Abstract
Interhospital transfer of emergency general surgery (EGS) patients is a common occurrence. Modern individual hospital practices for interhospital transfers have unknown variability. A retrospective review of the Maryland Health Services Cost Review Commission database was undertaken from 2013 to 2015. EGS encounters were divided into three groups: encounters not transferred, encounters transferred from a hospital, and encounters transferred to a hospital. In total, 380,405 EGS encounters were identified, including 12,153 (3.2%) encounters transferred to a hospital, 10,163 (2.7%) encounters transferred from a hospital, and 358,089 (94.1%) encounters not transferred. For individual hospitals, percentage of encounters transferred to a hospital ranged from 0 to 30.05 per cent, encounters transferred from a hospital from 0.02 to 14.62 per cent, and encounters not transferred from 69.25 to 99.95 per cent of total encounters at individual hospitals. Percentage of encounters transferred from individual hospitals was inversely correlated with annual EGS hospital volume (P < 0.001, r = −0.59), whereas percentage of encounters transferred to individual hospitals was directly correlated with annual EGS hospital volume (P < 0.001, r = 0.51). Individual hospital practices for interhospital transfer of EGS patients have substantial variability. This is the first study to describe individual hospital interhospital transfer practices for EGS.
EMERGENCY GENERAL SURGERY (EGS) patients comprise 26 percent of Maryland hospital encounters.1, 2 However, EGS patients do not always remain at one institution for the entirety of their care. Interhospital transfers commonly occur in EGS patients, with 7 per cent of patients at NSQIP hospitals admitted after interhospital transfer.3
Research on interhospital transfers is difficult. Large databases such as the NSQIP or National Inpatient Sample comprise a sample of patients and are not completely inclusive of all patients within a health-care system. For example, with the NSQIP database, the percentage of patients admitted from other hospitals to NSQIP hospitals after interhospital transfer can be calculated. However, when using databases that comprise a sample of patients, how hospitals participate in the system-wide transfer network is unknown, as is the overall burden of interhospital transfers. The Maryland State Health Services Cost Review Commission (HSCRC) presents a unique database which can be used to investigate the system-wide effects of interhospital transfer as the HSCRC database comprises all patient encounters within the state.4
Given this limitation in knowledge on interhospital transfers of EGS patients, the primary aim of this study was to describe the rates of interhospital transfer of EGS encounters for individual hospitals within the state of Maryland and the variability in these rate s. We hypothesized there would be variability in individual hospital practices for interhospital transfer of EGS patients.
Methods
Approval was first obtained from the University of Maryland Institutional Review Board. The American Association for the Surgery of Trauma has identified ICD-9 codes for EGS diagnoses, and these were used to identify eligible EGS encounters from the HSCRC database.5 A retrospective review of HSCRC encounters from 2013 to 2015 was then undertaken for the 45 Maryland state hospitals identified.
Three groups of encounters were created: encounters transferred to a hospital, encounters transferred from a hospital, and encounters not transferred. Encounters not transferred represented hospital encounters where treatment occurred at only one hospital without an interhospital transfer. Encounters transferred to a hospital represented hospital encounters after interhospital transfer. Encounters transferred from a hospital represented hospital encounters before interhospital transfer. Encounters transferred to a hospital and encounters transferred from a hospital would then represent the same group of patients before and after interhospital transfer.
Encounters were included if the record contained an EGS ICD-9 diagnosis code, if the patient was aged 20 years and older, and if admission and discharge locations within the encounter classified the encounter into one of the interhospital transfer categories. Encounters were excluded if the encounters were missing key data points, if the encounter was an intrahospital transfer, or if the encounter used ICD-10 coding after the change at the end of 2015. Age 20 years and older was used as a cutoff as exact ages were not available in the HSCRC dataset.
The HSCRC dataset uses multiple variables to stratify the degree of critical illness in each encounter. The severity of illness (SOI) score is based on the All Payer Related (APR) Diagnosis Related Groups from the Center for Medicaid and Medicare Services. The HSCRC dataset also includes the risk of mortality (ROM), which is based on APR Diagnosis Related Group codes, patient diagnoses, and surgical procedures performed.6 Both the ROM and SOI are divided into minor, moderate, major, and extreme.
Bivariate analysis was performed with chi-squared and Wilcoxin rank sum tests. Percentages of encounters transferred to a hospital, encounters transferred from a hospital, and encounters not transferred were calculated for each individual hospital. For example, if an individual hospital saw 1000 total EGS encounters over the study period with those encounters comprising 500 encounters not transferred, 300 encounters transferred from a hospital, and 200 encounters transferred to a hospital, the percentage of encounters not transferred for that individual hospital would be 50 per cent, the percentage of encounters transferred from a hospital would be 30 per cent, and the percentage of encounters transferred to a hospital would be 20 per cent.
