Harborview Burns – 1974 to 2009

Background Burn demographics, prevention and care have changed considerably since the 1970s. The objectives were to 1) identify new and confirm previously described changes, 2) make comparisons to the American Burn Association National Burn Repository, 3) determine when the administration of fluids in excess of the Baxter formula began and to identify potential causes, and 4) model mortality over time, during a 36-year period (1974–2009) at the Harborview Burn Center in Seattle, WA, USA. Methods and Findings 14,266 consecutive admissions were analyzed in five-year periods and many parameters compared to the National Burn Repository. Fluid resuscitation was compared in five-year periods from 1974 to 2009. Mortality was modeled with the rBaux model. Many changes are highlighted at the end of the manuscript including 1) the large increase in numbers of total and short-stay admissions, 2) the decline in numbers of large burn injuries, 3) that unadjusted case fatality declined to the mid-1980s but has changed little during the past two decades, 4) that race/ethnicity and payer status disparity exists, and 5) that the trajectory to death changed with fewer deaths occurring after seven days post-injury. Administration of fluids in excess of the Baxter formula during resuscitation of uncomplicated injuries was evident at least by the early 1990s and has continued to the present; the cause is likely multifactorial but pre-hospital fluids, prophylactic tracheal intubation and opioids may be involved. Conclusions 1) The dramatic changes include the rise in short-stay admissions; as a result, the model of burn care practiced since the 1970s is still required but is no longer sufficient. 2) Fluid administration in excess of the Baxter formula with uncomplicated injuries began at least two decades ago. 3) Unadjusted case fatality declined to ∼6% in the mid-1980s and changed little since then. The rBaux mortality model is quite accurate.


Introduction
The Harborview Burn Center in Seattle, WA opened in 1974 under the leadership of P. William Curreri, M.D., Janet Marvin, R.N., M.N., Verna Cain, R.N., and Leslie Einfeldt, R.N. Collection of the demographics, injury characteristics, and outcome from acute burn injuries began at that time. We have continued this database without interruption for 36 years (1974-2009) and it now provides a unique opportunity to review changes in burn care over a long period of time, at a single center, including both children and adults, with a stable surgical and rehabilitation staff.
There have been several prior reviews of burn injury demographics, prevention, and care since the mid-1990s [1,2,3,4,5,6,7, 8,9,10,11]. However, most were either focused on a select patient cohort or outcome, were discontinuous in time, or covered a relatively short time span. We herein report our results and include observations on changes in numbers of standard and short-stay admissions, age distribution, transport patterns, TBSA%, race/ethnicity, payer status, ''fluid creep'' [12], burn surgery, length of stay, mortality, trajectory to death, and ''unprecedented'' survival. The changes are highlighted at the end of the manuscript and the implications for burn care organization and delivery presented in the Discussion.

Definitions
We defined adults as $16 years of age and children as ,16 years of age. Incidence rate was defined as the number of events per 100,000 of the relevant population per year. All incidence rates were calculated using solely admissions and populations from WA (populations are shown in File S1) where WA State refers to Washington State. In this article we classified admissions into two types. Standard Admission (Type 1): Classic burn admission of any number of days for inpatient care of complex wounds until healed or grafted. Short-Stay Admission (Type 2): Admission for one or two days for wound care, education and pain control followed by outpatient management. The revised Baux score (rbaux) is calculated as age + TBSA% +17*(presence of inhalation injury) [13] where TBSA% is total body surface area percent).

Human Subjects
The Human Subjects Division of the University of Washington approved this study (#34999).

Data Sources
All patients (n = 14,266) admitted to the Harborview Burn Center with new cutaneous burn injuries from all etiologies between 8/2/1974 and 12/31/2009 were prospectively entered into a database. Those immediately placed on comfort care on arrival were included whereas persons with only smoke inhalation (no cutaneous injury) were excluded. The data was collected by various faculty/staff over the decades and the data points were those reported below. Prior to analysis, the entire database was searched for missing and out of range values and the data obtained or corrected from the medical records, if possible. The final result is that the maximum number of missing values was 305/14266 (2.1%) for payer status. The percent of missing values for all other parameters reported was ,2.1%.
Several other datasets were also used in this analysis. The Comprehensive Hospital Abstract Reporting System (CHARS) administered by the WA State Department of Health [14] was used to obtain the number of admissions in WA State. Populations for WA State were obtained from HistoryLink.org [15]. Populations of the several race/ethnicities were obtained from the Office of Financial Management of WA State [16] and HistoryLink.org [15]. WA [18].
Comparisons of our historical database were made to the ABA-NBR (American Burn Association National Burn Repository, American Burn Association, Chicago) using data obtained in April 2011 (version 6.0). The ABA-NBR includes data all persons admitted to 94 voluntarily participating U.S. burn centers. However, since not all hospitals participate, it is a convenience sample and does not include all persons or a random sample of all persons admitted to hospitals or burn centers for care of burn injuries [19].
The ABA-NBR dataset contained 280,664 records after excluding those from the Harborview Burn Center. Harborview records were excluded from our analysis to ensure comparison to other centers. As the ABA-NBR records included admissions for isolated inhalation injury, other skin diseases, and readmissions for complications or reconstruction, we used various filters (see File S2) to identify patients with new cutaneous burns, as recommended by Pavlovich [20].
We also created several subsets of the total ABA-NBR where all variables of interest were recorded, to maximize the number of records used for each comparison. For example, if a record contains race/ethnicity data but not smoke inhalation data, it can be used in race/ethnicity comparisons and ought not be removed from all analyses.
The Overall Set of the ABA-NBR included records for new burn admissions with known etiology, age, TBSA%, survival status and year of injury (File S3). There were too few records prior to 1995 to permit comparison of earlier periods, 1995

Fluid Resuscitation
For the fluid resuscitation analysis, a subset of the Harborview database was examined. Fluid requirements are substantially influenced by several variables including burn size, age, type of injury (e.g. electrical), presence of inhalation injury, etc. An extremely large sample size would be required to study first 24hour fluid resuscitation and associations with all of the important variables. Furthermore it is unclear if ''fluid creep'' [12] is present in resuscitation of uncomplicated injuries. Accordingly, we limited the fluid analysis to persons who survived to discharge with TBSA% from 20-60%; age $14 years; flame, scald or flash etiology; no inhalation injury and no other injuries. We asked three questions 1) was ''fluid creep'' present in successful resuscitation of uncomplicated injuries, and if so, 2) approximately what year did the practice begin, and finally, 3) what therapeutic events are associated with ''fluid creep''?'' There were 768 records that met these criteria from 1974 to 2006 and the list was submitted to Harborview Department of Health Information Management for retrieval. The timespan from 1974 to 2009 was divided into 5-year time periods and record retrieval continued until each 5-year time period contained at least fifteen complete records. Opioid equivalents were calculated as morphine 10 mg, hydromorphone 1.5 mg, fentanyl 100 mg, and methadone 5 mg.

