- Email: [email protected]
Effect of implementation of a paediatric neurocritical care programme on outcomes after severe traumatic brain injury: a retrospective cohort study
Articles
Effect of implementation of a paediatric neurocritical care programme on outcomes after severe traumatic brain injury: a retrospective cohort study Jose A Pineda*, Jeffrey R Leonard*, Ioanna G Mazotas, Michael Noetzel, David D Limbrick, Martin S Keller, Jeff Gill, Allan Doctor
Summary Background Outcomes after traumatic brain injury are worsened by secondary insults; modern intensive-care units address such challenges through use of best-practice pathways. Organisation of intensive-care units has an important role in pathway effectiveness. We aimed to assess the effect of a paediatric neurocritical care programme (PNCP) on outcomes for children with severe traumatic brain injury. Methods We undertook a retrospective cohort study of 123 paediatric patients with severe traumatic brain injury (Glasgow coma scale scores ≤8, without gunshot or abusive head trauma, cardiac arrest, or Glasgow coma scale scores of 3 with fixed and dilated pupils) admitted to the paediatric intensive-care unit of the St Louis Children’s Hospital (St Louis, MO, USA) between July 15, 1999, and Jan 15, 2012. The primary outcome was rate of categorised hospital discharge disposition before and after implementation of a PNCP on Sept 17, 2005. We developed an ordered probit statistical model to assess adjusted outcome as a function of initial injury severity. We assessed care-team behaviour by comparing timing of invasive neuromonitoring and scored intensity of therapies targeting intracranial hypertension. Findings Characteristics of treated patients (aged 3–219 months) were much the same between treatment periods. Before PNCP implementation, 33 (52%) of 63 patients had unfavourable disposition at hospital discharge (death or admission to an inpatient facility) and 30 (48%) had a favourable disposition (home with or without treatment); after PNCP implementation, 20 (33%) of 60 patients had unfavourable disposition and 40 (67%) had favourable disposition (p=0·01). Seven (11%) patients died before PNCP implementation compared with two (3%) deaths after implementation. The probit model indicated that outcome improved across the spectrum of Glasgow coma scale scores after resuscitation (p=0·02); this improvement progressed with increasing injury severity. Kaplan-Meier analysis suggested that neuromonitoring was started earlier and maintained longer after implementation of the PNCP (p=0·03). Therapeutic intensity scores were increased for the first 3 days of treatment after PNCP implementation (p=0·0298 for day 1, p=0·0292 for day 2, and p=0·0471 for day 3). The probit model suggested that increasing age (p=0·03), paediatric risk of mortality III scores (p=0·0003), and injury severity scores (p=0·02) were reliably associated with increased probability of unfavourable outcomes whereas white race (p=0·01), use of intracranial pressure monitoring (p=0·001), and increasing Glasgow coma scale scores after resuscitation (p=0·04) were associated with increased probability of favourable outcomes. Interpretation Outcomes for children with traumatic brain injury can be improved by altering the care system in a way that stably implements a cooperative programme of accepted best practice.
Lancet Neurol 2013; 12: 45–52 Published Online November 28, 2012 http://dx.doi.org/10.1016/ S1474-4422(12)70269-7 See Comment page 26 *Contributed equally Department of Pediatrics, Division of Critical Care Medicine (J A Pineda MD, A Doctor MD); Department of Neurology (J A Pineda, Prof M Noetzel MD), Department of Surgery (I G Mazotas MD, M S Keller MD, Prof J Gill PhD), Department of Neurosurgery (J R Leonard MD, D D Limbrick MD), Department of Pediatrics (D D Limbrick, J R Leonard), Department of Political Science (J Gill), Department of Biostatistics (J Gill), and Department of Biochemistry and Molecular Biophysics (A Doctor), Washington University School of Medicine, St Louis, MO, USA Correspondence to: Dr Jose A Pineda, Washington University School of Medicine, Campus Box 8116, 1 Children’s Place, St Louis, MO 63110, USA [email protected]
Funding St Louis Children’s Hospital and the Sean Glanvill Foundations.
Introduction Traumatic brain injury is a leading cause of death and disability in children.1 Outcomes are affected by the extent of the primary injury (original brain trauma) and the degree of secondary injuries, including maladaptive signalling caused by the primary trauma, perturbed cellular energetics, cerebral ischaemia (eg, perfusion deficits associated with dysregulated blood flow, intracranial hypertension or hypotension, and bleeding), cerebral hypoxia (eg, diminished blood oxygen content from respiratory failure or anaemia), and increased cerebral oxygen demand associated with seizures or fever.2 Because no treatments specific to primary brain trauma are available, the key goal for care is prevention of secondary insults. Achievement of this goal will require interpretation of rapidly changing, complex www.thelancet.com/neurology Vol 12 January 2013
information from several organ systems, coordinated decision making between services, time-sensitive and goal-directed interventions, and an iterative repeat of this cycle. We developed a paediatric neurocritical care programme (PNCP) to implement this approach and improve penetration and consistent retention of best practice into routine care. In paediatric health-care centres, specialised teams guided by evidence-based plans of care have improved outcomes for patients with severe trauma3 or congenital heart disease. Furthermore, implementation of rapid response teams has reduced mortality. Children with traumatic brain injury are especially likely to benefit from this approach to care because of their high susceptibility to secondary injury, the nature of complex multisystem pathological changes with narrow 45
Articles
See Online for appendix
therapeutic windows, and the robust potential for favourable functional recovery in young patients. Although management guidelines for children with traumatic brain injury have been created,4 effective implementation is challenged by weak evidence, cultural barriers to change, and inefficiencies inherent to complex, time-sensitive decision making across subspecialties. These factors all lead to inconsistent implementation and variability in care and outcomes.5 Here, we report the effect of a neurocritical care programme that was designed to implement a system of cross-specialty communication and an explicit plan of severity-based monitoring and intervention for children who have severe traumatic brain injury. We aimed to assess outcomes for patients and also the behaviour of the care team by comparing the timing of invasive neuromonitoring and intensity of therapies targeting intracranial hypertension and programme performance.
Methods Study design and participants In this cohort study, we screened trauma patients who were admitted to the St Louis Children’s Hospital (St Louis, MO, USA) emergency department between July 15, 1999, and Jan 15, 2012, with an initial Glasgow coma scale (GCS) score of 8 or less. Patients were screened for severe traumatic brain injury (GCS score ≤8 in the emergency department after resuscitation)6 and survival to paediatric intensive-care unit (PICU) admission. We excluded children with GCS scores of 597 patients with trauma admitted to emergency department with GCS score ≤8
316 without TBI or not admitted to PICU
281 met inclusion criteria
157 excluded 80 abusive head trauma 27 GCS 3 and fixed and dilated pupils in emergency department 28 cardiac arrest before admission 22 gunshot wound to the head
124 analysed
1 missing data from the medical record 63 before PNCP implementation
60 after PNCP implementation
Figure 1: Study profile GCS=Glasgow coma scale. TBI=traumatic brain injury. PICU=paediatric intensive-care unit. AHT=abusive head trauma. PNCP=paediatric neurocritical care programme.
46
3 with bilateral fixed and dilated pupils on admission to the emergency department, cardiac arrest before admission to the PICU, abusive head trauma, and gunshot wounds to the head.7,8 Research procedures were approved and exempted from informed consent by the Washington University institutional review board (St Louis, MO, USA).
Procedures The PNCP was first implemented on Sept 17, 2005. Using the 2003 Brain Trauma Foundation guidelines,4 we developed a time-sensitive, severity-based approach to monitoring and treatment of children with traumatic brain injury (appendix). This approach coordinated communication and activity in PICU staff and physician faculty and trainees in several disciplines (critical care, neurosurgery, surgery, anaesthesia, and radiology) and was implemented through a detailed training programme, an explicit process for maintenance of pathway fidelity, and continuous quality improvement. We examined raw outcomes for differences between study periods (ie, before implementation of the PNCP and after implementation of the PNCP). We defined favourable PNCP effects as reductions in population mortality and in the number of children with discharge status needing ongoing inpatient care, and also in terms of increases in the number of children with statuses that allowed discharge to home with or without outpatient treatment. Subsequently, we developed an ordered probit model to assess relations between patients’ characteristics, PNCP implementation, and disposition categories. Our primary outcome was categorised discharge disposition. Our secondary outcome analysis included examination of the relation between initial GCS scores after resuscitation and the Glasgow outcome scale (GOS) score9 at hospital discharge, and length of stay in the PICU. To assess the effect of the PNCP on outcomes, we analysed two scalable outcome variables: discharge disposition and GOS score at hospital discharge. Discharge disposition was prospectively recorded in the St Louis Children’s Hospital trauma registry and categorised as: medical examiner or morgue, different acute care hospital, inpatient rehabilitation facility, home with health care, home with outpatient rehabilitation, or home without assistance. Two experienced abstractors (based at Washington University, MO, USA), who were masked to study period, retrospectively extracted GOS score from medical records, which were categorised as death, persistent vegetative state, severe disability, moderate disability, or good recovery. Inter-rater variability was favourable for the GOS score (linearly weighted κ=0·82 [95% CI 0·65–0·99]; appendix). We extracted explanatory and secondary outcome variables by chart review and through query of the St Louis Children’s Hospital trauma registry and the virtual PICU Systems database (VPS LLC, Los Angeles, CA, USA),10 of www.thelancet.com/neurology Vol 12 January 2013
Articles
which St Louis Children’s Hospital is a participating centre. Explanatory variables included age, sex, race, injury mechanism, GCS scores after resuscitation (first GCS score documented in the emergency department after blood pressure and oxygenation levels were stabilised), injury severity score (entered in the trauma registry at hospital discharge),11 paediatric risk of mortality (PRISM III) 12 h score (collected during the first 12 h of admission to the PICU),12 and PICU length of stay. We assessed behaviour of the care teams by comparing the timing of invasive intracranial pressure monitoring and by assessment of scored intensity of therapies addressing intracranial hypertension with the paediatric intensity level of therapy (PILOT) scale (appendix).13 We assessed PICU length of stay as an outcome variable by determining the number of PICU-free days on day 28 after admission to hospital, with assignment of zero for deaths.
