surgical societies have advocated for including study of nonoperative management in surgical quality assessments. Our findings agree and support the need to include nonoperative management in assessments of pediatric surgical quality, as this population represents a significant proportion of children cared for by pediatric surgeons. A major strength of this study is the use of a large administrative database, which includes data on a large number of children treated at different hospitals across the U.S. Furthermore, in addition to assessing overall variation in nonoperative management across centers, we comprehensively evaluated both patient and hospital level factors to determine which factors are associated with nonoperative management. We also elected to evaluate four
Please cite this article as: M. Ingram, H.L. Short, C. Sathya, et al., Hospital-Level Factors Associated with Nonoperative Management in Common Pediatric Surgical Procedures, Journal of Pediatric Surgery, https://doi.org/10.1016/j.jpedsurg.2019.10.006
M. Ingram et al. / Journal of Pediatric Surgery xxx (xxxx) xxx
5
Table 2 Patient-level factors associated with undergoing a major operating room procedure by diagnosis of interest. Statistically significant results are noted in bold text. NONOP = nonoperative; OR = odds ratio; 95% CI = 95% confidence interval; Ref. = reference. Cholecystitis (N = 12,760)
Sex Male Female Race White Black Hispanic Asian/Pacific Islander Native American Other Primary Payer Medicare Medicaid Private Self-pay No charge Other Median Household Income (quartile) 1 2 3 4 Hospital Bed Size Small Medium Large Teaching Status Rural Urban nonteaching Urban teaching Region Northeast Midwest South West
Bowel Obstruction (N = 4655)
Perforated Appendix (N = 17,906)
Spontaneous Pneumothorax (N = 705)
NONOP (%)
OR (95% CI)
NONOP (%)
OR (95% CI)
NONOP (%)
OR (95% CI)
NONOP (%)
OR (95% CI)
566 (24.94) 2352 (22.42)
1.00 (ref.) 0.87 (0.77, 0.98)
1552 (51.73) 967 (58.43)
1.00 (ref.) 1.31 (1.16, 1.48)
938 (8.58) 703 (10.08)
1.00 (ref.) 1.20 (1.08, 1.32)
325 (55.37) 69 (58.47)
1.00 (ref.) 1.14 (0.77, 1.68)
1049 (20.38) 561 (31.55) 969 (22.04) 46 (34.85) 25 (14.71) 115 (21.50)
1.00 (ref.) 1.80 (1.56, 2.07) 1.10 (0.99, 1.23) 2.09 (1.41, 3.09) 0.67 (0.43, 1.06) 1.07 (0.86, 1.34)
1308 (54.43) 699 (202.02) 463 (59.82) 83 (57.24) 9 (50.00) 103 (43.83)
1.00 (ref.) 0.82 (0.68, 0.99) 1.25 (1.05, 1.48) 1.04 (0.88, 1.23) 0.99 (0.83, 1.18) 0.94 (0.80, 1.11)
704 (9.05) 134 (12.09) 530 (8.40) 60 (12.35) 15 (7.98) 796 (9.19)
1.00 (ref.) 1.38 (1.08, 1.76) 0.92 (0.72, 1.19) 1.41 (0.99, 2.02) 0.87 (0.52, 1.46) 1.16 (0.93, 1.45)
230 (54.76) 40 (62.50) 53 (57.61) 23 (52.27) 7 (100.00) 20 (62.50)
1.00 (ref.) 1.38 (0.80, 2.37) 1.12 (0.71, 1.77) 0.90 (0.49, 1.69) 0.00 (0.00, 0.00) 1.10 (0.55, 2.21)
17 (32.69) 1607 (25.16) 927 (20.52) 227 (18.82) 13 (20.63) 117 (23.59)
1.88 (1.00, 3.53) 1.30 (1.18, 1.44) 1.00 (ref.) 0.90 (0.75, 1.08) 1.01 (0.54, 1.88) 1.20 (0.92, 1.55)
13 (54.17) 910 (51.53) 1313 (54.66) 140 (64.52) 9 (60.00) 132 (58.41)
0.92 (0.78, 1.08) 0.88 (0.77, 1.00) 1.00 (ref.) 1.51 (1.15, 1.99) 1.24 (0.45, 3.43) 1.16 (0.83, 1.64)
5 (17.24) 727 (9.29) 771 (9.32) 80 (7.58) 2 (5.56) 52 (7.91)
2.03 (0.67, 6.14) 1.00 (0.86, 1.16) 1.00 (ref.) 0.80 (0.62, 1.02) 0.57 (0.13, 2.45) 0.84 (0.59, 1.19)
5 (83.33) 103 (59.54) 244 (52.47) 20 (68.97) 3 (60.00) 19 (70.37)
4.53 (0.53, 38.82) 1.33 (0.94, 1.89) 1.00 (ref.) 2.01 (0.95, 4.27) 1.36 (0.23, 8.10) 2.15 (0.93, 4.97)
1118 (24.81) 701 (21.68) 617 (21.60) 415 (21.60)
1.00 (ref.) 0.84 (0.75, 0.94) 0.83 (0.74, 0.94) 0.84 (0.72, 0.96)
637 (54.17) 551 (55.16) 605 (53.92) 666 (52.69)
1.00 (ref.) 1.04 (0.88, 1.23) 0.99 (0.83, 1.18) 0.94 (0.80, 1.11)
416 (8.19) 386 (8.87) 415 (9.53) 397 (10.53)
1.00 (ref.) 1.09 (0.93, 1.27) 1.18 (1.00, 1.40) 1.32 (1.04, 1.67)
98 (61.64) 81 (55.86) 98 (61.64) 111 (55.50)
1.00 (ref.) 0.79 (0.50, 1.24) 0.86 (0.56, 1.33) 0.63 (0.42, 0.95)
336 (24.82) 760 (24.53) 1822 (21.93)
1.00 (ref.) 0.98 (0.82, 1.18) 0.85 (0.72, 1.00)
233 (52.83) 578 (57.06) 1708 (52.86)
1.00 (ref.) 1.02 (0.74, 1.40) 0.86 (0.64, 1.14)
182 (8.54) 334 (7.76) 1125 (9.81)
1.00 (ref.) 0.90 (0.57, 1.43) 1.16 (0.79, 1.73)
35 (59.32) 100 (56.18) 259 (55.34)
1.00 (ref.) 0.88 (0.49, 1.58) 0.85 (0.49, 1.46)
301 (24.16) 973 (20.37) 1644 (24.40)
1.00 (ref.) 0.80 (0.68, 0.95) 1.01 (0.86, 1.19)
219 (138.61) 890 (166.67) 1827 (51.52)
1.00 (ref.) 0.58 (0.41, 0.81) 0.41 (0.30, 0.56)
86 (5.29) 229 (4.44) 1326(11.92)
1.00 (ref.) 0.83 (0.61, 1.13) 2.42 (1.78, 3.30)
35 (79.55) 120 (60.30) 239 (51.73)
1.00 (ref.) 0.39 (0.19, 0.82) 0.28 (0.14, 0.56)
602 (27.84) 530 (23.93) 1007 (21.98) 779 (20.49)
1.00 (ref.) 0.82 (0.69, 0.97) 0.73 (0.63, 0.85) 0.67 (0.57, 0.79)
493 (50.41) 608 (55.47) 801 (53.98) 617 (56.24)
1.00 (ref.) 1.23 (0.97, 1.54) 1.15 (0.93, 1.43) 1.26 (1.01, 1.59)
321 (11.86) 375 (10.40) 410 (7.53) 535 (8.71)
1.00 (ref.) 0.86 (0.61, 1.21) 0.86 (0.61, 1.21) 0.86 (0.61, 1.21)
88 (49.44) 97 (62.18) 127 (57.73) 88 (49.44)
1.00 (ref.) 1.68 (1.10, 2.57) 1.40 (0.94, 2.06) 1.22 (0.78, 1.90)
common pediatric surgical diagnoses, making our study applicable to a large population of children treated by pediatric surgeons. A limitation of the KID is that the appropriateness of nonoperative or operative management cannot be evaluated owing to a lack of clinically granular information surrounding each diagnosis and management detail. Lastly, as mentioned above, this is the first study of its kind and our findings open the door to a number of future analyses. We observed several trends relating to patient demographics, income levels, and insurance patterns. Females may be more likely to
