An exploratory analysis of transfer times in a rural trauma system.

BACKGROUND: Delays to definitive care are of particular concern in rural trauma systems, where prehospital times are significantly longer than average. AIMS: We evaluated for differences between transferring hospitals in the total time required to transport patients to definitive care, and analyzed for associations between transport times and outcomes. SETTINGS AND
DESIGN: We employed a cross-sectional design to analyze Level One Trauma Center registry data on interfacility transfer of 3,303 acute trauma patients.
MATERIALS AND METHODS: We calculated time in minutes from injury to definitive care (total elapsed time (TET)), and analyzed for associations between TET and both mortality and length of hospital stay at our center. We mapped hospitals and catchment areas to illustrate statistics by transferring hospital. STATISTICAL ANALYSIS: We employed analysis of covariance (ANCOVA) to analyze for the effect of TET and injury severity category upon hospital length of stay, and for the effects of TET and air transport as compared to ground transport. We evaluated for likelihood of in-hospital mortality using logistic regression.
RESULTS: TET had little or no effect upon length of hospital stay or in-hospital mortality. The effect of injury severity upon both length of stay and mortality was progressively greater with each categorical increase in severity. Air transport as compared to ground transport was associated with mild increases in length of stay and likelihood of mortality. Mapping revealed spatial patterns that were not evident by statistical analysis alone.
CONCLUSIONS: Mapping of geographic variations holds promise as a supplement to quantitative needs assessments of trauma systems in rural regions and developing countries.

INTRODUCTION

For patients with acute traumatic injuries in rural and developing regions, timely transfer to definitive care is likely to be a critical predictor of outcomes. Our Level One Trauma Center serves a largely rural region. Little is known about how well this system is serving the region's trauma patients, and whether limited healthcare resources are currently being allocated to the best advantage for population health. To provide insights for improvements in performance and resource allocation for the care of acute trauma in our region, we conducted an evaluation of referring hospitals and their associated catchment area populations.

Background

Rural trauma systems

The care of patients with acute traumatic injuries in rural and developing regions presents unique challenges.[12] In the US, designation of certain rural hospitals as Critical Access Hospitals (CAH) is intended to improve access to emergency and trauma care in rural areas.[3] Although CAHs face challenges imposed by inherent limitations in resources and personnel,[4] most patients injured in rural areas can be successfully treated in critical access and rural hospitals, and only 5-10% of such traumas require transport to a trauma center, but appropriate triage is critical.[5] In effectively regionalized trauma systems, patients with severe and critical injuries receive care in tertiary care hospitals.[6] For such patients, timely transfer to definitive care is likely to be a critical predictor of outcomes.[7] In the care of acute trauma, timeliness is essential to effectiveness, and excessive transport times are characteristic of immature trauma systems.[8] Delays to definitive care are of particular concern in rural trauma systems, where prehospital times are significantly longer than average;[9] and total elapsed time to definitive care is often measured in hours rather than minutes.[10] However, a recent study of patients transferred from rural nontrauma centers to a university hospital found that despite the development of a statewide trauma system and outreach education, transfer times failed to improve significantly over a 6-year period.[1]

Time from injury to definitive care in a rural setting

Our medical center is the only certified level one trauma center in state of New Hampshire (NH), USA and our center also serves a large proportion of the adjacent state of Vermont (VT) [Figure 1]. Located in the northeastern United States, both states are largely rural in character, with a combined land area of 49,000 km2 and combined population of approximately 2 million persons. Numerous hospitals in this rural bi-state region routinely transfer acutely injured patients to our trauma center. There are variations in the size of the geographic areas served by these facilities, in the distance between transferring facilities and our center, and in the mix of transport modes (helicopter or ground ambulance) utilized. Consequently, total time elapsed from injury to definitive care varies by transferring facility.

Figure 1

Selected catchment areas in Vermont and New Hampshire. Divisions within the two states represent town boundaries. The colored areas illustrate the geographic definition of four selected catchment areas and their associated hospitals (blue and white H). The green and white H symbol represents our level one trauma center.

