Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
Pediatric Injury Surveillance
ดร.นพ.ศักดา อาจองค์ วลัลภิากร, MD, PhD
MSIT,MA.(Information Science), BBA
CSIP(Child Safety & Injury Prevention) & TTPI(Thai Task Fore of Pediatric Injury)
Section for Clinical Epidemiology and Biostatistics
Ramathhibodi Hospital, Mahidol University
What is IS(Injury Surveillance) ?
• Etiology of injury Model
• Sorting of reason for contact code
Previous event Previous event Injury eventMechanism of
injury Injury
“Prevention action”
All ContactsReason for
Contact Code
Non Injury
(Disease)
Accidents, Violence, Self-harm
Place of occurrence
Mechanism
Activity Code
Transport
Accidents
Industry Sports Violence Self-harm Vehicles Accidents
Products or object related injuries
NOMESCO 1997
ICD 10 Integrated Concept
• 3 Aspects
– Separate contacts of injuries from contacts due
to diseases (e.g.. poisoning)
– Correct follow these answers
• Where (location/place ?) did the injury occur ?
• How did the injury occur ?
• What was activity(of victim) at the time of injury ?
• Which product (s) were involved in the accident or injury
process ?
– Provide a closer description of transport injury,
vehicle accident, occupational accidents, case of
violence, and self inflicted injury
NOMESCO 1997
NEISS
CHIRPP
• Place of Occurrence Code (Site Code)
• Mechanism of Injury Code
• Activity Code
– 2 dimensions : Purpose of victim x Pattern of movement at
moment of injury
• Transport Accident Classification Module (Several sub-dimension)
Transport-on land, sea, air and space
Transport sub-dimension-vehicle, railway, air- transport, transport
on animal
• Vehicle Accident classification Module
• Occupational/Industrial classification module
• Sport Activity classification Module
• Violence classification Module (perpetrators, counterparts and
specific of weather the act of violence etc)
• Self-inflicted Injury classification module
• Product Classification ****
NOMESCO
NOMESCO 1997
• ICD 10 Classification
Principle of diseases/injury + External causes
(Chap 20 V01-Y36) BR, AIS 85 (Chap 19)
• NOMESCO Classification of external causes of
injury » Reason of contact
» Place of occurrence
» Mechanism of injury
» Activity at time of injury
» Transport accident
» Occupational accident
» Sport accident
» Violence, Self harm
» Product involved
What is IS (Injury Surveillance) ?
NOMESCO 1997
IS in Thailand• ส านักระบาดวทิยาคลนิิก
http://www.ismis.wedev4u.com
• Rama PedISS
http://dti.mahidol.ac.th
ICD10 +External Cause <> NOMESCO
Interface Projects + Add on prompt
injury variables
http://www.ismis.wedev4u.com
Click to show a touring inside Rama PedISS Web application
Nation Registry of Thai Injury surveillance
Children Adult
Methodology Data Collection and Management
• Web-databases – Using PHP version 5.2.9 and MySQL client version 5.0.51a.
– The data from individual trauma care centers were real-time entered via the web-databases.
– The data qualities control program was constructed as a web-based application based on code of variables, feasibility, and cross-check in order to verify and validate data. All of these data were double check by database administration team
• Data cleaning and checking – DB are summarizing and cross-tabulate between the relevant
variables to check for completeness and data validation.
– The inquiries were made if there was any incorrect or missing of data. Medical records from each hospital was then retrieved to check and confirm data.
– Some of collaborated hospitals were visited, if they have a problem of delay of data flow & a lot of data missing.
MethodologyOutcome & Independent Variables
• The outcome of interest was death related to injury
within 30 days.
• The independent variables of 5 domains as follows:
– Demographic and general data consist of age, sex, weight, height,
occupational, and regions.
– Pre-hospital data consist of transport duration, prior communication,
and trauma care level
– Mechanism of injury and injury body regions that consist of velocity-
gravity related injury, injury mechanism (penetrating, blunt, both and
non-classify), sites of injury and injury body regions (brain-head and
neck, face, thoracic, abdomen-pelvis, musculoskeletal , and external
soft tissue region), wound types, fracture types and theirs severity.