Mortality Modeling
Many models to predict mortality for burn injury have been published over the years. Most include parameters not present in our historical database, such as pneumonia [21], percent fullthickness injury [22] and Abbreviated Burn Severity Index [23], and day one physiologic data FLAMES score [24]). In 2010 Colohan [25] reviewed previously described models of mortality and concluded that TBSA%, presence of inhalation injury, and age are the strongest predictors of mortality. The Belgian Outcome in Burn Injury Study Group studied 5,246 burn injured persons and published an elegant, simple method to predict burn mortality [26] based upon the method of Ryan [27] and the modified model was validated by Brusselaers [28]. However the definition of inhalation injury in the model is that mechanical ventilation was required. Our database includes a much broader definition of inhalation injury so the Belgian model could not be applied. The Osler [13] rBaux model fulfills the requirements of Colohan [25] and is applicable to our database.
Osler [13] published the rBaux Score using age, TBSA% and inhalation based upon 39,888 records from 2000-2007 in Version 4.0 of the ABA-NBR dataset. The rBaux score is calculated as age + TBSA% +17*(presence of inhalation injury). This means, for example, that an rBaux score of 117 might refer to a 50 year old person with 50% TBSA% and inhalation injury, a 25 year old person with 75% TBSA% and inhalation injury, a 50 year old person with 67% TBSA% and no inhalation injury, etc.

Predicted Mortality~e
Statistical Analysis As discussed above, the ABA-NBR is not a representative sample of the United States burn population. Neither is the Harborview data a representative sample of the larger population. Therefore, we considered both to be a population, for which we have complete data, not a sample of the United States population of burn patients. Therefore analyses, including changes over time, are done with descriptive, not inferential, statistics. Linear regression was used to assess trends in fluid administration, quantile regression to assess trends in length of stay, and logistic regression to assess risk factors for mortality. To further assess trends in mortality, we used a moving average of the observed/expected mortality ratio using a moving window of 2,000 patients, which corresponded to about two years of calendar time. Data analysis was accomplished with STATA version 11.1 (StataCorp, TX).

Results
Some of the results invite comment/explanation that does not warrant inclusion in the Discussion section; these are included here in Results.

Catchment Area
The majority of Harborview admissions were referred from WA State; other admissions came from Alaska, the panhandle of Idaho, and western Montana. Some persons from elsewhere sustained their injury in the Seattle region and were treated in the burn center. The catchment area did not change from 1974 to 2009. The zip code of residence of the 3,252 persons admitted in 2005-2009 is shown in Fig. 1.

Admissions
Upon analysis of the database, it became clear that there are two types of admissions to the Burn Center. Standard (Type 1) is the classic admission requiring complex inpatient care of any number of days. Short-stay (Type 2) is a short admission (defined as one or two days with survival) for initial wound care, pain control and education; the patient is then discharged to outpatient care.
For this analysis we defined Short-stay as one or two days determined after discharge, upon review of the data. This clearly means that some persons classified as Standard (Type 1) could have been Short-stay (Type 2) had they lived nearby the Harborview Burn Center or had sufficient family support. It also means that persons with ''simple'' injuries that require surgery might now be Short-stay (Type 2) whereas in the past they might have been Standard or Type 1. But neither of these variables is new and should not affect comparison of changes over time.
The concept of the two types of admissions is not new. Although the authors did not use these terms, Davey mentioned such a distribution over a decade ago [4] and more recently Anwar [29] and Ward [30]. The two types of admissions are very different, e.g. our overall mortality was 7.2% for standard admissions and zero for short-stay, median TBSA% was 8.0% for standard admissions and 2.5% for short-stay, and median rBaux score [13] was 42 for standard admissions and 25 for short-stay.
Nevertheless, studies do not typically differentiate between standard and short-stay admissions; an exception is the report by Onarheim [31]. This was unimportant in the 1970s and 1980s as the numbers of short-stay admissions were small. They did not become significant until the 1990s and have been increasing since then, now comprising greater than 50% of admissions to Harborview and ,40% of records in the Overall Set of the ABA-NBR. Combining the two types of admissions renders historical comparison to the 1970s and 1980s impossible and interpretation difficult. It seemed prudent to segregate the two types of admissions and exclude short-stay admissions from reports on mortality, length of stay, etc. This addresses the issue raised by Jeng [32], i.e. the need to filter the noise to see the signal. We will discuss the two types of admissions separately in the analyses that follow.
Standard admissions (Type 1). The number of standard admissions to Harborview was quite stable over the several decades ( Fig. 2, Panel A). However, the incidence rate of standard admissions peaked in the mid-1980s, declined to the approximately year 2000, and has been stable since then (Fig. 2, Panel B).
Standard admission records in the Overall Set of the ABA-NBR rose from 20,649 in 1995-1999 to 24,030 in 2000-2004 and to 33,677 in 2005-2009. Since the ABA-NBR is based on voluntary submission, these figures do not necessarily indicate an increasing incidence rate of burn injuries; more likely they indicate increasing submission.
Short-stay admissions (Type 2). On the contrary, the number of short-stay admissions to Harborview increased over time (Fig. 2, Panel A and D), as did the incidence rate (Fig. 2, Panel B). In 2005-2009 short-stay admissions made up 53% of admissions. The majority of short-stay admissions originated in eight counties surrounding Seattle (Fig. 2, Panel C) and could therefore be followed as outpatients, unlike those who traveled a great distance. Over the same period short-stay admissions to other WA State hospitals declined (Fig. 2, Panel D).
The proportion and absolute number of short-stay admissions has also risen in the Overall Set of the ABA-NBR. There were 7,242 in 1995-1999, 14,074 in 2000-2004 and 21,580 in 2005-2009. This constitutes 26%, 37% and 39% of records respectively in the Overall Set of the ABA-NBR.
Comment on admissions. It seems likely that the decline in incidence of standard admissions to Harborview up to approximately year 2000 was part of the national trend. HCUP data [18] reports ,50,000 hospital discharges for burn injury in 1994 declining to ,30,000 in1999 and remaining stable since then. That the incidence rate of standard admissions has been steady since the mid-1990s suggests no further improvement in the prevention of large injuries during the past 15 years. A similar pattern was found in Norway [31] in a study in which the investigators did attempt to limit the study to standard admissions. Å kerlund [33] reported a similar pattern in Sweden and Spinks [34] in Canada. These observations suggest that further reduction in serious burn injuries will require new prevention strategies.
In contrast to standard admissions, short-stay admissions have increased substantially at Harborview and in the ABA-NBR. One can only speculate on the reasons for the increasing numbers and incidence rate of short-stay admissions; they are probably many and complex. Anwar [29] and Ward [30] described similar changes in the UK. It is possible that the number of such injuries is increasing. More likely, it may be that less serious injuries are being referred to specialty centers in greater proportions, following the ''nationwide movement toward regionalized multidisciplinary care'' [35,36]. This hypothesis is supported by the decline in burn admissions to other WA State hospitals. Burn care has become quite complex and is an elective in many training programs. Therefore burn care may exceed the capability of many physicians and surgeons. In addition since Medicaid covers ,40% of those #15 years and 15-35% of those $16 years, it is possible that burn injuries are referred for financial considerations. Finally, the Harborview transfer center opened in the late 1990s and has increasingly facilitated referrals; this may have also played a role in the increased numbers of admissions.
Comment on TBSA%. TBSA% of standard admissions declined until the mid-to late-1980s but has been stable since the early 1990s. The number of admissions with large injuries has also been stable even though the incidence rate is declining. The incidence rate of standard admissions for children rose from ,5/100,000/year in the late 1970s to ,8 in 1990 and then declined to ,4 in the late 1990s and remained there. The incidence rate of standard admissions for persons .65 years was stable at 4-5 over the 36 years.
Patterns in the Overall Set of the ABA-NBR were quite similar including the decline in the proportion of standard admissions of children.
Short-stay admissions. The median age of short-stay admissions was stable over the entire 36-year period and in 2005-2009 was 21 (IQR 3-40), younger than standard admissions. Children accounted for ,40% of admissions since the mid-1980s, twice the percentage of standard admissions. The median age of children was 2 years, the same as for standard admissions, and 74% of the children were #5 years. Younger children were also at greater risk for injuries resulting in short-stay admissions than older children. Persons over 65 years made up ,2% of these admissions for the same period, fewer than for standard admissions.
Patterns in the Overall Set of the ABA-NBR were quite similar. The incidence rate for children declined from 1990 to 2000 but has changed little since. Bowman [37] found a similar pattern nationwide in the United States. The incidence rate of admission for persons .65 years was stable over the 36 years.