Evaluation of elapsed time and outcomes

We evaluated for differences in elapsed time and outcomes for acute trauma patients. We analyzed for associations between elapsed times and outcomes of mortality and length of hospital stay. Geographic analysis is a useful tool for the evaluation of regionalized trauma systems,[11] and we employed geographic analysis to illustrate the times and relative distances involved in the transfer of trauma patients to our level one trauma center. Geographic analysis of rural trauma systems may inform efforts at performance improvement. For example, if poorer outcomes are observed in patients transferred from hospitals in outlying areas with long transport times, allocation of resources to shorten transport times or upgrade capacities of transferring hospitals may be indicated. Alternatively, if poorer outcomes are tied to nearby hospitals with shorter transport times, exploration of facility-specific performance issues may be indicated. Development of trauma systems that deliver timely care in rural areas of industrialized countries may also inform the development of trauma systems in developing nations, where resources are likely to be limited,[12] and patient transport times in excess.[13]

MATERIALS AND METHODS

The authors’ institutional review board approved the study aims, design, and methods. We employed a cross-sectional design. Our trauma registry contains over 18,000 records of acute traumas transferred to and/or admitted to our center. We queried the trauma registry for data related to interfacility transfers of acute trauma patients that occurred over a 5-year period. We restricted the analysis to 26 referring hospitals in VT and NH with valid records of 10 or more transfers. The resultant analytic dataset included 3,303 observations.

We described subject characteristics, mapped selected characteristics by transferring hospital, and analyzed for associations between total elapsed time (TET) and patient outcomes. TET was calculated as the total time in minutes from time of injury to time of arrival at our center. The main outcome measures were in-hospital mortality and length of hospital stay. Mortality is an established and unambiguous measure of trauma outcomes. Length of hospital stay has been also been widely used as a proxy for overall outcomes in hospitalized patients.[14]

We identified outliers in TET, which were probably due to miscoding. To handle these outliers we removed any value for TET greater than 1,440 min (1 day), which resulted in removal of 44 patient records from the denominator population used for analysis of TET. Also, patient records with missing ISS grouping (96 subjects) or transportation mode (345 subjects) were excluded from the individual prediction models.

We used analysis of covariance (ANCOVA) to analyze for the effect of TET and injury severity category upon hospital length of stay. Injury severity scores (ISS) were categorized as mild (1-8), moderate (9-15), severe (16-24), and critical (>24). We analyzed for the effects of moderate, severe, and critical injury; with mild injury as the reference category. We also used ANCOVA to analyze for the effects of TET, and air transport as compared to ground transport. Likelihood of in-hospital mortality was examined using a logistic regression model that included the following exposure covariates: TET; air transport as compared to ground transport; and moderate, severe, and critical injury severity as compared to mild injury severity. All statistical analyses were performed using STATA 11.0 (StataCorp, College Station, TX).

To enhance our understanding of the relationship between outcomes and the geography of this rural trauma system, we mapped the hospitals as points and the catchment areas of transferring hospitals as polygonal areas. Figure 1 illustrates the boundaries of four selected catchment areas and the corresponding transferring hospitals. We defined the catchment area for each transferring hospital as the aggregation of every town that was the site of injury for at least one trauma transfer. The analytic dataset was merged (by transferring facility) to the attributes table associated with a digital geopolitical map of VT and NH towns. The merging of data to the digital map effectively joined the geographic location of each hospital and its catchment area to the associated trauma registry data. We then developed maps that depicted selected mean statistics for each transferring hospital. Mapping was performed using ArcGIS 9.3.1 (ESRI Inc. Redlands, CA).

RESULTS

The number of patients transferred per referring hospital ranged from 12 to 135. Mean ISS was 12.8 (standard deviation (SD) = 9.1, range = 1-75). Injuries were categorized as mild in 29% of patients, moderate in 35%, severe in 24%, and critical in 12%. Mean length of stay at our trauma center was 6.4 days (SD = 8.6, range = 1-112), and overall in-hospital mortality was 4.96%. Average TET was 346 minutes (5 h 46 m) (SD = 193.0, range = 53-1,394). Table 1 displays TET by ISS category, as well as mean ISS by severity category, mode of transport, and among those who survived and died.