– Airway management was categorized as no intervention, airway
adjuncts which included to oxygen supplement, positive ambulatory
bag, face mask ventilation and adjuncts airway with oropharyngeal or
nasopharyngeal airway, and intubation.
– GCS and vital signs domain (SBP, PR, and RR)
At least directed
Related outcome
Prediction Score
For Thai Children
Pediatric Injury
DB Registry
Descriptive
Data
Results
Results1.Base line characteristics
• During the study period, data of 43,561 injured
children who attended at emergency services of
34 studied hospitals were entered and retrieved
from the databases.
• 71 % were males, mean age was 11.4 + 5.5
years, and weight was 45, (7-76) kg
• From trauma level Care (N, %)
I 20,492 (47.0 %)
II 12,441 (28.6 %)
III 7,220 (17.6 %)
IV 3,408 (7.8 %)
Results1.Base line characteristics
• 92 % of patients injured mainly within their
residential areas,
• 39 % of cases were transferred by ambulance,
and the rests were transferred by non-
ambulance and their own transportations.
• 47 % had prior communication with the referral
hospital before transportation of patients.
• 51 % of patients did not need the initial first aid
at the scene of injury. Among 49% patients who
required the first aids, 87 % of them were
provided initial first aid
Results1.Base line characteristics
• Regions (N, %)
Bangkok 2,832 (6.50) Central 7,529 (17.28)
North 3,430 (7.87) North-East 13,382 (30.72)
East 4,638 (10.65) South 11,750 (26.97)
• Injury types (N, %)
Transportation 19,928 (45.75) *
Falling 7,902 (18.14) **
Poisoning 461 (1.06)
Animal Bite &Sting 1,641 (3.77)
Struck by or against 3,426 (7.86) ****
Cut & pierce 3,502 (8.04) ***
Burn & Scald 1,005 (2.31)
Fire gun & explosion 1,582 (3.63)
FB Aspiration & suffocation 1,359 (3.12)
Drowning & Submersion 355 (0.81)
Abuse, Assault & neglect 2,400 (5.51) *****
Results1.Base line characteristics
• Mechanism of injury, N (%)
Blunt 31,482 (72.27)
Penetrate 5,940 (13.64)
Both 1,943 (4.46)
Non Classify 4,196 (9.63)
• Product/Object related injury , N (%)
Chemical &Food product 801 (1.84)
Home & Office, Work place 12,730 (29.22) **
Sport equipment 639 (1.47)
Weapons 2,047 (4.70)
Transportation related 20,317 (46.64) *
Natural objects - animal 3,362 (7.72) ***
Miscellaneous 3,664 (8.41) ***
Results1.Base line characteristics
• Among the injury classification, Top 5 of common injuries were
Transportation 46 % Falling 18 % Cut and pierce 8 % Struck by/ or against 8 %Abuse, assault & neglect 6 %
• Causes of death was highest in Drowning 8.0 %Weapon, fire-gun, bomb-explosion injury 2.6 % Transportation 2.4 %Burn and scald 2.3 %Poisoning 1.3 %
• The overall death rate was 1.7 % (95% CI: 1.57-1.82)
• Death rate was highest in the East region (2.41 %, 95% CI: 1.97-2.85) but it was lowest in Bangkok (0.78 %, 95% CI: 0.45-1.10).
