- Email: [email protected]
Validation of international trauma scoring systems in urban trauma centres in India
G Model JINJ 6913 No. of Pages 6
Injury, Int. J. Care Injured xxx (2016) xxx–xxx
Contents lists available at ScienceDirect
Injury journal homepage: www.elsevier.com/locate/injury
Validation of international trauma scoring systems in urban trauma centres in India Nobhojit Roy, MS MPHa,b,* , Martin Gerdin, MD PhDa , Eric Schneider, PhDc , Deepa K. Kizhakke Veetil, MSd , Monty Khajanchi, DNBb , Vineet Kumar, DNB FNBe , Makhal Lal Saha, MSf , Satish Dharap, MSg , Amit Gupta, MSh , Göran Tomson, MD PhDi, Johan von Schreeb, MD PhDa a
Department of Public Health Sciences, Health Systems and Policy, Karolinska Institutet, Stockholm, Sweden BARC Hospital (Govt of India), HBNI University, Mumbai, India c Center for Surgery and Public Health, Brigham and Women's Hospital, Boston, MA 02120, USA d Department of General Surgery, King Edward Memorial Hospital, Mumbai, India e Department of General Surgery, Lokmanya Tilak Municipal Medical College and General Hospital, Mumbai, India f Department of Surgery, Institute of Post-Graduate Medical Education and Research and Seth Sukhlal Karnani Memorial Hospital, Kolkata, India g Department of Surgery, Lokmanya Tilak Municipal Medical College and General Hospital, Mumbai,India h Department of Surgery, Jai Prakash Narayan Apex Trauma Center, All India Institute of Medical Sciences, New Delhi, India i Department of Learning, Informatics, Management & Ethics (LIME) and Public Health Sciences, Karolinska Institutet, Stockholm, Sweden b
A R T I C L E I N F O
A B S T R A C T
Article history: Accepted 13 September 2016
Introduction: In the Lower-Middle Income Country setting, we validate trauma severity scoring systems, namely Injury Severity Score (ISS), New Injury Severity Scale (NISS) score, the Kampala Trauma Score (KTS), Revised Trauma Score (RTS) score and the TRauma Injury Severity Score (TRISS) using Indian trauma patients. Patients and methods: From 1 September 2013 to 28 February 2015, we conducted a prospective multicentre observational cohort study of trauma patients in four Indian university hospitals, in three megacities, Kolkata, Mumbai and Delhi. All adult patients presenting to the casualty department with a history of injury and who were admitted to inpatient care were included. The primary outcome was inhospital mortality within 30-days of admission. The sensitivity and specificity of each score to predict inpatient mortality within 30 days was assessed by the areas under the receiver operating characteristic curve (AUC). Model fit for the performance of individual scoring systems was accomplished by using the Akaike Information criterion (AIC). Results: In a registry of 8791 adult trauma patients, we had a cohort of 7197 patients eligible for the study. 4091 (56.8%)patients had all five scores available and was the sample for a complete case analysis. Over a 30-day period, the scores (AUC) was TRISS (0.82), RTS (0.81), KTS (0.74), NISS (0.65) and ISS (0.62). RTS was the most parsimonious model with the lowest AIC score. Considering overall mortality, both physiologic scores (RTS, KTS) had better discrimination and goodness-of-fit than ISS or NISS. The ability of all Injury scores to predict early mortality (24 h) was better than late mortality (30 day). Conclusion: On-admission physiological scores outperformed the more expensive anatomy-based ISS and NISS. The retrospective nature of ISS and TRISS score calculations and incomplete imaging in LMICs precludes its use in the casualty department of LMICs. They will remain useful for outcome comparison across trauma centres. Physiological scores like the RTS and KTS will be the practical score to use in casualty departments in the urban Indian setting, to predict early trauma mortality and improve triage. ã 2016 Elsevier Ltd. All rights reserved.
Keywords: India Injury Trauma Urban Injury scoring system TRISS RTS ISS NISS KTS
* Corresponding author at: Department of Public Health Sciences, Health Systems and Policy group, Karolinska Institutet, Widerströmska huset, Tomtebodavägen 18A, 171 77 Stockholm, Sweden. E-mail addresses: [email protected], [email protected] (N. Roy), [email protected] (M. Gerdin), [email protected] (E. Schneider), [email protected] (D.K. Kizhakke Veetil), [email protected] (M. Khajanchi), [email protected] (V. Kumar), [email protected] (M.L. Saha), [email protected] (S. Dharap), [email protected] (A. Gupta), [email protected] (G. Tomson), [email protected] (J. von Schreeb). http://dx.doi.org/10.1016/j.injury.2016.09.027 0020-1383/ã 2016 Elsevier Ltd. All rights reserved.
Please cite this article in press as: N. Roy, et al., Validation of international trauma scoring systems in urban trauma centres in India, Injury (2016), http://dx.doi.org/10.1016/j.injury.2016.09.027
G Model JINJ 6913 No. of Pages 6
2
N. Roy et al. / Injury, Int. J. Care Injured xxx (2016) xxx–xxx
Introduction Lower-middle income countries (LMICs) pay the price of a growing volume of trauma, as collateral damage for development, rapid urbanization and sociodemographic transition [1]. In High Income Countries (HICs), the 30-day fatality criteria (dying within 30 days of injury) as recommended by the Global Road Safety report, is used to compare countries and their commitment to road safety [2]. The current decade of action for road safety (2011– 2020), advocates for a ‘post-crash response’ with appropriate care, including prehospital care, in-hospital care and post-discharge rehabilitation, for individuals who have sustained trauma. Inhospital trauma mortality is proportionally higher in LMICs as compared to HICs [3]. Understanding the factors associated with the observed differences in mortality between patients undergoing in-hospital trauma care in LMIC and HIC presents substantial challenges. The lack of a standard and uniform system for quantification of trauma severity is a drawback for researchers seeking to study systems of trauma care in multiple settings. A standardized measure of trauma severity is needed to ensure a fair and meaningful comparison of outcomes and effectiveness of trauma care among different settings with differing case-mix, within and across countries. Trauma severity scoring systems have been designed to estimate the mortality risk based upon specific combinations of factors associated with patient injuries. The scoring systems used to quantifying injury severity in HIC settings have historically relied upon anatomical measures of derangement while those predominant in LMICs have relied more heavily on measures of physiological derangement which limits comparability across HIC and LMIC trauma centres [4,5]. The anatomy-based scoring systems rely on imaging technologies, such as Computed Axial Tomography (CT) scans, prehospital care information, access to surgical intervention, quality intraoperative documentation and autopsy reports which may be unavailable in many LMIC trauma care facilities, thereby preventing the accurate calculation of Injury severity scores (ISS). Neither consensus-based scoring systems nor data-driven trauma severity scoring systems have been well-researched in the LMIC setting. Therefore, our aim was to validate commonly used trauma severity scoring systems ranging from the purely anatomy-based Injury Severity Score and New Injury Severity Scale (NISS) score, to more physiology-focused scores, including the Kampala Trauma Score (KTS) and the Revised Trauma Score (RTS) score, as well as the combined score TRauma Injury Severity Score (TRISS), both within and across facilities treating substantial numbers of trauma patients in India.