Table 1

Injury severity score and total elapsed time

Overall, TET had little or no effect upon length of hospital stay [Table 2]. Moderate, severe, and critical injury severity; as compared with mild severity; were all associated with increased length of hospital stay. The effect of injury severity upon length of stay was progressively greater with each categorical increase in severity. R2 for the model was 0.1393.

Table 2

Effect of total elapsed time and injury severity upon hospital length of stay

Our analysis for the effects of TET and air transport as compared with ground transport also revealed that overall TET had little or no effect upon hospital stay, but air transport was associated with increased length of stay as compared with ground transport [Table 3]. The value of R2 for the model was 0.0291.

Table 3

Effect of total elapsed time and transport mode upon hospital length of stay

Finally, analysis for likelihood of in-hospital mortality showed that TET had no effect upon the likelihood of patient death in the hospital [Table 4]. Air transport as compared to ground transport was associated with mildly increased likelihood of mortality, and categorical injury severity predicted increased likelihood of mortality. The regression model had a c2 of 113.5 and a pseudo R2 of 0.18.

Table 4

Likelihood of mortality by TET, mode of transport, and ISS

Mapping revealed spatial patterns that were not evident by statistical analysis alone. In this trauma system region, northern and western areas are generally more rural, and urban centers are concentrated in the southeast. Despite the fact that elapsed minutes from time of injury to ambulance arrival at scene is likely to be significantly longer in rural areas, TET appeared to be unrelated to rurality or distance between transferring hospital and level one trauma center [Figure 2a]. Many of the transferring hospitals located far from the level one trauma center had relatively shorter mean transport times, and not all of the referring hospitals with long mean transport times were located far from the destination hospital. Mean ISS appeared to be higher among transferring hospitals in the southeast, which is the most highly populated part of the trauma system region [Figure 2b].

Figure 2

(a) Total elapsed time and (b) injury severity score: Divisions within the two states represent town boundaries. Colored areas represent towns where injuries occurred. Circles represent transferring hospitals. The green and white H symbol represents our level one trauma center

DISCUSSION

Average time from injury to definitive care is very lengthy in this setting: Even among the most critically injured patients mean transfer times averaged well over 3 h. However, despite lengthy delays in delivering even the most severely injured patients to a setting with the capacity for delivering definitive trauma care we found no significant association between TET and subsequent length of hospital stay or in-hospital mortality. However, our use of data mapping allowed us to immediately visualize geographic variations in TET and ISS that were not readily apparent through statistical analysis alone. TET appears to be unrelated to distance between transport origin and destination. However, it is likely that ambulance and referring facility personnel spend more time caring for severely injured patients than mildly injured patients, thus prolonging time to definitive care. TET is therefore unlikely to be a function of distance alone, and any effect of TET may be expected to be confounded by injury severity. Delays in transport may be attributable to variation by referring facility in the delivery of acute trauma care.

Limitations

This study was subject to several limitations that should be acknowledged. The trauma program at our center employs numerous routine measures intended to assure the accuracy of trauma registry data; however, the data is subject to the possibility of human error. Also, data reported from transferring hospitals may vary in quality. This analysis does not account for trauma patients who may have been injured within the described catchment areas but transferred for definitive care to other hospitals. Transfer of trauma patients to other hospitals was likely higher in the extreme northwest, east, and southeast areas of the region. Relative homogeneity with regard to population demographics in the bi-state region that was the subject of this study lends internal validity to this study, but the results may not be generalizable to other regional trauma systems. Similar study methods may have some applicability to the study of trauma systems in rural and developing regions, but high variability among the multiple components of TET appears to limit the usefulness of TET as a measure for evaluation of trauma systems. The visual information provided by mapping of quantitative data may ultimately prove more valuable.

CONCLUSIONS

Total elapsed time to definitive care was associated with neither length of hospital stay nor in-hospital mortality. Furthermore, the visual information provided by mapping suggests that total elapsed time to definitive care may be unrelated to rurality or distance of transferring hospital to level one trauma center. Geographic analysis may reveal spatial patterns that are not appreciable through statistical analysis, and may prove useful as a supplement to quantitative assessments of trauma systems of rural regions and developing countries.