behaviour during the two study periods was compared. Initiation and maintenance of invasive intracranial pressure monitoring and the intensity of therapies addressing intracranial hypertension (PILOT scale; appendix) during the first 7 PICU days were compared by Kaplan-Meier analysis and difference of daily PILOT means, respectively. Finally, we estimated the mortality trend rate for children with severe traumatic brain injury in the USA during our study period by use of annual death rates in 2002–11 in a sample extracted from the VPS database. The sample query returned estimates for patients with head trauma and GCS scores of 8 or less who were admitted to participating PICUs (70 institutions from 39 states). We used the generalised estimating
Before implementation After implementation p value* of PNCP (n=63) of PNCP (n=60)
Statistical analysis
Demographic characteristics
We compared demographics and injury severity of patients between study periods with a difference of means test. We assessed stability of the relation between discharge disposition category and GOS score with the Mantel test of the tabular matrix. We selected disposition category at hospital discharge as our primary regression outcome variable of interest. We used an ordered probit model to link informative explanatory variables to this ordinal outcome measure. This model was developed with the proportional odds logistic regression function (polr) in the R statistical environment (replication data and code available on request; appendix). The model is presented in standard tabular form; furthermore, we assessed the effect of PNCP as a function of initial injury severity by generation of predicted outcomes across ranges of important explanatory variable values from the estimated coefficients. Because our data show an entire population of interest, rather than a sample from one population, we regarded the results as reliable (rather than significant) if the standard threshold of p<0·05 was met. Separately, we assigned GCS scores after resuscitation (ie, estimates of severity for primary brain injury) and discharge GOS scores (ie, estimates of neurological recovery) as input-output variables showing care system performance as a function of primary injury severity. We used regression analysis to examine the fit between GCS scores after resuscitation (input) and discharge GOS scores (output); we also did this analysis for predicted GOS adjusted by the explanatory variables employed in the ordered probit model. We used the fitted ordered probit model to predict each of the five discharge dispositions for values along the full range of initial injury severity, and plotted graphs of predicted destination on the basis of GCS and PRISM III scores. As a putative mechanism for altered outcomes following PNCP deployment, quantifiable care team
Age, months
www.thelancet.com/neurology Vol 12 January 2013
For the R project for statistical computing see http://www.r-project.org
Mean (SD)
127 (60)
145 (54)
Median (IQR)
144·0 (99·0–173·5)
157·0 (115·0–194·5)
0·08
Sex, male
36 (57%)
34 (57%)
0·96
Race, white
45 (71%)†
43 (72%)
0·98
Severity of injury GCS‡ score after resuscitation Mean (SD)
5 (2)
4 (2)
Median (IQR)
5 (3–6)
4 (3–6)
0·06
ISS Mean (SD)
28 (12)
30 (12)
Median (IQR)
26·0 (17·5–38·0)
29·0 (21·0–38·0)
0·30
PRISM III§ Mean (SD)
6·5 (6·9)
6·9 (6·3)
Median (IQR)
5·0 (2·0–9·0)
5·0 (2·0–8·0)
0·61
Injury mechanism¶ Motor vehicle accident
39 (62%)
42 (70%)
Pedestrian accident
9 (14%)
10 (17%)
Fall
9 (14%)
3 (5%)
Other
6 (10%)
5 (8%)
Glasgow outcome scale score at hospital discharge 1: death
7 (11%)
2: persistent vegetative state
0
2 (3%) 1 (2%)
3: severe disability
18 (29%)
21 (35%)
4: moderate disability
31 (49%)
26 (43%)
7 (11%)
10 (17%)
5: good recovery Disposition at hospital discharge|| Unfavourable (death or inpatient facility)
33 (52%)
20 (33%)
Medical examiner or morgue
7 (11%)
2 (3%)
Different acute care hospital
2 (3%)
2 (3%)
24 (38%)
16 (27%)
30 (48%)
40 (67%)
Inpatient rehabilitation facility Favourable (home, with or without therapy) Home with health care
0
Home with outpatient rehabilitation
3 (5%)
14 (23%)
27 (43%)
25 (42%)
Home without assistance
1 (2%)
(Continues on next page)
47
Articles
Before implementation After implementation p value* of PNCP (n=63) of PNCP (n=60) (Continued from previous page) Length of stay, days PICU length of stay Mean (SD)
10·8 (7·8)
16·8 (17·4)
Median (IQR)
11·7 (4·0–14·6)
14·6 (6·3–22·0)
Mean (SD)
14·7 (8·6)
12·6 (8·7)
Median (IQR)
15·0 (10·0–22·6)
15·0 (10·0–22·6)
Mean (SD)
33·2 (35·7)
45·2 (44·8)
Median (IQR)
21·0 (12·0–39·5)
35·0 (16·8–53·0)
0·01
0·195
Hospital length of stay 0·10
Data are n (%), unless otherwise stated. PNCP=paediatric neurocritical care programme. GCS=Glasgow coma scale. ISS=injury severity score. PRISM=paediatric risk of mortality score. PICU=paediatric intensive-care unit. *Difference of means test for continuous variables and difference of proportions test for dichotomous variables. †Data available for 60 patients. ‡Influence of sedation and neuromuscular blockade on Glasgow coma scale scores did not differ between groups (37 [59%] children received sedation or neuromuscular blockade before GCS scoring in the period before the PNCP started compared with 40 [66%] after the PNCP started; p=0·37). §Data missing for one patient in the period before implementation of PNCP. ¶p=0·41 for injury mechanism. ||p=0·01 for unfavourable vs favourable outcomes.
Table 1: Demographic characteristics, severity of injury, injury mechanisms, and raw outcomes
Coefficient
Standard error
p value
0·02
Variable coefficients (β) Age, months Race, white (vs non-white)
0·482477
0·216061
–0·004674
0·002127
0·03
0·637847
0·257294
0·01
Length of stay in PICU
–0·003776
0·007839
0·60
Sex, male (vs female)
–0·223968
0·215839
0·30
ICP monitoring
0·997479
0·299579
0·001
GCS score after resuscitation
0·125677
0·060159
0·04
PRISM III
–0·065137
0·018125
0·0003
ISS
–0·000315
0·000134
0·02
Fall (vs other)
0·291087
0·268258
0·30
Motor vehicle accident (vs other)
0·197797
0·191271
0·30
Pedestrian accident (vs other)
0·147976
0·241442
0·50 0·02
Threshold estimates ( ) Medical examiner or morgue vs different acute care hospital
–0·647
0·188
Different acute care hospital vs inpatient rehabilitation facility
–0·377
0·226
0·40
Inpatient rehabilitation facility vs home with health care
0·979
0·262
<0·0001
Home with health care vs home with outpatient rehabilitation
1·005
0·262
<0·0001
Home with outpatient rehabilitation vs home without assistance
1·433
0·266
<0·0001
Positive coefficients show that increasing corresponding explanatory variables predict increasingly favourable outcomes. Residual deviance was 287·18 and null deviances were 330·79. PNCP=paediatric neurocritical care programme. PICU=paediatric intensive-care unit. ICP=intracranial pressure. GCS=Glasgow coma scale. PRISM=paediatric risk of mortality score. ISS=injury severity score.
Table 2: Parameter estimates from the ordered probit model for discharge destination of 123 patients
equations model to analyse mortality over time; deidentified sites were controlled as a random effect factor (two-tailed p≤0·05). We used SAS version 9.1 to analyse the sample extracted from the VPS national database. 48
The sponsor of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. All authors had full access to all the data in the study and JAP had final responsibility for the decision to submit for publication.