Variability was defined as heterogeneity in individual hospital practices for interhospital transfer. Variability in this study was different from the statistical concept of “variance.” Rates of percentages of encounters transferred to a hospital and encounters transferred from a hospital were compared with annual EGS hospital volume with Pearson’s testing. Length of stay (LOS) was calculated as well for each individual hospital to again evaluate variability in individual hospital practices. Significance was considered as α < 0.001. Statistical analysis was completed with SAS software, version 9.4, of the SAS system for Windows® 7 operating system (Copyright © 2014, SAS Institute Inc., Cary, NC).
Results
There were 380,405 EGS encounters in the HSCRC dataset (Fig. 1). Overall, of the 380,405 EGS encounters, 172,475 (45.3%) were patients aged 65 years and older, 48,086 (12.6%) carried an extreme SOI, and 37,586 (9.9%) carried an extreme ROM. ICU admission was required in 46,553/380,405 (12.2%) and mortality was seen in 17,003/380,405 (4.5%) of EGS encounters.
Fig. 1.
Patient group selection algorithm.
Encounters not transferred comprised 358,089/380,405 (94.1%), encounters transferred from a hospital 10,163/380,405 (2.7%), and encounters transferred to a hospital 12,153/380,405 (3.2%) of all EGS encounters. Encounters not transferred were less likely to have an extreme SOI (11.6% vs 26.7% vs 30.8%) or an extreme ROM (9.1% vs 21.7% vs 23.2%) compared with encounters transferred from a hospital and encounters transferred to a hospital, respectively (Table 1).
Table 1.
Demographic Variables for Encounters Transferred from an Institution, Transferred to an Institution, and Not Transferred
| Total (n = 380,405) | Encounters not Transferred (a) (n = 358,089) | Encounters Transferred from a Hospital (b) (n = 10,163) | Encounters Transferred to a Hospital (c) (n = 12,153) | Column a vs b (P Value) | Column a vs c (P Value) | |
|---|---|---|---|---|---|---|
| Age group (years), n (%) | – | – | – | – | <0.001 | <0.001 |
| 2CM-4 | 72,592 (19.1%) | 68,779 (19.2%) | 1,546 (15.2%) | 2,267 (18.7%) | – | – |
| 45–64 | 135,338 (35.6%) | 126,458 (35.3%) | 3,989 (39.3%) | 4,891 (40.3%) | – | – |
| 65 and older | 172,475 (45.3%) | 162,852 (45.5%) | 4,628 (45.5%) | 4,995 (41.1%) | – | – |
| Gender, n (%) | – | – | – | – | <0.001 | <0.001 |
| Male | 180,424 (47.4%) | 168,560 (47.1%) | 5,372 (52.9%) | 6,492 (53.4%) | – | – |
| Female | 199,981 (52.6%) | 189,529 (52.9%) | 4,791 (47.1%) | 5,661 (46.6%) | – | – |
| Race, n (%) | – | – | – | – | 0.002 | <0.001 |
| Caucasian | 217,529 (57.2%) | 204,509 (57.1%) | 5,923 (58.3%) | 7,097 (58.4%) | – | – |
| African American | 110,121 (29%) | 103,580 (28.9%) | 2,961 (29.1%) | 3,580 (29.5%) | – | – |
| Hispanic | 11,570 (3%) | 11,041 (3.1%) | 265 (2.6%) | 264 (2.2%) | – | – |
| Asian/Pacific Islander | 5,928 (1.6%) | 5,598 (1.6%) | 145 (1.4%) | 185 (1.5%) | ||
| Other/unknown | 35,257 (9.3%) | 33,361 (9.3%) | 869 (8.6%) | 1,027 (8.5%) | – | – |
| Patient insurance, n (%) | – | – | – | – | <0.001 | <0.001 |
| Commercial | 103,158 (27.1%) | 97,121 (27.1%) | 2,726 (26.8%) | 3,311 (27.2%) | – | – |
| Public (Medicare/Medicaid) | 253,925 (66.8%) | 238,894 (66.7%) | 6,835 (67.3%) | 8,196 (67.4%) | – | – |
| Self-pay/charity | 16,638 (4.4%) | 15,868 (4.4%) | 371 (3.7%) | 399 (3.3%) | – | – |
| Other/unknown | 6,684 (1.8%) | 6,206 (1.7%) | 231 (2.3%) | 247 (2%) | – | – |
| SOI, n (%) | – | – | – | – | <0.001 | <0.001 |
| Minor | 44,541 (11.7%) | 43,600 (12.2%) | 364 (3.6%) | 577 (4.6%) | – | – |
| Moderate | 133,108 (35%) | 128,214 (35.8%) | 2,141 (21.1%) | 2,753 (22.7%) | – | – |
| Major | 154,670 (40.7%) | 144,646 (40.4%) | 4,941 (48.6%) | 5,083 (41.8%) | – | – |
| Extreme | 48,086 (12.6%) | 41,629 (11.6%) | 2,717 (26.7%) | 3,740 (30.8%) | – | – |