Results
• All of variables Univariate Analysis found 20 variables Colinearity checked 10 variables Multivariate analysis suggest equation
• Logistic Equation ;
𝐼𝑛𝑃
1 − 𝑃= −7.82 + 0.65 𝑥 𝐴𝑔𝑒 ≤ 5 𝑦𝑟𝑠 + 1.09 𝑥 𝐴𝑔𝑒 6 − 12 𝑦𝑟𝑠+ 1.20 𝑥 𝐴𝑑𝑗𝑢𝑛𝑐𝑡𝑠 𝐴𝑖𝑟𝑤𝑎𝑦 + 2.39 𝑥 𝐼𝑛𝑡𝑢𝑏𝑎𝑡𝑖𝑜𝑛+ 0.24 𝑥 𝑛𝑜𝑛𝑒 𝑉𝐺 𝑟𝑒𝑙𝑎𝑡𝑒𝑑 𝑚𝑒𝑐ℎ𝑎𝑛𝑖𝑠𝑚+ 0.71 𝑥 𝑃𝑢𝑟𝑒 𝑉𝑒𝑙𝑜𝑐𝑖𝑡𝑦 𝑟𝑒𝑙𝑎𝑡𝑒𝑑 𝑚𝑒𝑐ℎ𝑎𝑛𝑖𝑠𝑚+ 0.36 𝑥 𝑃𝑢𝑟𝑒 𝐺𝑟𝑎𝑣𝑖𝑡𝑦 𝑟𝑒𝑙𝑎𝑡𝑒𝑑 𝑚𝑒𝑐ℎ𝑎𝑛𝑖𝑠𝑚+ 1.61 𝑥 𝐻𝑒𝑎𝑑 − 𝑛𝑒𝑐𝑘 𝑖𝑛𝑗𝑢𝑟𝑦 + 1.52 𝑥 𝑇ℎ𝑜𝑟𝑎𝑐𝑖𝑐 𝑖𝑛𝑗𝑢𝑟𝑦
+ 1.62 𝑥 𝐴𝑏𝑑𝑜𝑚𝑖𝑛𝑎𝑙 − 𝑃𝑒𝑙𝑣𝑖𝑠 𝑖𝑛𝑗𝑢𝑟𝑦 + 1.40 𝑥 (GCS < 9)+ 1.61 𝑥 𝐴𝑏𝑛𝑜𝑟𝑚𝑎𝑙 𝑆𝐵𝑃 ∗ + 0.80 𝑥 𝑇𝑎𝑐ℎ𝑦𝑐𝑎𝑟𝑑𝑖𝑎 ∗+ 2.42 𝑥 𝐵𝑟𝑎𝑑𝑦𝑐𝑎𝑟𝑑𝑖𝑎 ∗ + 0.79 𝑥 (𝐷𝑦𝑠𝑝𝑛𝑒𝑎 ∗)
*Abnormal vital signs : Systolic blood pressure(SBP), Pulse rate(PR; tachy- bradycardia), Respiratory rate(RR; Dyspnea) reference abnormality cut off values from Pediatric Advanced Life Support(PALS), American Heart
Association(AHA), 2010)
• 10 Significant variables : Age, Airway, Physical mechanism, Head-neck injury, Thoracic injury, Abdominal-pelvis injury, GCS, SBP, PR, RR
ResultsDerivation Phase
• 15 variables (age, sex, weight, airway intervention, velocity-gravity related mechanism, mechanism of injury, trauma body regions (head-neck injury, thorax injury, abdomen-pelvis injury), wound types, fracture types, GCS, SBP, PR, RR) were thus simultaneously included in the multivariate logit model
• 10 variables were significant and thus kept in the final model, The performances of risk score was evaluated by explore calibration and discrimination properties. The goodness of fit of the mode was assessed and found it fitted well with our data (Hosmer-Lemeshow Chi square = 13.64, d.f.= 5 p =0.092).
• The model was also well in discriminate death from survive subjects with the C-statistic of 0.938 (95 % CI: 0.929-0.947)
Table 4. , Figure 1 and The logit equation was described in the appendix I.
Figure 1- Receiver-operator characteristic (ROC) curve (plotting sensitivity versus 1-specificity) for
predicting the presence of a death outcome according to point of scores in the derivation data set. By
using different thresholds for point of score, which maximize sensitivity or specificity, and Area under
the curve= 0.9376
0.0
00.2
50.5
00.7
51.0
0
Sen
sitiv
ity
0.00 0.25 0.50 0.75 1.001 - Specificity
Area under ROC curve = 0.9376
Calibration performance
GOF : Hosmer-Lemeshow Chi square = 13.64,
d.f., = 5, p-value =0.092
O/E ratio : 0.86 (95%CI: 0.70-1.02)
Discrimination performance
C-statistic of 0.938 (95 % CI: 0.929-0.947)
Thanks You for
Your Attentions
Acknowledgement
• Child Safety Promotion and Injury Prevention Research Center
(CSIP)/ Safe Kids Thailand.