operate trauma-specific units providing trauma services as a part of a general hospital. All facilities were classified as ‘free-to-public’ indicating nominal fees to users facilitating access to care to the lower socio-economic strata of the population. Eligibility criteria All adult patients ( 15 years of age) presenting to the casualty department during the study period with a history of injury associated with a mechanism of road traffic, railway, fall, assault or burns and who were admitted to inpatient care at the index hospital were included. Patients who were dead on arrival were not included. Dependent and independent variables The primary outcome of interest was in-hospital mortality within 30-days of admission [7]. The secondary outcomes were 1. Mortality within 24-h, 2. Mortality between 24 h and 7 days, 3. Mortality between 7 and 30 days. For each included individual, patient-specific demographic factors including age and sex, as well as injury and pre-hospital treatment factors such as transfer status, mode of transport, and mechanism of injury were collected. Physiological measures at the time of initial assessment were recorded, including Systolic Blood Pressure (SBP), Glasgow Coma Scale (GCS) score, respiratory rate (RR), heart rate (HR) and oxygen saturation. Data To ensure consistency in data quality at each participating site, one data collector prospectively gathered data during the initial assessment using a standardized intake form for eight hours per day by directly observing the providers who were delivering trauma care and recording vital signs. This data collector served only as an observer and was not involved in patient care. At each site, the data collector, who was funded externally by TITCO-India, rotated through all 8-h shifts (morning, evening, night), and also on public holidays. For patients admitted outside of the 8-h ‘directly observed’ shift, data were retrieved from patient case records within days of initial presentation. Collected data were uploaded to a central database on a weekly basis and the authors (NR, MG, VK, MK) conducted weekly data review meetings. All data collectors had at minimum a health science master degree and were continuously trained and supervised throughout the study period. Injury severity scores
Patients and methods Study design, context and setting This prospective multi-centre observational cohort study was conducted under the guidance of the collaborative research consortium “Towards improving trauma care outcomes” (TITCOIndia) [6] from 1 September 2013 to 28 February 2015 in four Indian teaching and referral hospitals, each of which operate trauma units that receive citywide referral of trauma patients. The megacities (populations of more than 10 million) were geographically representative of urban India, namely Kolkata, Mumbai (2centres) and Delhi. The centres were the Apex Trauma Centre of the All-India Institute of Medical Sciences (AIIMS), New Delhi; Lokmanya Tilak Municipal General Hospital (LTMGH), Mumbai; KEM hospital, Mumbai; and the Seth Sukhlal Karnani Memorial Hospital (SSKM), Kolkata. The Apex trauma centre in Delhi is a standalone trauma-care facility while the other three centres
Baker et al., in 1974 [8], proposed ISS as an anatomical score that incorporates Abbreviated Injury Scale (AIS) in six body regions. In our study a single surgeon (DKV) accredited by the Association for the Advancement of Automotive Medicine (AAAM) calculated the ISS. The ISS has the advantage of converting all injuries in a trauma patient into a single number. However, though ISS appears to be a continuous variable (0–75), due to its sum of squares calculation formula, some integers are mathematically not possible. The impossible integers are 7, 15, 23, 28, 31, 37, 39, 40, 44, 46, 47, 49, 52, 53, 55, 56, 58, 60–65, 67–74 [9]. The NISS, a modification of ISS, was computed from the three worst injuries, regardless of the body region in which they occurred. The Revised Trauma Score (RTS), is a physiologic score that a patient's systemic response to injury measured through GCS, SBP and RR. It is the current standard physiologic scoring system used in trauma research and quality improvement in both high-income countries and low- and middle-income countries. The TRISS
Please cite this article in press as: N. Roy, et al., Validation of international trauma scoring systems in urban trauma centres in India, Injury (2016), http://dx.doi.org/10.1016/j.injury.2016.09.027
G Model JINJ 6913 No. of Pages 6
N. Roy et al. / Injury, Int. J. Care Injured xxx (2016) xxx–xxx
system combines RTS, ISS, age and type of injury (blunt or penetrating) to estimate the probability of survival (Ps) for each individual patient [10]. The Ps ranges from 0 to 1, where 0 represents a 0% and 1 a 100% probability of survival. The original TRISS coefficients were derived from data submitted by 139 North American hospitals between 1982 and 1987 and was published as the Major Trauma Outcome Study (MTOS) in 1990 [11]. In 2009, an updated version of the TRISS coefficients was published, and this time the coefficients were derived from the US National Trauma Data Bank (NTDB)[12,13]. The Kampala Trauma Score (KTS) is a simplified injury scoring system that uses patient age, systolic blood pressure, respiratory rate, neurologic status and number of serious injuries as variables and was developed in Uganda for use in resource-limited settings [14]. The components of Kampala Trauma Score as described by MacLeod, J.B.A., et al. is available as a Supplementary information online.
Adult t rauma victims admitted at 4 Indian urban University referral hospitals Sept 2013 - Feb 2015 (n=8791) Excluded:
Included: Mechanism of Injury: road traffic, railway, fall, assault or burns
Death status unknown=7 Did not complete 30 in-patient days of follow up =1587
7197 Adult patients in study had outcomes: 1) death or discharge within 30 days
Died n=1754 (24.3%)
Statistical methods For each patient KTS, ISS, NISS, RTS, TRISS and GCS were calculated. The following minor modifications were made to the KTS score so that it could be calculated retrospectively, as per Weeks et al’s work [15]. The number of serious injuries for each patient was determined based on a list of final diagnoses by a member of the research team with expertise in trauma care. Because no standardised conversion from GCS to AVPU (“alert, voice, pain, unresponsive”) score exists, an estimated AVPU score was assigned based on GCS using data from the original validation study of the KTS (GCS 14–15 = “alert”, GCS 10–13 = “responds to voice”, GCS 5–9 = “responds to pain”, GCS 3–4 = “unresponsive”) [16]. A note was made when insufficient data prevented calculation of any of the scores. Because not all values required to compute each score were available for every patient, the primary outcomes reported are the results of a complete-case analysis for the subset of patients for whom each of the scores was available. The differences in means of continuous variables were examined using Student’s t-tests. Non-parametric tests (e.g. Wilcoson rank-sum) were employed when continuous variables were not normally distributed. Pearson’s chi-squared test was used to compare mortality rates among patients who did and did not have sufficient data to calculate each score recorded. Association between injury scoring systems and in-hospital mortality was evaluated using logistic regression. The sensitivity and specificity associated with the ability of each score to predict inpatient mortality within 30 days was assessed by analyzing the areas under the receiver operating characteristic (ROC) curve (AUC). The discrimination of each of the scoring systems (a measure of how well a model distinguishes survivors from non-survivors) was compared using these ROC curves. The AUCs along with 95% confidence intervals (CI) were compared between the injury scores. A sample size of 2350 patients, was calculated based on a difference of 0.05 in the AUC with a reference AUC of 0.75 to compare the ability of Injury severity scores to predict mortality. The confidence interval was set at 95% with 90% power and a 0.05. However, all patients in the available dataset were included for the analysis. Overlapping of confidence intervals was considered to indicate the absence of any statistically significant difference between score discrimination. Model fit for the models used to examine the performance of individual scoring systems was accomplished by using the Akaike Information criterion (AIC) and the Pearson’s v2 goodness-of-fit test. We used Stata Release 13 (StataCorp LP, College station, Texas) for all statistical analysis.
3
Survived n=5443 (75.6%)
Fig. 1. Trauma patient inclusion.