undergo operative intervention for abdominal surgical diagnoses (in this study, BO and PA) as the differential diagnosis for surgical abdominal pain in young females is not easily deduced using imaging, and includes certain emergent conditions for which nonoperative management is not feasible. Similarly, children of black and Asian/Pacific Islander race are at higher predisposition for acquiring gallbladder disease secondarily from hemopathies than the general population, which could influence providers' judgment toward performing cholecystectomy for presentations of cholecystitis in children of these
Table 3 Hospital-level characteristics associated with nonoperative management. Statistically significant results are noted in bold text. OR = odds ratio; 95% CI = 95% confidence interval. Ref. = reference.
Bed size Small Medium Large Teaching Status Rural Urban, nonteaching Urban, teaching Region Northeast Midwest South West
All Hospitals N = 4179, (%)
Hospitals with N 5 cases of 1 diagnosis N = 1102, (%)
Non-operative Hospitals N = 208, (%)
Operative Hospitals N = 894, (%)
OR (95% CI)
1734 (41.49) 1087 (26.01) 1358 (32.50)
111 (10.07) 281 (25.50) 710 (64.43)
21 (10.10) 51 (24.52) 136 (65.38)
90 (10.07) 230 (25.73) 574 (64.21)
1.00 (ref.) 0.95 (0.54, 1.67) 1.02 (0.61, 1.69)
1567 (37.71) 1736 (41.54) 867 (20.75)
101 (9.17) 484 (43.92) 517 (46.91)
23 (11.06) 66 (31.73) 119 (57.21)
78 (8.72) 418 (46.76) 398 (44.52)
1.00 (ref.) 0.54 (0.31, 0.91) 1.01 (0.61, 1.69)
518 (12.40) 1310 (31.35) 1520 (36.37) 831 (19.89)
191 (17.33) 226 (20.51) 342 (31.03) 343 (31.13)
48 (23.08) 61 (29.33) 52 (25.00) 47 (22.06)
143 (16.00) 165 (18.46) 290 (32.44) 296 (33.11)
1.00 (ref.) 1.10 (0.71, 1.71) 0.53 (0.34, 0.83) 0.47 (0.30, 0.74)
Please cite this article as: M. Ingram, H.L. Short, C. Sathya, et al., Hospital-Level Factors Associated with Nonoperative Management in Common Pediatric Surgical Procedures, Journal of Pediatric Surgery, https://doi.org/10.1016/j.jpedsurg.2019.10.006
6
M. Ingram et al. / Journal of Pediatric Surgery xxx (xxxx) xxx
ethnicities. Additionally, our findings of regional and academic variability in use of nonoperative management for bowel obstruction, pneumothorax, and cholecystitis suggest that practice or cultural preferences may exist that contribute to these observed trends in management. Further studies are needed to understand the influence of provider bias, patient factors, location, and possible disparities on the decision to elect nonoperative management. Our study has a number of limitations. We were unable to analyze particular clinical factors (i.e, duration of symptoms prior to presentation, findings on imaging, severity of presentation) that may have impacted clinical judgment toward operative or nonoperative management and appropriateness of that decision. Second, there may be some element of selection bias based on the cases captured in the KID and after our selection of diagnoses of interest and exclusion of transferred patients. Additionally, certain variables of possible interest (including whether the operating surgeon is pediatric vs adulttrained) were not available for this analysis. Nevertheless, the KID contains data on a large number of children treated across different types of hospitals in the majority of U.S. states, hence our findings of proportion of non-operatively managed patients are generalizable. Another limitation of the study is introduced when using ORPROC for analyses of interest to identify cases where operative intervention occurred. This variable is provided by the KID based on reporting of as many as 15 procedure codes that capture major operative intervention. While we expect that most procedures requiring a major operating room would be accurately coded in the discharge data, it is theoretically possible that some procedures could be missed if this process was not performed accurately. Another limitation is confounding, which was not controlled for in our univariate analyses or in our evaluation of variation across centers. Lastly, our classification of hospitals as nonoperative vs. operative based on median operative rates may be too crude a marker for this definition. This study has identified additional factors that may be beneficial to examine in order to understand more about nonoperative management. Some examples of these variables include examination of impact on length of stay, particular nuances in presentation or management, and possible associations with adult compared to pediatric treating surgeon. Future studies should aim at more focused analyses within each of these disease states to help identify institutional and patient factors that are directly contributing to variation in nonoperative care across centers. Multivariable analyses should be performed to control for confounding. Once key factors influencing operative vs nonoperative management have been identified, targeted quality improvement initiatives can developed. National pediatric protocols for indications for surgical intervention and metrics surrounding nonoperative management of cholecystitis, bowel obstruction, perforated appendicitis, and spontaneous pneumothorax management are needed to help reduce variation in practice patterns.