• Staff of Section of Clinical Epidemiology and Biostatistics,
Faculty of Medicine, Ramathibodi Hospital, Mahidol University.
• Members and committee of Thai Taskforce of Pediatric Injury,
Thailand (34 nationwide collaborating Hospital).
• Bureau of Epidemiology, Non-Communicable Diseases Ministry
of Public Health (MOPH), Thailand.
• Co-operation Research funding supported by Thai Research
Funding (TRF), and Mahidol University, Thailand,
• Co-operation Research funding supported by Thai Health
Promotion Foundation (ThaiHealth).
How about height & weight
And Temperature ?
Pediatric intravenous contrast material reaction scenario.
Gaca A M et al. Radiology 2007;245:236-244
©2007 by Radiological Society of North America
Summary of Important Factors
• Age
• Weight/Height * need in future
• Airway : Intubation vs. Adjunct Airway
• Conscious level (GCS, AVPU )
• Physical Mechanism (V-G Related)
• Injury Sites (1, 2 , > 3 sites)
• Brain-Spinal Cord- Thoracic-Abdominal –injured regions
• Hct *
• VS : SBP * PR *- RR
• Type of Injuries
• Object Related Injury
• High Risk of Injury : Car Accident(Inside/Outsite)
IS MOPH-RamaPedISS
Pre-lim Analysis by Multiple Logistic Model
ResultsDerivation Phase
For demographic domain
• Only age was significant associated with death by
younger subjects were poorer prognoses than
older subjects.
• This suggested that patients aged 1-5 and 6-12
years were 1.9 (95 % CI: 1.4-2.6) and 3.0 times
(95% CI:2.0-4.3) respectively higher risk of death
than patients aged 13-19 years.
ResultsDerivation Phase
Airway management domain
• Airway management affected on patient
survival, subjects who were intubated were
about 10.9 (95 % CI: 8.6-13.7) higher odds of
death than patients who were not intubated.
• An injured subjects with adjunct airway and
support ventilation were about 3.3 (95 % CI:
2.4-4.6) times higher odds of death than
subjects who did not require the airway
management.
ResultsDerivation Phase
The mechanism and region of injury domains
• Physical mechanism related injury (velocity and gravity forces) and injured region. For physical mechanism, gravity, velocity, and both related injuries were 2.0 (95 % CI: 1.4-3.0), 1.3 (95% CI: 1.0-1.7), 1.4 (95% CI: 1.1-1.9) respectively compared to none velocity and gravity related injuries.
• Among injured regions, injured at head and abdomen were the most affected on death followed by thorax with the ORs of 5.0 (95% CI: 4.1-6.1), 5.0 (95% CI: 4.0-6.5) and 4.6 (95% CI: 3.5-6.0).
ResultsDerivation Phase
GCS
• Subjects with GCS less than 9 were about 4.0
(95 % CI: 3.2-5.1) times higher odds of death
than subjects with GCS 9 or higher
ResultsDerivation Phase
Vital signs domain
• PR was significantly associated with death. Injured subjects with bradycardia and tachycardia were respectively 11.3 (95% CI: 7.5-17.0) and 2.2 (95% CI: 1.8-2.8) significantly higher odds of death when compared to subjects with normal PR.
• The injured subjects with abnormal SBP and RR were 5.0 (95 % CI: 3.9-6.4) and 2.2 (95 % CI: 1.5-3.1) times higher odds of death than those subjects with normal range of RR and SBP.
Results
Result
1.Score was classified into 4 categories
By coefficient for risk strategies
2.Sum of coefficient was 0.15 -15.16
3.For easy and simplified of use, we
classified according to its performance
and distribution by LR + test, i.e.,
< 1.02,
1.02 to 1.96,
1.96 to 3.06, and
> 3.06
which represented to:
very low,
low,
intermediate,
and high risk of death.