Results
PICU-free days (at hospital day 28)
Group (before vs after PCNP implementation)
Role of the funding source
123 assessable patients (aged 3–219 months) met study entry criteria (figure 1, table 1). Baseline demographics, injury severity, and injury mechanisms did not differ reliably between treatment periods. Table 1 shows the raw outcomes for differences in injury mechanism, injury severity, and treatment outcomes between study periods. Absolute mortality and rates of unfavourable outcomes for survivors was reduced by implementation of the PNCP (table 1). Notably, the proportion of children discharged home with or without rehabilitation treatment increased, suggesting that increased survival was accompanied by improved functional outcome (table 1). Regression analysis and modelling of GOS as a function of initial injury severity suggested an improvement after PNCP implementation (appendix). We considered the possibility that discharge decision making for children with similar function was biased by PNCP implementation or other factors differing between the two periods; however, the Mantel test showed a consistent relation between discharge disposition and GOS (appendix). We also considered the possibility that improved mortality reflected a secular trend. A national sample of 1193 children with severe traumatic brain injury extracted from the VPS database estimated that annual mortality in 2002–11 was 25% (SD 7), and increased over time (p=0·0012; appendix). Table 2 shows the results of the ordered probit model; p values did not all need to be less than 0·05 to show overall model quality. For numerical stability, software packages fix one threshold near zero on the latent dimension (in this case the category of different acute care hospital to inpatient rehabilitation facility). Variables reliably associated with increased probability of unfavourable outcomes (negative β) were older age, higher PRISM III scores, and higher injury severity score. Alternatively, white race, use of intracranial pressure monitoring, and high GCS scores after resuscitation were associated with increased probability of favourable outcomes (positive β). Coefficients for PICU length of stay, sex, and mechanism of injury were not statistically reliable. With regard to overall model fit, the difference in the residual deviance and the null deviance (from a model without explanatory variables) showed substantial improvement in fit (χ² tail=1·7831×10–05 on 12 degrees of freedom). The key result in table 2 is the positive coefficient for group, providing evidence that implementation of the PNCP increased the probability of positive categorical outcomes. To show the strength of this relation, we can www.thelancet.com/neurology Vol 12 January 2013
Articles
www.thelancet.com/neurology Vol 12 January 2013
A
Before After PNCP PNCP
0·6
Discharge destination probability
0·5
Home without assistance Inpatient rehabilitation facility Home with outpatient rehabilitation Medical examiner or morgue Different acute care hospital
0·4
0·3
0·2
0·1
0 0
5
10
15
20
25
30
35
PRISM III
B 0·6
0·5 Discharge destination probability
consider a white male patient who had a motor vehicle accident, with all other explanatory variables set at the data mean. For such a patient, the probability of death declined after initiation of the PNCP from 21·0% to 9·9% and the probability for discharge to home without assistance increased from 10·1% to 21·4% (appendix). Overall, predicted outcome probabilities after initiation of the PNCP improved across the spectrum of initial injury (figure 2). Because only one patient was in the category of “home with health care”, we have insufficient information to produce a prediction line across PRISM III or GCS for graphical purposes. Therefore, figure 2 provides predictive lines for five categories only. Specific improvement varied by injury severity measured by PRISM III, creating a curvilinear effect for the category of “inpatient rehabilitation facility” (figure 2). After PNCP implementation, the probability of this outcome decreased for children with lower PRISM III scores, but increased for more severely ill children; this effect seems centred between PRISM III scores of 15–20. Several findings at this injury level suggest improved system performance after implementation of the PNCP. First, probabilities for “home without assistance” and “medical examiner or morgue” equalised, shown by the level of injury at which the curves cross; after PNCP implementation, the PRISM III threshold for this balance point increased from 16 to 21. Second, for children with less severe PRISM III scores (scores <15), the PNCP shifted outcomes from “inpatient rehabilitation” to “home without assistance”. Third, for children with more severe PRISM III scores (scores >20), PNCP redistributed probability from “medical examiner or morgue” to all other outcomes; the magnitude of this effect progressed with increasing PRISM III scores. From regression approaches, implementation of PNCP was not reliably associated with improvements in GOS score at hospital discharge (p=0·1023). However, after implementation of the PNCP, the model predicted more favourable GOS values across the full spectrum of categories of GCS after resuscitation (appendix). When GCS category after resuscitation was treated as an input variable and discharge GOS was treated as an output variable to assess care team effectiveness, PNCP deployment tended to improve GOS output per GCS input, with or without adjustment for covariates identified in our model (appendix). Notably, the difference between initial injury severity and outcome for the two periods increased as GCS scores after resuscitation worsened, reinforcing the aforementioned finding that PNCP deployment improved outcome most for the most severely injured. We assessed the influence of the PNCP on behaviour of the care team by analysing the initiation and maintenance of invasive intracranial pressure monitoring and the intensity of therapies addressing intracranial hypertension (PILOT scale). After initiation of the PNCP,
0·4
0·3
0·2
0·1
0 8
7
6 5 GCS after resuscitation
4
3
Figure 2: Outcome probabilities by injury severity, by PRISM III (A) and GCS (B) scores We created these figures by using the fitted ordered probit model to predict each of the five outcomes for values along the full range of initial injury severity, before and after implementation of the PNCP. Outcomes after implementation improved comprehensively across the full range of PRISM III and GCS severities: note the robust, favourable effect of the PNCP on probability for “medical examiner or morgue’ and ‘home without assistance’ outcome categories (A), and the striking increase after implementation of the PNCP in probability for ‘home without assistance’ and diminished probability for ‘inpatient rehabilitation’ (B). Outcome trends in (B) appear more linear than in (A); this data-driven linear relation between initial GCS and functionally influenced outcomes lends credence to the broad superiority of the predictions after PNCP implementation. PRISM=paediatric risk of mortality score. GCS=Glasgow coma scale.
intracranial pressure monitoring was started earlier and maintained for a greater duration than it was before the PNCP was used (figure 3). Moreover, the intensity of therapies directed against intracranial hypertension was reliably higher on PICU days 1, 2, and 3 after 49
Articles
Probability of ICP monitoring
A
1·00
After PNCP Before PNCP
0·75 0·50
Discussion
0·25 0 0
Number at risk After PNCP 60 Before PNCP 63
B 10
1
2
3
4
5 6 PICU day
57 57
55 52
48 41
46 31
39 26
34 24
7 32 20
8
9
10
11
28 17
25 12
20 9
18 4
p=0·0298
Before PNCP After PNCP
Mean pilot scale
p=0·0292 8
p=0·0471
p=0·1346
6
p=0·1678
p=0·4016
p=0·0698
4 2 0
Day 1
Day 2
Day 3
Day 4
Day 5
Day 6
C
Day 7 Before PNCP After PNCP
20
Pilot scale
15
10
5
0
Day 1
Day 2
Day 3
Day 4
Day 5
Day 6
Day 7
Figure 3: Use of invasive neuromonitoring and intensity of ICP-directed therapies Kaplan-Meier estimates of probability of ICP monitoring before and after establishment of the PNCP (Cox proportional hazards likelihood ratio p=0·03); vertical bars show one standard error for probability of ICP monitoring (A). Daily PILOT scores shown as means and p values (B) and boxplots of median (IQR) and whiskers of 1·5 times the length of the box (C). Therapeutic intensity increased after PNCP when p<0·05. ICP=intracranial pressure. PNCP=paediatric neurocritical care programme.
implementation of the PNCP than before implementation (figure 3). Notably, decompressive craniectomy for increased intracranial pressure was included in the PILOT score (appendix). This treatment was done for one child treated in the PICU before PNCP was implemented and seven children after implementation, all of whom survived. All these children underwent decompression on day 1, apart from two patients in the later time period who underwent decompression on day 2. The patient treated before implementation of PNCP was discharged home without assistance, and four out of the seven children treated after implementation of PNCP were discharged home with or without outpatient care (appendix). Length of stay in the PICU increased after implementation of the PNCP (p=0·01; table 1). However, 50
when the mortality-attributable reduction in stay was controlled, PICU-free days during the first 28 hospital days were not reliably different (p=0·195). Furthermore, overall hospital length of stay did not differ between the two periods (p=0·10).
Mortality for children who have had severe traumatic brain injury ranges between 9% and 25%.7,8,14,15 No outcome improvements have been reported in the past few decades and most survivors have long-term intellectual deficits and reduced quality of life. Our data show that outcomes might be improved by changing the process of care in a way that stably implements a cooperative programme of accepted best practice (panel). Our primary outcome was categorised disposition at hospital discharge, a measure with known correlations to functional neurological outcomes.21–23 Although we initially considered use of a dichotomous survival model, because of the small number of fatalities and our desire to discriminate within the full spectrum of a meaningful outcome range, we abandoned this approach. We therefore analysed two scalable variables: categorised discharge disposition and GOS at hospital discharge. Although these variables share content, they enabled assessment of the PNCP in different ways. Disposition proved to be more robust for modelling, because this measure is more categorically distinct and qualitatively informed than is GOS, and thus better shows overall wellbeing. GOS more narrowly shows neurological recovery than does disposition and, as such, was used to assess limitations of secondary brain injury as a function of primary injury severity. Although discharge GOS was the best means available to retrospectively estimate functional neurological recovery, its predictive validity has been shown only when outpatients are prospectively scored at a timepoint remote from discharge.9,24 Although discharge disposition improved after PNCP was implemented at our centre, several variables affected outcomes in both periods. Increased age, PRISM III score, and injury severity score at entry to the PICU were associated with unfavourable outcome, as reported elsewhere.7,25 White race, increased post-resuscitation GCS score, and use of intracranial pressure monitoring were associated with favourable outcome, also as reported elsewhere.26–28 Notably, patients treated before and after implementation did not differ demographically or in terms of injury severity (generally and neurologically). Despite limitations related to sedative administration for GCS scoring and concern for leadtime bias with PRISM III scoring, both strategies have been used reliably to measure severity and robustly predict outcome for critically ill children.6,12 Our finding of PNCP effectiveness reinforces other evidence suggesting that traumatic brain injury outcome www.thelancet.com/neurology Vol 12 January 2013
Articles
is improved by timely, integrated medical and surgical interventions designed to mitigate secondary brain insults.29 Use of the PNCP in our study improved timing and intensity of intracranial hypertension-targeted therapies. Whereas most studies of guidelines directed at targeting improvements in intracranial pressure in adults are favourable, increased intensity of therapy does not universally improve outcomes.30 Although our PILOT data are important, several changes probably led to our findings. The PNCP provided timely, multisystem monitoring of factors affecting cerebral oxygen delivery and consumption, and allowed integration of aggressive medical and surgical interventions in a tiered plan designed to reduce secondary brain injury;15,27 both changes to our approach were achieved by repeated education, attention to interdisciplinary collaboration, and continuous quality improvement. The influence of PNCP deployment is also consistent with the general observation that organisational patterns in intensive-care units affect outcomes.31 Specifically, disease-focused teams dedicated to explicit management approaches improve outcomes after trauma, acute lung injury, severe sepsis, and acute myocardial infarction.31,32 Moreover, implementation of neurocritical-care teams in adult intensive-care units improve outcomes after stroke, intracerebral haemorrhage, subarachnoid haemorrhage, and traumatic brain injury.29,33 Specialised services have been described for critically ill children and neonates with neurological diagnoses.19,20 This care model exploits evidence-based best practice guidelines and complementary expertise inherent to multidisciplinary teams, which achieve synergy as members work concurrently, rather than serially, on problems.34 Although such guidelines are composed of specific interventions (which individually are often supported by weak evidence), application of complementary interventions by expert teams seems to achieve outcomes exceeding the efficacy predicted when individual guideline components are studied alone.35 Moreover, stable application of integrated algorithms in intensive-care units needs a comprehensive implementation programme (with coordinated guideline rollout, staff education, compliance monitoring, and quality improvement cycles).36 Our single-centre study was limited by its small sample size, use of short-term outcome measures, retrospective design, slow patient accrual, and exclusion of patients with abusive head trauma. However, considering these issues, we designed a robust model and analysis plan that enabled us to report a meaningfully improved outcome for children with severe traumatic brain injury; the magnitude of this effect progressed with initial injury severity and was associated with timely, aggressive application of integrated medical and surgical interventions addressing intracranial hypertension. These outcomes, compellingly linked to programmatic (PNCP) change in intensive-care units, emphasise the need for novel www.thelancet.com/neurology Vol 12 January 2013
Panel: Research in context Systematic review We searched PubMed for articles published without language restriction up to October, 2012, with the following MeSH terms: “brain injuries”[MeSH terms] OR (“brain”[all fields] AND “injuries”[all fields]) OR “brain injuries”[all fields] OR (“traumatic”[all fields] AND “brain”[all fields] AND “injury”[all fields]) OR “traumatic brain injury”[all fields] AND (“guideline”[publication type] OR “guidelines as topic” [MeSH terms] OR “guidelines” [all fields]) AND outcome[all fields]; traumatic brain injury guidelines[all fields] AND neurocritical care[all fields] AND (Clinical Trial[ptyp]). The search was repeated adding the terms “children” and “pediatric”. Interpretation We identified no clinical trials or meta-analyses comparing implementation strategies of guidelines for severe traumatic brain injury in children. Several studies in adults16 support aggressive management and implementation of guidelines for traumatic brain injury to improve outcome in adults with severe traumatic brain injury. However, the substantial variability that exists in practice restricts favourable effects of guidelines. We identified two reviews17,18 in children; although guideline publication was associated with change in relevant clinical endpoints (such as hypocapnia), effects on patient functional outcomes were lacking. Two paediatric studies19,20 described implementation of a neurocritical care service for children and neonates, but the effect on team behaviour or clinical outcomes was not reported. Our study is the first to report an association between implementation of traumatic brain injury guidelines driven by a paediatric neurocritical care programme and a beneficial effect on team behaviour and clinical outcomes.
