Inter-Hospital Variation of Inpatient Versus Outpatient Pediatric Burn Treatment After Emergency Department Evaluation

Abstract

Background:

Approaches to burn care in the pediatric population are highly variable and can be targeted as a potential measure in cost-reduction. We hypothesized that institutions vary significantly in treatment allocation of non-severe burns to either inpatient or outpatient care.

Methods:

We queried the PHIS database for fiscal year 2017 to quantify small pediatric burn admissions and Emergency Department visits(ED). The ICD-10 code T31.0 was used to identify burns involving ≤10% of total body surface area (TBSA). Centers were categorized by burn center status and length of stay, readmissions, and charges were compared.

Results:

Inpatient versus outpatient management distribution was significantly different across the included pediatric children’s hospitals (n=34, p<0.00001). When data was analyzed with respect to outpatient care, a bimodal distribution distinguished two groups: high hospital-utilizers with an average of 30% outpatient burn care and low-utilizers averaging 87%. Median inpatient charge per patient was greater than 31-fold compared to ED burn management (p<0.0001).

Conclusions:

Variability of inpatient versus outpatient pediatric burn management in small burns was significant. Compared to outpatient burn care, inpatient care is significantly more costly. Implementing protocols and personnel to provide adequate attention to small burns in the ED could be an important cost-saving measure.

Introduction:

Burn injury remains a major threat to public health, affecting nearly half a million Americans each year annually^1,2. Children and adolescents account for approximately 120,000 (24%) of these burns, with a disproportionate majority occurring in those under 6 years of age³. The Pediatric Health Information System (PHIS) database is a national database that focuses on de-identified data collection of clinical and resource information for the purposes of comparative analysis and quality improvement. Patient populations reported in the database include inpatient, emergency department and observation.

Guidelines for referral of pediatric burn patients to a burn center for admission have been established that recommend referral of all burns that are: >10% TBSA, full thickness, cross joints, chemical, electric, associated with inhalational injury, or include burns on hands, face, and/or genitalia. This criterion also recommends transfer of children with burns if the hospital does not have qualified personnel or supplies to treat the burn⁴. Interestingly, a set of guidelines targeting trauma centers that treat burns recommend that an adult burn center may serve as an alternative if no regional pediatric burn center is available⁵. While guidelines suggest referral of pediatric burns from adult centers or children’s hospitals without burn centers, this does not suggest these patients should be admitted. In fact, literature suggests that up to 90% of pediatric burns can be treated with outpatient management based on the above stated criteria. This has been successful provided patients have adequate home support, minimal co-morbidities, adequate pain control, and reliable access to ambulatory care facilities⁶.

Despite these guidelines institutions have reported high referral and admission rates. One of study was recent single-center retrospective review by Anderson et al that evaluated admission rates in relation to burn severity. The study revealed 80% of low acuity (<5% TBSA) burns were admitted to observation⁷.

With increasing focus on containing healthcare costs, the outpatient care setting may serve as an extension to inpatient care with attempts to decrease inpatient length of stay. We sought to better understand the national trends in admission rates for non-severe pediatric burn patients and hypothesized that admission rates for non-severe pediatric burn patients could provide a potential area of improvement for institutional cost-efficacy.

1. Methods:

1.1. Pediatric Health Information System (PHIS) database query:

We performed a query of the PHIS database for fiscal year 2017 to quantify small pediatric burn admissions and Emergency Department visits and analyze associated charges. The PHIS database is a national administrative comparative database with de-identified information and is an extremely robust administrative dataset in pediatrics. It focuses on resource data for inpatient care as well as emergency department and observation encounters in pediatric hospitals, but does not capture clinic encounters. The PHIS database utilizes standardized algorithms based on temporal data to determine admission status of patients and captures parameters like charges, length of stay (LOS) complications and readmissions pertaining to the inpatient, observation, or ED encounters. This is a retrospective analysis of information from the de-identified PHIS database and thus, in accordance with the Medical University of South Carolina Institutional Review Board (IRB), was exempted from IRB approval. The ICD-10 code T31.0 was chosen to identify burns involving less than 10% of TBSA. This specific ICD-10 code only includes non-severe partial thickness burns. Patients included were pediatric patients <18 years old with small non-severe burns of less than 10% total body surface area (TBSA) at any center that accepts and treats pediatric patients.

The key performance indicator reports were generated within this date range that included patient numbers, and charges for Emergency Department (ED) visits, inpatient admissions and observation admissions. The We included centers with medium to high-volume of burns, which we defined as ≥ 15 pediatric burn patient visits per year based on a recent study on burn center volume⁸. These institutions were required to have data for inpatient, observation and emergency department management of pediatric burns, but were not required to be burn centers or free-standing children’s hospitals. Although hospitals have been de-identified for this study we also analyzed each of the above performance indicators for significance between the following subsets of centers: free-standing children’s hospitals (FS) versus combined adult and pediatric centers(CC). Length of stay (LOS) was measured by ratio of expected versus actual length of stay. Readmission data was available only for patients previously admitted as either inpatient or observation. For this reason patients treated as outpatients in the ED were not included in this analysis as they were not captured as a readmission if admitted following an ED visit.