Results In a registry of 8791 patients, we had a cohort of 7197 adult trauma patients eligible for the study (Fig. 1). 4091 (56.8%)patients had all five scores available and was the sample for a complete case analysis. The mean age of the adult trauma patient admitted to the urban trauma units was 31.8 years and 75% were male. The timing of in-hospital deaths into early, delayed and late are summarized in Table 2 alongwith the comparability of all cases with the subset of complete cases, with no missing data. The participating hospitals displayed a case-mix (Table 1) in keeping with the city’s policies for referral from other smaller hospitals or the police department. Two-thirds of the admitted patients were transferred from other hospitals (range 48–82%). The difference in performance of the severity scores between the directly admitted and the larger group of patients transferred from other facilities, was not significant (p = 0.29). Burns and railway injuries had a higher mean ISS score than assaults and falls. In the subset of patients with all five scores
Table 1 Demography and case-mix in participating urban trauma facilities. Mean (SD)
Allcentres (n = 7197)
Kolkata n = 2398
Mumbai 1 Mumbai 2 Delhi n = 2580 (KEM) (Sion) n = 1191 n = 1028
Age Males
37.2 (16) 81%
40 (17) 72%
37.5 (17) 80%
34.5 (15) 93%
35 (14) 85%
Mechanism of Injury (mean ISS) Fall (11) 30% 33% Railway (14) 6.5% 2% Road Traffic (12) 47.3% 45% Assault (8) 9.6% 6% Burns (29) 6.6% 14%
42% 5% 36% 7% 10%
22% 28.8% 35% 14% 0.2%
25% 3% 59% 13% 0%
Injury Severity GCS (n = 6280) RTS (n = 4299) ISS (n = 6732) NISS (n = 6732) KTS (n = 4091) TRISS (n = 4091) Ps Transferred
11 (4) 7 (1.1) 11.5 (10) 17 (14) 12 (1.1) 0.94 (0.1) 68%
11(4) 6.7 (1.5) 13 (8) 19 (11) 11 (1.6) 0.92 (0.1) 72%
12 (3) 7 (1.2) 11.5 (8) 16 (10) 12 (1.2) 0.96 (0.1) 47%
11 (4) 7 (1.3) 12 (10) 17 (12) 12 (1.3) 0.95 (0.1) 65%
10 (4) 6.7 (1.4) 14 (13) 18 (13) 11.5 (1.4) 0.94 (0.1) 82%
Please cite this article in press as: N. Roy, et al., Validation of international trauma scoring systems in urban trauma centres in India, Injury (2016), http://dx.doi.org/10.1016/j.injury.2016.09.027
G Model JINJ 6913 No. of Pages 6
4
N. Roy et al. / Injury, Int. J. Care Injured xxx (2016) xxx–xxx
Table 2 Mortality with time comparison of ‘all cases’ in dataset and ‘complete cases’ analysis (with complete trauma scoring data and no missing data). All cases (%)
Complete cases (%)
p-value
n Males Age mean Age 16–55 years Age > 55 years
7197 5825 (81%) 37.1 1401 (80%) 353 (20%)
4091 (56.8%) 3596 (83%) 36.5 731 (81%) 169 (19%)
0.2 0.03 0.41
In-hospital mortality 24-h mortality 1–7 day mortality 8–30 day mortality
1754 (100%) 595 (34%) 752 (43%) 407 (23%) 1754 (100%)
900 (100%) 303 (34%) 381 (42%) 216 (24%) 900 (100%)
0.9 0.8 0.7
available (n = 4091), RTS, KTS, had similar discrimination—characterised by the area under the ROC curve and goodness-of-fit characterised by the Akaike information criterion, as shown in Table 3. The low AIC score supports RTS as the most parsimonious model, as the best predictive regression model, with the least number of variables. Considering overall mortality, both physiologic scores (RTS, KTS) had better discrimination and goodness-offit than ISS or NISS (Fig. 2a). The ability of all Injury scores to predict early mortality (24 h) was better than late mortality (30 day mortality), as seen in Fig. 2b–d. The performance of the trauma scores did not improve in the subset of severely injured patients (ISS > 15). Discussion This study tests the international injury scoring systems on a large multicentre cohort of Indian trauma patients. The main finding was that TRISS and KTS were comparable and both were significantly associated with mortality in the setting of the Indian trauma centres. While KTS is cheaper and can be calculated on arrival of the trauma patient, TRISS is a retrospective score, and requires much expertise and expense. TRISS has been the benchmark for trauma scoring and research for more than 20 years [17,18]. The performance of KTS is consistent with similar studies in the LMIC settings, like Cameroon [4], and may even outperform TRISS in the HIC, such as the USA [15]. The on-admission vital signs were consistently strongly associated with early (0–24 h) and delayed mortality (1–7 days), but not late mortality (7–30 days). Respiratory rate is easy to collect but was missing in a third (36%) of patients in our dataset, as is common worldwide [19]. GCS is a 3-part score, more difficult to collect and calculate, but surprisingly, was available in 86% patients and more often than systolic blood pressure readings (80%). This is in contrast to trauma databases worldwide [19] and is hard to explain, except by attribution to the surgery residency training culture [20]. We handled the missing data of our dataset by doing a complete case analysis (patients with complete data for calculating scores and no missing data) and an all-case analysis. The complete case analysis was comparable to the all-cases dataset and the
differences were not significant as shown in Table 2. The complete case analysis graphs are presented in Fig. 2a–d. Though the mechanism of thermal injury is different, we found that the outcome of severely burned patients were better predicted than for road traffic injuries, by the various scoring systems in the Indian dataset. Cassidy et al. [21] found that RTI had a better mortality concordance in patients with ISS > 15, as compared to burns. They suggested that adding age and burns body surface area would improve the ISS model for burn patients. The insignificant difference in the injury severity scores of the patients directly admitted to the trauma centre, when compared to the larger group of patients referred from other facilities, may be attributed to the lack of formal prehospital care in most LMICs, suggesting that the anatomic and physiologic status of both groups were similar and therefore the scores [22,23]. Though trauma scores are known to perform better in the severely injured subgroup (ISS > 15) [10], we could not find this correlation. ISS has a disadvantage that it is a retrospective calculation, as the exact anatomic injuries are not known at admission. Therefore ISS cannot be used to predict probable outcome or the risk of adverse outcome on arrival at the hospital [9]. Also, ISS requires much resource for data collection in a standardized manner, expertise, training of accredited coders in scoring and also detailed investigations (preferably CT scans), intra-operative case notes and autopsy reports. It is important to note that these are not commonly available in the LMIC setting, often because patients do not get CT scans due to affordability and resource limitations. This in turn, limits the use of ISS, which can predict mortality only with complete information. We found that NISS did not achieve any better discrimination and performed only marginally better than ISS. In a previous work, the authors tested a newer next generation score, the ICISS (ICD based Injury severity score) in the Indian setting, with better discrimination and statistical properties, but this did not prove to be a better performer than the first generation ones, besides having no real use in the clinical setting [24]. Our study shows that the prediction of large database generated-scores (like TRISS) is not superior to expert or consensus-based scores, like GCS and KTS. Also, since TRISS was developed primarily on US data, it may lack validity outside the North American context [25] of trauma centres. Japanese and Thai researchers showed that a modified TRISS, with context-adapted coefficients, resulted in more accurate predictions when used on their data [26]. Similarly, calculating updated coefficients for the Indian trauma patient, would be an area for further research. Comparisons between trauma centres using TRISS will remain important for trauma outcomes researcher, while it will be more practical and cost effective to use KTS in the LMIC clinical setting. If researchers continue to retrospectively calculate KTS based on trauma registry data, it will be important to formalize a methodology for calculating each patient’s number of serious injuries and to establish a conversion from GCS to AVPU score [15]. An easy to collect scoring system like the KTS will remain the practitioner's tool, to triage salvageable trauma patients, and to notify healthcare workers to trigger action-interventions.
Table 3 Analysis of area under ROC curves for individual trauma scores for 30-day mortality. Predictive regression model
AUC
[95% CI]
p(Wald test)
Akaike Information Criteria (AIC)
ISS NISS RTS KTS TRISS
0.62 0.65 0.81 0.74 0.82
0.60–0.64 0.63–0.67 0.79–0.82 0.72–0.76 0.81–0.84
0.001 0.001 0.001 0.001 0.001
3929 3889 3221 3429 3388
Please cite this article in press as: N. Roy, et al., Validation of international trauma scoring systems in urban trauma centres in India, Injury (2016), http://dx.doi.org/10.1016/j.injury.2016.09.027
G Model JINJ 6913 No. of Pages 6
N. Roy et al. / Injury, Int. J. Care Injured xxx (2016) xxx–xxx
5
at fixed intervals after the injury. In the absence of a prehospital system of care, there is little or no intervention at the field level, unlike in mature trauma systems where intubation, analgesia and sedation are more common. However, implementing complex scoring protocols across India, in the absence of a trauma system, would be expensive besides yielding incorrect data with incorrect AIS estimates. We handled this limitation of interrater reliability and accuracy with single co-author (DKV) AIS coding all injuries of all patients in the cohort. In addition, our use of Akaike information criterion to compare the goodness-of-fit of each score, we had to assume a linear relationship between score values and the probability of in-hospital mortality on the logistic scale. Our graphs of 7–30 mortality indicate that outcome prediction can never be very certain and all scores fail to predict complications which cause late mortality. Conclusion Given the realities of incomplete information in LMICs, it is reasonable to favour simplified injury scoring systems such as the KTS and RTS. Further, we found no statistically significant differences in the performance of these scores compared to the more expensive AIS-based TRISS. In India, it will be imperative for academic trauma research centres to maintain trauma database registries, code and characterize injuries and calculate the anatomic injury scores. Though the retrospective nature of ISS and TRISS score calculations and incomplete imaging in LMICs precludes its use in the casualty department, it will remain useful for comparison across trauma centres. Physiological scores like the RTS remain very useful in casualty departments of the urban Indian setting to predict early trauma mortality. All trauma scores predicted early mortality at 24 h better than late mortality at 30-days. Conflict of interest This study was funded by grants from the Swedish National Board of Health and Welfare and the Laerdal Foundation for Acute Care Medicine, Norway. The funding agencies had no influence on the conduct of the study, the contents of the manuscript, or the decision to send the manuscript for publication. Ethical considerations The institutional ethics committee of all participating hospitals LTMGH (IEC/11/13 dated 26 Jul 2013), KEM (IEC(I)/out/222/14 dated 4 Mar 2014), SSKM (IEC/279 dated 21 Mar 2013) and Apex Delhi (IEC/NP-327/2013 RP-24/2013 dated 25 Sep 2013) individually approved the collation of the database and analysis. Acknowledgements
Fig. 2. (a) 30a-day mortality by TRISS, KTS and ISS, NISS, RTS—Complete case analysis (n = 4084) (AUC in parenthesis). (b) 0–24-h mortality by TRISS, KTS and ISS, NISS, RTS (Complete case analysis). (c) 24h–7-day mortality by TRISS, KTS and ISS, NISS, RTS (Complete case analysis). (d) 7–30-day mortality by TRISS, KTS and ISS, NISS, RTS (Complete case analysis).