“T3” implementation research methodologies37 designed to assess and improve four factors: processes for crafting best practice algorithms, optimum care team composition and communication, algorithm efficacy and cost-effectiveness, and implementation and knowledge transfer methodologies specific to the intensive-care unit. Contributors JAP, JRL, IGM, DDL, MSK, JG, and AD designed the study. JAP, JRL, JG, and AD did the data analysis. JAP, JRL, MN, DDL, MSK, JG, and AD interpreted the data and wrote the report. JAP, JG, and AD drew the figures. IGM collected data. Conflicts of interest We declare that we have no conflicts of interest. Acknowledgments Tina Day contributed to data collection; Barbara Miller contributed to traumatic brain injury pathway development, implementation, and monitoring, and data collection; Daniel Martin contributed to manuscript editing (all Washington University School of Medicine, St Louis, MO, USA); all received salary support from the paediatric neurocritical care programme. We gratefully acknowledge the outstanding care delivered by the nurses who staff the St Louis Children’s Hospital paediatric intensive-care unit. References 1 US Centers for Disease Control and Prevention. http://www.cdc. gov/TraumaticBrainInjury/ (accessed March 29, 2012). 2 Scaife ER, Statler KD. Traumatic brain injury: preferred methods and targets for resuscitation. Curr Opin Pediatr 2010; 22: 339–45. 3 Potoka DA, Schall LC, Gardner MJ, Stafford PW, Peitzman AB, Ford HR. Impact of pediatric trauma centers on mortality in a statewide system. J Trauma 2000; 49: 237–45. 4 Guidelines for the acute medical management of severe traumatic brain injury in infants, children, and adolescents. Crit Care Med 2003; 31 (suppl): S407–91.
51
Articles
5
6 7
8
9
10 11 12 13
14
15
16
17
18
19 20 21
22
52
Dean NP, Boslaugh S, Adelson PD, Pineda JA, Leonard JR. Physician agreement with evidence-based recommendations for the treatment of severe traumatic brain injury in children. J Neurosurg 2007; 107 (suppl): 387–91. Maas AI, Harrison-Felix CL, Menon D, et al. Standardizing data collection in traumatic brain injury. J Neurotrauma 2011; 28: 177–87. Ducrocq SC, Meyer PG, Orliaguet GA, et al. Epidemiology and early predictive factors of mortality and outcome in children with traumatic severe brain injury: experience of a French pediatric trauma center. Pediatr Crit Care Med 2006; 7: 461–67. Figaji AA, Zwane E, Thompson C, et al. Brain tissue oxygen tension monitoring in pediatric severe traumatic brain injury. Part 1: relationship with outcome. Childs Nerv Syst 2009; 25: 1325–33. McCauley SR, Wilde EA, Anderson VA, et al. Recommendations for the use of common outcome measures in pediatric traumatic brain injury research. J Neurotrauma 2012; 29: 678–705. Wetzel RC, Sachedeva R, Rice TB. Are all ICUs the same? Paediatr Anaesth 2011; 21: 787–93. American College of Surgeons. National trauma data bank. http:// www.facs.org/trauma/ntdb/index.html (accessed May 30, 2012). Pollack MM, Patel KM, Ruttimann UE. PRISM III: an updated pediatric risk of mortality score. Crit Care Med 1996; 24: 743–52. Shore PM, Hand LL, Roy L, Trivedi P, Kochanek PM, Adelson PD. Reliability and validity of the Pediatric Intensity Level of Therapy (PILOT) scale: a measure of the use of intracranial pressure-directed therapies. Crit Care Med 2006; 34: 1981–87. Tasker RC, Fleming TJ, Young AE, Morris KP, Parslow RC. Severe head injury in children: intensive care unit activity and mortality in England and Wales. Br J Neurosurg 2011; 25: 68–77. Chiaretti A, Piastra M, Pulitano S, et al. Prognostic factors and outcome of children with severe head injury: an 8-year experience. Childs Nerv Syst 2002; 18: 129–36. Stein SC, Georgoff P, Meghan S, Mirza KL, El Falaky OM. Relationship of aggressive monitoring and treatment to improved outcomes in severe traumatic brain injury. J Neurosurg 2010; 112: 1105–12. Guerguerian AM, Lo TY, Hutchison JS. Clinical management and functional neuromonitoring in traumatic brain injury in children. Curr Opin Pediatr 2009; 21: 737–44. Stocchetti N, Conte V, Ghisoni L, Canavesi K, Zanaboni C. Traumatic brain injury in pediatric patients. Minerva Anestesiol 2010; 76: 1052–59. Bell MJ, Carpenter J, Au AK, et al. Development of a pediatric neurocritical care service. Neurocrit Care 2009; 10: 4–10. Glass HC, Bonifacio SL, Peloquin S, et al. Neurocritical care for neonates. Neurocrit Care 2010; 12: 421–29. Shi J, Xiang H, Wheeler K, et al. Costs, mortality likelihood and outcomes of hospitalized US children with traumatic brain injuries. Brain Inj 2009; 23: 602–11. Bulger EM, Nathens AB, Rivara FP, Moore M, MacKenzie EJ, Jurkovich GJ. Management of severe head injury: institutional variations in care and effect on outcome. Crit Care Med 2002; 30: 1870–76.
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
Keller JM, Sciadini MF, Sinclair E, O’Toole RV. Geriatric trauma: demographics, injuries, and mortality. J Orthop Trauma 2012; 26: e161–65. Beers SR, Wisniewski SR, Garcia-Filion P, et al. Validity of a pediatric version of the Glasgow Outcome Scale-Extended. J Neurotrauma 2012; 29: 1126–39. Crowe LM, Catroppa C, Babl FE, Rosenfeld JV, Anderson V. Timing of traumatic brain injury in childhood and intellectual outcome. J Pediatr Psychol 2012; 37: 745–54. Forsyth RJ, Parslow RC, Tasker RC, Hawley CA, Morris KP. Prediction of raised intracranial pressure complicating severe traumatic brain injury in children: implications for trial design. Pediatr Crit Care Med 2008; 9: 8–14. Salorio CF, Slomine BS, Guerguerian AM, et al. Intensive care unit variables and outcome after pediatric traumatic brain injury: a retrospective study of survivors. Pediatr Crit Care Med 2008; 9: 47–53. Haider AH, Efron DT, Haut ER, DiRusso SM, Sullivan T, Cornwell EE 3rd. Black children experience worse clinical and functional outcomes after traumatic brain injury: an analysis of the National Pediatric Trauma Registry. J Trauma 2007; 62: 1259–62. Fakhry SM, Trask AL, Waller MA, Watts DD. Management of brain-injured patients by an evidence-based medicine protocol improves outcomes and decreases hospital charges. J Trauma 2004; 56: 492–99. Cremer OL, van Dijk GW, van Wensen E, et al. Effect of intracranial pressure monitoring and targeted intensive care on functional outcome after severe head injury. Crit Care Med 2005; 33: 2207–13. Randolph AG, Pronovost P. Reorganizing the delivery of intensive care could improve efficiency and save lives. J Eval Clin Pract 2002; 8: 1–8. Cuschieri J, Johnson JL, Sperry J, et al. Benchmarking outcomes in the critically injured trauma patient and the effect of implementing standard operating procedures. Ann Surg 2012; 255: 993–99. Knopf L, Staff I, Gomes J, McCullough L. Impact of a neurointensivist on outcomes in critically ill stroke patients. Neurocrit Care 2012; 16: 63–71. Sutton G, Liao J, Jimmieson NL, Restubog SL. Measuring multidisciplinary team effectiveness in a ward-based healthcare setting: development of the team functioning assessment tool. J Healthc Qual 2011; 33: 10–23. Logani S, Green A, Gasperino J. Benefits of high-intensity intensive care unit physician staffing under the affordable care act. Crit Care Res Pract 2011; 2011: 170814. Davis DA, Taylor-Vaisey A. Translating guidelines into practice. A systematic review of theoretic concepts, practical experience and research evidence in the adoption of clinical practice guidelines. Can Urol Assoc J 1997; 157: 408–16. Dougherty D, Conway PH. The “3T’s” road map to transform US health care: the “how” of high-quality care. JAMA 2008; 299: 2319–21.