1.2. Classification of institutions by burn center status and hospital utilization:

The ABA burn center listing was utilized to identify centers as either verified burn centers (VBC), self-identified burn centers (SIBC), or having no official burn status. We also classified centers as either high hospital-utilizers (HH) which we defined as >70% inpatient burn care or low hospital-utilizers (LH) defined as <30% inpatient burn care.

1.3. Statistical analysis:

Parameters of the data analyzed were median and mean charge data for all inpatient, observation and Emergency Department visits. A breakdown of ED, versus observational, versus inpatient management strategies within each institution was depicted in graph form for comparison purposes. Data was analyzed using a student’s t-test or ANOVA test for analysis of data sets with two groups or three groups, respectively, to determine significant differences in length of stay and readmissions between the groups. Significance defined as p < 0.05. A histogram was generated to show patterns in burn management admission across 34 centers with medium to high-volume of burn patients.

2. Results:

2.1. Burn patient admission patterns across institutions:

Of the 39 centers identified 24 provided both inpatient and ED management of burns, 7 provided exclusively ED management, and 3 institutions provided inpatient only management of non-severe burns. A total of 34 centers treating at least 15 burns per year from the PHIS database were included in the review. Four of the included institutions were verified burn centers (VBC) and five were self-identified burn centers (SIBC) according to ABA registry. The remaining 25 did not have any official burn status. The included institutions consisted of 22 free-standing children’s hospitals (FS) and 12 combined pediatric and adult centers (CC).

Overall, included centers performed an average of 86±14 ED burn evaluations per hospital (range: 6–371). Management of non-severe burns was highly variable across institutions. Overall inpatient admissions of non-severe burns across institutions ranged from 0% to 100% (mean of 16%). In comparison, observation admissions ranged from 0% to 70% with a mean of 8.0% and emergency department management of non-severe burns ranged from 0 to 100% (p <0.001) with an mean of 82%.

The total collective number of burns treated in all hospitals was 3,797 with a mean number of 111 patients per hospital. The total number of cases treated on an inpatient basis was 624 (16.4%). The total number of cases with observational management was 429 (11.3%), and the total number of cases treated in the emergency department was 2,744 (72.3%). While the majority of total burns were treated in the ED as outpatients, the proportion of ED-only burns was not evenly distributed across institutions. Across the 34 institutions there was a statistically significant difference between distribution of management strategy between inpatient admissions (n=20), observation admissions (n=22), and ED management (n=31) (p<0.00001) with not every institution having patients in each category. When analyzed with respect to emergency department care, a bimodal distribution was observed distinguishing two groups: high hospital-utilizers (HH) with an average of 8% ED burn care and low-utilizers (LH) averaging 87% ED care (Fig. 1). The HH group was distinguished as having <30% ED management of small burns, whereas the LH group was defined as having > 70% of burn care through the ED. When management of non-severe burns was further broken down into three categories: ED only, observation and inpatient, distribution was highly variable (Fig. 2).

Figure 1:

Histogram demonstrating bimodal distribution of institutions as high hospital utilizers with <30% ED only management and admission of 70–100% of small burns versus >70% of ED management and only 0–30% of patients admitted.

Figure 2:

Bar graph showing distribution of burn management allocation: inpatient (gray lines), observation (white), or outpatient (black), within each hospital (n=34) ranging from 0 to 100% outpatient. High hospital utilizers (>70% admission rate) and low hospital utilizers (<30% ◆ admission rate).

= Verified Burn Center

2.2. Comparative charge analysis of non-severe pediatric burns:

The mean of total ED charges per institution (n=31) was 122,024 (range: $25,9316–501,444), with median charges per patient of $1,586. This was lower than the mean observation admission charge (n=22) of $221,901 (range: $10,004 – $1,697,973) with median charge per patient of $10,214. Further increase in charge was noted with inpatient total charges (n=20) with a mean of $1,514,560 (range: $9,475 – 5,509,286). Median inpatient charge per patient was $49,961 and almost 5-fold more than observational management and 31 times that of the emergency department treatment (Table 1). On statistical analysis comparing the three different types of management with ANOVA, inpatient charges were significantly increased compared to observation and outpatient (ED only) management (p<0.0001) (Table 1). Analysis of hospital utilization revealed that the ED charges of the HH group (>30% inpatient care, n=10) were significantly increased compared to LH groups (<30% in patient, n=24) (p<0.05). There was, however no statistical difference in inpatient or observation charges between high and low hospital utilizers. Additionally no statistical difference was observed between institutions with regard to burn center status or between free-standing children’s hospitals compared to combined pediatric and adult centers (Table 2).

