Measuring trauma outcomes in India

Injury, Int. J. Care Injured (2004) 35, 386—390 Measuring trauma outcomes in India An analysis based on TRISS methodology in a Mumbai university hospital V. Murlidhara,*, Nobhojit Royb,1 a Lokmanya Tilak Municipal General Hospital, Sion, Mumbai 400022, India Bhabha Atomic Research Centre Hospital, Anushaktinagar, Mumbai 400094, India b Accepted 23 May 2003 KEYWORDS Injury Severity Score; Revised Trauma Score; Abbreviated Injury Score; Developing countries; Trauma; Outcome; Probability of survival; Major Trauma Outcome Study; Survival Analysis; India Summary Background: In this prospective study, the TRISS methodology is used to compare trauma care at a university hospital (Lokmanya Tilak Municipal General (LTMG) Hospital) in Mumbai, India, with the standards reported in the Major Trauma Outcome Study (MTOS). Methods: Between 1 August 2001 and 31 May 2002, 1074 severely injured patients were included in the study. Survival analysis was completed for 98.3% of the patients. Results: The majority of the patients were men (84%) and the average age was 31 years. 90.4% were blunt injuries, with road traffic crashes (39.2%) being the most common cause. The predicted mortality was 10.89% and the observed mortality was 21.26%. The mean Revised Trauma Score (RTS) was 6:61
1:65 and the mean Injury Severity Score (ISS) was 16:7
10:67. The average probability of survival (Ps) was 89.14. The M and Z statistics were 0.84 and 14.1593, respectively. Conclusion: The injured in India were found to be older, the injuries more severe and with poorer outcomes, than in the MTOS study. ß 2003 Elsevier Ltd. All rights reserved. Introduction Injury in a developing country like India is a leader together with non-communicable diseases, when measured in terms of disability adjusted life years (DALYs) lost. The younger population is more prone to injury and the resulting mortality accounts for a higher number of life years lost. The severity and the resulting disability is higher than in any other disease.8 It is recognised that there is a lack of information on the quality of trauma care in India. One of the causes of injury not being acknowledged as a serious *Corresponding author. Tel.: þ91-22-24098402. E-mail addresses: [email protected] (V. Murlidhar), [email protected], [email protected] (N. Roy). 1 Tel.: þ91-9821291225. public health problem is due to its association with the word ‘‘accident’’ which offsets any attempt at planning and prevention. It has yet to be accepted in India that accidental deaths can be measured, predicted and thereafter prevented by setting up injury prevention systems. These would eventually replace the ad hoc treatment scenario of the present with comprehensive and cost-effective care thereby preventing the loss of lives of those who are in their most productive years. Trauma systems have been shown to decrease the number of preventable deaths caused by trauma.5 The Lokmanya Tilak Municipal General (LTMG) Hospital (1416-bed) is a Level I trauma centre, which caters to the megapolis of Mumbai (formerly, Bombay). Its location at the termination of two major arterial roadways (the Eastern and the Western 0020–1383/$ — see front matter ß 2003 Elsevier Ltd. All rights reserved. doi:10.1016/S0020-1383(03)00214-6 Measuring trauma outcomes in India Method A total of 1074 patients were treated in the Trauma Ward of the Lokmanya Tilak General Hospital between August 2001 and May 2002. Registrars in surgery provide standardised care in the Trauma Ward, round the clock. Only trauma patients with multiple or severe injuries (Abbreviated Injury Score ðAISÞ > 2) are admitted in the Trauma Ward and were included in the study. Only those patients with the complete set of parameters required to calculate the RTS were included in the study. The parameters of every patient admitted to the Trauma Ward were entered into the EpiInfo 6 software (CDC Statistical package), on admission. One of the authors (VM) coded all the ISS and RTS injury scorings. The data collected included: demographics, cause of injury, physiologic status (RTS) at admission; definitive anatomic injury diagnoses (ISS) after stabilising the patient, obtained from the trauma chart, roentgenogram or CT