This study has several limitations. On-admission time is not a fixed time interval after injury, as there are prehospital delays in the LMIC setting. Therefore, vitals collected ‘on-admission’ are not
In addition to the authors, the following are members of the TITCO Consortium (listed alphabetically by first name): Amit Gupta, Additional Professor, Jai Prakash Narayan Apex Trauma Center, All India Institute of Medical Sciences, New Delhi, Ashish Jhakal, Project Officer Supervisor, Jai Prakash Narayan Apex Trauma Center, All India Institute of Medical Sciences, New Delhi Debojit Basak, Project Officer, Institute of Post-Graduate Medical Education and Research and Seth Sukhlal Karnani Memorial Hospital, Kolkata Deen Mohamed Ismail, Professor, Department of Orthopaedics, Madras Medical College, Chennai Dusu Yabo, Project Officer, Jai Prakash Narayan Apex Trauma Center, All India Institute of Medical Sciences, New Delhi Jegadeesan K, Project Officer Supervisor, Madras Medical College, Chennai Jyoti Kamble, Project Officer, King Edward Memorial Hospital, Mumbai Makhan Lal Saha,
Please cite this article in press as: N. Roy, et al., Validation of international trauma scoring systems in urban trauma centres in India, Injury (2016), http://dx.doi.org/10.1016/j.injury.2016.09.027
G Model JINJ 6913 No. of Pages 6
6
N. Roy et al. / Injury, Int. J. Care Injured xxx (2016) xxx–xxx
Professor, Department of Surgery, Institute of Post-Graduate Medical Education and Research and Seth Sukhlal Karnani Memorial Hospital, Kolkata Mangesh Nitnaware, Project Officer, Lokmanya Tilak Municipal Medical College and General Hospital, Mumbai Monty Khajanchi, Associate professor, General Surgery, Seth GS Medical College & King Edward Memorial Hospital, Mumbai Ranganathan Jothi, Professor, Department of Neurosurgery, Madras Medical College, Chennai Samarendra Nath Ghosh, Professor, Department of Neurosurgery, Institute of Post-Graduate Medical Education and Research and Seth Sukhlal Karnani Memorial Hospital, Kolkata Sanjeev Bhoi, Additional Professor, Jai Prakash Narayan Apex Trauma Center, All India Institute of Medical Sciences, New Delhi Santosh Mahindrakar, Project Officer, Jai Prakash Narayan Apex Trauma Center, Santosh Tirlotkar, Administrative officer, TITCO project, TISS, All India Institute of Medical Sciences, New Delhi Satish Dharap, Professor, Department of Surgery, Lokmanya Tilak Municipal Medical College and General Hospital, Mumbai Shilpa Rao, Professor, Department of Surgery, King Edward Memorial Hospital, Mumbai Veera Kamal, Project Officer, Madras Medical College, Chennai Vineet Kumar, Associate Professor, Department of Surgery, Lokmanya Tilak Municipal Medical College and General Hospital, Mumbai Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j. injury.2016.09.027. References [1] Global Burden of Disease Study Collaborators. Global, regional, and national incidence, prevalence, and years lived with disability for 301 acute and chronic diseases and injuries in 188 countries, 1990–2013: a systematic analysis for the Global Burden of Disease Study 2013. Lancet 2015. [2] World Health Organization. WHO Global Status Report on Road Safety 2013: Supporting a Decade of Action. World Health Organization; 2013. [3] Murlidhar V, Roy N. Measuring trauma outcomes in India. Injury 2004;35 (4):386–90. [4] Weeks SR, et al. Is the Kampala trauma score an effective predictor of mortality in low-resource settings? A comparison of multiple trauma severity scores. World J Surg 2014;38(8):1905–11.
[5] Laytin AD, et al. Choice of injury scoring system in low- and middle-income countries: lessons from Mumbai. Injury 2015. [6] Roy N, et al. 30-Day in-hospital trauma mortality in four urban university hospitals using an Indian trauma registry. World J Surg 2016;40(6):1299–307. [7] Shafi S, et al. Trauma quality improvement using risk-adjusted outcomes. J Trauma 2008;64(3)599–604 discussion 604–606. [8] Baker SP, et al. The injury severity score: a method for describing patients with multiple injuries and evaluating emergency care. J Trauma Acute Care Surg 1974;14(3):187–96. [9] Rozenfeld M, et al. ISS groups: are we speaking the same language? Inj Prev 2014. [10] Meredith JW, et al. A comparison of the abilities of nine scoring algorithms in predicting mortality. J Trauma 2002;53(4)621–8 discussion 628–629. [11] Champion HR, et al. The Major Trauma Outcome Study: establishing national norms for trauma care. J Trauma 1990;30(11):1356–65. [12] Schluter PJ. The Trauma and Injury Severity Score (TRISS) revised. Injury 2011;42(1):90–6. [13] Schluter PJ, et al. Trauma and Injury Severity Score (TRISS) coefficients 2009 revision. J Trauma 2010;68(4):761–70. [14] MacLeod JBA, et al. A Comparison of the Kampala Trauma Score (KTS) with the Revised Trauma Score (RTS), Injury Severity Score (ISS) and the TRISS method in a Ugandan Trauma Registry. Eur J Trauma 2003;29(6):392–8. [15] Weeks SR, et al. A modified Kampala trauma score (KTS) effectively predicts mortality in trauma patients. Injury 2015. [16] McNarry A, Goldhill D. Simple bedside assessment of level of consciousness: comparison of two simple assessment scales with the Glasgow Coma scale. Anaesthesia 2004;59(1):34–7. [17] Gabbe BJ. TRISS: does it get better than this? Acad Emerg Med 2004;11(2):181–6. [18] Glance LG, et al. The relation between trauma center outcome and volume in the National Trauma Databank. J Trauma 2004;56(3):682–90. [19] O'Reilly GM, Cameron PA, Jolley DJ. Which patients have missing data? An analysis of missingness in a trauma registry. Injury 2012;43(11):1917–23. [20] O'Reilly GM, Gabbe B, Cameron PA. Trauma registry methodology: a survey of trauma registry custodians to determine current approaches. Injury 2015;46 (2):201–6. [21] Cassidy JT, et al. Developing a burn injury severity score (BISS): adding age and total body surface area burned to the injury severity score (ISS) improves mortality concordance. Burns 2014;40(5):805–13. [22] Kumar V, et al. Ready: steady, go or just go?—The question of stabilization before transport for trauma victims. Injury 2013;44(11):1654–5. [23] Veetil DK, et al. Direct versus transferred severely injured: different context, similar findings. Injury 2016;47(2):510. [24] Gerdin M, et al. Early hospital mortality among adult trauma patients significantly declined between 1998 and 2011: three single-centre cohorts from Mumbai, India. PLoS One 2014;9(3):e90064. [25] Rogers FB, et al. Has TRISS become an anachronism? A comparison of mortality between the National Trauma Data Bank and major trauma outcome study databases. J Trauma Acute Care Surg 2012;73(2)326–31 discussion 331. [26] Kimura A, Chadbunchachai W, Nakahara S. Modification of the Trauma and Injury Severity Score (TRISS) method provides better survival prediction in Asian blunt trauma victims. World J Surg 2012;36(4):813–8.