ResultsScoring Scheme
• The LR+ of these corresponding risk strategies were 1, 1.45
(95% CI: 1.34-1.57), 2.52 (95% CI: 2.09-3.04), and 4.72 (95%
CI: 4.57-4.88), respectively
• The PV+ of these corresponding risk groups were 0.01 % (95
% CI: 0.00-0.07), 0.15 % (95 % CI:0.09-0.22), 0.51 % (95%
CI:0.37-0.69) and 6.73 % (95% CI: 6.20-7.29), respectively.
• The scoring scheme should be easy to apply in practice or
easy help aid by today technologies
ResultsScoring Scheme
Application of scoring scheme; • A child aged 6 years was transferred by the ambulance to the ER of a 30
bed-hospital due to head injury from car accident. His PR was 140/min, RR was only 6/min, SBP was 100 mmHg, and thus he was on endotracheal intubation during transit and the GCS was 8 at ER arrival.
• He was scored
• 1.09 for aged 6 years, 2.39 for intubation, 0.36 for physical mechanism both V-G related injury, 1.61 for head injury 0 for the remaining regions1.40 for in-hospital GCS < 90.8 for PR 140 /min0 for SBP 100 mmHg0.89 for dyspnea, with spontaneous respiration 6/min.
The total score was 8.54, which was classified as high risk of death with the LR+ of 4.72 and PV+ of 6.73 %.
• This patient requires intensive care, aggressive management, and close monitoring on arrival at the hospital.
ResultsComparison of prediction models
• We compared our model to other previous models including Tepas score 1987, Tepas & Ramenofsky core 1988, Rosso score 2012, and pediatric polytrauma scores 2012
• The C-statistics were
Thai-PedISS 2013 0.938 (95% CI 0.929-0.947) ***
Tepas score 1987 0.876 (95% CI 0.862-0.891)
Tepas score 1988 0.876 (95 % CI 0.861-0.891)
Rosso score 2012 0.893 (95% CI 0.879-0.908)
Polytrauma scores 2012 0.874 (95 % CI 0.860-0.888)
• NRI was estimated and suggested that our model could improve in classification of subjects in both death and survival groups. The rate of improvements ranged from
4.4 % to 13.6 % in the death group
2.9 % to 7.4 % in the survival group.
• Overall NRIs were 19.7%, 18.0%, 16.2%, and 1.5 % when we compared ours to the Russo BD (2012), Tepas & Ramenofsky (1988), Tepas (1987), and Polytrauma score (2012), respectively
• Finally demonstrated that our model also gained in discrimination when compared to preexisting models as previously mentioned with the IDI of 0.06 (95% CI: 0.03-0.09) or increasing of 6 % discrimination and robustness from pre-existing scores which have been used.
Figure 2-The demonstration of C-statistics (ROC Area) of Thai-PedISS was statistically significant when compared to
preexisting pediatric injury scores, Tepas (1987), Tepas &Ramenofsky (1988), Rosso DB (2012), Polytrauma score (2012).
Our score
TRISS GOF : not fitted well
AUC 0.88
O/E ratio 0.80 (95 % CI 0.71-0.89)
GOF : fitted well
AUC 0.94
O/E ratio 0.86 (95 % CI 0.70-1.02)
Discussion
• The study was conducted to development of
Risk prediction score of death in Pediatric injury,
Thailand
• The risk prediction score of death was derived and
indicated 10 variables significantly related with death
(i.e.. age, intubation, physical mechanism, injured at
head, abdomen, and thorax, GCS, PR, RR, and SBP).
• The derived model fitted well with the data and also
gave good discrimination of death from survival subjects
with the C statistic of 0.938 (95 % CI 0.929-0.947) and
0.873 (95 % CI 0.872-0.875) in the derive and internal
validate data, respectively.
Discussion• This developed risk prediction score of death in pediatric
injury have demonstrated a good calibration and discrimination performance, superb pre-existing injury scores with these reasons;• Directly specific Children DB and concerned variable to
children
• Suitable of DB matched with Eastern norms, of the developing country with different theirs background of regional geographic, socioeconomic and healthcare status.