Sex
The sex distribution did not change over time for either standard or short-stay admissions. However, there was difference in the sex distribution by age group. Of those persons #5 and .65 years, ,60% were male; ,75% were male between 6 and 65 years. The same pattern exists in the Overall Set of the ABA-NBR.

Race/Ethnicity
Rivara [38], co-author of this manuscript, has pointed out that race and ethnicity should not be used as explanatory variables when the underlying constructs (e.g. education, income) can be measured directly. However, he also states that, when considering health disparities including demographics and outcomes, it is appropriate to use race and ethnicity, not as explanatory variables, but to examine the underlying sociocultural reasons for the disparities. It is in this sense that we include self-declared race and ethnicity.
The categorization of race/ethnicity changed in 2000 when the United States Census Bureau subdivided each race into Hispanic and non-Hispanic ethnicities and more recently when the Department of Health and Human Services adopted new standards for recording race/ethnicity [39]. The Harborview database and the ABA-NBR do not include this detail, rather they use the categories previously defined by the Census Bureau including White, Black, Pacific Islander, Asian, Native American, Hispanic and Multiracial. This introduces some discrepancy but the changes were not sufficiently large to alter conclusions.
For standard admissions to Harborview, the proportion of Whites and Blacks decreased over time whereas that of Hispanic, Asian, and non-White increased for both children and adults (Table 1). For standard admissions of children the incidence rate for Whites, Blacks and non-Whites declined whereas the rate rose for Hispanics and Asians ( Table 1). The incidence rate was higher for Black, Hispanic, Asian, and non-White children than White. For standard admissions of adults, the incidence rate rose for Hispanics but declined for the other groups ( Table 1). The incidence rate was higher for Blacks than the other groups.
The trends were similar for Harborview short-stay admissions. On the contrary, in the Overall Set of the ABA-NBR from 1995 to the present for both standard and short-stay admissions, the proportions remained stable. For standard admissions of children in the ABA-NBR the proportions are White ,40%, Black ,20%, Hispanic ,20%, and Asian ,5%. For standard admissions of Figure 3. TBSA% Over Time for Standard Admissions. The median TBSA% declined to the mid-1980s and subsequently varied from 7% to 10%. The 75 th percentile TBSA% also declined to the mid-1980s and then varied from 12% to 20%. The 25 th percentile was stable between 3% and 5%. doi:10.1371/journal.pone.0040086.g003 adults the proportions are White ,60%, Black ,20%, Hispanic ,10%, and Asian ,5%.
Comment on race/ethnicity. The race/ethnicity distribution is strikingly different from other centers/regions. Garner [8] reported 56% Hispanic and McGwin [40] ,30% Blacks. This regional difference could have implications for multi-center trials.
The incidence rate was higher for persons of color. Karr previously reported this discrepancy for Hispanic children [41] as did Rimmer [42]. There continues to be a disparity in the proportions and incidence rates for persons of color.  The proportion of Whites and Blacks decreased over time whereas that of Hispanic, Asian, and non-White increased for both children and adults. For children the incidence rate for Whites, Blacks and non-Whites declined whereas the rate rose for Hispanics and Asians. The incidence rate was higher for Black, Hispanic, Asian, and non-White children than White. For adults, the incidence rate rose for Hispanics but declined for the other groups. The incidence rate was higher for Blacks than the other groups. (The row totals do not equal 100% since some categories with small n were omitted from the table Payer Status on Admission (relevant only to USA readers) Children. The proportions of payer status on admission for standard and short-stay admissions were similar and shown overall in Table 2. Medicaid coverage rose to nearly 50% of admissions in the early 1990s, then decreased, and has now risen again to similar levels. The current distribution (2005-2009) is ,45% Medicaid, ,45% Insurance and ,5% uninsured.
The ABA-NBR proportions of payer status for standard and short-stay admissions of children followed a similar, but less striking, pattern. Medicaid coverage declined from 25% in 1995-1999 to 22% in 2000-2004, and then rose to 32% in 2005-2009. However, the incidence rate of children admitted to the Burn Center with Medicaid coverage was less than for those with other coverage for both standard (Type 1) and short-stay (Type 2) admissions (Fig. 5).
Adults. For adults with standard and short-stay admissions, Medicaid coverage increased to 40% in the early 1990s and then declined ( Table 2). Commercial insurance coverage followed the reverse pattern. The proportion of Labor & Industries coverage declined over time but has changed little since the early 1990s. The incidence rate of Medicaid coverage for standard admissions is greater than for non-Medicaid coverage (Fig. 5). There was little difference for short-stay admissions.
By zip code of residence. King County is the location of Seattle and the Burn Center. The proportion of WA State residents residing in King County and the proportion of WA State admissions from King County are shown in File S4. In summary, the percentage of WA State residents residing in King County decreased from 33% in 1974 to 29% in 2009 and the proportion of WA State admissions from King County declined from over 70% in the late 1970s to ,40% in 2009. The distribution of payer status on admission from King County did not differ substantially from other WA counties except perhaps for Medicaid coverage of short-stay admissions in the 1980s and early 1990s ( Table 3). The distribution of payer on admission for admissions from outside WA State generally included fewer persons with Medicaid coverage and more with commercial insurance (Table 4).
Comment on payer status on admission. There continues to be disparity in the income status of persons with burn injuries both in developed and developing countries. Mistri [43] demonstrated economic disparities in burn admissions in New Zealand. Ahuja [44] did likewise in India and van Niekerk [45] in Capetown and Park [46] and Edelman [47] in the United States.