www.thelancet.com/neurology Vol 12 January 2013
Effect of implementation of a paediatric neurocritical care programme on outcomes after severe traumatic brain injury: a retrospective cohort study Jose A Pineda*, Jeffrey R Leonard*, Ioanna G Mazotas, Michael Noetzel, David D Limbrick, Martin S Keller, Jeff Gill, Allan Doctor
Summary Background Outcomes after traumatic brain injury are worsened by secondary insults; modern intensive-care units address such challenges through use of best-practice pathways. Organisation of intensive-care units has an important role in pathway effectiveness. We aimed to assess the effect of a paediatric neurocritical care programme (PNCP) on outcomes for children with severe traumatic brain injury. Methods We undertook a retrospective cohort study of 123 paediatric patients with severe traumatic brain injury (Glasgow coma scale scores ≤8, without gunshot or abusive head trauma, cardiac arrest, or Glasgow coma scale scores of 3 with fixed and dilated pupils) admitted to the paediatric intensive-care unit of the St Louis Children’s Hospital (St Louis, MO, USA) between July 15, 1999, and Jan 15, 2012. The primary outcome was rate of categorised hospital discharge disposition before and after implementation of a PNCP on Sept 17, 2005. We developed an ordered probit statistical model to assess adjusted outcome as a function of initial injury severity. We assessed care-team behaviour by comparing timing of invasive neuromonitoring and scored intensity of therapies targeting intracranial hypertension. Findings Characteristics of treated patients (aged 3–219 months) were much the same between treatment periods. Before PNCP implementation, 33 (52%) of 63 patients had unfavourable disposition at hospital discharge (death or admission to an inpatient facility) and 30 (48%) had a favourable disposition (home with or without treatment); after PNCP implementation, 20 (33%) of 60 patients had unfavourable disposition and 40 (67%) had favourable disposition (p=0·01). Seven (11%) patients died before PNCP implementation compared with two (3%) deaths after implementation. The probit model indicated that outcome improved across the spectrum of Glasgow coma scale scores after resuscitation (p=0·02); this improvement progressed with increasing injury severity. Kaplan-Meier analysis suggested that neuromonitoring was started earlier and maintained longer after implementation of the PNCP (p=0·03). Therapeutic intensity scores were increased for the first 3 days of treatment after PNCP implementation (p=0·0298 for day 1, p=0·0292 for day 2, and p=0·0471 for day 3). The probit model suggested that increasing age (p=0·03), paediatric risk of mortality III scores (p=0·0003), and injury severity scores (p=0·02) were reliably associated with increased probability of unfavourable outcomes whereas white race (p=0·01), use of intracranial pressure monitoring (p=0·001), and increasing Glasgow coma scale scores after resuscitation (p=0·04) were associated with increased probability of favourable outcomes. Interpretation Outcomes for children with traumatic brain injury can be improved by altering the care system in a way that stably implements a cooperative programme of accepted best practice.
Lancet Neurol 2013; 12: 45–52 Published Online November 28, 2012 http://dx.doi.org/10.1016/ S1474-4422(12)70269-7 See Comment page 26 *Contributed equally Department of Pediatrics, Division of Critical Care Medicine (J A Pineda MD, A Doctor MD); Department of Neurology (J A Pineda, Prof M Noetzel MD), Department of Surgery (I G Mazotas MD, M S Keller MD, Prof J Gill PhD), Department of Neurosurgery (J R Leonard MD, D D Limbrick MD), Department of Pediatrics (D D Limbrick, J R Leonard), Department of Political Science (J Gill), Department of Biostatistics (J Gill), and Department of Biochemistry and Molecular Biophysics (A Doctor), Washington University School of Medicine, St Louis, MO, USA Correspondence to: Dr Jose A Pineda, Washington University School of Medicine, Campus Box 8116, 1 Children’s Place, St Louis, MO 63110, USA [email protected]
Funding St Louis Children’s Hospital and the Sean Glanvill Foundations.
Introduction Traumatic brain injury is a leading cause of death and disability in children.1 Outcomes are affected by the extent of the primary injury (original brain trauma) and the degree of secondary injuries, including maladaptive signalling caused by the primary trauma, perturbed cellular energetics, cerebral ischaemia (eg, perfusion deficits associated with dysregulated blood flow, intracranial hypertension or hypotension, and bleeding), cerebral hypoxia (eg, diminished blood oxygen content from respiratory failure or anaemia), and increased cerebral oxygen demand associated with seizures or fever.2 Because no treatments specific to primary brain trauma are available, the key goal for care is prevention of secondary insults. Achievement of this goal will require interpretation of rapidly changing, complex www.thelancet.com/neurology Vol 12 January 2013
information from several organ systems, coordinated decision making between services, time-sensitive and goal-directed interventions, and an iterative repeat of this cycle. We developed a paediatric neurocritical care programme (PNCP) to implement this approach and improve penetration and consistent retention of best practice into routine care. In paediatric health-care centres, specialised teams guided by evidence-based plans of care have improved outcomes for patients with severe trauma3 or congenital heart disease. Furthermore, implementation of rapid response teams has reduced mortality. Children with traumatic brain injury are especially likely to benefit from this approach to care because of their high susceptibility to secondary injury, the nature of complex multisystem pathological changes with narrow 45
Articles
See Online for appendix
therapeutic windows, and the robust potential for favourable functional recovery in young patients. Although management guidelines for children with traumatic brain injury have been created,4 effective implementation is challenged by weak evidence, cultural barriers to change, and inefficiencies inherent to complex, time-sensitive decision making across subspecialties. These factors all lead to inconsistent implementation and variability in care and outcomes.5 Here, we report the effect of a neurocritical care programme that was designed to implement a system of cross-specialty communication and an explicit plan of severity-based monitoring and intervention for children who have severe traumatic brain injury. We aimed to assess outcomes for patients and also the behaviour of the care team by comparing the timing of invasive neuromonitoring and intensity of therapies targeting intracranial hypertension and programme performance.
Methods Study design and participants In this cohort study, we screened trauma patients who were admitted to the St Louis Children’s Hospital (St Louis, MO, USA) emergency department between July 15, 1999, and Jan 15, 2012, with an initial Glasgow coma scale (GCS) score of 8 or less. Patients were screened for severe traumatic brain injury (GCS score ≤8 in the emergency department after resuscitation)6 and survival to paediatric intensive-care unit (PICU) admission. We excluded children with GCS scores of 597 patients with trauma admitted to emergency department with GCS score ≤8
316 without TBI or not admitted to PICU
281 met inclusion criteria
157 excluded 80 abusive head trauma 27 GCS 3 and fixed and dilated pupils in emergency department 28 cardiac arrest before admission 22 gunshot wound to the head
124 analysed
1 missing data from the medical record 63 before PNCP implementation
60 after PNCP implementation
Figure 1: Study profile GCS=Glasgow coma scale. TBI=traumatic brain injury. PICU=paediatric intensive-care unit. AHT=abusive head trauma. PNCP=paediatric neurocritical care programme.
46
3 with bilateral fixed and dilated pupils on admission to the emergency department, cardiac arrest before admission to the PICU, abusive head trauma, and gunshot wounds to the head.7,8 Research procedures were approved and exempted from informed consent by the Washington University institutional review board (St Louis, MO, USA).
Procedures The PNCP was first implemented on Sept 17, 2005. Using the 2003 Brain Trauma Foundation guidelines,4 we developed a time-sensitive, severity-based approach to monitoring and treatment of children with traumatic brain injury (appendix). This approach coordinated communication and activity in PICU staff and physician faculty and trainees in several disciplines (critical care, neurosurgery, surgery, anaesthesia, and radiology) and was implemented through a detailed training programme, an explicit process for maintenance of pathway fidelity, and continuous quality improvement. We examined raw outcomes for differences between study periods (ie, before implementation of the PNCP and after implementation of the PNCP). We defined favourable PNCP effects as reductions in population mortality and in the number of children with discharge status needing ongoing inpatient care, and also in terms of increases in the number of children with statuses that allowed discharge to home with or without outpatient treatment. Subsequently, we developed an ordered probit model to assess relations between patients’ characteristics, PNCP implementation, and disposition categories. Our primary outcome was categorised discharge disposition. Our secondary outcome analysis included examination of the relation between initial GCS scores after resuscitation and the Glasgow outcome scale (GOS) score9 at hospital discharge, and length of stay in the PICU. To assess the effect of the PNCP on outcomes, we analysed two scalable outcome variables: discharge disposition and GOS score at hospital discharge. Discharge disposition was prospectively recorded in the St Louis Children’s Hospital trauma registry and categorised as: medical examiner or morgue, different acute care hospital, inpatient rehabilitation facility, home with health care, home with outpatient rehabilitation, or home without assistance. Two experienced abstractors (based at Washington University, MO, USA), who were masked to study period, retrospectively extracted GOS score from medical records, which were categorised as death, persistent vegetative state, severe disability, moderate disability, or good recovery. Inter-rater variability was favourable for the GOS score (linearly weighted κ=0·82 [95% CI 0·65–0·99]; appendix). We extracted explanatory and secondary outcome variables by chart review and through query of the St Louis Children’s Hospital trauma registry and the virtual PICU Systems database (VPS LLC, Los Angeles, CA, USA),10 of www.thelancet.com/neurology Vol 12 January 2013
Articles
which St Louis Children’s Hospital is a participating centre. Explanatory variables included age, sex, race, injury mechanism, GCS scores after resuscitation (first GCS score documented in the emergency department after blood pressure and oxygenation levels were stabilised), injury severity score (entered in the trauma registry at hospital discharge),11 paediatric risk of mortality (PRISM III) 12 h score (collected during the first 12 h of admission to the PICU),12 and PICU length of stay. We assessed behaviour of the care teams by comparing the timing of invasive intracranial pressure monitoring and by assessment of scored intensity of therapies addressing intracranial hypertension with the paediatric intensity level of therapy (PILOT) scale (appendix).13 We assessed PICU length of stay as an outcome variable by determining the number of PICU-free days on day 28 after admission to hospital, with assignment of zero for deaths.