4. Conclusion Nonoperative management of common pediatric surgical diagnoses occurs in more than 20% of children and literature on this subject is scarce. There is significant variation in use of nonoperative versus operative management for common pediatric surgical diseases. Quality improvement initiatives aimed at understanding what causes this variability and its potential impact on outcomes are required to help reduce heterogeneity.
References [1] Szatmary P, Arora S, Sevdalis N. To operate or not to operate? A multi-method analysis of decision-making in emergency surgery. Am J Surg 2010;200(2):298–304. [2] Eddy DM. Clinical decision making: from theory to practice. Anatomy of a decision. JAMA 1990;263(3):441–3. [3] Yule S, Flin R, Paterson-Brown S, et al. Non-technical skills for surgeons in the operating room: a review of the literature. Surgery 2006;139(2):140–9. [4] Sacks GD, Dawes AJ, Ettner SL, et al. Surgeon perception of risk and benefit in the decision to operate. Ann Surg 2016;264(6):896–903. [5] Rutkow IM, Gittelsohn AM, Zuidema GD. Surgical decision making. The reliability of clinical judgment. Ann Surg 1979;190(3):409–19. [6] Rutkow IM. Surgical decision making. The reproducibility of clinical judgement Arch Surg 1982;117(3):337–40. [7] Rutkow IM. The reliability and reproducibility of the surgical decision-making process. Surg Clin North Am 1982;62(4):721–35. [8] Short HL, Zhu W, McCracken C, et al. Exploring regional variability in utilization of antireflux surgery in children. J Surg Res 2017;214:49–56. [9] Wennberg J, Gittelsohn. Small area variations in health care delivery. Science 1973; 182(4117):1102-8. [10] Birkmeyer JD, Reames BN, McCulloch P, et al. Understanding of regional variation in the use of surgery. Lancet 2013;382(9898):1121–9. [11] Kraemer K, Cohen ME, Liu Y, et al. Development and evaluation of the American College of Surgeons NSQIP pediatric surgical risk calculator. J Am Coll Surg 2016;223 (5):685–93. [12] Bilimoria KY, Liu Y, Paruch JL, Zhou L, Kmiecik TE, Ko CY, et al. Development and evaluation of the universal ACS NSQIP surgical risk calculator: a decision aid and informed consent tool for patients and surgeons. J Am Coll Surg 2013;217(5):833-42 e1-3. [13] Wilson NP, Wilson FP, Neuman M, Epstein A, Bell R, Armstrong K, et al. Determinants of surgical decision making: a national survey. Am J Surg 2013;206(6):970-7; discussion 7-8. [14] Introduction to the HCUP KID’s Inpatient Database (KID). Healthcare Cost and Utilization Project (HCUP). Agency for Healthcare Research and Quality. MD: Rockville; 2012. https://www.hcup-us.ahrq.gov/db/nation/kid/KID_2012_Introduction.pdf. [18] Tian Y, Heiss KF, Wulkan ML, et al. Assessment of variation in care and outcomes for pediatric appendicitis at children’s and non-children’s hospitals. J Pediatr Surg 2015: 1885–92. [19] Rice-Townsend S, Chen C, Barnes JN, et al. Variation in practice patterns and resource utilization surrounding management of intussusception at freestanding children’s hospitals. J Pediatr Surg 2012;48(1):104–10. [20] Cuenca AG, Islam S (2012) Pediatric pancreatic trauma: trending toward nonoperative management? Am Surg 78: 1204-1210. [21] Wandling MW, et al. Expanding the scope of quality measurement in surgery to include nonoperative care. Results from the American College of Surgeons NSQIP emergency general surgery pilot J Trauma Acute Care Surg 2017;83(5):837–45.
Please cite this article as: M. Ingram, H.L. Short, C. Sathya, et al., Hospital-Level Factors Associated with Nonoperative Management in Common Pediatric Surgical Procedures, Journal of Pediatric Surgery, https://doi.org/10.1016/j.jpedsurg.2019.10.006
Contents lists available at ScienceDirect
Journal of Pediatric Surgery journal homepage: www.elsevier.com/locate/jpedsurg
Original Investigation
Hospital-Level Factors Associated with Nonoperative Management in Common Pediatric Surgical Procedures☆,☆☆ Martha Ingram a, Heather L. Short b, Chethan Sathya a, Helene Fevrier b, Mehul V. Raval a,⁎ a b
Division of Pediatric Surgery, Department of Surgery, Northwestern University Feinberg School of Medicine, Ann & Robert H. Lurie Children's Hospital of Chicago, Chicago, IL, USA Division of Pediatric Surgery, Department of Surgery, Emory University School of Medicine, Children's Healthcare of Atlanta, Atlanta, GA, USA
a r t i c l e
i n f o
Article history: Received 27 February 2019 Received in revised form 18 September 2019 Accepted 19 October 2019 Available online xxxx Key words: Pediatric surgery Nonoperative management Variation in care Healthcare Utilization
a b s t r a c t Purpose: Our purpose was to examine patient- and hospital-level factors associated with nonoperative management in common pediatric surgical diagnoses. Methods: Using the 2012 Kid's Inpatient Database (KID), we identified patients b 20 years old diagnosed with cholecystitis (CHOL), bowel obstruction (BO), perforated appendicitis (PA), or spontaneous pneumothorax (SPTX). Logistic regression models were used to identify factors associated with nonoperative management. Results: Of 36,026 admissions for the diagnoses of interest, 7472 (20.7%) were managed nonoperatively. SPTX had the highest incidence of NONOP (55.9%; n = 394), while PA had the lowest incidence (9.2%; n = 1641). Utilization of operative management varied significantly between hospitals. Patients diagnosed with BO (OR 0.41; 95% CI 0.30–0.56) and SPTX (OR 0.28; 95% CI 0.14–0.56) had decreased odds of operative management when treated at an urban, teaching hospital compared to a rural hospital. Patients with PA had increased odds of operative management when treated at an urban, teaching hospital (OR 2.42; 95% CI 1.78–3.30). Hospitallevel factors associated with decreased odds of nonoperative management included urban, nonteaching status (OR 0.54; 95% CI 0.31–0.91) and location in the South (OR 0.53; 95% CI 0.34–83) and West (OR 0.47; 95% CI 0.30–0.74). Conclusions: Despite representing more than 20% of pediatric surgical care for several conditions, nonoperative management is an understudied aspect of care with significant variation that warrants further research. Level of evidence: III © 2019 Published by Elsevier Inc.