Table 1:

Median charges compared between inpatient admission burn management and observational and Emergency Department (ED) management.

Management Charges	Median Charge ($)	Range ($)
ED Charges	1,420 ± 20.5	471–2,540**
Observation Charges	10,455 ± 235	5,626 – 1,697,973**
Inpatient Charges	47,766 ± 2,743	9,475 – 5,509,286**

ED and observation p values are compared to inpatient charges.

^**p<0.00001.

Table 2:

Comparison of ED charges between different groups of hospitals.

Institution type	Median ED Charge ($)	Mean ED Charge ($)	P-value
Burn Center Status	VBC: 1,708 SIBC: 1,567 Neither: 1,586	VBC: 2,146 ± 726 SIBC: 1,576 ± 214 Neither: $1,724 ± 119	p = 0.48
Children’s hospital status	FS: 1,677 CC: 1,520	FS: 1,750 ± 126 CC: 1,712 ±226	p = 0.88
Hospital utilization	HH: 2,316 LH: 1,454	HH: 2,202 ± 314 LH: 1,587 ± 99	p < 0.05

VBC= verified burn center, SIBC = self-identified burn center, neither =no official burn status, FS = Free-standing children’s hospital, CC = combined pediatric and adult center, HH = high hospital-utilizers, LH = low hospital-utilizers.

2.3. Admission practices between institutions based on hospital utilization, burn center status and children’s hospital status:

Percent admission was calculated for all institutions by adding total number of observation and inpatient admissions and dividing by total treated patients. We observed a significant difference in percent admission between our HH and LH group (p<0.000001) confirming that our cut-off dividing these two groups was relevant. There was, however, no significant difference between percent admissions regardless of burn center status: verified, self-identified, or neither (p=0.09). Although, interestingly, all four of the verified burn centers were in the high hospital-utilizer group (Fig. 2). When compared to combined pediatric and adult centers (n=12), free-standing children’s centers (n=22) had significantly decreased admissions (p<0.05) (Table 3).

Table 3:

Percent admission was calculated by division of the sum of inpatient and observation encounters by the total number of non-severe burns treated at a center and multiplied by 100.

Institution type	Median Percent Admission	Mean Percent admission	P-value
Burn Center Status	VBC: 74.4 SIBC: 14.9 Neither: 8.25	VBC: 70.7 ± 12 SIBC: 31.4 ±15.8 Neither: 18.6 ± 5.7	p = 0.09
Children’s hospital status	FS: 11.6 CC: 30.9	FS: 18.5 ± 5.9 CC: 41.3 ± 10.6	p < 0.05
Hospital utilization	HH: 80 LH: 1,454	HH: 78.6 ± 5.5 LH: 8.2 ± 1.7	p < 0.000001

2.4. Analysis of readmission and length of stay with respect to ED, observation, or inpatient management.

33 of the 34 included centers had readmission data and all 34 centers had length of stay data. Between the 22 FS hospitals and 12 combined pediatric and adult centers (CC) however no significant difference was seen in readmissions (p=0.99). Included in this study were four verified burn centers that had gone through formal accreditation process through the ABA, five self-identified burn centers which are recognized on the ABA directory, and 25 non-burn centers. No significant difference was observed between the three categories of small burn management with respect to readmissions (p=0.6).

Analysis of length of stay was measured by ratio of expected to observed length of stay. Between FS (n=22) and CC (n=12) there was no difference in LOS in any type of burn management. However when we compared LOS between the highest (>30% inpatient) (n=8) and lowest hospital utilizers (>80% outpatient) (n=21), there was a significant difference in LOS. There was no statistical difference in LOS in the ED management setting in this analysis, but however in the inpatient and observations settings, the high hospital-utilizers had a shorter LOS compared to the low hospital utilizers (p<0.05). Data revealed no significant difference in inpatient LOS when analyzed with respect to burn center status (p=0.53) and there was not enough data across all groups to compare ED or observations LOS in this category.

3. Discussion:

In this multicenter analysis of 39 children’s burn centers, we found a significant discordance in utilization of inpatient services in pediatric patients who suffered minor (<10% BSA) burns. Additionally these burn management practices were, not surprisingly, directly related increased mean charges. We identified significant variation in resource utilization for a relatively homogenous population with just 28% of the patients who were either admitted or observed accounting for 79% ($37 million) of the total charges for burn care in the PHIS database over this time period. Institutions that successfully managed these patients in the ED only, did so at a mean charge of $1,586 which is 6 times less than observation and 31 times less inpatient admissions. Given that healthcare management is transitioning from fee for service to bundled accountable care reimbursement, evaluation of burn admission is a potential area for cost-improvement^9,10. Under a managed care model admission for less severe burns would result in minimal reimbursement and could potentially result in loss of revenue for hospitals.