behaviour during the two study periods was compared. Initiation and maintenance of invasive intracranial pressure monitoring and the intensity of therapies addressing intracranial hypertension (PILOT scale; appendix) during the first 7 PICU days were compared by Kaplan-Meier analysis and difference of daily PILOT means, respectively. Finally, we estimated the mortality trend rate for children with severe traumatic brain injury in the USA during our study period by use of annual death rates in 2002–11 in a sample extracted from the VPS database. The sample query returned estimates for patients with head trauma and GCS scores of 8 or less who were admitted to participating PICUs (70 institutions from 39 states). We used the generalised estimating
Before implementation After implementation p value* of PNCP (n=63) of PNCP (n=60)
Statistical analysis
Demographic characteristics
We compared demographics and injury severity of patients between study periods with a difference of means test. We assessed stability of the relation between discharge disposition category and GOS score with the Mantel test of the tabular matrix. We selected disposition category at hospital discharge as our primary regression outcome variable of interest. We used an ordered probit model to link informative explanatory variables to this ordinal outcome measure. This model was developed with the proportional odds logistic regression function (polr) in the R statistical environment (replication data and code available on request; appendix). The model is presented in standard tabular form; furthermore, we assessed the effect of PNCP as a function of initial injury severity by generation of predicted outcomes across ranges of important explanatory variable values from the estimated coefficients. Because our data show an entire population of interest, rather than a sample from one population, we regarded the results as reliable (rather than significant) if the standard threshold of p<0·05 was met. Separately, we assigned GCS scores after resuscitation (ie, estimates of severity for primary brain injury) and discharge GOS scores (ie, estimates of neurological recovery) as input-output variables showing care system performance as a function of primary injury severity. We used regression analysis to examine the fit between GCS scores after resuscitation (input) and discharge GOS scores (output); we also did this analysis for predicted GOS adjusted by the explanatory variables employed in the ordered probit model. We used the fitted ordered probit model to predict each of the five discharge dispositions for values along the full range of initial injury severity, and plotted graphs of predicted destination on the basis of GCS and PRISM III scores. As a putative mechanism for altered outcomes following PNCP deployment, quantifiable care team
Age, months
www.thelancet.com/neurology Vol 12 January 2013
For the R project for statistical computing see http://www.r-project.org
Mean (SD)
127 (60)
145 (54)
Median (IQR)
144·0 (99·0–173·5)
157·0 (115·0–194·5)
0·08
Sex, male
36 (57%)
34 (57%)
0·96
Race, white
45 (71%)†
43 (72%)
0·98
Severity of injury GCS‡ score after resuscitation Mean (SD)
5 (2)
4 (2)
Median (IQR)
5 (3–6)
4 (3–6)
0·06
ISS Mean (SD)
28 (12)
30 (12)
Median (IQR)
26·0 (17·5–38·0)
29·0 (21·0–38·0)
0·30
PRISM III§ Mean (SD)
6·5 (6·9)
6·9 (6·3)
Median (IQR)
5·0 (2·0–9·0)
5·0 (2·0–8·0)
0·61
Injury mechanism¶ Motor vehicle accident
39 (62%)
42 (70%)
Pedestrian accident
9 (14%)
10 (17%)
Fall
9 (14%)
3 (5%)
Other
6 (10%)
5 (8%)
Glasgow outcome scale score at hospital discharge 1: death
7 (11%)
2: persistent vegetative state
0
2 (3%) 1 (2%)
3: severe disability
18 (29%)
21 (35%)
4: moderate disability
31 (49%)
26 (43%)
7 (11%)
10 (17%)
5: good recovery Disposition at hospital discharge|| Unfavourable (death or inpatient facility)
33 (52%)
20 (33%)
Medical examiner or morgue
7 (11%)
2 (3%)
Different acute care hospital
2 (3%)
2 (3%)
24 (38%)
16 (27%)
30 (48%)
40 (67%)
Inpatient rehabilitation facility Favourable (home, with or without therapy) Home with health care
0
Home with outpatient rehabilitation
3 (5%)
14 (23%)
27 (43%)
25 (42%)
Home without assistance
1 (2%)
(Continues on next page)
47
Articles
Before implementation After implementation p value* of PNCP (n=63) of PNCP (n=60) (Continued from previous page) Length of stay, days PICU length of stay Mean (SD)
10·8 (7·8)
16·8 (17·4)
Median (IQR)
11·7 (4·0–14·6)
14·6 (6·3–22·0)
Mean (SD)
14·7 (8·6)
12·6 (8·7)
Median (IQR)
15·0 (10·0–22·6)
15·0 (10·0–22·6)
Mean (SD)
33·2 (35·7)
45·2 (44·8)
Median (IQR)
21·0 (12·0–39·5)
35·0 (16·8–53·0)
0·01
0·195
Hospital length of stay 0·10
Data are n (%), unless otherwise stated. PNCP=paediatric neurocritical care programme. GCS=Glasgow coma scale. ISS=injury severity score. PRISM=paediatric risk of mortality score. PICU=paediatric intensive-care unit. *Difference of means test for continuous variables and difference of proportions test for dichotomous variables. †Data available for 60 patients. ‡Influence of sedation and neuromuscular blockade on Glasgow coma scale scores did not differ between groups (37 [59%] children received sedation or neuromuscular blockade before GCS scoring in the period before the PNCP started compared with 40 [66%] after the PNCP started; p=0·37). §Data missing for one patient in the period before implementation of PNCP. ¶p=0·41 for injury mechanism. ||p=0·01 for unfavourable vs favourable outcomes.
Table 1: Demographic characteristics, severity of injury, injury mechanisms, and raw outcomes
Coefficient
Standard error
p value
0·02
Variable coefficients (β) Age, months Race, white (vs non-white)
0·482477
0·216061
–0·004674
0·002127
0·03
0·637847
0·257294
0·01
Length of stay in PICU
–0·003776
0·007839
0·60
Sex, male (vs female)
–0·223968
0·215839
0·30
ICP monitoring
0·997479
0·299579
0·001
GCS score after resuscitation
0·125677
0·060159
0·04
PRISM III
–0·065137
0·018125
0·0003
ISS
–0·000315
0·000134
0·02
Fall (vs other)
0·291087
0·268258
0·30
Motor vehicle accident (vs other)
0·197797
0·191271
0·30
Pedestrian accident (vs other)
0·147976
0·241442
0·50 0·02
Threshold estimates ( ) Medical examiner or morgue vs different acute care hospital
–0·647
0·188
Different acute care hospital vs inpatient rehabilitation facility
–0·377
0·226
0·40
Inpatient rehabilitation facility vs home with health care
0·979
0·262
<0·0001
Home with health care vs home with outpatient rehabilitation
1·005
0·262
<0·0001
Home with outpatient rehabilitation vs home without assistance
1·433
0·266
<0·0001
Positive coefficients show that increasing corresponding explanatory variables predict increasingly favourable outcomes. Residual deviance was 287·18 and null deviances were 330·79. PNCP=paediatric neurocritical care programme. PICU=paediatric intensive-care unit. ICP=intracranial pressure. GCS=Glasgow coma scale. PRISM=paediatric risk of mortality score. ISS=injury severity score.
Table 2: Parameter estimates from the ordered probit model for discharge destination of 123 patients
equations model to analyse mortality over time; deidentified sites were controlled as a random effect factor (two-tailed p≤0·05). We used SAS version 9.1 to analyse the sample extracted from the VPS national database. 48
The sponsor of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. All authors had full access to all the data in the study and JAP had final responsibility for the decision to submit for publication.