| ROM, n (%) | – | – | – | – | <0.001 | <0.001 |
| Minor | 131,712 (34.6%) | 127,629 (35.6%) | 1,675 (16.5%) | 2,408 (19.8%) | – | – |
| Moderate | 113,028 (29.7%) | 107,071 (29.9%) | 2,703 (26.6%) | 3,254 (26.8%) | – | – |
| Major | 98,079 (25.8%) | 90,830 (25.4%) | 3,576 (35.2%) | 3,673 (30.2%) | – | – |
| Extreme | 37,586 (9.9%) | 32,559 (9.1%) | 2,209 (21.7%) | 2,818 (23.2%) | – | – |
| ICU admission, n (%) | 46,553 (12.2%) | 40,697 (11.4%) | 1,858 (18.3%) | 3,998 (32.9%) | <0.001 | <0.001 |
| ICU LOS, median (IQR) | 2(1–5) | 2 (1–5) | 2 (1–4) | 4 (2–10) | 0.06* | <0.001* |
| LOS, median (IQR) | 4 (2–7) | 4 (2–7) | 3 (1–7) | 7 (4–14) | <0.001* | <0.001* |
| Total hospital charges, median (IQR) | $10,974 (6,459–20,514) | $10,799 (6,423–19,801) | $10,539 (5,765–20,954) | $27,185 (10,762–64,929) | <0.001* | <0.001* |
| In-hospital mortality, n (%) | 17,003 (4.5%) | 15,604 (4.4%) | 0 (0%) | 1,399 (11.5%) | <0.001 | <0.001 |
Calculated by Wilcoxon Mann-Whitney U test, all others by Pearson’s chi-squared test.
Encounters not transferred were less critically ill than encounters transferred from a hospital and encounters transferred to a hospital, with 11.4 per cent of encounters not transferred requiring ICU admission compared with 18.3 per cent for encounters transferred from a hospital (P < 0.001) and 32.9 per cent for encounters transferred to a hospital (P < 0.001). Encounters not transferred had an equivalent median ICU hospital LOS to encounters transferred from a hospital at two days (P = 0.06), but a shorter median ICU LOS for encounters transferred to a hospital at four days (P < 0.001). Mortality was lower in encounters not transferred at 4.4 per cent compared with 0 per cent for encounters transferred from a hospital (P < 0.001) and 11.5 per cent for encounters transferred to a hospital (P < 0.001).
Percentages of encounters not transferred, encounters transferred from a hospital, and encounters transferred to a hospital were calculated for individual hospitals. All 45 hospitals identified had EGS encounters not transferred and encounters transferred from those hospitals. Thirty-six of the 45 hospitals (80%) had encounters transferred to those hospitals. Ten of the individual hospitals (22.2%) had more encounters transferred to those hospitals than encounters transferred from those hospitals. Thirty-five of the individual hospitals (77.8%) had more encounters transferred from those hospitals than encounters transferred to those hospitals. These 10 hospitals with more encounters transferred to that hospital than encounters transferred from that hospital comprised 11,236/12,153 (92.5%) of the total encounters transferred to a hospital.
Variability was seen for the percentages of encounters not transferred, encounters transferred from a hospital, and encounters transferred to a hospital at individual hospitals. The percentage of encounters not transferred ranged from 69.25 to 99.95 per cent, the percentage of encounters transferred from a hospital ranged from 0.02 to 14.62 per cent, and the percentage of encounters transferred to a hospital ranged from 0 to 30.05 per cent for individual hospitals.
This variability in the percentage of encounters not transferred, encounters transferred from a hospital, and encounters transferred to a hospital at each institution continued when stratified by SOI subsets (Table 2). The greatest ranges of percentages of encounters not transferred, encounters transferred from a hospital, and encounters transferred to a hospital for individual hospitals were within extreme SOI encounters. Conversely, the narrowest ranges of percentages of encounters not transferred, encounters transferred to a hospital, and encounters transferred from a hospital for individual hospitals were within the minor SOI encounters.