Site of Event
The majority of injuries in children with both standard and short-stay injuries (81 and 90%) occurred in the home; this did not change over time. The majority of adults with standard and shortstay admissions were also injured at home although the proportion is less than in children, 53 and 71%. Again, there was no trend over time.
Considering persons age 16-65 from WA State, the proportion injured at work declined from 30% in 1975-1979 to 15% in 2005-2009; the incidence rate of work-related burn injuries in persons age 16-65 also declined from 1.8 in 1975-1979 to 1.3 in 1995-1999 and has been stable since then. When these injuries are segregated into type of injury, the absolute numbers changed little except for electrical injuries, which declined over time. The incidence rates of all declined but the majority of the decline occurred prior to year 2000 (Table 5). Mandelcorn [48] also found no reduction in work-related injuries during the past 10 years.
A similar decline in the proportion of work-related injuries exists in the ABA-NBR data. In the Overall Set for standard and shortstay injuries in adults, there were 27% work-related injuries in 1990-1994, 27% 1995-1999, 21% 2000-2004, and 17% 2005-2009.
Comment on site of event. The majority of injuries occur at home, unchanged over time. The absolute numbers of electrical injuries at Harborview declined until year 2000. The incidence rate of work-related burn injuries in persons age 16-65 in WA State declined until approximately year 2000 but has been stable since then suggesting no further improvement. Furthermore there was a decline in work-related injuries in the ABA-NBR.

Method of Transport
For standard admissions, the use of air transportation increased considerably until the early 1990s when it stabilized at ,35% of admissions; the median TBSA% transported declined to ,15% ( Table 6). The use of ground transport emergency medical services declined to ,50% of admissions and the TBSA% of those transported by ground declined to ,10% (Table 6). Self-transport also declined to ,10% (Table 6).
For short-stay admissions, air transport increased to ,10% of admissions and TBSA% transported was stable at ,5%. The use of ground ambulance transport increased from ,40% to ,55% of admissions for injuries of ,5%. Self-transport of short-stay injuries declined from ,60% to ,35% for injuries ,5%. Comment on method of transport. The use of air transport increased over time and is now utilized for 30-40% of standard admissions and ,10% of short-stay. Air transport is now used for relatively small injuries, 10-15% TBSA% for standard injuries and ,5% TBSA% for short-stay. It seems prudent to find alternatives to air transport for short-stay admissions, particularly since the cost of air transport may not be justified for the time savings involved in small burns reaching definitive care. We note that the Harborview catchment area is large and contains expanses of water and mountains so ground transport is not always efficient. This situation is not unique to Harborview as other burn centers also serve large catchment areas [49] and long distance air travel is safe [50].