Results
PICU-free days (at hospital day 28)
Group (before vs after PCNP implementation)
Role of the funding source
123 assessable patients (aged 3–219 months) met study entry criteria (figure 1, table 1). Baseline demographics, injury severity, and injury mechanisms did not differ reliably between treatment periods. Table 1 shows the raw outcomes for differences in injury mechanism, injury severity, and treatment outcomes between study periods. Absolute mortality and rates of unfavourable outcomes for survivors was reduced by implementation of the PNCP (table 1). Notably, the proportion of children discharged home with or without rehabilitation treatment increased, suggesting that increased survival was accompanied by improved functional outcome (table 1). Regression analysis and modelling of GOS as a function of initial injury severity suggested an improvement after PNCP implementation (appendix). We considered the possibility that discharge decision making for children with similar function was biased by PNCP implementation or other factors differing between the two periods; however, the Mantel test showed a consistent relation between discharge disposition and GOS (appendix). We also considered the possibility that improved mortality reflected a secular trend. A national sample of 1193 children with severe traumatic brain injury extracted from the VPS database estimated that annual mortality in 2002–11 was 25% (SD 7), and increased over time (p=0·0012; appendix). Table 2 shows the results of the ordered probit model; p values did not all need to be less than 0·05 to show overall model quality. For numerical stability, software packages fix one threshold near zero on the latent dimension (in this case the category of different acute care hospital to inpatient rehabilitation facility). Variables reliably associated with increased probability of unfavourable outcomes (negative β) were older age, higher PRISM III scores, and higher injury severity score. Alternatively, white race, use of intracranial pressure monitoring, and high GCS scores after resuscitation were associated with increased probability of favourable outcomes (positive β). Coefficients for PICU length of stay, sex, and mechanism of injury were not statistically reliable. With regard to overall model fit, the difference in the residual deviance and the null deviance (from a model without explanatory variables) showed substantial improvement in fit (χ² tail=1·7831×10–05 on 12 degrees of freedom). The key result in table 2 is the positive coefficient for group, providing evidence that implementation of the PNCP increased the probability of positive categorical outcomes. To show the strength of this relation, we can www.thelancet.com/neurology Vol 12 January 2013
Articles
www.thelancet.com/neurology Vol 12 January 2013
A
Before After PNCP PNCP
0·6
Discharge destination probability
0·5
Home without assistance Inpatient rehabilitation facility Home with outpatient rehabilitation Medical examiner or morgue Different acute care hospital
0·4
0·3
0·2
0·1
0 0
5
10
15
20
25
30
35
PRISM III
B 0·6
0·5 Discharge destination probability
consider a white male patient who had a motor vehicle accident, with all other explanatory variables set at the data mean. For such a patient, the probability of death declined after initiation of the PNCP from 21·0% to 9·9% and the probability for discharge to home without assistance increased from 10·1% to 21·4% (appendix). Overall, predicted outcome probabilities after initiation of the PNCP improved across the spectrum of initial injury (figure 2). Because only one patient was in the category of “home with health care”, we have insufficient information to produce a prediction line across PRISM III or GCS for graphical purposes. Therefore, figure 2 provides predictive lines for five categories only. Specific improvement varied by injury severity measured by PRISM III, creating a curvilinear effect for the category of “inpatient rehabilitation facility” (figure 2). After PNCP implementation, the probability of this outcome decreased for children with lower PRISM III scores, but increased for more severely ill children; this effect seems centred between PRISM III scores of 15–20. Several findings at this injury level suggest improved system performance after implementation of the PNCP. First, probabilities for “home without assistance” and “medical examiner or morgue” equalised, shown by the level of injury at which the curves cross; after PNCP implementation, the PRISM III threshold for this balance point increased from 16 to 21. Second, for children with less severe PRISM III scores (scores <15), the PNCP shifted outcomes from “inpatient rehabilitation” to “home without assistance”. Third, for children with more severe PRISM III scores (scores >20), PNCP redistributed probability from “medical examiner or morgue” to all other outcomes; the magnitude of this effect progressed with increasing PRISM III scores. From regression approaches, implementation of PNCP was not reliably associated with improvements in GOS score at hospital discharge (p=0·1023). However, after implementation of the PNCP, the model predicted more favourable GOS values across the full spectrum of categories of GCS after resuscitation (appendix). When GCS category after resuscitation was treated as an input variable and discharge GOS was treated as an output variable to assess care team effectiveness, PNCP deployment tended to improve GOS output per GCS input, with or without adjustment for covariates identified in our model (appendix). Notably, the difference between initial injury severity and outcome for the two periods increased as GCS scores after resuscitation worsened, reinforcing the aforementioned finding that PNCP deployment improved outcome most for the most severely injured. We assessed the influence of the PNCP on behaviour of the care team by analysing the initiation and maintenance of invasive intracranial pressure monitoring and the intensity of therapies addressing intracranial hypertension (PILOT scale). After initiation of the PNCP,
0·4
0·3
0·2
0·1
0 8
7
6 5 GCS after resuscitation
4
3
Figure 2: Outcome probabilities by injury severity, by PRISM III (A) and GCS (B) scores We created these figures by using the fitted ordered probit model to predict each of the five outcomes for values along the full range of initial injury severity, before and after implementation of the PNCP. Outcomes after implementation improved comprehensively across the full range of PRISM III and GCS severities: note the robust, favourable effect of the PNCP on probability for “medical examiner or morgue’ and ‘home without assistance’ outcome categories (A), and the striking increase after implementation of the PNCP in probability for ‘home without assistance’ and diminished probability for ‘inpatient rehabilitation’ (B). Outcome trends in (B) appear more linear than in (A); this data-driven linear relation between initial GCS and functionally influenced outcomes lends credence to the broad superiority of the predictions after PNCP implementation. PRISM=paediatric risk of mortality score. GCS=Glasgow coma scale.
intracranial pressure monitoring was started earlier and maintained for a greater duration than it was before the PNCP was used (figure 3). Moreover, the intensity of therapies directed against intracranial hypertension was reliably higher on PICU days 1, 2, and 3 after 49
Articles
Probability of ICP monitoring
A
1·00
After PNCP Before PNCP
0·75 0·50
Discussion
0·25 0 0
Number at risk After PNCP 60 Before PNCP 63
B 10
1
2
3
4
5 6 PICU day
57 57
55 52
48 41
46 31
39 26
34 24
7 32 20
8
9
10
11
28 17
25 12
20 9
18 4
p=0·0298
Before PNCP After PNCP
Mean pilot scale
p=0·0292 8
p=0·0471
p=0·1346
6
p=0·1678
p=0·4016
p=0·0698
4 2 0
Day 1
Day 2
Day 3
Day 4
Day 5
Day 6
C
Day 7 Before PNCP After PNCP
20
Pilot scale
15
10
5
0
Day 1
Day 2
Day 3
Day 4
Day 5
Day 6
Day 7
Figure 3: Use of invasive neuromonitoring and intensity of ICP-directed therapies Kaplan-Meier estimates of probability of ICP monitoring before and after establishment of the PNCP (Cox proportional hazards likelihood ratio p=0·03); vertical bars show one standard error for probability of ICP monitoring (A). Daily PILOT scores shown as means and p values (B) and boxplots of median (IQR) and whiskers of 1·5 times the length of the box (C). Therapeutic intensity increased after PNCP when p<0·05. ICP=intracranial pressure. PNCP=paediatric neurocritical care programme.
implementation of the PNCP than before implementation (figure 3). Notably, decompressive craniectomy for increased intracranial pressure was included in the PILOT score (appendix). This treatment was done for one child treated in the PICU before PNCP was implemented and seven children after implementation, all of whom survived. All these children underwent decompression on day 1, apart from two patients in the later time period who underwent decompression on day 2. The patient treated before implementation of PNCP was discharged home without assistance, and four out of the seven children treated after implementation of PNCP were discharged home with or without outpatient care (appendix). Length of stay in the PICU increased after implementation of the PNCP (p=0·01; table 1). However, 50
when the mortality-attributable reduction in stay was controlled, PICU-free days during the first 28 hospital days were not reliably different (p=0·195). Furthermore, overall hospital length of stay did not differ between the two periods (p=0·10).
Mortality for children who have had severe traumatic brain injury ranges between 9% and 25%.7,8,14,15 No outcome improvements have been reported in the past few decades and most survivors have long-term intellectual deficits and reduced quality of life. Our data show that outcomes might be improved by changing the process of care in a way that stably implements a cooperative programme of accepted best practice (panel). Our primary outcome was categorised disposition at hospital discharge, a measure with known correlations to functional neurological outcomes.21–23 Although we initially considered use of a dichotomous survival model, because of the small number of fatalities and our desire to discriminate within the full spectrum of a meaningful outcome range, we abandoned this approach. We therefore analysed two scalable variables: categorised discharge disposition and GOS at hospital discharge. Although these variables share content, they enabled assessment of the PNCP in different ways. Disposition proved to be more robust for modelling, because this measure is more categorically distinct and qualitatively informed than is GOS, and thus better shows overall wellbeing. GOS more narrowly shows neurological recovery than does disposition and, as such, was used to assess limitations of secondary brain injury as a function of primary injury severity. Although discharge GOS was the best means available to retrospectively estimate functional neurological recovery, its predictive validity has been shown only when outpatients are prospectively scored at a timepoint remote from discharge.9,24 Although discharge disposition improved after PNCP was implemented at our centre, several variables affected outcomes in both periods. Increased age, PRISM III score, and injury severity score at entry to the PICU were associated with unfavourable outcome, as reported elsewhere.7,25 White race, increased post-resuscitation GCS score, and use of intracranial pressure monitoring were associated with favourable outcome, also as reported elsewhere.26–28 Notably, patients treated before and after implementation did not differ demographically or in terms of injury severity (generally and neurologically). Despite limitations related to sedative administration for GCS scoring and concern for leadtime bias with PRISM III scoring, both strategies have been used reliably to measure severity and robustly predict outcome for critically ill children.6,12 Our finding of PNCP effectiveness reinforces other evidence suggesting that traumatic brain injury outcome www.thelancet.com/neurology Vol 12 January 2013
Articles
is improved by timely, integrated medical and surgical interventions designed to mitigate secondary brain insults.29 Use of the PNCP in our study improved timing and intensity of intracranial hypertension-targeted therapies. Whereas most studies of guidelines directed at targeting improvements in intracranial pressure in adults are favourable, increased intensity of therapy does not universally improve outcomes.30 Although our PILOT data are important, several changes probably led to our findings. The PNCP provided timely, multisystem monitoring of factors affecting cerebral oxygen delivery and consumption, and allowed integration of aggressive medical and surgical interventions in a tiered plan designed to reduce secondary brain injury;15,27 both changes to our approach were achieved by repeated education, attention to interdisciplinary collaboration, and continuous quality improvement. The influence of PNCP deployment is also consistent with the general observation that organisational patterns in intensive-care units affect outcomes.31 Specifically, disease-focused teams dedicated to explicit management approaches improve outcomes after trauma, acute lung injury, severe sepsis, and acute myocardial infarction.31,32 Moreover, implementation of neurocritical-care teams in adult intensive-care units improve outcomes after stroke, intracerebral haemorrhage, subarachnoid haemorrhage, and traumatic brain injury.29,33 Specialised services have been described for critically ill children and neonates with neurological diagnoses.19,20 This care model exploits evidence-based best practice guidelines and complementary expertise inherent to multidisciplinary teams, which achieve synergy as members work concurrently, rather than serially, on problems.34 Although such guidelines are composed of specific interventions (which individually are often supported by weak evidence), application of complementary interventions by expert teams seems to achieve outcomes exceeding the efficacy predicted when individual guideline components are studied alone.35 Moreover, stable application of integrated algorithms in intensive-care units needs a comprehensive implementation programme (with coordinated guideline rollout, staff education, compliance monitoring, and quality improvement cycles).36 Our single-centre study was limited by its small sample size, use of short-term outcome measures, retrospective design, slow patient accrual, and exclusion of patients with abusive head trauma. However, considering these issues, we designed a robust model and analysis plan that enabled us to report a meaningfully improved outcome for children with severe traumatic brain injury; the magnitude of this effect progressed with initial injury severity and was associated with timely, aggressive application of integrated medical and surgical interventions addressing intracranial hypertension. These outcomes, compellingly linked to programmatic (PNCP) change in intensive-care units, emphasise the need for novel www.thelancet.com/neurology Vol 12 January 2013
Panel: Research in context Systematic review We searched PubMed for articles published without language restriction up to October, 2012, with the following MeSH terms: “brain injuries”[MeSH terms] OR (“brain”[all fields] AND “injuries”[all fields]) OR “brain injuries”[all fields] OR (“traumatic”[all fields] AND “brain”[all fields] AND “injury”[all fields]) OR “traumatic brain injury”[all fields] AND (“guideline”[publication type] OR “guidelines as topic” [MeSH terms] OR “guidelines” [all fields]) AND outcome[all fields]; traumatic brain injury guidelines[all fields] AND neurocritical care[all fields] AND (Clinical Trial[ptyp]). The search was repeated adding the terms “children” and “pediatric”. Interpretation We identified no clinical trials or meta-analyses comparing implementation strategies of guidelines for severe traumatic brain injury in children. Several studies in adults16 support aggressive management and implementation of guidelines for traumatic brain injury to improve outcome in adults with severe traumatic brain injury. However, the substantial variability that exists in practice restricts favourable effects of guidelines. We identified two reviews17,18 in children; although guideline publication was associated with change in relevant clinical endpoints (such as hypocapnia), effects on patient functional outcomes were lacking. Two paediatric studies19,20 described implementation of a neurocritical care service for children and neonates, but the effect on team behaviour or clinical outcomes was not reported. Our study is the first to report an association between implementation of traumatic brain injury guidelines driven by a paediatric neurocritical care programme and a beneficial effect on team behaviour and clinical outcomes.
