Introduction

Traumatic brain injury (TBI) is a major source of morbidity and mortality, accounting for 2.8 million ED visits, hospitalizations, and deaths in 2013 [1]. Acute neurosurgical management of patients with severe TBI (sTBI), which is defined by a Glasgow Coma Scale (GCS) < 9, includes craniotomy for evacuation of hemorrhage producing mass effect or placement of an intracranial pressure (ICP) monitor. The latter is recommended for all patients with sTBI and an abnormal head computed tomography (CT) scan by the Brain Trauma Foundation (BTF) [2]. Increased ICP is associated with poor outcomes [3, 4] due to its propensity for causing secondary brain injury, mandating accurate diagnosis, and prompt treatment.

Although several studies exist to support the placement of ICP monitors in patients with sTBI, little data are available to inform the optimal timing of their placement. Thus, the BTF does not provide a recommendation of timing on ICP monitor placement. One study in pediatric patients did not identify a difference in outcome between early and late ICP monitor placement [5], but this has not been investigated in adults. The goals of this study were to (1) determine the effect of early compared to late placement of ICP monitors on outcomes in patients with sTBI and (2) to identify the patient and injury-specific factors associated with late ICP monitor placement. We hypothesized that early ICP monitor placement would be associated with improved outcomes by achieving more rapid identification of intracranial hypertension (IH) so that medical and surgical therapies for its treatment could be initiated sooner.

Methods

Description of the Dataset

The National Trauma Data Bank (NTDB) from 2013 through 2017 was used for this study. The NTDB is operated by the American College of Surgeons Trauma Quality Improvement Program and includes trauma registry data from over 700 facilities in the USA. It employs the National Trauma Data Standard, which defines the reporting of specific data elements. Data abstractors are trained to report data in standardized forms, and interrater reliability audits are performed to ensure the consistency of the data [6, 7].

Inclusion Criteria

Patients aged 18 years or older with a head abbreviated injury score (AIS) of 3, 4, or 5 were identified. Patients that were not treated at a hospital with ACS trauma verification level 1 or 2 were excluded, because these hospitals were not required to report data regarding the placement of ICP monitors. Those who were treated at an outside hospital before being transferred to the TQIP-participating facility were excluded. We subsequently identified those who experienced a severe TBI (GCS 3–8). Patients with GCS scores that were obtained from examinations limited by pharmacological sedation or neuromuscular blockade were excluded. We also excluded patients with a hospital length of stay (LOS) < 2 days because these patients were more likely to have suffered catastrophic, non-survivable injuries. Patients with advanced directives to withhold life-sustaining treatments were also removed.

Exposure Variables

The injury severity score (ISS) was calculated for each patient using AIS scores and included as a covariate to adjust for overall injury severity. Patients were considered to have an isolated intracranial injury if they did not have a non-head AIS ≥ 3. Hypotension was defined as a systolic blood pressure < 90 mmHg in the Emergency Department. Missing data were present for race, insurance, and injury mechanism variables; these patients were combined with those classified as “other” for each variable.

Outcome

The primary outcome of interest was in-hospital mortality. Secondary outcomes included non-routine discharge disposition, which was defined as any discharge other than home. Total LOS, LOS in the intensive care unit (ICU), and number of days mechanically ventilated were also evaluated after excluding patients who died during hospitalization, since these patients could have had artificially lowered LOS.

Statistical Analysis

Propensity score-matched cohorts were developed to compare outcomes in patients undergoing early or late ICP monitor placement using both 6 and 12 h from arrival to the hospital as thresholds. Logistic regression models were used to develop propensity scores for the groups with the following independent variables: age, sex, total GCS, ISS, and presence of hypotension. Patients receiving early or late ICP monitoring were matched using a 1:1 matching technique with a match tolerance of 0.001. McNemar’s test was used to compare the differences in the incidences of in-hospital mortality and non-routine discharge. Wilcoxon signed-rank test was used to compare total LOS, ICU LOS, and number of days ventilated between patients who underwent early or late ICP monitoring. We also evaluated the effect of time to ICP monitor placement as a continuous variable on each outcome. Binary logistic regression was used for dichotomous outcomes, and linear regression was used for continuous outcomes with adjustment for propensity scores. Finally, Cox regression was used to identify factors associated with early ICP monitor placement and craniotomy. Statistical analysis was performed with IBM SPSS Statistics 24.0 (IBM Corp., Armonk, NY).