Etiology
Flash and flame injuries may be similar in biology but have been segregated historically and were so in the Harborview database; therefore they were separated in this analysis.
For standard admissions (Type 1) of children, the current proportions are flame ,30%, scald ,50%, contact ,10%, grease ,5% and flash ,5%; there were no large changes over time. The incidence rates for children generally rose for all mechanisms until ,1990, declined until ,2000 and may have stabilized. The increase and decrease was most profound for scald burns (Fig. 6, Panel A).
For short-stay admissions of children, the current proportions are flame ,10%, scald ,50%, contact ,30%, grease ,5%, and flash ,5%, representing a shift from flame injuries in standard admissions to scald and contact in short-stay admissions.
For standard admissions (Type 1) of adults, there was an increase in the proportion of flame burns and decrease in the proportion of scald injuries; the current proportions are ,50% flame, ,10% scald, ,10% contact, ,10% grease, and ,15% flash. The incidence rates for adults of flame, contact, flash, grease, and electrical (including both work and not work-related) burns were largely stable over the years (Fig. 6, Panel B). The incidence rate of scald burns declined in the 1990s and has stabilized since then. Figure 5. Incidence Rate of Medicaid Coverage. Panel A -The incidence rate of standard (Type 1) admissions of children covered by Medicaid was similar to those covered by other payers. Panel B -The incidence rate of short-stay (Type 2) admissions of children covered by Medicaid was lower than for other payers but rose to similar levels. Panel C -The incidence rate of standard (Type 1) admissions of adults covered by Medicaid was higher than for those covered by other payers. Panel D -The incidence rate of short-stay (Type 2) admissions of adults covered by Medicaid was similar to those covered by other payers. (WA State data of Medicaid eligibility was not available prior to 2000. Incidence rates are small so are expressed in scientific notation. WA State segregates children into #18 so the age breakdown is slightly different from the remainder of the manuscript.). doi:10.1371/journal.pone.0040086.g005 For short-stay admissions of adults, the current proportions are flame ,30%, scald ,20%, contact ,10%, grease ,15% and flash ,20%, again a shift from flame to scald injuries compared to standard admissions.
In the Overall Set of the ABA-NBR, there was no decline in the proportion of scald injuries for standard (Type 1) admissions of adults.
Comment on etiology. In children, scald burns had the highest incidence rate overall but the incidence rate of scald, flame and contact all began to decline in the late 1980s or early 1990s. In adults, flame burns had the highest incidence rate and did not decline as did scald and flash. For both children and adults, the incidence rate of scald injuries declined after about 1990. This would support a conclusion that the tap water legislation of 1980-1984 was effective in preventing accidental scald injuries in adults as reported by Erdmann [51]. However this database does not   divide scald injuries into tap water and other types of scald injuries. Peck reviewed tap water injuries and found that they make up only 12-14% of scald injuries [52].

Body Part Injured
The involvement of any single body part decreased along with the declining TBSA%. However the distribution of body parts injured did not change over time. At present, for standard injuries for all ages the distribution is ,50% with involvement of the head and neck, arm, hand, and trunk, ,40% with involvement of the leg and ,20% for the foot.

Inhalation Injury
Inhalation injury was a clinical diagnosis made by the treating physicians based on various factors including history of enclosed space fire, estimated/measured carbon monoxide levels .10% in the fire, carbonaceous sputum, and in some instances, PaO2/ FiO 2 ,400. Not all findings were required to make the diagnosis. Fiberoptic bronchoscopy was rarely performed prior to 2008.
There was no trend in the proportion of admissions with smoke inhalation in conjunction with cutaneous injury for either children or adults, with a mean of 4% of the standard admissions for children and 10% for the adults over time. Similarly there was no For standard admissions, the use of airlift increased considerably until the mid-1990s and the median TBSA% transported decreased. The use of ground and selftransport declined, as did the TBSA% transported. The pattern was different for short-stay admissions where air and ground transport increased but self-transport declined. doi:10.1371/journal.pone.0040086.t006 trend in the incidence rate, with a mean of 0.2 for children and 0.5 for adults. There was likewise no trend over time (1995)(1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)) in the ABA-NBR but the proportions were slightly higher at 5% and 14%. Only 702/5978 (12%) of persons with burn injury of the head and neck also sustained inhalation injury as defined.

Burn Surgery
The percent of standard admissions (Type 1) treated surgically was ,50% over the entire time period (Fig. 7, Panel A). If treated surgically, the mean number of procedures was approximately two for all years, except the partial year 1974 (Fig. 7, Panel B). The total number of burn procedures rose when excision became standard procedure and then varied between 200 and 300 per year until 2009 when it surged to 400 (Fig. 7

Length of Stay for Standard (Type 1) Admissions
Unadjusted median length of stay at Harborview for survivors of standard admissions declined from 17 days in 1975-1979 to 11 in 1985-1989, was then stable for two decades, and then declined further to 9 days in 2005-2009 (Table 7). There was a similar decline in the length of stay in the ABA-NBR data from 11 days in 1995-1999 to 9 days in 2005-2009 (Table 7). To examine factors associated with length of stay, we used age groups #5, 6-15, 16-45, 46-65 and .65 years as modified from Moreau [53] and TBSA% groups #20%, 21-40%, 41-60% and .60% as modified from Galeiras [54]. This analysis was limited to those persons with standard admissions who survived to discharge (File S5).
Adjusting for other variables, age, TBSA%, and inhalation injury were associated with increased length of stay (p,0.001). Compared to patients 5 years and younger and adjusted for all other variables, the median length of stay was 4.2 days higher for patients aged 46-65 years (95%CI 3.3, 5.1) and 6.6 days higher for patients older than 65 (95%CI 5. Johnson [55] reported that one day per percent TBSA remains a good ''rule of thumb'' to evaluate burn care. At Harborview ( Table 8) the mean length of stay/TBSA% for children with standard admissions, TBSA% ,20 and residing in King Co. declined over time and was 1.7 in 2005-2009. For adults, the mean length of stay/TBSA% also declined over time and was 3.1 in 2005-2009. For children and adults with TBSA% $20 the mean length of stay/TBSA% was slightly greater than one throughout the 36 years. There was little difference between those residing in King County and those residing further away.
Using the Length of Stay Set of the ABA-NBR (Table 8)  Comment on length of stay. As length of stay declined for smaller injuries, it is possible that the number of readmissions for infection, skin grafting, and inpatient occupational and physical therapy increased. This detailed data is not available in either the Harborview database or the ABA-NBR. However, data is available in the Harborview database for the period 1994-2002. The number of readmissions ranged from 9-24 with no trend over time suggesting that the number of readmissions did not increase as length of stay decreased. Similarly it is not possible to know precisely from these databases exactly why length of stay declined, although excision of burn injuries and expanded outpatient care of many medical conditions are commonly proposed.
There was no decrease in length of stay/TBSA% for injuries with TBSA $20%. However for injuries ,20% the ratio declined from 1975 to 1989; this decline could be associated with the increased application of excision and grafting. It seems, however, that early excision did not decrease length of stay for the larger injuries and that length of stay has changed little since ,1990.
Johnson [55] concluded that ''anticipating patient LOS to be 1 day for every TBSA% is still a useful exercise''. We cannot actually compare data since the authors did not publish which ABA-NBR records were used nor did they exclude short-stay admissions, but these data suggest that that criterion may be too stringent and sets a goal that may not be realistic for standard admissions. Of the sixty-eight time points listed in Table 7, only two times was the value of one day/TBSA% achieved. A more realistic goal for standard admissions might be ,2 days/TBSA% for injuries $20 TBSA% and ,3 days/TBSA% for injuries ,20 TBSA%.
It is intriguing that the regression indicated that location of residence was associated with increased length of stay whereas the length of stay/TBSA% data did not. This may be because length of stay/TBSA% does not adjust for age, inhalation injury, location of residence, etc.