“T3” implementation research methodologies37 designed to assess and improve four factors: processes for crafting best practice algorithms, optimum care team composition and communication, algorithm efficacy and cost-effectiveness, and implementation and knowledge transfer methodologies specific to the intensive-care unit. Contributors JAP, JRL, IGM, DDL, MSK, JG, and AD designed the study. JAP, JRL, JG, and AD did the data analysis. JAP, JRL, MN, DDL, MSK, JG, and AD interpreted the data and wrote the report. JAP, JG, and AD drew the figures. IGM collected data. Conflicts of interest We declare that we have no conflicts of interest. Acknowledgments Tina Day contributed to data collection; Barbara Miller contributed to traumatic brain injury pathway development, implementation, and monitoring, and data collection; Daniel Martin contributed to manuscript editing (all Washington University School of Medicine, St Louis, MO, USA); all received salary support from the paediatric neurocritical care programme. We gratefully acknowledge the outstanding care delivered by the nurses who staff the St Louis Children’s Hospital paediatric intensive-care unit. References 1 US Centers for Disease Control and Prevention. http://www.cdc. gov/TraumaticBrainInjury/ (accessed March 29, 2012). 2 Scaife ER, Statler KD. Traumatic brain injury: preferred methods and targets for resuscitation. Curr Opin Pediatr 2010; 22: 339–45. 3 Potoka DA, Schall LC, Gardner MJ, Stafford PW, Peitzman AB, Ford HR. Impact of pediatric trauma centers on mortality in a statewide system. J Trauma 2000; 49: 237–45. 4 Guidelines for the acute medical management of severe traumatic brain injury in infants, children, and adolescents. Crit Care Med 2003; 31 (suppl): S407–91.
51
Articles
5
6 7
8
9
10 11 12 13
14
15
16
17
18
19 20 21
22
52
Dean NP, Boslaugh S, Adelson PD, Pineda JA, Leonard JR. Physician agreement with evidence-based recommendations for the treatment of severe traumatic brain injury in children. J Neurosurg 2007; 107 (suppl): 387–91. Maas AI, Harrison-Felix CL, Menon D, et al. Standardizing data collection in traumatic brain injury. J Neurotrauma 2011; 28: 177–87. Ducrocq SC, Meyer PG, Orliaguet GA, et al. Epidemiology and early predictive factors of mortality and outcome in children with traumatic severe brain injury: experience of a French pediatric trauma center. Pediatr Crit Care Med 2006; 7: 461–67. Figaji AA, Zwane E, Thompson C, et al. Brain tissue oxygen tension monitoring in pediatric severe traumatic brain injury. Part 1: relationship with outcome. Childs Nerv Syst 2009; 25: 1325–33. McCauley SR, Wilde EA, Anderson VA, et al. Recommendations for the use of common outcome measures in pediatric traumatic brain injury research. J Neurotrauma 2012; 29: 678–705. Wetzel RC, Sachedeva R, Rice TB. Are all ICUs the same? Paediatr Anaesth 2011; 21: 787–93. American College of Surgeons. National trauma data bank. http:// www.facs.org/trauma/ntdb/index.html (accessed May 30, 2012). Pollack MM, Patel KM, Ruttimann UE. PRISM III: an updated pediatric risk of mortality score. Crit Care Med 1996; 24: 743–52. Shore PM, Hand LL, Roy L, Trivedi P, Kochanek PM, Adelson PD. Reliability and validity of the Pediatric Intensity Level of Therapy (PILOT) scale: a measure of the use of intracranial pressure-directed therapies. Crit Care Med 2006; 34: 1981–87. Tasker RC, Fleming TJ, Young AE, Morris KP, Parslow RC. Severe head injury in children: intensive care unit activity and mortality in England and Wales. Br J Neurosurg 2011; 25: 68–77. Chiaretti A, Piastra M, Pulitano S, et al. Prognostic factors and outcome of children with severe head injury: an 8-year experience. Childs Nerv Syst 2002; 18: 129–36. Stein SC, Georgoff P, Meghan S, Mirza KL, El Falaky OM. Relationship of aggressive monitoring and treatment to improved outcomes in severe traumatic brain injury. J Neurosurg 2010; 112: 1105–12. Guerguerian AM, Lo TY, Hutchison JS. Clinical management and functional neuromonitoring in traumatic brain injury in children. Curr Opin Pediatr 2009; 21: 737–44. Stocchetti N, Conte V, Ghisoni L, Canavesi K, Zanaboni C. Traumatic brain injury in pediatric patients. Minerva Anestesiol 2010; 76: 1052–59. Bell MJ, Carpenter J, Au AK, et al. Development of a pediatric neurocritical care service. Neurocrit Care 2009; 10: 4–10. Glass HC, Bonifacio SL, Peloquin S, et al. Neurocritical care for neonates. Neurocrit Care 2010; 12: 421–29. Shi J, Xiang H, Wheeler K, et al. Costs, mortality likelihood and outcomes of hospitalized US children with traumatic brain injuries. Brain Inj 2009; 23: 602–11. Bulger EM, Nathens AB, Rivara FP, Moore M, MacKenzie EJ, Jurkovich GJ. Management of severe head injury: institutional variations in care and effect on outcome. Crit Care Med 2002; 30: 1870–76.
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
Keller JM, Sciadini MF, Sinclair E, O’Toole RV. Geriatric trauma: demographics, injuries, and mortality. J Orthop Trauma 2012; 26: e161–65. Beers SR, Wisniewski SR, Garcia-Filion P, et al. Validity of a pediatric version of the Glasgow Outcome Scale-Extended. J Neurotrauma 2012; 29: 1126–39. Crowe LM, Catroppa C, Babl FE, Rosenfeld JV, Anderson V. Timing of traumatic brain injury in childhood and intellectual outcome. J Pediatr Psychol 2012; 37: 745–54. Forsyth RJ, Parslow RC, Tasker RC, Hawley CA, Morris KP. Prediction of raised intracranial pressure complicating severe traumatic brain injury in children: implications for trial design. Pediatr Crit Care Med 2008; 9: 8–14. Salorio CF, Slomine BS, Guerguerian AM, et al. Intensive care unit variables and outcome after pediatric traumatic brain injury: a retrospective study of survivors. Pediatr Crit Care Med 2008; 9: 47–53. Haider AH, Efron DT, Haut ER, DiRusso SM, Sullivan T, Cornwell EE 3rd. Black children experience worse clinical and functional outcomes after traumatic brain injury: an analysis of the National Pediatric Trauma Registry. J Trauma 2007; 62: 1259–62. Fakhry SM, Trask AL, Waller MA, Watts DD. Management of brain-injured patients by an evidence-based medicine protocol improves outcomes and decreases hospital charges. J Trauma 2004; 56: 492–99. Cremer OL, van Dijk GW, van Wensen E, et al. Effect of intracranial pressure monitoring and targeted intensive care on functional outcome after severe head injury. Crit Care Med 2005; 33: 2207–13. Randolph AG, Pronovost P. Reorganizing the delivery of intensive care could improve efficiency and save lives. J Eval Clin Pract 2002; 8: 1–8. Cuschieri J, Johnson JL, Sperry J, et al. Benchmarking outcomes in the critically injured trauma patient and the effect of implementing standard operating procedures. Ann Surg 2012; 255: 993–99. Knopf L, Staff I, Gomes J, McCullough L. Impact of a neurointensivist on outcomes in critically ill stroke patients. Neurocrit Care 2012; 16: 63–71. Sutton G, Liao J, Jimmieson NL, Restubog SL. Measuring multidisciplinary team effectiveness in a ward-based healthcare setting: development of the team functioning assessment tool. J Healthc Qual 2011; 33: 10–23. Logani S, Green A, Gasperino J. Benefits of high-intensity intensive care unit physician staffing under the affordable care act. Crit Care Res Pract 2011; 2011: 170814. Davis DA, Taylor-Vaisey A. Translating guidelines into practice. A systematic review of theoretic concepts, practical experience and research evidence in the adoption of clinical practice guidelines. Can Urol Assoc J 1997; 157: 408–16. Dougherty D, Conway PH. The “3T’s” road map to transform US health care: the “how” of high-quality care. JAMA 2008; 299: 2319–21.
www.thelancet.com/neurology Vol 12 January 2013