There are operative and nonoperative treatment options for many pediatric surgical conditions. The decision to operate is multifactorial and reflects patient and surgeon preferences but also the institutional and/or practice familiarity with use of nonoperative surgical management. Clinical decision making, namely knowing when to operate and when not to operate, is an essential skill required of surgeons [1–3]. The right answer is not always readily apparent, however, and both options often carry risks and benefits, which need to be carefully weighed by the surgeon [4]. Previous studies have demonstrated wide variation in the reliability and reproducibility of clinical surgical
Abbreviations: KID, Kid's Inpatient Database; CHOL, Cholecystitis; BO, Bowel obstruction; PA, Perforated appendicitis; SPTX, Spontaneous pneumothorax; HCUP, Healthcare Cost and Utilization Project. ☆ Disclosures: None ☆☆ IRB approval: Exempt ⁎ Corresponding author at: Division of Pediatric Surgery, Department of Surgery, Faculty, Surgical Outcomes and Quality Improvement Center (SOQIC), Center for Healthcare Studies, Institute for Public Health And Medicine (IPHAM), Northwestern University Feinberg School of Medicine, Surgical Quality Improvement System Lead, Ann & Robert H. Lurie Children's Hospital of Chicago, 633 N. St. Clair St., 20th Floor, Chicago, IL 60611, USA E-mail address: [email protected] (M.V. Raval).
judgment when surgeons are surveyed regarding how they would manage several clinical vignettes [5–7]. The apparent discrepancies in when to use operative versus nonoperative treatment modalities may, in part, explain previously described regional variation in surgical utilization for some procedures [8–10]. Surgeons have been shown to be more likely to recommend operative management if they perceived the risks of operating to be low and the benefits of operating to be high and to be more likely to recommend nonoperative management if they perceived the risks of nonoperative management to be low and the benefits to be high [4]. These perceptions are dependent upon an assessment of the patient's operative risk factors [11,12]. Additionally, surgeon characteristics have been shown to have a small influence on the decision to operate or not [1,13]. This previous research has focused on overall surgical decision-making. Despite persistent controversy over the use of nonoperative versus operative management for certain diagnoses (i.e. small bowel obstruction), there are no studies examining factors associated with operative versus nonoperative management for specific pediatric diagnoses of interest. The purpose of this study was to examine patient- and hospital-level factors associated with use of operative versus nonoperative management in four common pediatric surgical diagnoses.
https://doi.org/10.1016/j.jpedsurg.2019.10.006 0022-3468/© 2019 Published by Elsevier Inc.
Please cite this article as: M. Ingram, H.L. Short, C. Sathya, et al., Hospital-Level Factors Associated with Nonoperative Management in Common Pediatric Surgical Procedures, Journal of Pediatric Surgery, https://doi.org/10.1016/j.jpedsurg.2019.10.006
2
M. Ingram et al. / Journal of Pediatric Surgery xxx (xxxx) xxx
1. Methods
were designated operative hospitals. This process was performed for all 4 diagnoses of interest.
1.1. Data source 1.3. Statistical analysis A cross-sectional analysis of the 2012 release of the Agency for Healthcare Research and Quality (AHRQ) sponsored Healthcare Cost and Utilization Project (HCUP) Kids' Inpatient Database (KID) was performed to identify children b20 years old with a diagnosis of cholecystitis (CHOL), bowel obstruction (BO), perforated appendicitis (PA), or spontaneous pneumothorax (STPX). The KID is a national administrative dataset containing hospital discharge-level data for patients aged less than 21 years. The dataset contains data from 44 states and more than 4100 US hospitals including information for more than 3 million pediatric discharges per year sampled from all community, nonrehabilitation hospitals of states participating in HCUP. National estimates of discharges were generated using discharge weights provided by HCUP that are based on the poststratification of hospitals according to characteristics including ownership, bed size, teaching status, rural/ urban location, and region in the United States. Records from each hospital are randomly sampled to include 10% of uncomplicated births and 80% of complicated births because less variation in outcomes among uncomplicated births is expected. More than 100 clinical and nonclinical data elements are available for each discharge record including patient demographics, hospital charges, and diagnosis and procedure codes [14]. Using the International Classification of Diseases, Ninth Revision, Clinical Manifestation (ICD-9CM), all patients (age b 21 years old) with 1 of the 4 diagnoses of interest were identified. CHOL cases included ICD-9-CM codes 574.0 (calculus of gallbladder with acute cholecystitis), 574.1 (calculus of gallbladder with other cholecystitis), 574.3 (calculus of bile duct with acute cholecystitis), 574.4 (calculus of bile duct with other cholecystitis), 575.0 (acute cholecystitis), 575.10 (cholecystitis, unspecified), and 575.11 (chronic cholecystitis). BO cases included ICD-9-CM code 560 (intestinal obstruction without mention of hernia). PA cases included ICD-9-CM codes 540.0 (acute appendicitis with generalized peritonitis) and 540.1 (acute appendicitis with peritoneal abscess). SPTX cases included ICD-9-CM codes 512.0 (spontaneous tension pneumothorax) and 512.8 (other spontaneous pneumothorax). Using the KID major operating room procedure indicator (ORPROC), individual cases were classified as operative if an associated major operating room procedure was performed and nonoperative if there was no major operating room procedure performed. Based on KID documentation, the OPROPC variable is present for all cases. Patient-level demographics obtained from the database included age, sex, race, expected primary payer, and median household income. Hospital-related information obtained from the database included bed size, teaching hospital status, and census region (Northeast, South, Midwest, West). Cases with invalid values for sex, death, insurance, or race were excluded. Additionally, any case that was transferred in or out of a hospital facility and those that carried more than 1 of the diagnoses of interest were excluded. 1.2. Hospital-level analysis Hospitals were characterized as favoring operative or nonoperative management based on utilization of nonoperative management for each diagnosis of interest after applying disease specific minimum case thresholds. Only hospitals that saw greater than 5 cases of any diagnosis of interest were included in the hospital-level analysis. It was assumed that, if any hospital reported fewer than 5 cases of any diagnosis of interest, the hospital experienced rare to never occurrence of decision-making around nonoperative management for the diagnosis of interest, or too much information would be missing from records to include in analysis. The median percentage of time that a major OR procedure was performed for each diagnosis was determined for all hospitals. Hospitals performing less than the median were designated nonoperative hospitals, and hospitals performing above the median