Table 2.
Range of Percentage of Hospital Encounters Not Transferred, Transferred from a Hospital, and Transferred to a Hospital for Individual Hospitals Stratified by SOI Category
| Range(%) | |
|---|---|
| Minor SOI | |
| Encounters not transferred | 88.5–100 |
| Encounters transferred to a hospital | 0–11.3 |
| Encounters transferred from a hospital | 0–6.5 |
| Moderate SOI | |
| Encounters not transferred | 79.4–99.9 |
| Encounters transferred to a hospital | 0–20.2 |
| Encounters transferred from a hospital | 0–15.1 |
| Major SOI | |
| Encounters not transferred | 67.6–100 |
| Encounters transferred to a hospital | 0–31.8 |
| Encounters transferred from a hospital | 0–14.8 |
| Extreme SOI | |
| Encounters not transferred | 53.1–99.7 |
| Encounters transferred to a hospital | 0–45.7 |
| Encounters transferred from a hospital | 0.3–37.5 |
Variability in individual hospital practices also existed for hospital LOS. The median LOS for encounters not transferred ranged between three and five days for individual hospitals. The median LOS for encounters transferred from a hospital ranged from two to seven days for individual hospitals, and was non-normally distributed. The median LOS for encounters transferred to a hospital had the broadest range, from 3 to 14 days for individual hospitals (Fig. 2 A–C).
Fig. 2.
Median LOS distribution for encounters not transferred (A), encounters transferred from a hospital (B), and encounters transferred to a hospital (C).
The percentage of encounters transferred from a hospital at individual hospitals was inversely correlated with annual EGS hospital volume (P < 0.001, r = −0.59), with a broad range of percentage of encounters transferred from a hospital correlating with lower annual EGS hospital volume (Fig. 3). The percentage of encounters transferred to a hospital at individual hospitals was directly correlated with annual EGS hospital volume (P < 0.001, r = 0.51).
Fig. 3.
Annual EGS hospital volume compared with percentage of encounters transferred from a hospital.
Discussion
Greater than 1 in 20 EGS encounters are involved in interhospital transfer in the state of Maryland, supporting estimates of the commonality of interhospital transfers in EGS from previous work.3 Although all hospitals identified in this study participated in interhospital transfers, there was substantial variability in practices for interhospital transfer when examined at the individual hospital level. To the authors’ knowledge, this study is the first description of individual hospital practices in interhospital EGS transfers.
The need for interhospital transfer will depend on individual hospital resources, surgeon experience, and need for specialized care. Similarly, acceptance of interhospital transfers will depend on the ability of each individual hospital to provide the care required for each individual patient. When the percentages of encounters transferred from a hospital and encounters transferred to a hospital were compared with the resources available at each hospital (approximated with annual EGS hospital volume), appropriate allocation of EGS encounters was noted in the interhospital transfer system overall. Patients undergoing interhospital transfer moved from low-volume hospitals to high-volume hospitals where presumably more resources exist for patient care, especially for the critically ill.
Although all hospitals participated in interhospital transfers, variability was seen in the percentages of EGS encounters transferred from and transferred to individual hospitals. Some hospitals transferred from their hospital almost no EGS encounters initially presenting to their institution, whereas other hospitals transferred from their hospital almost one of seven EGS encounters initially presenting to their institution. Multiple hospitals did not accept any EGS encounters transferred to their hospital, whereas for other hospitals EGS encounters transferred to their hospital comprised one of three of their total EGS encounters.
There was also significant variability in LOS of EGS encounters undergoing interhospital transfer, with the median LOS in encounters transferred to another hospital having a substantial and broad range of days. This variability in LOS for receiving institutions after interhospital transfer suggests there may be differences in the complexity and degree of critical illness of patients that individual hospitals accept after interhospital transfer.
Similar variability between individual institutions was seen in the LOS of encounters transferred from a hospital, with a broad range in median LOS in these encounters before interhospital transfer. This suggests that there is variability not only in individual hospital rates of interhospital transfer, but also in their practice of time until interhospital transfer from that institution. This variability in LOS before interhospital transfer suggests that patients who undergo interhospital transfer may have become more ill at the first institution, thus necessitating interhospital transfer to a higher resource institution, although this cannot be directly concluded from the data available in this administrative dataset.