Results

Description of the Cohort

A total of 807,372 patients with TBI of any severity who were admitted between 2013 and 2017 were identified in the NTDB. After applying the exclusion criteria, 5057 patients with sTBI were included in the study group. There were 2690 (53.2%) patients who had an EVD placed, 3027 (59.9%) patients who had an intraparenchymal monitor placed, and 660 (13.1%) patients who had both types of monitors placed. Baseline demographic and injury data are shown in Table 1. The median time to ICP monitor placement for the entire cohort was 3.73 h (IQR 2.18–8.67). The majority of patients had their ICP monitors placed within 6 or 12 h (66.2% and 81.5%, respectively) of hospital arrival.

Table 1 Baseline demographic and injury data stratified by time to intracranial pressure monitor placement
Full size table

In-hospital Mortality

There were a total of 1665 (32.9%) patients who died during hospitalization. The unadjusted in-hospital mortality rates were 34.0% in the within 6 h group compared to 30.7% in the > 6 h group (p = 0.018). Likewise, the unadjusted in-hospital mortality rates were 33.8% in the within 12 h group compared to 29.0% in the > 12 h group (p = 0.005). Propensity score matching generated 1617 matched pairs of patients undergoing ICP monitor placement within or greater than 6 h after arrival. For these two groups, the incidences of mortality were 33.6% and 30.4%, respectively (p = 0.049). When analyzing time to ICP monitor placement as a continuous variable and adjusting for propensity score, there was no association with time to placement and mortality (p = 0.776). We also created 908 propensity-matched pairs of patients undergoing ICP monitor placement within or greater than 12 h after arrival. The incidences of mortality in these two groups were 35.6% and 29.0%, respectively (p = 0.003).

Discharge Disposition

The majority of patients experienced a non-routine discharge (93.5%). In the propensity score-matched pairs, the incidence of non-routine disposition was 92.6% in the within 6 h group and 94.4% in the > 6 h group (p = 0.037). When comparing ICP monitor placement before or after 12 h, non-routine disposition was more common in the latter (93.9% compared to 96.0%, respectively; p = 0.042). In the binary logistic regression evaluating time to ICP monitor as a continuous variable and adjusting for propensity score, greater time was associated with non-routine discharge disposition (OR = 1.007, 95% CI 1.000–1.013). This approached but did not reach statistical significance (p = 0.058).

Length of Stay and Number of Days Ventilated

After removing patients who died during hospitalization, the median hospital LOS was 24.0 days (IQR 16.0–35.0), the median ICU LOS was 16.0 days (IQR 11.0–23.0), and the median number of days mechanically ventilated was 12.0 days (IQR 7.0–18.0). Propensity score matching generated 1124 matched pairs of patients undergoing ICP monitor placement within or greater than 6 h after arrival. As shown in Fig. 1a, hospital LOS, ICU LOS, and number of days mechanically ventilated were significantly greater in the late ICP monitoring group. Similar results were seen when using a 12-h cutoff for late ICP monitor placement. When analyzing time to ICP monitor placement as a continuous variable in linear regressions adjusting for propensity score, greater time to monitor placement was associated with longer hospital LOS (p < 0.001), ICU LOS (p < 0.001), and days mechanically ventilated (p = 0.042).

Fig. 1
figure 1

Mean duration of hospital length of stay (LOS), intensive care unit LOS, and mechanical ventilation among patients with early compared to late intracranial pressure monitor placement using a 6-h threshold (a) and a 12-h threshold (b). (ICP intracranial pressure, ICU intensive care unit, LOS length of stay)

Full size image

Predictors of Early ICP Monitor Placement and Craniotomy

In the Cox proportional hazards model, there was no association between age, sex, GCS, or ISS and ICP monitor placement within 6 h. As shown in Fig. 2, craniotomy (HR 1.097, 95% CI 1.037–1.160) and isolated intracranial injury (HR 1.128, 95% CI 1.055–1.207) were associated with early ICP monitor placement, while hypotension was negatively associated with early ICP monitor placement (HR 0.801, 95% CI 0.725–0.884). As shown in Fig. 3, a separate Cox regression indicated that early ICP monitor placement was associated with craniotomy when using both 6-h (HR 1.103, 95% CI 1.012–1.203) and 12-h thresholds (HR 1.121, 95% CI 1.010–1.244).