Disposition of Persons Surviving
The majority of persons were discharged to home, ranging from 91-100% for children and 82-98% for adults; there was no trend over time. Other discharge options used infrequently include AMA (Against Medical Advice), Other Home, Transitional, e.g. Nursing Home, Extended Care, Other Acute Care, Outside Rehabilitation, Institution, e.g. Jail, Alcohol or Drug Rehabilitation, Shelter, and Street.
Persons requiring inpatient rehabilitation were transferred to the rehabilitation services at Harborview (adults) or Seattle   Table 9). The percent of the first eight-hour fluids delivered prior to admission to the Burn Center increased from ,50% to ,80%. Fluids administered in the Burn Center increased from 3.2 to 4.3 cc/kg/TBSA%. Urine volume averaged over the first 24-hours did not change over time, ranging from 0.9 to 1.2 cc/kg/hr.
The unadjusted data (Table 10) suggest increased use of prophylactic tracheal intubation, placement of arterial lines and pulmonary artery catheters, and administration of opioids and decreasing use of central venous pressure lines and colloid may be related to fluids in excess of the Baxter formula during the first 24 hours. To examine the adjusted data, we conducted a regression examining opioid equivalents/24 hrs, benzodiazepine administration, tracheal intubation; central venous pressure and arterial lines; pulmonary arterial catheters; and colloid administration (File S6).
Comment on fluid resuscitation and ''fluid creep [12]''. ''Fluid creep'' refers to the increasing administration of fluids in excess of the Baxter formula over the decades. Fluid resuscitation in excess of the Baxter formula was reported in 1996 [56]. There appears to be consensus that supraBaxter fluid resuscitation is associated with complications [57]. But fifteen years later, the cause and prevention of ''fluid creep'' [12] are still unclear [58] and unresolved [59]. We therefore reviewed fluid therapy over the 36 years in a subset of persons in an attempt to answer the questions 1) was ''fluid creep'' present in successful resuscitation of uncomplicated injuries, and if so, 2) approximately what year did the phenomenon begin, and finally, 3) what therapeutic events are associated with ''fluid creep''?''.
Although a retrospective study cannot definitively answer these questions, it seems that the process is present in resuscitation of uncomplicated injuries, began in the late 1980s or early 1990s, and is not associated with increased urine volumes. Furthermore the data support the notion that the etiology is multifactorial including increasing pre-hospital fluids, pre-hospital prophylactic tracheal intubation, in-hospital fluids, opioid administration and perhaps (although not supported by the regression) increasing use of arterial lines and decreasing use of colloid. It seems that the   correction must also be multifactorial and searching for a single cause will be futile. Alvarado [60] and later Chung [61] have recommended a new formula, ''The Rule of 10'', but also mention that the old rules are often misapplied. It seems unlikely that simply swapping rules will solve the issue. Eastman [62] reported that ,40% of persons with tracheal intubation in the field were extubated in 1-2 days. Since tracheal intubation is associated with increased fluids, this is another issue that might be addressed. Sullivan [63] and Wibbenmeyer [64] also suggested that large doses of opioids might be involved. The extreme rise in opioid administration in 1998-2002 likely followed the events described by Peck [65]. Lawrence [66] uses a different definition of ''fluid creep'' than Pruitt [12]. Lawrence refers to the increasing fluids requirement of a single individual whereas Pruitt referred to the increasing requirement of populations over time. Perhaps the exact definition does not matter; Lawrence reported that colloid administration rapidly reduces hourly fluid requirements. Blumetti [67] reported that urine output is the important parameter. However in our data urine volumes were stable at approximately 1 cc/kg/hr as fluids administered increased. This suggests that urine output is too far downstream to monitor fluid input, at least after other events have occurred, i.e. excessive administration by first responders, prophylactic tracheal intubation, etc. With the other variables in play, it might be a ''deadly pitfall'' as suggested by Hartford [68].
Again, it appears that a multifactorial response will be required.
It is possible that with a one page, prospective data collection protocol and 10 centers with sufficient admissions, in one year's time the data would be sufficient and likely clear. In 1999, the U.S. Department of Veterans Affairs (VA) launched an initiative called ''Pain as a Fifth Vital Sign''. If pain is the 5 th vital sign, perhaps the first 24-hour fluids should be the 6 th vital sign in burn-injured persons. The data collection form might be similar to the document described by Chung [69] for use in the military. These forms might be required for verification and be stored in the ABA-NBR.