Please cite this article in press as: N. Roy, et al., Validation of international trauma scoring systems in urban trauma centres in India, Injury (2016), http://dx.doi.org/10.1016/j.injury.2016.09.027
Injury, Int. J. Care Injured xxx (2016) xxx–xxx
Contents lists available at ScienceDirect
Injury journal homepage: www.elsevier.com/locate/injury
Validation of international trauma scoring systems in urban trauma centres in India Nobhojit Roy, MS MPHa,b,* , Martin Gerdin, MD PhDa , Eric Schneider, PhDc , Deepa K. Kizhakke Veetil, MSd , Monty Khajanchi, DNBb , Vineet Kumar, DNB FNBe , Makhal Lal Saha, MSf , Satish Dharap, MSg , Amit Gupta, MSh , Göran Tomson, MD PhDi, Johan von Schreeb, MD PhDa a
Department of Public Health Sciences, Health Systems and Policy, Karolinska Institutet, Stockholm, Sweden BARC Hospital (Govt of India), HBNI University, Mumbai, India c Center for Surgery and Public Health, Brigham and Women's Hospital, Boston, MA 02120, USA d Department of General Surgery, King Edward Memorial Hospital, Mumbai, India e Department of General Surgery, Lokmanya Tilak Municipal Medical College and General Hospital, Mumbai, India f Department of Surgery, Institute of Post-Graduate Medical Education and Research and Seth Sukhlal Karnani Memorial Hospital, Kolkata, India g Department of Surgery, Lokmanya Tilak Municipal Medical College and General Hospital, Mumbai,India h Department of Surgery, Jai Prakash Narayan Apex Trauma Center, All India Institute of Medical Sciences, New Delhi, India i Department of Learning, Informatics, Management & Ethics (LIME) and Public Health Sciences, Karolinska Institutet, Stockholm, Sweden b
A R T I C L E I N F O
A B S T R A C T
Article history: Accepted 13 September 2016
Introduction: In the Lower-Middle Income Country setting, we validate trauma severity scoring systems, namely Injury Severity Score (ISS), New Injury Severity Scale (NISS) score, the Kampala Trauma Score (KTS), Revised Trauma Score (RTS) score and the TRauma Injury Severity Score (TRISS) using Indian trauma patients. Patients and methods: From 1 September 2013 to 28 February 2015, we conducted a prospective multicentre observational cohort study of trauma patients in four Indian university hospitals, in three megacities, Kolkata, Mumbai and Delhi. All adult patients presenting to the casualty department with a history of injury and who were admitted to inpatient care were included. The primary outcome was inhospital mortality within 30-days of admission. The sensitivity and specificity of each score to predict inpatient mortality within 30 days was assessed by the areas under the receiver operating characteristic curve (AUC). Model fit for the performance of individual scoring systems was accomplished by using the Akaike Information criterion (AIC). Results: In a registry of 8791 adult trauma patients, we had a cohort of 7197 patients eligible for the study. 4091 (56.8%)patients had all five scores available and was the sample for a complete case analysis. Over a 30-day period, the scores (AUC) was TRISS (0.82), RTS (0.81), KTS (0.74), NISS (0.65) and ISS (0.62). RTS was the most parsimonious model with the lowest AIC score. Considering overall mortality, both physiologic scores (RTS, KTS) had better discrimination and goodness-of-fit than ISS or NISS. The ability of all Injury scores to predict early mortality (24 h) was better than late mortality (30 day). Conclusion: On-admission physiological scores outperformed the more expensive anatomy-based ISS and NISS. The retrospective nature of ISS and TRISS score calculations and incomplete imaging in LMICs precludes its use in the casualty department of LMICs. They will remain useful for outcome comparison across trauma centres. Physiological scores like the RTS and KTS will be the practical score to use in casualty departments in the urban Indian setting, to predict early trauma mortality and improve triage. ã 2016 Elsevier Ltd. All rights reserved.
Keywords: India Injury Trauma Urban Injury scoring system TRISS RTS ISS NISS KTS
* Corresponding author at: Department of Public Health Sciences, Health Systems and Policy group, Karolinska Institutet, Widerströmska huset, Tomtebodavägen 18A, 171 77 Stockholm, Sweden. E-mail addresses: [email protected], [email protected] (N. Roy), [email protected] (M. Gerdin), [email protected] (E. Schneider), [email protected] (D.K. Kizhakke Veetil), [email protected] (M. Khajanchi), [email protected] (V. Kumar), [email protected] (M.L. Saha), [email protected] (S. Dharap), [email protected] (A. Gupta), [email protected] (G. Tomson), [email protected] (J. von Schreeb). http://dx.doi.org/10.1016/j.injury.2016.09.027 0020-1383/ã 2016 Elsevier Ltd. All rights reserved.
Please cite this article in press as: N. Roy, et al., Validation of international trauma scoring systems in urban trauma centres in India, Injury (2016), http://dx.doi.org/10.1016/j.injury.2016.09.027
G Model JINJ 6913 No. of Pages 6
2
N. Roy et al. / Injury, Int. J. Care Injured xxx (2016) xxx–xxx
Introduction Lower-middle income countries (LMICs) pay the price of a growing volume of trauma, as collateral damage for development, rapid urbanization and sociodemographic transition [1]. In High Income Countries (HICs), the 30-day fatality criteria (dying within 30 days of injury) as recommended by the Global Road Safety report, is used to compare countries and their commitment to road safety [2]. The current decade of action for road safety (2011– 2020), advocates for a ‘post-crash response’ with appropriate care, including prehospital care, in-hospital care and post-discharge rehabilitation, for individuals who have sustained trauma. Inhospital trauma mortality is proportionally higher in LMICs as compared to HICs [3]. Understanding the factors associated with the observed differences in mortality between patients undergoing in-hospital trauma care in LMIC and HIC presents substantial challenges. The lack of a standard and uniform system for quantification of trauma severity is a drawback for researchers seeking to study systems of trauma care in multiple settings. A standardized measure of trauma severity is needed to ensure a fair and meaningful comparison of outcomes and effectiveness of trauma care among different settings with differing case-mix, within and across countries. Trauma severity scoring systems have been designed to estimate the mortality risk based upon specific combinations of factors associated with patient injuries. The scoring systems used to quantifying injury severity in HIC settings have historically relied upon anatomical measures of derangement while those predominant in LMICs have relied more heavily on measures of physiological derangement which limits comparability across HIC and LMIC trauma centres [4,5]. The anatomy-based scoring systems rely on imaging technologies, such as Computed Axial Tomography (CT) scans, prehospital care information, access to surgical intervention, quality intraoperative documentation and autopsy reports which may be unavailable in many LMIC trauma care facilities, thereby preventing the accurate calculation of Injury severity scores (ISS). Neither consensus-based scoring systems nor data-driven trauma severity scoring systems have been well-researched in the LMIC setting. Therefore, our aim was to validate commonly used trauma severity scoring systems ranging from the purely anatomy-based Injury Severity Score and New Injury Severity Scale (NISS) score, to more physiology-focused scores, including the Kampala Trauma Score (KTS) and the Revised Trauma Score (RTS) score, as well as the combined score TRauma Injury Severity Score (TRISS), both within and across facilities treating substantial numbers of trauma patients in India.