Fig. 2
figure 2

Kaplan–Meier curves demonstrating time to intracranial pressure monitor placement for patients who underwent a craniotomy (a), presented with hypotension (b), and suffered an isolated intracranial injury (c). (ICP intracranial pressure, TBI traumatic brain injury)

Full size image
Fig. 3
figure 3

Kaplan–Meier curves demonstrating time to craniotomy in patients who underwent early compared to late intracranial pressure monitor placement using a 6-h threshold (a) and a 12-h threshold (b). (ICP intracranial pressure)

Full size image

Discussion

Previous studies have demonstrated improved outcomes when ICP monitoring is performed in patients with sTBI [8,9,10,11]. IH leads to second brain injury by decreasing cerebral perfusion pressure, which limits blood flow and oxygenation to otherwise viable brain tissue. In severe cases, cerebral herniation occurs, which can rapidly become fatal. Appropriate management of ICP to avoid IH has been shown to decrease mortality rate and improve functional outcomes in patients with sTBI [12]. Therefore, placement of a device for measuring ICP is indicated. The 4th Edition of the BTF guidelines provided a level IIB recommendation that ICP should be monitored in all salvageable patients with sTBI. Although numerous studies have evaluated the outcomes associated with ICP monitoring, none have systematically evaluated the effect of timing of ICP monitor placement in adults. In the present study, we found that although there was an association between increased mortality and early ICP monitor placement, outcomes such as hospital LOS, ICU LOS, and days ventilated were significantly better when ICP monitors were placed earlier. We also found that hypotension was negatively associated with early ICP monitor placement. The results of this study suggest that ICP monitors should be placed within 6 h of presentation when possible, but prospective data are required to verify this.

Although there are limited data available to direct timing of ICP monitor placement, a considerable amount of literature has amassed that suggests ICP monitoring is associated with decreased mortality and improved outcomes [13, 14]. While most studies support the use of ICP monitoring in patients with sTBI, there are some that have found conflicting results. Chesnut et al. conducted a multicenter randomized controlled trial that investigated the effect of ICP monitor placement and maintaining an ICP of 20 mmHg or less compared to imaging and clinical examination-based management. The former was not associated with improved mortality, functional outcome, or LOS [15]. Similarly, Cremer et al. found no difference in functional outcome or mortality between patients who underwent ICP monitoring and those who did not [16]. This study evaluated the intervention retrospectively by comparing two level I trauma centers that differed according to utilization of ICP monitors. Differences in staff education, a comprehensive ICU team, and hospital funding could have confounded the outcomes, however [16]. In contrast to the prior studies, Shafi et al. found that ICP monitoring was associated with a 45% reduction in survival when controlling for AIS and ISS [17]. To definitively establish the effects of ICP monitoring on mortality and functional outcomes, large multicenter randomized controlled trials are required. It is unlikely that this will be realized, since there is a perceived lack of equipoise when withholding ICP monitoring from patients with sTBI. The current study suggests that ICP monitoring should be initiated rapidly in order to optimize outcomes. Thus, time to ICP monitor placement may be an important parameter to control for in future studies investigating its effects.