For each procedure of interest, logistic regression models were used to determine patient- and hospital-level characteristics associated with operative versus nonoperative management. At the individual patientlevel, a univariate analysis was performed by individual diagnosis with use of a major OR procedure as the outcome and individual demographic variables and hospital characteristics (i.e. bed size, teaching status, and census region) as the exposure. If an observation was missing information for a particular variable, it was excluded from that univariate analysis. The observation was reincorporated into univariate analysis for the patient factor information that it could provide. In order to determine hospital-level factors associated with the use of nonoperative management, a separate analysis was performed for each of the hospital with at least 5 cases for a diagnosis with nonoperative management as the outcome and hospital characteristics as the exposure. Statistical significance was set to two-tailed p b 0.05. All statistical analyses were performed using SAS 9.4 (Cary, NC). 2. Results Demographics for the overall cohort and by individual diagnoses are included in Table 1. In 2012, there were 36,026 admissions for the diagnoses of interest with nearly a quarter representing nonoperative management (20.7%, n = 7472). The highest incidence of nonoperative management was seen with spontaneous pneumothorax (55.9%, n = 394), while the lowest incidence was seen with perforated appendicitis (9.2%, n = 1641). The proportion of cases undergoing a major OR procedure was similar for all diagnoses based on census region and hospital bed size and for PA and CHOL based on hospital teaching status. However, there was more variation in the use of a major OR procedure for BO and SPTX based on teaching status with urban, teaching hospitals having a higher proportion of cases with a major OR procedure than rural or urban, nonteaching hospitals (Fig. 1). Among hospitals of any type, the utilization of a major OR procedure varied significantly between hospitals for all diagnoses ranging from a 10-fold difference for BO and a 5-fold difference for all other diagnoses (Fig. 2). Table 2 details the results of the univariate analysis and odds ratios (ORs) represent patient-level factors associated with the use of a major OR procedure as treatment for each of the 4 diagnoses of interest. Patients with a diagnosis of CHOL had increased odds of operative management if they were Black or Asian and decreased odds of operative management if they were female, had a median household income in the second through fourth quartile, or treated in an urban, nonteaching hospital. Additionally, CHOL patients treated in the West had the lowest odds of operative management compared to other census regions. BO patients had increased odds of undergoing a major OR procedure if they were female, Hispanic, self-pay or treated in the West. Odds of undergoing a major OR procedure were decreased if BO patients were Black, or treated at an urban hospital (teaching or nonteaching) compared to a rural hospital. Patients with PA had increased odds of operative management if they were female, Black, in the fourth quartile for median household income, or treated at an urban, teaching hospital. Lastly, decreased odds of operative management were observed for SPTX patients who were in the highest median household income quartile, and those treated at urban (teaching and nonteaching) hospitals compared to rural hospitals, while SPTX patients treated in the Midwest had increased odds of operative management compared to other census regions. Hospitals were designated as favoring operative or nonoperative management based on the median percentage of time that a major OR procedure was performed for each diagnosis as described in the Methods section. Of the 4179 hospitals included in the database, 1102 (26%) met the criteria of seeing N 5 cases of at least one of the
Please cite this article as: M. Ingram, H.L. Short, C. Sathya, et al., Hospital-Level Factors Associated with Nonoperative Management in Common Pediatric Surgical Procedures, Journal of Pediatric Surgery, https://doi.org/10.1016/j.jpedsurg.2019.10.006
M. Ingram et al. / Journal of Pediatric Surgery xxx (xxxx) xxx
3
Table 1 Demographics for all procedures and by individual diagnoses of interest. Freq = frequency. SD = standard deviation * = cells reporting b10 patients cannot be reported as per instruction from Healthcare Cost and Utilization Project (HCUP) Kids' Inpatient Database (KID).
Patient Characteristic Age Male Race Missing White Black Hispanic Asian/Pacific Islander Native American Other Length of Stay (days) Mortality Major OR Procedure (Y) Primary Payer Missing Medicare Medicaid Private insurance Self-pay No charge Other Quartile Median Household Income Missing 1 ($1–$38,999) 2 ($39,000–$47,999) 3 ($48,000–$62,999) 4 ($63,000+)
All Diagnoses N = 36,026
Cholecystitis N = 12,760 (35.42%)
Bowel Obstruction N = 4655 (12.92%)
Perforated Appendix N = 17,906 (49.70%)
Spontaneous Pneumothorax N = 705 (1.96%)
Mean / Freq (SD/%)
Mean / Freq (SD/%)
Mean / Freq (SD/%)
Mean / Freq (SD/%)
Mean / Freq (SD/%)
13.38 (5.70) 16,790 (46.61)
17.19 (3.36) 2269 (17.78)
10.92 (7.62) 3000 (64.45)
11.27 (4.90) 10,934 (61.06)
14.42 (6.68) 587 (83.26)
2179 (6.05) 15,745 (43.70) 3649 (10.13) 11,569 (32.11) 807 (2.24) 383 (1.06) 2884 (4.70) 4.46 (5.40) 42 (0.12) 28,554 (79.26)
602 (4.72) 5146 (40.33) 1778 (13.93) 4397 (34.46) 132 (1.03) 170 (1.33) 535 (4.19) 2.97 (2.80) b10 * 9842 (77.13)
381 (8.18) 2403 (51.62) 699 (15.02) 774 (16.63) 145 (3.11) 18 (0.39) 235 (5.05) 5.42 (8.64) b10 * 2136 (45.89)
1153 (6.44) 7776 (43.43) 1108 (6.19) 6306 (35.22) 486 (2.71) 188 (1.05) 889 (4.96) 5.05 (3.61) b10* 16,265 (90.84)
43 (6.10) 420 (59.57) 64 (9.08) 92 (13.05) 44 (6.24) b10* 35 (4.96) 10.23 (20.92) 32 (4.54) 311 (44.11)
75 (0.21) 111 (0.31) 16,152 (44.83) 15,656 (43.46) 2507 (6.96) 119 (0.33) 1406 (3.90)
37 (0.29) 52 (0.41) 6388 (50.06) 4518 (35.41) 1206 (9.45) 63 (0.49) 496 (3.89)
b10* 24 (0.52) 1766 (37.94) 2402 (51.60) 217 (4.66) 15 (0.32) 226 (4.85)
33 (0.18) 29 (0.16) 7825 (43.70) 8271 (46.19) 1055 (5.89) 36 (0.20) 657 (3.67)
0 (0.00) b10* 173 (24.54) 465 (65.96) 29 (4.11) b10* 27 (3.83)
705 (1.96) 10,920 (30.31) 8731 (24.24) 8494 (23.58) 7176 (19.92)
242 (1.90) 4506 (35.31) 3234 (25.34) 2857 (22.39) 1921 (15.05)
94 (2.02) 1176 (25.26) 999 (21.46) 1122 (24.10) 1264 (27.15)
350 (1.95) 5079 (28.36) 4353 (24.31) 4353 (24.31) 3771 (21.06)
19 (2.70) 159 (22.55) 145 (20.57) 162 (22.98) 220 (31.21)
Fig. 1. Proportion of cases with a major operating room (OR) procedure by diagnosis. (A) Overall cases managed operatively versus nonoperatively by diagnosis. (B) Cases with a major OR procedure by census region. (C) Cases with a major OR procedure by bed size. (D) Cases with a major OR procedure by teaching status. OP=operative; NONOP=nonoperative
Please cite this article as: M. Ingram, H.L. Short, C. Sathya, et al., Hospital-Level Factors Associated with Nonoperative Management in Common Pediatric Surgical Procedures, Journal of Pediatric Surgery, https://doi.org/10.1016/j.jpedsurg.2019.10.006
4
M. Ingram et al. / Journal of Pediatric Surgery xxx (xxxx) xxx
Fig. 2. Variation in operative management by hospital by diagnosis of interest. Only hospitals with 5 or more cases of a diagnosis are included.