operate trauma-specific units providing trauma services as a part of a general hospital. All facilities were classified as ‘free-to-public’ indicating nominal fees to users facilitating access to care to the lower socio-economic strata of the population. Eligibility criteria All adult patients ( 15 years of age) presenting to the casualty department during the study period with a history of injury associated with a mechanism of road traffic, railway, fall, assault or burns and who were admitted to inpatient care at the index hospital were included. Patients who were dead on arrival were not included. Dependent and independent variables The primary outcome of interest was in-hospital mortality within 30-days of admission [7]. The secondary outcomes were 1. Mortality within 24-h, 2. Mortality between 24 h and 7 days, 3. Mortality between 7 and 30 days. For each included individual, patient-specific demographic factors including age and sex, as well as injury and pre-hospital treatment factors such as transfer status, mode of transport, and mechanism of injury were collected. Physiological measures at the time of initial assessment were recorded, including Systolic Blood Pressure (SBP), Glasgow Coma Scale (GCS) score, respiratory rate (RR), heart rate (HR) and oxygen saturation. Data To ensure consistency in data quality at each participating site, one data collector prospectively gathered data during the initial assessment using a standardized intake form for eight hours per day by directly observing the providers who were delivering trauma care and recording vital signs. This data collector served only as an observer and was not involved in patient care. At each site, the data collector, who was funded externally by TITCO-India, rotated through all 8-h shifts (morning, evening, night), and also on public holidays. For patients admitted outside of the 8-h ‘directly observed’ shift, data were retrieved from patient case records within days of initial presentation. Collected data were uploaded to a central database on a weekly basis and the authors (NR, MG, VK, MK) conducted weekly data review meetings. All data collectors had at minimum a health science master degree and were continuously trained and supervised throughout the study period. Injury severity scores
Patients and methods Study design, context and setting This prospective multi-centre observational cohort study was conducted under the guidance of the collaborative research consortium “Towards improving trauma care outcomes” (TITCOIndia) [6] from 1 September 2013 to 28 February 2015 in four Indian teaching and referral hospitals, each of which operate trauma units that receive citywide referral of trauma patients. The megacities (populations of more than 10 million) were geographically representative of urban India, namely Kolkata, Mumbai (2centres) and Delhi. The centres were the Apex Trauma Centre of the All-India Institute of Medical Sciences (AIIMS), New Delhi; Lokmanya Tilak Municipal General Hospital (LTMGH), Mumbai; KEM hospital, Mumbai; and the Seth Sukhlal Karnani Memorial Hospital (SSKM), Kolkata. The Apex trauma centre in Delhi is a standalone trauma-care facility while the other three centres
Baker et al., in 1974 [8], proposed ISS as an anatomical score that incorporates Abbreviated Injury Scale (AIS) in six body regions. In our study a single surgeon (DKV) accredited by the Association for the Advancement of Automotive Medicine (AAAM) calculated the ISS. The ISS has the advantage of converting all injuries in a trauma patient into a single number. However, though ISS appears to be a continuous variable (0–75), due to its sum of squares calculation formula, some integers are mathematically not possible. The impossible integers are 7, 15, 23, 28, 31, 37, 39, 40, 44, 46, 47, 49, 52, 53, 55, 56, 58, 60–65, 67–74 [9]. The NISS, a modification of ISS, was computed from the three worst injuries, regardless of the body region in which they occurred. The Revised Trauma Score (RTS), is a physiologic score that a patient's systemic response to injury measured through GCS, SBP and RR. It is the current standard physiologic scoring system used in trauma research and quality improvement in both high-income countries and low- and middle-income countries. The TRISS
Please cite this article in press as: N. Roy, et al., Validation of international trauma scoring systems in urban trauma centres in India, Injury (2016), http://dx.doi.org/10.1016/j.injury.2016.09.027
G Model JINJ 6913 No. of Pages 6
N. Roy et al. / Injury, Int. J. Care Injured xxx (2016) xxx–xxx
system combines RTS, ISS, age and type of injury (blunt or penetrating) to estimate the probability of survival (Ps) for each individual patient [10]. The Ps ranges from 0 to 1, where 0 represents a 0% and 1 a 100% probability of survival. The original TRISS coefficients were derived from data submitted by 139 North American hospitals between 1982 and 1987 and was published as the Major Trauma Outcome Study (MTOS) in 1990 [11]. In 2009, an updated version of the TRISS coefficients was published, and this time the coefficients were derived from the US National Trauma Data Bank (NTDB)[12,13]. The Kampala Trauma Score (KTS) is a simplified injury scoring system that uses patient age, systolic blood pressure, respiratory rate, neurologic status and number of serious injuries as variables and was developed in Uganda for use in resource-limited settings [14]. The components of Kampala Trauma Score as described by MacLeod, J.B.A., et al. is available as a Supplementary information online.
Adult t rauma victims admitted at 4 Indian urban University referral hospitals Sept 2013 - Feb 2015 (n=8791) Excluded:
Included: Mechanism of Injury: road traffic, railway, fall, assault or burns
Death status unknown=7 Did not complete 30 in-patient days of follow up =1587
7197 Adult patients in study had outcomes: 1) death or discharge within 30 days
Died n=1754 (24.3%)
Statistical methods For each patient KTS, ISS, NISS, RTS, TRISS and GCS were calculated. The following minor modifications were made to the KTS score so that it could be calculated retrospectively, as per Weeks et al’s work [15]. The number of serious injuries for each patient was determined based on a list of final diagnoses by a member of the research team with expertise in trauma care. Because no standardised conversion from GCS to AVPU (“alert, voice, pain, unresponsive”) score exists, an estimated AVPU score was assigned based on GCS using data from the original validation study of the KTS (GCS 14–15 = “alert”, GCS 10–13 = “responds to voice”, GCS 5–9 = “responds to pain”, GCS 3–4 = “unresponsive”) [16]. A note was made when insufficient data prevented calculation of any of the scores. Because not all values required to compute each score were available for every patient, the primary outcomes reported are the results of a complete-case analysis for the subset of patients for whom each of the scores was available. The differences in means of continuous variables were examined using Student’s t-tests. Non-parametric tests (e.g. Wilcoson rank-sum) were employed when continuous variables were not normally distributed. Pearson’s chi-squared test was used to compare mortality rates among patients who did and did not have sufficient data to calculate each score recorded. Association between injury scoring systems and in-hospital mortality was evaluated using logistic regression. The sensitivity and specificity associated with the ability of each score to predict inpatient mortality within 30 days was assessed by analyzing the areas under the receiver operating characteristic (ROC) curve (AUC). The discrimination of each of the scoring systems (a measure of how well a model distinguishes survivors from non-survivors) was compared using these ROC curves. The AUCs along with 95% confidence intervals (CI) were compared between the injury scores. A sample size of 2350 patients, was calculated based on a difference of 0.05 in the AUC with a reference AUC of 0.75 to compare the ability of Injury severity scores to predict mortality. The confidence interval was set at 95% with 90% power and a 0.05. However, all patients in the available dataset were included for the analysis. Overlapping of confidence intervals was considered to indicate the absence of any statistically significant difference between score discrimination. Model fit for the models used to examine the performance of individual scoring systems was accomplished by using the Akaike Information criterion (AIC) and the Pearson’s v2 goodness-of-fit test. We used Stata Release 13 (StataCorp LP, College station, Texas) for all statistical analysis.
3
Survived n=5443 (75.6%)
Fig. 1. Trauma patient inclusion.