Duration of IH has been associated with poorer outcomes [18], and early craniectomy in patients with sTBI and refractory IH has been associated with improved long-term functional outcomes [19]. Although one may expect that similarly early ICP monitor placement would improve outcomes by allowing more rapid detection of IH and initiation of goal-directed therapy, there is little empiric evidence to support this. In theory, early coagulopathy, which is prevalent in patients with TBI, could lead to hemorrhagic complications relating to placement of the intracranial device. Additionally, it is possible that patients could benefit from an initial period of resuscitation and stabilization before exposing them to the stress of an invasive intracranial procedure, which often requires increased sedation and analgesia as well as a post-procedure head CT. Previously, Balakrishnan et al. investigated timing of ICP monitor placement in pediatric patients using a national pediatric ICU database [5]. Similar to our study, they defined early ICP monitor placement as occurring within 6 h of admission. The majority of the patients in the early ICP monitoring group had significantly worse injury and illness severity, but even when controlling for these factors early ICP monitoring placement was significantly associated with mortality. Although they did not find an association between early ICP monitoring and decreased ICU LOS like we did, there was an association with fewer number of days mechanically ventilated [5]. Similar to our results, Davidson et al. found that early ICP monitor placement was associated with reduced hospital and ICU LOS in pediatric patients [20]. However, they did not identify a difference in mortality. Their study differed from ours in that they used a 4-h threshold to define early ICP monitor placement. It is difficult to compare our results with the pediatric literature given that children have greater neuroplasticity and physiologic reserve to survive sTBI and experience functional recovery. This may explain why the two aforementioned studies did not identify an association between timing of ICP monitor placement and mortality.

The association between early ICP monitoring and increased mortality, although only marginally statistically significant, was unexpected. While we attempted to homogenize the early and late groups with propensity score matching, the retrospective nature of the study prevented us from accounting for all confounders. It is possible that patients with more severe injuries and poorer neurologic statuses underwent ICP monitoring sooner due to the perceived likelihood of IH. This is supported by the fact that patients with early ICP monitor placement underwent surgical intervention earlier, which could have been due to the presence of a large hematoma. However, the secondary outcomes all favored early ICP monitor placement, which would suggest that there is benefit to this approach. Discharge disposition functioned as a surrogate for functional outcome, which is likely to benefit from more rapid identification of increased ICP. The threshold of 6 h to define early ICP monitor placement was arbitrary but has been used before [5]. Other authors have used 4 h as a threshold [20]. Therefore, we included analyses that included time as a continuous measure. When evaluating time to ICP monitor placement continuously, there was no correlation with mortality or discharge disposition. Such an association would be expected if a true time-dependent effect on outcome existed. Conversely, there was a continuous time-dependent effect of ICP monitor placement on hospital LOS, ICU LOS, and number of days mechanically ventilated.

Patients with isolated intracranial injuries were more likely to undergo early ICP monitor placement, which is not surprising given that some with multisystem trauma could have been too hemodynamically unstable or required emergent surgical intervention for a different injury. Indeed, patients with hypotension were less likely to undergo early ICP monitor placement. Craniotomy was associated with early ICP monitor placement, which may be related to the common practice of placing one in the operating room following the surgery if there is concern for persistent IH.

Our study was limited by several factors, many of which are inherent to its retrospective nature and use of a database. The ICU care of patients could have varied considerably based on whether or not a dedicated neurological ICU or neurointensivist was present. The volume of the treating hospital also could have impacted patients’ outcomes, which we were unable to control for. Detailed descriptions of each patient’s hospital course were not available, so it was impossible to determine why patients underwent ICP monitoring early or late. We also did not have access to ICP measurements or data regarding the application of medical interventions for IH, which would have been instructive in understanding how information from ICP monitoring was utilized. Since it is not the ICP monitoring placement itself that improves outcomes, but rather the proper management of data obtained from it, this is relevant. As mentioned previously, there could have been a tendency toward placing ICP monitors earlier in more severely injured patients. Although we attempted to address this by creating propensity scores that accounted for GCS and ISS, the study’s retrospective nature prevented us from accounting for all confounding factors. Finally, a validated functional outcome measure such as the Glasgow Outcome Scale at 30- or 90-day follow-up would have been a more reliable determinant of functional status than discharge disposition, but these data were not available.

While the utility of ICP monitor placement has been studied extensively, little information exists regarding its optimal timing. Our results suggest that placing ICP monitors within 6 h of patient presentation leads to shorter hospital and ICU LOS as well as more routine discharges. These findings are highly generalizable, given that data were obtained from trauma centers across the country. However, only ACS level 1 and 2 centers were included, so extrapolation to facilities with other levels of verification may not be valid. Our results merit further investigation in a multicenter prospective study.

Conclusion

Despite an association between mortality and early ICP monitor placement, early ICP monitor placement yielded more routine discharges, fewer days mechanically ventilated, and shorter ICU as well as total hospital LOS. The favorable outcomes may be due to earlier identification and treatment of IH.