• We categorized the standardize variables for children, e.g. pediatric GCS, Vital signs (PALS, AHA 2010), Age ranges, Types of injury etc.
• Number of sites of injury not proper to estimate in children especially generalized to all injury, but may be proper for only accidental injury.
• Thus, Multisite injury was replaced by significant injured body region instead, these variables was automatic working to weight the different region of injuries which are severe.
• From these dominantly reasons support to enhanced of discrimination performance of these score, and improve to more accurate predictions to enable all children injuries to be more realistically assessed than previous injury prediction scores.
Discussion
• Strengthens
• We looked back and focus on more individual variables
(except GCS, which is accepted worldwide)
• Not too many variable, Covered the vital signs as well as
the critical variables in both the physiologic and anatomic
variants in the same system
• Concerned about interaction among injury mechanisms, in
both physical perspective and specific types of mechanism
in the same time.
• The principles of development looked at each single
variable rather than the combined ratings system or ready-
made model which already exists like a combined-scores
development, as which has previously been developed
Discussion
Strengthens• Not need to separately calculate the ISS
or additionally calculate the other injury coefficients of each sub-type of injury outside the model to prevent any mistake of calculation and inconvenience of usage
• The score cut-offs are not sophisticated, they are easy to calculate and provide the proper result to aid decision based on severity and facility of treatment.
• The score easy to use, define the important aspects and be readily available data to support facilitation of the medical decision-making process
• In future should provide the possibility for retrospective evaluation of the quality of pediatric trauma care and to compare result to those obtained at other centers/regions
Discussion
Limitations
• Not yet performed external validation in other general
subjects needs to be assessed in various types of
accidents and injuries in children
• The other important aspects are trauma level care, which is
currently being evaluated for its level of preservation of
injury or accident to levels from I-IV to be universal.
• This trauma level assignment and evaluation is currently in
progress and being evaluated across the country to be able
to tell that when assessed with high severity or mortality
should be forwarded to the higher level
Discussion
Limitations
Transportation duration.
• Thailand lacks standardization of transportation policy,
which should guarantee transport time based on the
severity of the injury. However, this is limited many reasons
such as limitation
of resources, no standard transportation policy, staff
shortages, insufficient emergency medical & ambulance
service for children (EMSC), medical instruments, and
knowledge of medical personnel.
• These are the factors which need to be developed and
studied in the future.
Opportunities of future Studies• Construct the regimes of referring criterions and
integrated with actual assigned trauma care level
to conduct of cluster RCT study (Score’s based
intervention vs. Conventional intervention) with
MOPH to validate outline of transfer of pediatric
injury
• Going on NPIRT(National Pediatric Injury
Registry, Thailand) by process the MOU with
MOPH and transferring model concepts.
• Advanced web application development to
universal platform and Artificial intelligent(AI) ;
Bayesian network and neural network model to
help aid decision making in the future.
Conclusion• The risk prediction score of death in Thai injured
children has been developed and validated using
significant 10 variables (i.e. age, intubation, physical
mechanism, injured at head, abdomen, and thorax,
GCS, PR, RR, and SBP).
• The developed score is superiorly performances
when compare to pre-existing injury scores from
dominant reasoning and proper selected variables as
our previous mentioned.
• These variables are easy to assess and measured
form routine practice. However, the risk model needs
to be externally validated in general Thai population
or outside the country in the future.
Acknowledgement
• Staff of Section of Clinical Epidemiology and Biostatistics,
Faculty of Medicine, Ramathibodi Hospital, Mahidol University.
• Members and committee of Thai Taskforce of Pediatric Injury,
Thailand (34 nationwide collaborating Hospital).
• Bureau of Epidemiology, Non-Communicable Diseases Ministry
of Public Health (MOPH), Thailand.
• Child Safety Promotion and Injury Prevention Research Center
(CSIP)/ Safe Kids Thailand.
• Co-operation Research funding supported by Thai Research
Funding (TRF), and Mahidol University, Thailand,
• Co-operation Research funding supported by Thai Health
Promotion Foundation (ThaiHealth).
Thanks You for
Your Attentions