diagnoses of interest. More than 80% (n = 894) of the included hospitals were designated as operative hospitals. Hospital-level factors associated with decreased odds of nonoperative management included urban, nonteaching status and location in the South and West. Hospital bed size was not associated with the use of nonoperative management for any diagnosis (Table 3).
3. Discussion To our knowledge, this is the first paper to assess patient and hospital level factors associated with operative versus nonoperative management in children or to assess variability in nonoperative management across centers. We found that 21% of cases of four common pediatric surgical diagnoses were managed nonoperatively and there was considerable variation in the utilization of operative management among individual hospitals for all diagnoses. A prior study evaluated children undergoing appendectomy using the KID and found significant variation in different processes of care and outcomes among children treated at children vs. nonchildren hospitals [18], however this study did not evaluate nonoperative management. Another study evaluated children with intussusception using the PHIS database and found significant variation in practice patterns (operative management) and resource utilization across freestanding children's hospitals [19]. The majority of studies evaluating variability in nonoperative vs. operative management in children have focused on trauma, particularly solid organ injury and pancreatic injury management [20]. Hence, data with respect to
variation in nonoperative management of the common pediatric surgical conditions that we evaluated are lacking. Our study demonstrated that nonoperative management is utilized for 1 out of every 5 children with a common surgical diagnosis, representing a significant portion of the surgical population for which there are currently no quality indicators. This finding highlights that there is significant variation in application of nonoperative versus operative management for several common pediatric surgical conditions. We also identified particular populations of children and hospital types that were associated with nonoperative management of these pediatric conditions. As a result, quality improvement initiatives can be targeted to these populations and hospitals types. In adults, a significant number of patients with surgical diagnoses are managed nonoperatively and demonstrate wide variations in associated morbidities, length of stay, and mortality outcomes [21]. Given the wide variation in surgical practice of nonoperative management, major surgical societies have advocated for including study of nonoperative management in surgical quality assessments. Our findings agree and support the need to include nonoperative management in assessments of pediatric surgical quality, as this population represents a significant proportion of children cared for by pediatric surgeons. A major strength of this study is the use of a large administrative database, which includes data on a large number of children treated at different hospitals across the U.S. Furthermore, in addition to assessing overall variation in nonoperative management across centers, we comprehensively evaluated both patient and hospital level factors to determine which factors are associated with nonoperative management. We also elected to evaluate four
Please cite this article as: M. Ingram, H.L. Short, C. Sathya, et al., Hospital-Level Factors Associated with Nonoperative Management in Common Pediatric Surgical Procedures, Journal of Pediatric Surgery, https://doi.org/10.1016/j.jpedsurg.2019.10.006
M. Ingram et al. / Journal of Pediatric Surgery xxx (xxxx) xxx
5
Table 2 Patient-level factors associated with undergoing a major operating room procedure by diagnosis of interest. Statistically significant results are noted in bold text. NONOP = nonoperative; OR = odds ratio; 95% CI = 95% confidence interval; Ref. = reference. Cholecystitis (N = 12,760)
Sex Male Female Race White Black Hispanic Asian/Pacific Islander Native American Other Primary Payer Medicare Medicaid Private Self-pay No charge Other Median Household Income (quartile) 1 2 3 4 Hospital Bed Size Small Medium Large Teaching Status Rural Urban nonteaching Urban teaching Region Northeast Midwest South West
Bowel Obstruction (N = 4655)
Perforated Appendix (N = 17,906)
Spontaneous Pneumothorax (N = 705)
NONOP (%)
OR (95% CI)
NONOP (%)
OR (95% CI)
NONOP (%)
OR (95% CI)
NONOP (%)
OR (95% CI)
566 (24.94) 2352 (22.42)