Interhospital transfers of EGS patients should be undertaken if another institution can better provide resources needed for patient care7 and should not be dissuaded if medically appropriate. However, interhospital transfer of EGS patients could influence measurement of quality metrics8 as EGS patients who have undergone interhospital transfer are known to have higher rates of mortality, complications, and reoperation.3 Risk stratification of patients after transfer has the potential to mitigate the reporting impact of accepting patients in transfer.3, 8–10 A noteworthy example from this study is the 0 per cent mortality of encounters transferred from a hospital despite an 11.5 per cent mortality for the encounters transferred to another hospital.
A limitation of this study is that it entails analysis of a large administrative database. Analysis is limited to the variables present within the dataset. Reasons for interhospital transfer are unknown. Hospital resources were inferred from annual EGS hospital volume, with higher annual EGS hospital volume assumed to represent higher resources. Only one interhospital transfer could be accounted for, and patients do sometimes undergo more than one interhospital transfer.11 The rates of request (or decline) of interhospital transfer are unknown. Generalizability of these data to other states is also unknown, as this study used a single-state database.
In addition, it merits discussion that the encounters transferred from a hospital and encounters transferred to a hospital represent the same patients but in different hospital encounters, as these were the encounters before interhospital transfer and after interhospital transfer. The number of encounters in these groups was not the same. Given that the number of encounters transferred from a hospital is less than the number of encounters transferred to a hospital, it is likely that an EGS diagnosis was unknown before interhospital transfer or occurred as a progression of disease. The authors acknowledge this discrepancy as a limitation.
The network of interhospital transfers for EGS patients is used by all hospitals in the state of Maryland, and transfer of patients seems appropriate from lower resource institutions to higher resource institutions. However, substantial variability exists for interhospital transfer practices, especially when examining individual hospital rates of interhospital transfer, LOS before and after interhospital transfer, and SOI of encounters undergoing interhospital transfer. This is the first study to our knowledge describing the interhospital transfer patterns in EGS patients.
Acknowledgments
We would like to thank Gordon Smith for his assistance with this study.
Funding: Jennifer S. Albrecht was awarded AHRQ grant 1K01HS024560. Anthony V. Herrera was training under the T32 training grant T32 AG00262.
REFERENCES
- 1.Narayan M, Tesoriero R, Bruns BR, et al. Acute care surgery: defining mortality in emergency general surgery in the state of Maryland. J Am Coll Surg 2015;220:762–70. [DOI] [PubMed] [Google Scholar]
- 2.Bruns BR, Tesoriero R, Narayan M, et al. Emergency general surgery: defining burden of disease in the state of Maryland. Am Surg 2015;81:829–34. [PubMed] [Google Scholar]
- 3.Lucas DJ, Ejaz A, Haut ER, et al. Interhospital transfer and adverse outcomes after general surgery: implications for pay for performance. J Am Coll Surg 2014;218:393–400. [DOI] [PubMed] [Google Scholar]
- 4.HSCRC Overview. Available at: http://www.hscrc.state.md.us/aboutHSCRC.cfm. Accessed March 15, 2016.
- 5.Shafi S, Aboutanos MB, Agarwal S Jr, et al. Emergency general surgery: definition and estimated burden of disease. J Trauma Acute Care Surg 2013;74:1092–7. [DOI] [PubMed] [Google Scholar]
- 6.Averill R, Goldfield N, Hughes J, et al. All patient refined diagnosis related groups. Available at: https://www.hcup-us.ahrq.gov/db/nation/nis/APR-DRGsV20MethodologyOverviewandBibliography.pdf. Updated 2003. Accessed March 15, 2016.
- 7.Santry HP, Janjua S, Chang Y, et al. Interhospital transfers of acute care surgery patients: should care for nontraumatic surgical emergencies be regionalized? World J Surg 2011;35:2660–7. [DOI] [PubMed] [Google Scholar]
- 8.Combes A, Luyt CE, Trouillet JL, et al. Adverse effect on a referral intensive care unit’s performance of accepting patients transferred from another intensive care unit. Crit Care Med 2005; 33:705–10. [DOI] [PubMed] [Google Scholar]
- 9.Rosenberg AL, Hofer TP, Strachan C, et al. Accepting critically ill transfer patients: adverse effect on a referral center’s outcome and benchmark measures. Ann Intern Med 2003;138: 882–90. [DOI] [PubMed] [Google Scholar]
- 10.Golestanian E, Scruggs JE, Gangnon RE, et al. Effect of interhospital transfer on resource utilization and outcomes at a tertiary care referral center. Crit Care Med 2007;35:1470–6. [DOI] [PubMed] [Google Scholar]
- 11.Iwashyna TJ, Christie JD, Moody J, et al. The structure of critical care transfer networks. Med Care 2009;47:787–93. [DOI] [PMC free article] [PubMed] [Google Scholar]