Results In a registry of 8791 patients, we had a cohort of 7197 adult trauma patients eligible for the study (Fig. 1). 4091 (56.8%)patients had all five scores available and was the sample for a complete case analysis. The mean age of the adult trauma patient admitted to the urban trauma units was 31.8 years and 75% were male. The timing of in-hospital deaths into early, delayed and late are summarized in Table 2 alongwith the comparability of all cases with the subset of complete cases, with no missing data. The participating hospitals displayed a case-mix (Table 1) in keeping with the city’s policies for referral from other smaller hospitals or the police department. Two-thirds of the admitted patients were transferred from other hospitals (range 48–82%). The difference in performance of the severity scores between the directly admitted and the larger group of patients transferred from other facilities, was not significant (p = 0.29). Burns and railway injuries had a higher mean ISS score than assaults and falls. In the subset of patients with all five scores
Table 1 Demography and case-mix in participating urban trauma facilities. Mean (SD)
Allcentres (n = 7197)
Kolkata n = 2398
Mumbai 1 Mumbai 2 Delhi n = 2580 (KEM) (Sion) n = 1191 n = 1028
Age Males
37.2 (16) 81%
40 (17) 72%
37.5 (17) 80%
34.5 (15) 93%
35 (14) 85%
Mechanism of Injury (mean ISS) Fall (11) 30% 33% Railway (14) 6.5% 2% Road Traffic (12) 47.3% 45% Assault (8) 9.6% 6% Burns (29) 6.6% 14%
42% 5% 36% 7% 10%
22% 28.8% 35% 14% 0.2%
25% 3% 59% 13% 0%
Injury Severity GCS (n = 6280) RTS (n = 4299) ISS (n = 6732) NISS (n = 6732) KTS (n = 4091) TRISS (n = 4091) Ps Transferred
11 (4) 7 (1.1) 11.5 (10) 17 (14) 12 (1.1) 0.94 (0.1) 68%
11(4) 6.7 (1.5) 13 (8) 19 (11) 11 (1.6) 0.92 (0.1) 72%
12 (3) 7 (1.2) 11.5 (8) 16 (10) 12 (1.2) 0.96 (0.1) 47%
11 (4) 7 (1.3) 12 (10) 17 (12) 12 (1.3) 0.95 (0.1) 65%
10 (4) 6.7 (1.4) 14 (13) 18 (13) 11.5 (1.4) 0.94 (0.1) 82%
Please cite this article in press as: N. Roy, et al., Validation of international trauma scoring systems in urban trauma centres in India, Injury (2016), http://dx.doi.org/10.1016/j.injury.2016.09.027
G Model JINJ 6913 No. of Pages 6
4
N. Roy et al. / Injury, Int. J. Care Injured xxx (2016) xxx–xxx
Table 2 Mortality with time comparison of ‘all cases’ in dataset and ‘complete cases’ analysis (with complete trauma scoring data and no missing data). All cases (%)
Complete cases (%)
p-value
n Males Age mean Age 16–55 years Age > 55 years
7197 5825 (81%) 37.1 1401 (80%) 353 (20%)
4091 (56.8%) 3596 (83%) 36.5 731 (81%) 169 (19%)
0.2 0.03 0.41
In-hospital mortality 24-h mortality 1–7 day mortality 8–30 day mortality
1754 (100%) 595 (34%) 752 (43%) 407 (23%) 1754 (100%)
900 (100%) 303 (34%) 381 (42%) 216 (24%) 900 (100%)
0.9 0.8 0.7
available (n = 4091), RTS, KTS, had similar discrimination—characterised by the area under the ROC curve and goodness-of-fit characterised by the Akaike information criterion, as shown in Table 3. The low AIC score supports RTS as the most parsimonious model, as the best predictive regression model, with the least number of variables. Considering overall mortality, both physiologic scores (RTS, KTS) had better discrimination and goodness-offit than ISS or NISS (Fig. 2a). The ability of all Injury scores to predict early mortality (24 h) was better than late mortality (30 day mortality), as seen in Fig. 2b–d. The performance of the trauma scores did not improve in the subset of severely injured patients (ISS > 15). Discussion This study tests the international injury scoring systems on a large multicentre cohort of Indian trauma patients. The main finding was that TRISS and KTS were comparable and both were significantly associated with mortality in the setting of the Indian trauma centres. While KTS is cheaper and can be calculated on arrival of the trauma patient, TRISS is a retrospective score, and requires much expertise and expense. TRISS has been the benchmark for trauma scoring and research for more than 20 years [17,18]. The performance of KTS is consistent with similar studies in the LMIC settings, like Cameroon [4], and may even outperform TRISS in the HIC, such as the USA [15]. The on-admission vital signs were consistently strongly associated with early (0–24 h) and delayed mortality (1–7 days), but not late mortality (7–30 days). Respiratory rate is easy to collect but was missing in a third (36%) of patients in our dataset, as is common worldwide [19]. GCS is a 3-part score, more difficult to collect and calculate, but surprisingly, was available in 86% patients and more often than systolic blood pressure readings (80%). This is in contrast to trauma databases worldwide [19] and is hard to explain, except by attribution to the surgery residency training culture [20]. We handled the missing data of our dataset by doing a complete case analysis (patients with complete data for calculating scores and no missing data) and an all-case analysis. The complete case analysis was comparable to the all-cases dataset and the
differences were not significant as shown in Table 2. The complete case analysis graphs are presented in Fig. 2a–d. Though the mechanism of thermal injury is different, we found that the outcome of severely burned patients were better predicted than for road traffic injuries, by the various scoring systems in the Indian dataset. Cassidy et al. [21] found that RTI had a better mortality concordance in patients with ISS > 15, as compared to burns. They suggested that adding age and burns body surface area would improve the ISS model for burn patients. The insignificant difference in the injury severity scores of the patients directly admitted to the trauma centre, when compared to the larger group of patients referred from other facilities, may be attributed to the lack of formal prehospital care in most LMICs, suggesting that the anatomic and physiologic status of both groups were similar and therefore the scores [22,23]. Though trauma scores are known to perform better in the severely injured subgroup (ISS > 15) [10], we could not find this correlation. ISS has a disadvantage that it is a retrospective calculation, as the exact anatomic injuries are not known at admission. Therefore ISS cannot be used to predict probable outcome or the risk of adverse outcome on arrival at the hospital [9]. Also, ISS requires much resource for data collection in a standardized manner, expertise, training of accredited coders in scoring and also detailed investigations (preferably CT scans), intra-operative case notes and autopsy reports. It is important to note that these are not commonly available in the LMIC setting, often because patients do not get CT scans due to affordability and resource limitations. This in turn, limits the use of ISS, which can predict mortality only with complete information. We found that NISS did not achieve any better discrimination and performed only marginally better than ISS. In a previous work, the authors tested a newer next generation score, the ICISS (ICD based Injury severity score) in the Indian setting, with better discrimination and statistical properties, but this did not prove to be a better performer than the first generation ones, besides having no real use in the clinical setting [24]. Our study shows that the prediction of large database generated-scores (like TRISS) is not superior to expert or consensus-based scores, like GCS and KTS. Also, since TRISS was developed primarily on US data, it may lack validity outside the North American context [25] of trauma centres. Japanese and Thai researchers showed that a modified TRISS, with context-adapted coefficients, resulted in more accurate predictions when used on their data [26]. Similarly, calculating updated coefficients for the Indian trauma patient, would be an area for further research. Comparisons between trauma centres using TRISS will remain important for trauma outcomes researcher, while it will be more practical and cost effective to use KTS in the LMIC clinical setting. If researchers continue to retrospectively calculate KTS based on trauma registry data, it will be important to formalize a methodology for calculating each patient’s number of serious injuries and to establish a conversion from GCS to AVPU score [15]. An easy to collect scoring system like the KTS will remain the practitioner's tool, to triage salvageable trauma patients, and to notify healthcare workers to trigger action-interventions.
Table 3 Analysis of area under ROC curves for individual trauma scores for 30-day mortality. Predictive regression model
AUC
[95% CI]
p(Wald test)
Akaike Information Criteria (AIC)
ISS NISS RTS KTS TRISS
0.62 0.65 0.81 0.74 0.82
0.60–0.64 0.63–0.67 0.79–0.82 0.72–0.76 0.81–0.84
0.001 0.001 0.001 0.001 0.001
3929 3889 3221 3429 3388
Please cite this article in press as: N. Roy, et al., Validation of international trauma scoring systems in urban trauma centres in India, Injury (2016), http://dx.doi.org/10.1016/j.injury.2016.09.027
G Model JINJ 6913 No. of Pages 6
N. Roy et al. / Injury, Int. J. Care Injured xxx (2016) xxx–xxx
5
at fixed intervals after the injury. In the absence of a prehospital system of care, there is little or no intervention at the field level, unlike in mature trauma systems where intubation, analgesia and sedation are more common. However, implementing complex scoring protocols across India, in the absence of a trauma system, would be expensive besides yielding incorrect data with incorrect AIS estimates. We handled this limitation of interrater reliability and accuracy with single co-author (DKV) AIS coding all injuries of all patients in the cohort. In addition, our use of Akaike information criterion to compare the goodness-of-fit of each score, we had to assume a linear relationship between score values and the probability of in-hospital mortality on the logistic scale. Our graphs of 7–30 mortality indicate that outcome prediction can never be very certain and all scores fail to predict complications which cause late mortality. Conclusion Given the realities of incomplete information in LMICs, it is reasonable to favour simplified injury scoring systems such as the KTS and RTS. Further, we found no statistically significant differences in the performance of these scores compared to the more expensive AIS-based TRISS. In India, it will be imperative for academic trauma research centres to maintain trauma database registries, code and characterize injuries and calculate the anatomic injury scores. Though the retrospective nature of ISS and TRISS score calculations and incomplete imaging in LMICs precludes its use in the casualty department, it will remain useful for comparison across trauma centres. Physiological scores like the RTS remain very useful in casualty departments of the urban Indian setting to predict early trauma mortality. All trauma scores predicted early mortality at 24 h better than late mortality at 30-days. Conflict of interest This study was funded by grants from the Swedish National Board of Health and Welfare and the Laerdal Foundation for Acute Care Medicine, Norway. The funding agencies had no influence on the conduct of the study, the contents of the manuscript, or the decision to send the manuscript for publication. Ethical considerations The institutional ethics committee of all participating hospitals LTMGH (IEC/11/13 dated 26 Jul 2013), KEM (IEC(I)/out/222/14 dated 4 Mar 2014), SSKM (IEC/279 dated 21 Mar 2013) and Apex Delhi (IEC/NP-327/2013 RP-24/2013 dated 25 Sep 2013) individually approved the collation of the database and analysis. Acknowledgements
Fig. 2. (a) 30a-day mortality by TRISS, KTS and ISS, NISS, RTS—Complete case analysis (n = 4084) (AUC in parenthesis). (b) 0–24-h mortality by TRISS, KTS and ISS, NISS, RTS (Complete case analysis). (c) 24h–7-day mortality by TRISS, KTS and ISS, NISS, RTS (Complete case analysis). (d) 7–30-day mortality by TRISS, KTS and ISS, NISS, RTS (Complete case analysis).