1.00 (ref.) 0.87 (0.77, 0.98)
1552 (51.73) 967 (58.43)
1.00 (ref.) 1.31 (1.16, 1.48)
938 (8.58) 703 (10.08)
1.00 (ref.) 1.20 (1.08, 1.32)
325 (55.37) 69 (58.47)
1.00 (ref.) 1.14 (0.77, 1.68)
1049 (20.38) 561 (31.55) 969 (22.04) 46 (34.85) 25 (14.71) 115 (21.50)
1.00 (ref.) 1.80 (1.56, 2.07) 1.10 (0.99, 1.23) 2.09 (1.41, 3.09) 0.67 (0.43, 1.06) 1.07 (0.86, 1.34)
1308 (54.43) 699 (202.02) 463 (59.82) 83 (57.24) 9 (50.00) 103 (43.83)
1.00 (ref.) 0.82 (0.68, 0.99) 1.25 (1.05, 1.48) 1.04 (0.88, 1.23) 0.99 (0.83, 1.18) 0.94 (0.80, 1.11)
704 (9.05) 134 (12.09) 530 (8.40) 60 (12.35) 15 (7.98) 796 (9.19)
1.00 (ref.) 1.38 (1.08, 1.76) 0.92 (0.72, 1.19) 1.41 (0.99, 2.02) 0.87 (0.52, 1.46) 1.16 (0.93, 1.45)
230 (54.76) 40 (62.50) 53 (57.61) 23 (52.27) 7 (100.00) 20 (62.50)
1.00 (ref.) 1.38 (0.80, 2.37) 1.12 (0.71, 1.77) 0.90 (0.49, 1.69) 0.00 (0.00, 0.00) 1.10 (0.55, 2.21)
17 (32.69) 1607 (25.16) 927 (20.52) 227 (18.82) 13 (20.63) 117 (23.59)
1.88 (1.00, 3.53) 1.30 (1.18, 1.44) 1.00 (ref.) 0.90 (0.75, 1.08) 1.01 (0.54, 1.88) 1.20 (0.92, 1.55)
13 (54.17) 910 (51.53) 1313 (54.66) 140 (64.52) 9 (60.00) 132 (58.41)
0.92 (0.78, 1.08) 0.88 (0.77, 1.00) 1.00 (ref.) 1.51 (1.15, 1.99) 1.24 (0.45, 3.43) 1.16 (0.83, 1.64)
5 (17.24) 727 (9.29) 771 (9.32) 80 (7.58) 2 (5.56) 52 (7.91)
2.03 (0.67, 6.14) 1.00 (0.86, 1.16) 1.00 (ref.) 0.80 (0.62, 1.02) 0.57 (0.13, 2.45) 0.84 (0.59, 1.19)
5 (83.33) 103 (59.54) 244 (52.47) 20 (68.97) 3 (60.00) 19 (70.37)
4.53 (0.53, 38.82) 1.33 (0.94, 1.89) 1.00 (ref.) 2.01 (0.95, 4.27) 1.36 (0.23, 8.10) 2.15 (0.93, 4.97)
1118 (24.81) 701 (21.68) 617 (21.60) 415 (21.60)
1.00 (ref.) 0.84 (0.75, 0.94) 0.83 (0.74, 0.94) 0.84 (0.72, 0.96)
637 (54.17) 551 (55.16) 605 (53.92) 666 (52.69)
1.00 (ref.) 1.04 (0.88, 1.23) 0.99 (0.83, 1.18) 0.94 (0.80, 1.11)
416 (8.19) 386 (8.87) 415 (9.53) 397 (10.53)
1.00 (ref.) 1.09 (0.93, 1.27) 1.18 (1.00, 1.40) 1.32 (1.04, 1.67)
98 (61.64) 81 (55.86) 98 (61.64) 111 (55.50)
1.00 (ref.) 0.79 (0.50, 1.24) 0.86 (0.56, 1.33) 0.63 (0.42, 0.95)
336 (24.82) 760 (24.53) 1822 (21.93)
1.00 (ref.) 0.98 (0.82, 1.18) 0.85 (0.72, 1.00)
233 (52.83) 578 (57.06) 1708 (52.86)
1.00 (ref.) 1.02 (0.74, 1.40) 0.86 (0.64, 1.14)
182 (8.54) 334 (7.76) 1125 (9.81)
1.00 (ref.) 0.90 (0.57, 1.43) 1.16 (0.79, 1.73)
35 (59.32) 100 (56.18) 259 (55.34)
1.00 (ref.) 0.88 (0.49, 1.58) 0.85 (0.49, 1.46)
301 (24.16) 973 (20.37) 1644 (24.40)
1.00 (ref.) 0.80 (0.68, 0.95) 1.01 (0.86, 1.19)
219 (138.61) 890 (166.67) 1827 (51.52)
1.00 (ref.) 0.58 (0.41, 0.81) 0.41 (0.30, 0.56)
86 (5.29) 229 (4.44) 1326(11.92)
1.00 (ref.) 0.83 (0.61, 1.13) 2.42 (1.78, 3.30)
35 (79.55) 120 (60.30) 239 (51.73)
1.00 (ref.) 0.39 (0.19, 0.82) 0.28 (0.14, 0.56)
602 (27.84) 530 (23.93) 1007 (21.98) 779 (20.49)
1.00 (ref.) 0.82 (0.69, 0.97) 0.73 (0.63, 0.85) 0.67 (0.57, 0.79)
493 (50.41) 608 (55.47) 801 (53.98) 617 (56.24)
1.00 (ref.) 1.23 (0.97, 1.54) 1.15 (0.93, 1.43) 1.26 (1.01, 1.59)
321 (11.86) 375 (10.40) 410 (7.53) 535 (8.71)
1.00 (ref.) 0.86 (0.61, 1.21) 0.86 (0.61, 1.21) 0.86 (0.61, 1.21)
88 (49.44) 97 (62.18) 127 (57.73) 88 (49.44)
1.00 (ref.) 1.68 (1.10, 2.57) 1.40 (0.94, 2.06) 1.22 (0.78, 1.90)
common pediatric surgical diagnoses, making our study applicable to a large population of children treated by pediatric surgeons. A limitation of the KID is that the appropriateness of nonoperative or operative management cannot be evaluated owing to a lack of clinically granular information surrounding each diagnosis and management detail. Lastly, as mentioned above, this is the first study of its kind and our findings open the door to a number of future analyses. We observed several trends relating to patient demographics, income levels, and insurance patterns. Females may be more likely to
undergo operative intervention for abdominal surgical diagnoses (in this study, BO and PA) as the differential diagnosis for surgical abdominal pain in young females is not easily deduced using imaging, and includes certain emergent conditions for which nonoperative management is not feasible. Similarly, children of black and Asian/Pacific Islander race are at higher predisposition for acquiring gallbladder disease secondarily from hemopathies than the general population, which could influence providers' judgment toward performing cholecystectomy for presentations of cholecystitis in children of these
Table 3 Hospital-level characteristics associated with nonoperative management. Statistically significant results are noted in bold text. OR = odds ratio; 95% CI = 95% confidence interval. Ref. = reference.