When the data was analyzed based on burn center status we found no significant differences charges, readmission, or inpatient length of stay, Despite all four verified burn centers (VBC) falling within the high hospital utilizers group there was also no statistical difference in percent admission. Although there was an impressive trend of increased percent admission in verified burn centers especially in the median percent admission (Table 3). The VBC group has a median percent admission of 74.4 compared to SIBC (14.9) and non-burn centers (8.25) but the mean did not quite reach significance. However there was only a small subset of VBCs that were also children’s hospitals included in the PHIS database, and so this is certainly not a representative comparison of verified burn centers and is a limitation of this study.

Conversely the comparison between free-standing children’s hospitals and combined pediatric and adult centers showed a significant difference between the two groups in percent of admissions. It is interesting that FS hospitals had a lower percent admission compared to CC. This may reflect the inability of PHIS to capture specific circumstances related to patient encounters such as social needs and circumstances.

Analysis of differences with respect to hospital utilization identified a link between increased percent admissions, charges with a decreased LOS as these were all statistically significant differences in the high hospital-utilizers compared to low hospital-utilizers. The finding of HH having a shorter LOS could make the argument that the institutions with increased admitting practices may have shorter LOS because many of these non-severe burns do not actually require admission and are therefore discharged quickly. This is a pattern that has been reported in both a 2011 study by Vercruyyse et al² and a 2018 study by Anderson et al⁷ that demonstrated high utilization of very short duration inpatient services for even low acuity burns despite efficacy of outpatient services for many of these burns. Interestingly, uninsured burn patients have been found to have increased hospital length of stay (LOS), increased charges, and more concerning, increased mortality compared to insured patients¹¹. However when uninsured burn patients were compared to patients with Medicare or Medicaid, uninsured patients with similar injury severity were found to have more complications and shorter LOS¹². It could be that the uninsured patient volume is much higher at these HH centers and therefore admission with a short LOS is more likely.

Although we did not find any particular similarity between hospitals with high readmission rates, there were 7 hospitals with readmission rates greater than 10%, with one as high as 35%. One recent study assessed pediatric surgery procedures that involve high readmission rates as a means of improving care and costs¹³. It focused on care of non-severe burns following ED discharge is as a potential mode of cost-reduction. A more detailed investigation into the cause of high readmission rates was not possible within the confines of the accessible information gathered from the PHIS database.

3.1. Study Limitations

This study has several limitations. One limitation of using an administrative database is the inability to compare individual patient-level outcomes data. While we could look at readmissions of inpatient or observation patients, the complication rates for the ED managed patients were not captured as PHIS is limited to in-hospital encounters only. The PHIS database is also unable to distinguish between patients admitted at one hospital and then readmitted to another, and so any patients in these circumstances would not have been captured by the database. It is also important to note that PHIS database is a large billing dataset that relies on accurate healthcare coding for fidelity¹⁴ but other studies have validated the accuracy of coding using the PHIS database which undergoes a robust analysis of data for validity and reliability^14,15. Despite this, it is possible that, if improperly coded, more complex burns such as full thickness or electrical burns could be included in the T31.0 cohort. As a retrospective review of a database our access to information is restricted to the data collected at the time. Additionally, while cost would be a much better parameter to analyze than charge, given lack of transparency on actual hospital cost to the patient, the best surrogate to cost we have available for comparison is charge. However, despite these limitations the differential charges as well as identification of inter-hospital variability in management were extremely compelling.

4. Conclusions

Several interventions can be utilized to help optimize ED initiated outpatient management of these burns. First, the availability of pediatric sedation services 24/7 in the emergency department is necessary to adequately control anxiety and pain surrounding initial debridement. Parental education and guidance with dressing changes helps facilitate home compliance with the treatment plan and is necessary before discharge. While direct interaction with a burn clinician would be optimal, the use of telemedicine has been found to be effective in both parental education and wound assessment in patients receiving treatment for ED initiated outpatient burns¹⁶.

The US healthcare system is moving towards accountable care organizations with population-based reimbursement models and a focus on bundling of care. A need for focus cost-saving measures in healthcare is becoming more apparent. Based on our study, non-severe pediatric burn treatment is an area that can greatly improve cost-efficacy for pediatric centers with medium to high-volume burn care.

Abbrieviations:

TBSA

total body surface area

ED

Emergency Department

PHIS

Pediatric Health Information System

HH

High Hospital-utilizers

LH

Low Hospital-utilizers

VBC

Verified Burn Center

LOS

Length of Stay

FS

Free-standing

CC

combined pediatric and adult hospital