This study has several limitations. On-admission time is not a fixed time interval after injury, as there are prehospital delays in the LMIC setting. Therefore, vitals collected ‘on-admission’ are not
In addition to the authors, the following are members of the TITCO Consortium (listed alphabetically by first name): Amit Gupta, Additional Professor, Jai Prakash Narayan Apex Trauma Center, All India Institute of Medical Sciences, New Delhi, Ashish Jhakal, Project Officer Supervisor, Jai Prakash Narayan Apex Trauma Center, All India Institute of Medical Sciences, New Delhi Debojit Basak, Project Officer, Institute of Post-Graduate Medical Education and Research and Seth Sukhlal Karnani Memorial Hospital, Kolkata Deen Mohamed Ismail, Professor, Department of Orthopaedics, Madras Medical College, Chennai Dusu Yabo, Project Officer, Jai Prakash Narayan Apex Trauma Center, All India Institute of Medical Sciences, New Delhi Jegadeesan K, Project Officer Supervisor, Madras Medical College, Chennai Jyoti Kamble, Project Officer, King Edward Memorial Hospital, Mumbai Makhan Lal Saha,
Please cite this article in press as: N. Roy, et al., Validation of international trauma scoring systems in urban trauma centres in India, Injury (2016), http://dx.doi.org/10.1016/j.injury.2016.09.027
G Model JINJ 6913 No. of Pages 6
6
N. Roy et al. / Injury, Int. J. Care Injured xxx (2016) xxx–xxx
Professor, Department of Surgery, Institute of Post-Graduate Medical Education and Research and Seth Sukhlal Karnani Memorial Hospital, Kolkata Mangesh Nitnaware, Project Officer, Lokmanya Tilak Municipal Medical College and General Hospital, Mumbai Monty Khajanchi, Associate professor, General Surgery, Seth GS Medical College & King Edward Memorial Hospital, Mumbai Ranganathan Jothi, Professor, Department of Neurosurgery, Madras Medical College, Chennai Samarendra Nath Ghosh, Professor, Department of Neurosurgery, Institute of Post-Graduate Medical Education and Research and Seth Sukhlal Karnani Memorial Hospital, Kolkata Sanjeev Bhoi, Additional Professor, Jai Prakash Narayan Apex Trauma Center, All India Institute of Medical Sciences, New Delhi Santosh Mahindrakar, Project Officer, Jai Prakash Narayan Apex Trauma Center, Santosh Tirlotkar, Administrative officer, TITCO project, TISS, All India Institute of Medical Sciences, New Delhi Satish Dharap, Professor, Department of Surgery, Lokmanya Tilak Municipal Medical College and General Hospital, Mumbai Shilpa Rao, Professor, Department of Surgery, King Edward Memorial Hospital, Mumbai Veera Kamal, Project Officer, Madras Medical College, Chennai Vineet Kumar, Associate Professor, Department of Surgery, Lokmanya Tilak Municipal Medical College and General Hospital, Mumbai Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j. injury.2016.09.027. References [1] Global Burden of Disease Study Collaborators. Global, regional, and national incidence, prevalence, and years lived with disability for 301 acute and chronic diseases and injuries in 188 countries, 1990–2013: a systematic analysis for the Global Burden of Disease Study 2013. Lancet 2015. [2] World Health Organization. WHO Global Status Report on Road Safety 2013: Supporting a Decade of Action. World Health Organization; 2013. [3] Murlidhar V, Roy N. Measuring trauma outcomes in India. Injury 2004;35 (4):386–90. [4] Weeks SR, et al. Is the Kampala trauma score an effective predictor of mortality in low-resource settings? A comparison of multiple trauma severity scores. World J Surg 2014;38(8):1905–11.
[5] Laytin AD, et al. Choice of injury scoring system in low- and middle-income countries: lessons from Mumbai. Injury 2015. [6] Roy N, et al. 30-Day in-hospital trauma mortality in four urban university hospitals using an Indian trauma registry. World J Surg 2016;40(6):1299–307. [7] Shafi S, et al. Trauma quality improvement using risk-adjusted outcomes. J Trauma 2008;64(3)599–604 discussion 604–606. [8] Baker SP, et al. The injury severity score: a method for describing patients with multiple injuries and evaluating emergency care. J Trauma Acute Care Surg 1974;14(3):187–96. [9] Rozenfeld M, et al. ISS groups: are we speaking the same language? Inj Prev 2014. [10] Meredith JW, et al. A comparison of the abilities of nine scoring algorithms in predicting mortality. J Trauma 2002;53(4)621–8 discussion 628–629. [11] Champion HR, et al. The Major Trauma Outcome Study: establishing national norms for trauma care. J Trauma 1990;30(11):1356–65. [12] Schluter PJ. The Trauma and Injury Severity Score (TRISS) revised. Injury 2011;42(1):90–6. [13] Schluter PJ, et al. Trauma and Injury Severity Score (TRISS) coefficients 2009 revision. J Trauma 2010;68(4):761–70. [14] MacLeod JBA, et al. A Comparison of the Kampala Trauma Score (KTS) with the Revised Trauma Score (RTS), Injury Severity Score (ISS) and the TRISS method in a Ugandan Trauma Registry. Eur J Trauma 2003;29(6):392–8. [15] Weeks SR, et al. A modified Kampala trauma score (KTS) effectively predicts mortality in trauma patients. Injury 2015. [16] McNarry A, Goldhill D. Simple bedside assessment of level of consciousness: comparison of two simple assessment scales with the Glasgow Coma scale. Anaesthesia 2004;59(1):34–7. [17] Gabbe BJ. TRISS: does it get better than this? Acad Emerg Med 2004;11(2):181–6. [18] Glance LG, et al. The relation between trauma center outcome and volume in the National Trauma Databank. J Trauma 2004;56(3):682–90. [19] O'Reilly GM, Cameron PA, Jolley DJ. Which patients have missing data? An analysis of missingness in a trauma registry. Injury 2012;43(11):1917–23. [20] O'Reilly GM, Gabbe B, Cameron PA. Trauma registry methodology: a survey of trauma registry custodians to determine current approaches. Injury 2015;46 (2):201–6. [21] Cassidy JT, et al. Developing a burn injury severity score (BISS): adding age and total body surface area burned to the injury severity score (ISS) improves mortality concordance. Burns 2014;40(5):805–13. [22] Kumar V, et al. Ready: steady, go or just go?—The question of stabilization before transport for trauma victims. Injury 2013;44(11):1654–5. [23] Veetil DK, et al. Direct versus transferred severely injured: different context, similar findings. Injury 2016;47(2):510. [24] Gerdin M, et al. Early hospital mortality among adult trauma patients significantly declined between 1998 and 2011: three single-centre cohorts from Mumbai, India. PLoS One 2014;9(3):e90064. [25] Rogers FB, et al. Has TRISS become an anachronism? A comparison of mortality between the National Trauma Data Bank and major trauma outcome study databases. J Trauma Acute Care Surg 2012;73(2)326–31 discussion 331. [26] Kimura A, Chadbunchachai W, Nakahara S. Modification of the Trauma and Injury Severity Score (TRISS) method provides better survival prediction in Asian blunt trauma victims. World J Surg 2012;36(4):813–8.
Please cite this article in press as: N. Roy, et al., Validation of international trauma scoring systems in urban trauma centres in India, Injury (2016), http://dx.doi.org/10.1016/j.injury.2016.09.027