Hospital Mortality
Persons placed on comfort care (see the following section on Comfort Care) immediately upon arrival were included in mortality analyses because our goal is to study overall burn mortality, not treatment efficacy. Unadjusted case fatality at Harborview declined from ,12% in 1974-1979 to ,6% in 1985-1994, rose to ,8% in 1995-2004, and declined to ,6% in 2005-2009 (Table 11). Case fatality with the Overall Set of the ABA-NBR and standard admissions declined from 6.9% 1995-1999 to 6.0% 2005-2009 (Table 11).  The data suggest increased use of tracheal intubation, arterial lines, and pulmonary artery catheters and a decline in the administration of colloid. The administration of opioids steadily increased and then surged in the late 1990s. (n refers to the number of charts studied during the time period. Colloid refers to the number of persons to whom colloid was administered The odds of death also increased with female sex, OR 1.68 (95%CI 1.30, 2.16). There was no evidence that race/ethnicity was associated with mortality.
The previous findings merely confirm many other published observations. In this study examining change over time, adjusting for other variables and compared to 1975-1979, the odds of death were 17% lower in 1980-1984, 33% lower in 1985-1989, and 38% lower in 1990-1994. During the period 1995-1999, the odds of death were similar to that in 1975-79. The smallest odds of death were for the most recent time period, with OR = 0.39 (95%CI 0.25, 0.59) compared to the earliest time period (61% lower odds of death).
We also examined adjusted changes in mortality over time for standard admissions separately for children and the elderly. There Pham [71] also studied mortality in the elderly in the ABA-NBR and, although the authors did not segregate standard and shortstay admissions, found improved survival. With standard admissions and TBSA% limited to .60% (n = 2,368), unlike the Harborview data, in 2005-2009 there was improved survival compared to 1995-1999 with OR 0.5 (95%CI 0.4, 0.65).
Mortality generally declined over time in the Harborview data and in 2005-2009 matched rBaux prediction except for Group 2, which was less than predicted, and Group 5, which was slightly more than predicted (Table 12). Using the Mortality Model Set of ABA-NBR records, there was little change over time and mortality matched the rBaux prediction in all groups as expected since the ABA-NBR data from 2000-2007 was used to build the rBaux system. However the ABA-NBR data from 1995-1999 also matched, suggesting that rBaux scores are robust over time and that there has been little improvement in survival since ,1995.
Comparison to the rBaux model in Table 12 also adds detail to the regression mortality findings. There was improved survival in groups 1, 2, 3, 4, and 6 over time, with the less severe groups exhibiting an increase in mortality in 1995-1999. These results indicate that the improved survival was in the small to medium rBaux scores. Further clarity of the regression findings may be seen in a graph of mortality standardized to that predicted by the rBaux model (Fig. 8). To assess trends in the observed/expected ratio over time, we computed a moving average of the observed/expected ratio using a moving window of 2,000 patients, which corresponded to about two years of calendar time. The moving average was created by calculating the observed/expected ratio on a series of different subsets of the full dataset. In our case, we computed the ratio for the first 2,000 patients in the dataset, then for patient #2 and on to patient #2001, and so on. Observed mortality was about 1.5 times higher than rBaux predicted mortality prior to 1980, and was about 1.25 times rBaux predicted mortality from ,1980 to ,2000. Since ,2005 observed mortality has been less than rBaux predicted mortality. The adjusted odds of death was higher for all half-decades compared to 1975-1979 except for the most recent half decade.
To Regression of mortality over time with the rBaux score, Harborview data and standard admissions. After confirming the validity of the rBaux model, we reran the mortality regression using the rBaux score rather than the age, TBSA% and inhalation groups (File S9). The odds of death for the rBaux score were 1.09 (95%CI 1.082, 1.095). This, however, renders the association with age, TBSA% and inhalation injury individually not apparent. The regression confirmed the findings by time period. It differed in that for race/ethnicity the odds of death were 1.59 (95%CI 1.16, 2.17) suggesting that perhaps there is an association with race/ethnicity, as suggested by the ABA-NBR data.
Comment on mortality. Unadjusted case fatality declined to 5.7% in 1990-1994, relapsed in 1995-2004, and returned to 5.5% in 2005-2009, i.e. the largest improvement occurred two decades ago, much as reported by Danilla Enei [72], Muller [73] and Jaskille [74]. Nor did we find improvement in the ABA-NBR data since 1995. This suggests that burn care has achieved the floor of survival, as also noted by Blaisdell [75]. The regression revealed improved survival for the elderly as also reported by Macrino [76] Gomez [77] and Lionelli [78] although none of the authors clearly segregated standard and short-stay admissions. The association with gender is clear but the association with race/ethnicity is not. Current Harborview survival statistics match the rBaux model quite well as do ABA-NBR survival statistics for 1995-1999, suggesting that the model will be useful.
It is not possible to compare the three groups above to the ABA-NBR data, as the ABA-NBR data has no reliable field indicating comfort care. However, the case fatality rate of Group 3 was 0.056   ''Unprecedented Survival'' and ''Unsurvivable Burn Injury'' One of the parameters sometimes used in predicting whether or not resuscitation is futile is the concept of ''unprecedented survival'' or ''unsurvivable burn injury''. Both concepts are flawed because, even if ''unprecedented'', survival might be achieved and ''unsurvivable'' requires great foresight. Nevertheless, it is important to know what has been achieved when discussing resuscitation plans with injured persons and families.
The highest rBaux scores with survival achieved at Harborview are shown in Table 13. However, the ABA-NBR data with the Mortality Model Set of the ABA-NBR lists survivors with higher rBaux scores, suggesting that survival with higher scores is possible.
For this investigation we required that the record be in the Overall Set of the ABA-NBR, that the person be discharged from care in the burn center, and that numbers of operations and length of stay be reasonable, e.g. a record with length of stay of 479 days would not be considered a valid record. The highest rBaux scores we found are shown in Table 13.
It is difficult to determine rBaux scores in published manuscripts as the detailed information by patient is usually not reported. However Lumenta [80] reported a 47 year old with 80% TBSA% and inhalation injury who survived (rBaux = 144) and a 36 year old with 90% TBSA% and inhalation injury who survived (rBaux score = 143).

Comfort Care
The frequency with which comfort care was selected and the rBaux scores of the persons treated in this fashion are shown in Fig. 10. There was no trend over time in to the number of persons placed on comfort care (Fig. 10, Panel A) or the rBaux scores of those persons (Fig. 10, Panel B).
Comment on comfort care. In 2005-2009 there were 84 deaths in the Harborview Burn Center and 31 (37%) were placed on comfort care on arrival (Fig. 9). Hemington-Gorse [81] reported on 32 deaths at St. Andrew's Centre for Burns in Chelmsford, UK between 1/1/2008 and 12/31/2009. Eleven (34%) were place on comfort care on arrival. The percentages are similar, suggesting that the decisions for comfort care at Harborview may be similar to other centers.