scan findings, intra-operative 0 2 RTS Express Highways) makes it the recipient hospital of virtually all vehicular crash victims in Mumbai. Also, since it is located close to the two of the three rail mass transit networks (the Harbour line and the Central line), it caters for a large number of railway casualties. Another pool of trauma patients, mainly of assault and riot victims, is received from the communally sensitive areas around the hospital (Dharavi and Koliwada). The aim of this study was to evaluate this developing country trauma centre in terms of treatment and outcome and compare it with centres around the world. An attempt at auditing the casualties received over a period of 10 months has been performed based on the injury scoring systems and the Major Trauma Outcome Study (MTOS) initiated in the US by Champion et al. in 1990.6 This study measured the overall severity of the injury, recorded management and outcome, provided a database for audit in the individual patient and now allows for comparison of performance over time and between hospitals. The TRISS methodology is often used for outcome analysis in the injured patient. The TRISS system combines both the physiologic (Revised Trauma Score (RTS) and anatomic Injury Severity Score (ISS))11 assessments of injury, with age and mechanism of injury (blunt versus penetrating), to quantify the probability of survival (Ps) for each individual patient.4 Also, the Z and M statistics were calculated to compare the number of survivors in this institution to the number expected on the basis of the MTOS norm. 387 4 6 8 0 10 20 Survivors 30 40 ISS Non-Survivors 50 60 70 Ps 50 Isobar Figure 1 Scatter plot of RTS and ISS of survivors and non-survivors. findings and procedures. Both physiological and anatomic indices are required to effectively characterise injury severity. To evaluate the trauma care in the LTMG Hospital Trauma Ward, we used two methodologies. The first method is the preliminary outcome-based evaluation (PRE chart).7 The PRE chart (Fig. 1) is a scatter diagram of RTS versus ISS in the patient group, both survivors (represented as diamonds) and non-survivors (shown as squares). The diamonds above and squares below the line represent patients with mathematically unexpected outcome. The second method is the definitive outcomebased evaluation (DEF).9 In DEF, Flora’s Z score quantifies the difference in the actual number of deaths (or survivors) in the test subset and the predicted number of deaths (or survivors) on the basis of the baseline population (MTOS norm). The formula for calculating Z is: Z ¼ D  qi /piQi, where D is the actual number of deaths, Qi the predicted probability of death for a patient i, qi the predicted number of death and pi the predicted Ps for patient i. When mortality is studied, a negative value of Z is desired, since it implies that the number of deaths predicted from the baseline exceeds the number observed in the test. Therefore, a positive value of Z is desired in case survival is studied. Although the formula for calculation and the sign of Z changes, the absolute value does not. An absolute value of Z, which exceeds 1.96, is required for a significance level of 0.005. The injury severity match between the study and the baseline patient set can affect the Z score. The M score is a measure for that match. Values for M range from 0 to 1 and the closer the value is to 1, the better the match of injury severity. Limitations and exclusions Firstly, 18 records were excluded due to incomplete data entry. Secondly, deaths due to railway accidents are not routinely subjected to post-mortems, 388 V. Murlidhar, N. Roy as per a government order; therefore, the autopsy findings were excluded from all cases in this study. Railway accidents were not collisions, but injuries sustained by falling off overcrowded trains, sometimes in a state of inebriation or while crossing the tracks. Results Between 1 August 2001 and 31 May 2002, 1074 severely injured patients admitted to the Trauma Intensive Care Unit of the LTMG Hospital were studied. ‘Severely injured’ was defined as ISS > 25 or an AIS > 4 in one body region. Eighteen patients were eliminated because the data was incomplete or the patient was transferred