Bed size Small Medium Large Teaching Status Rural Urban, nonteaching Urban, teaching Region Northeast Midwest South West
All Hospitals N = 4179, (%)
Hospitals with N 5 cases of 1 diagnosis N = 1102, (%)
Non-operative Hospitals N = 208, (%)
Operative Hospitals N = 894, (%)
OR (95% CI)
1734 (41.49) 1087 (26.01) 1358 (32.50)
111 (10.07) 281 (25.50) 710 (64.43)
21 (10.10) 51 (24.52) 136 (65.38)
90 (10.07) 230 (25.73) 574 (64.21)
1.00 (ref.) 0.95 (0.54, 1.67) 1.02 (0.61, 1.69)
1567 (37.71) 1736 (41.54) 867 (20.75)
101 (9.17) 484 (43.92) 517 (46.91)
23 (11.06) 66 (31.73) 119 (57.21)
78 (8.72) 418 (46.76) 398 (44.52)
1.00 (ref.) 0.54 (0.31, 0.91) 1.01 (0.61, 1.69)
518 (12.40) 1310 (31.35) 1520 (36.37) 831 (19.89)
191 (17.33) 226 (20.51) 342 (31.03) 343 (31.13)
48 (23.08) 61 (29.33) 52 (25.00) 47 (22.06)
143 (16.00) 165 (18.46) 290 (32.44) 296 (33.11)
1.00 (ref.) 1.10 (0.71, 1.71) 0.53 (0.34, 0.83) 0.47 (0.30, 0.74)
Please cite this article as: M. Ingram, H.L. Short, C. Sathya, et al., Hospital-Level Factors Associated with Nonoperative Management in Common Pediatric Surgical Procedures, Journal of Pediatric Surgery, https://doi.org/10.1016/j.jpedsurg.2019.10.006
6
M. Ingram et al. / Journal of Pediatric Surgery xxx (xxxx) xxx
ethnicities. Additionally, our findings of regional and academic variability in use of nonoperative management for bowel obstruction, pneumothorax, and cholecystitis suggest that practice or cultural preferences may exist that contribute to these observed trends in management. Further studies are needed to understand the influence of provider bias, patient factors, location, and possible disparities on the decision to elect nonoperative management. Our study has a number of limitations. We were unable to analyze particular clinical factors (i.e, duration of symptoms prior to presentation, findings on imaging, severity of presentation) that may have impacted clinical judgment toward operative or nonoperative management and appropriateness of that decision. Second, there may be some element of selection bias based on the cases captured in the KID and after our selection of diagnoses of interest and exclusion of transferred patients. Additionally, certain variables of possible interest (including whether the operating surgeon is pediatric vs adulttrained) were not available for this analysis. Nevertheless, the KID contains data on a large number of children treated across different types of hospitals in the majority of U.S. states, hence our findings of proportion of non-operatively managed patients are generalizable. Another limitation of the study is introduced when using ORPROC for analyses of interest to identify cases where operative intervention occurred. This variable is provided by the KID based on reporting of as many as 15 procedure codes that capture major operative intervention. While we expect that most procedures requiring a major operating room would be accurately coded in the discharge data, it is theoretically possible that some procedures could be missed if this process was not performed accurately. Another limitation is confounding, which was not controlled for in our univariate analyses or in our evaluation of variation across centers. Lastly, our classification of hospitals as nonoperative vs. operative based on median operative rates may be too crude a marker for this definition. This study has identified additional factors that may be beneficial to examine in order to understand more about nonoperative management. Some examples of these variables include examination of impact on length of stay, particular nuances in presentation or management, and possible associations with adult compared to pediatric treating surgeon. Future studies should aim at more focused analyses within each of these disease states to help identify institutional and patient factors that are directly contributing to variation in nonoperative care across centers. Multivariable analyses should be performed to control for confounding. Once key factors influencing operative vs nonoperative management have been identified, targeted quality improvement initiatives can developed. National pediatric protocols for indications for surgical intervention and metrics surrounding nonoperative management of cholecystitis, bowel obstruction, perforated appendicitis, and spontaneous pneumothorax management are needed to help reduce variation in practice patterns.
4. Conclusion Nonoperative management of common pediatric surgical diagnoses occurs in more than 20% of children and literature on this subject is scarce. There is significant variation in use of nonoperative versus operative management for common pediatric surgical diseases. Quality improvement initiatives aimed at understanding what causes this variability and its potential impact on outcomes are required to help reduce heterogeneity.
References [1] Szatmary P, Arora S, Sevdalis N. To operate or not to operate? A multi-method analysis of decision-making in emergency surgery. Am J Surg 2010;200(2):298–304. [2] Eddy DM. Clinical decision making: from theory to practice. Anatomy of a decision. JAMA 1990;263(3):441–3. [3] Yule S, Flin R, Paterson-Brown S, et al. Non-technical skills for surgeons in the operating room: a review of the literature. Surgery 2006;139(2):140–9. [4] Sacks GD, Dawes AJ, Ettner SL, et al. Surgeon perception of risk and benefit in the decision to operate. Ann Surg 2016;264(6):896–903. [5] Rutkow IM, Gittelsohn AM, Zuidema GD. Surgical decision making. The reliability of clinical judgment. Ann Surg 1979;190(3):409–19. [6] Rutkow IM. Surgical decision making. The reproducibility of clinical judgement Arch Surg 1982;117(3):337–40. [7] Rutkow IM. The reliability and reproducibility of the surgical decision-making process. Surg Clin North Am 1982;62(4):721–35. [8] Short HL, Zhu W, McCracken C, et al. Exploring regional variability in utilization of antireflux surgery in children. J Surg Res 2017;214:49–56. [9] Wennberg J, Gittelsohn. Small area variations in health care delivery. Science 1973; 182(4117):1102-8. [10] Birkmeyer JD, Reames BN, McCulloch P, et al. Understanding of regional variation in the use of surgery. Lancet 2013;382(9898):1121–9. [11] Kraemer K, Cohen ME, Liu Y, et al. Development and evaluation of the American College of Surgeons NSQIP pediatric surgical risk calculator. J Am Coll Surg 2016;223 (5):685–93. [12] Bilimoria KY, Liu Y, Paruch JL, Zhou L, Kmiecik TE, Ko CY, et al. Development and evaluation of the universal ACS NSQIP surgical risk calculator: a decision aid and informed consent tool for patients and surgeons. J Am Coll Surg 2013;217(5):833-42 e1-3. [13] Wilson NP, Wilson FP, Neuman M, Epstein A, Bell R, Armstrong K, et al. Determinants of surgical decision making: a national survey. Am J Surg 2013;206(6):970-7; discussion 7-8. [14] Introduction to the HCUP KID’s Inpatient Database (KID). Healthcare Cost and Utilization Project (HCUP). Agency for Healthcare Research and Quality. MD: Rockville; 2012. https://www.hcup-us.ahrq.gov/db/nation/kid/KID_2012_Introduction.pdf. [18] Tian Y, Heiss KF, Wulkan ML, et al. Assessment of variation in care and outcomes for pediatric appendicitis at children’s and non-children’s hospitals. J Pediatr Surg 2015: 1885–92. [19] Rice-Townsend S, Chen C, Barnes JN, et al. Variation in practice patterns and resource utilization surrounding management of intussusception at freestanding children’s hospitals. J Pediatr Surg 2012;48(1):104–10. [20] Cuenca AG, Islam S (2012) Pediatric pancreatic trauma: trending toward nonoperative management? Am Surg 78: 1204-1210. [21] Wandling MW, et al. Expanding the scope of quality measurement in surgery to include nonoperative care. Results from the American College of Surgeons NSQIP emergency general surgery pilot J Trauma Acute Care Surg 2017;83(5):837–45.
Please cite this article as: M. Ingram, H.L. Short, C. Sathya, et al., Hospital-Level Factors Associated with Nonoperative Management in Common Pediatric Surgical Procedures, Journal of Pediatric Surgery, https://doi.org/10.1016/j.jpedsurg.2019.10.006