Summarized & Highlighted Results
Since the Results section is extensive, we have summarized and highlighted the findings.
1. The number of standard admissions has been stable for two decades and the incidence rate declined in the 1990s but has since stabilized. The number of short-stay admissions increased markedly as did the incidence rate; in 2005-2009 short-stay admissions made up.50% of all admissions. The proportion of short-stay admissions increased in the ABA-NBR as well. 2. The number of admissions with TBSA .20% was stable over time; the incidence rate declined to 2005-2009. The median TBSA% and 75 th percentile declined to the mid-1980s but no further thereafter. The number of records in the ABA-NBR with TBSA .20% declined over time. 3. The median age increased due to increasing proportions of persons over 45 years of age. Increasing age was also evident in the ABA-NBR. 4. There was disparity in race/ethnicity and payer status. 5. Most injuries occurred at home. The incidence rate of injuries on the job declined to the mid-1990s but no further thereafter. This was also observed in the ABA-NBR. 6. The use of air transport increased, even for injuries of small TBSA%. 7. In children, scald remained the most common etiology and the incidence rate did not decline during the past decade. 8. The incidence rate of inhalation injury did not change over time and affected 4% of standard admissions of children and 10% of adults. Proportions in the ABA-NBR were slightly higher. 9. Fluids in excess of the Baxter formula did occur with uncomplicated resuscitations, was present in the early 1990s, was not associated with increased urine volume, and continued to [2005][2006][2007][2008][2009]

Use of the ABA-NBR
We found several manuscripts using the ABA-NBR data but none reported which records were studied and how they were selected. The National Burn Repository 2011 Report (Dataset Version 7.0) (http://www.ameriburn.org) likewise does not include the record numbers and the record selection method. This makes comparison very difficult. It would seem helpful for authors using ABA-NBR data to include a file containing the record numbers used in the study as done with File S3. Furthermore it is true that culling the ABA-NBR of irrelevant records is required as noted by Pavlovich [20]. Therefore the methods used to filter records should also be provided as in File S2.

Implications for the Organization and Delivery of Burn Care at Harborview
The current burn center model emerged in the 1940s and surged in the 1970s [82,83] to serve the needs of persons that could not be efficiently and properly managed in unspecialized acute care hospitals. It is fundamentally an inpatient model, an intensive care unit with an attached ward, operating room and clinic, staffed primarily by general surgeons; Heimbach described the attributes of the model [35]. For the purpose for which it was designed, the model has been and still is, a success [82].
However our analysis confirms that many changes have occurred at Harborview over the past three decades, some of which have been described in other burn centers. The total number of admissions increased; the number of large burns declined; the number of burn related operative procedures increased; the number of short-stay and other non-operative, non-intensive care unit admissions increased; the length of stay declined; and mortality declined. In addition to these burn changes, the wound care methods that have been found to be beneficial for burn injuries are often applicable to other types of wounds. Therefore as Kastenmeier [36], Imahara [84], and Pham [85] have reported, persons with various other wounds are being referred to and treated at Harborview and other burn centers, further increasing the numbers of persons needing care and making resource allocation even more complex. And finally the Federal requirements for chart documentation have grown enormously. Ten minutes per day per patient with fifteen patients is 2.5 hours/day.
As a result of these changes, the Burn Center is no longer a homogeneous intensive care unit with an attached burn operating room, ward and burn clinic treating injuries of large TBSA%. Rather it is a voluminous operating room, heterogeneous ward, and large clinic with an attached intensive care unit. The old Heimbach/Engrav model may still be required, but is no longer sufficient. In fact, Gamelli called for change in 2006 [86].
What can be done to adapt to the updated set of needs, to the shift from large injuries in the intensive care unit to smaller outpatient injuries? We proffer several ideas for consideration. First, we could simply expand everything, more surgeons, more nurses, more therapists, more outpatient facilities, etc. A second option is to selectively expand outpatient care (medical, nursing, therapy). Third, we could develop satellite burn centers in Montana, Idaho, Eastern Washington, and Alaska to care for the smaller, simpler injuries. Blaisdell [75] has already suggested that burn centers be identified with a ''level'' designation; these new satellites could be ''level 2''. A fourth option is to offload various Burn Center activities to ''burn team extenders'', ICU care to intensivists trained in the needs of burn injury, outpatient nonoperative care to other physicians trained in non-operative burn management, outpatient occupational and physical therapy to non-Burn Center therapists trained in the needs of burn injury. Finally, we could fully develop ''Smartphone Medicine'' to treat ''simple'' injuries over great distances.
Some systems have already been reported elsewhere to address similar needs and situations. Sagraves [87] described burn care provided for short-stay injuries by rural trauma surgeons and a dedicated burn nurse. Burn surgeons provide consultation via email and telephone. The authors reported clinical success with this system, which lessens the burden on the burn center and minimizes patient travel time and family lodging expenses. Blaisdell [75] described ''de facto'' regionalized care in Maine with no difference in mortality when compared to the ABA-NBR.

The Proposed Outcome Paradigm Shift
In 1992 Salisbury [88] and 1993 Warden [89] suggested it was time to consider outcomes other than mortality. In 1993 the National Institute on Disability and Rehabilitation Research began to fund burn rehabilitation research. In 2003 Shakespeare [90] called for a paradigm shift away from assessment of injury severity towards quality of outcome. Later Pereira [70] and Jaskille [74] reiterated the call.
The data from Harborview and the ABA-NBR confirms that mortality and length of stay have changed little since the mid-1980s, reinforcing the call for other outcome measures. But still, fifteen years after the first call for change, the Guidelines for the Operation of Burn Centers [91] do not speak to these matters in any detail. Richard and twenty-two other authors [92] described the situation in some detail but suggestions to remedy the situation were sketchy. There are many possibilities but several warrant discussion.
The ABA maintains the ABA-NBR, an acute burn database, and the National Institute on Disability and Rehabilitation Research (NIDRR) Burn Model Systems maintain a burn rehabilitation database, and both are decades old with large numbers of records. However the two are separate and distinct and it is difficult or impossible to study the contents of both simultaneously. It seems timely to mix/match/merge the two databases.
It would also seem timely to rewrite the Guidelines for the Operation of Burn Centers to recognize the shift from mortality to psychological, social, vocational, and functional outcome, even to include the core of current outcomes research, i.e. how people function and their experiences with their care (http://www.ahrq. gov/clinic/outfact.htm).
And finally, the present NIDRR Burn Model System funding ends 9/30/2011 and may or may not be renewed. If it is renewed, a call for comments will likely be published in the Federal Register. It will be important for the burn community to comment in order to shape the structure of the renewal.

Accuracy of Mortality Models
It is a fair question ''can mortality be predicted on admission?'' Pereira [70] stated ''clinical decisions …using variables obtained at admission are essentially inaccurate''. It is true that the longer one waits, the more accurate the prediction. However the rBaux model, based upon three variables known on admission, is quite accurate.

Supporting Information
File S1 Populations WA State. This file includes the populations of Washington State from 1974 to 2009 by age and race.

(XLS)
File S2 ABA-NBR Filters. The ABA-NBR includes many records not relevant to this study. This file includes the filters used to select relevant records.