to another hospital and therefore outcome could not be established. Sufficient data was available to perform survival analysis in 1056 patients. Demography shows the majority to be male patients 84%. The male to female ratio was 5:1. The average age was 31 years, with no significant difference in age between genders. The age distribution is shown in Fig. 2. The majority (90.4%) of the injuries were blunt. The mechanism of injury is shown in Table 1. Road traffic crashes accounted for 39.2% of the injuries, 26.7% due to falls (excluding those which were railway-related), while 22.7% of injuries were due to railway injuries. The highest mortality (30%) was found in the patients who sustained railway-related Figure 2 Table 1 Age distribution of casualties. Mechanisms of injury (%) Road traffic crashes Railway Fall from heights Assault Fall of objects Occupational Other 39.2 22.7 26.7 8.3 1.8 0.9 0.5 injuries. The vast majority (82%) of all the cases in this study had no pre-hospital care. The time between the injury and admission was an average of 6 h. A fifth of the survivors underwent a major operation. The mean RTS was 6:61
1:65 and the mean ISS was 16:7
10:67. The average Ps was 89.14. The number of predicted survivors were 941 and the actual survivors were 831. The standard error of Ps was 7.207458. Severe head injuries ðAIS > 4Þ accounted for 47% of all patients with head injury. The severe thoracic and abdominal injuries ðAIS > 4Þ were 0.6 and 8.2% of all the thoracic and abdominal injuries, respectively. Head trauma was present in a majority (76%) of the patients in this study. In the severe head injuries that underwent surgery, there was a 38% mortality. The mortality in the AIS  3 group was 11.9% (the MTOS subset had 5% mortality) and AIS > 4 was 37.9% (MTOS subset was 40%).6 Overall mortality was 21%. The respective mortality for male and female patients was 23 and 22%, with no significant difference. There were 167 unexpected deaths and 47 unexpected survivors. The misclassification ratio is 1:6 (167 of the total 1056). Disparity is 0.27, with an average Ps of 0.95 for survivors and an average Ps of 0.68 for non-survivors. The sensitivity of the TRISS methodology in the LTMG Trauma ICU is 68% and the specificity is 86%. Comparisons of survivors with non-survivors was done by using the Chi square test it was found significant with the odds ratio at 0.30, with a 95% confidence interval ð0:19 < OR < 0:46Þ and P < 0:05. Table 2 outlines the proportion of study patients in each of the calculated TRISS Ps categories as compared to the MTOS study. The M statistic19 compares the distribution of the predicted survival probabilities in a sample population with that of the main database. The nearer M is to 1.0, the better the fit. The M statistic for the LTMG Trauma ICU patients was 0.84. Table 2 Proportion of patients in each TRISS Ps category Ps interval No. of patients US fraction LTMGH fraction Minimum 0.96—1.00 0.91—0.95 0.76—0.90 0.51—0.75 0.26—0.50 0.00—0.25 0.828 0.045 0.044 0.029 0.017 0.036 0.999 0.668 0.112 0.077 0.068 0.032 0.043 0.999 0.668 0.045 0.044 0.029 0.017 0.036 0.84 632 106 73 64 30 41 Measuring trauma outcomes in India 389 The W statistic is the difference between the actual and the predicted numbers of survivors per 100 patients (on the basis of MTOS norm). W¼ 831  941 ¼ 10:416 1056=100 The significance of the W statistic was quantified by the Z statistic. The Z statistic for the LTMG Hospital Trauma ICU was 14.1593. There was no influence of pre-hospital delay, time from injury to operation and gender on survival. Discussion In this study, the quality of trauma care in a Level I trauma centre was evaluated using the TRISS methodology. The outcome of these patients were compared to those of the MTOS patients studied by Champion et al.6 This is represented in Table 3. The majority of the victims were young male subjects. While past studies have shown that 40% of the trauma victims are between 16 and 30 years of age,18 we found 32% patients in this age group. Most of the injuries encountered were blunt, resulting from motor vehicle crashes. However, most of the victims were pedestrians as opposed to occupants of the motor vehicle, an important Table 3 Comparison of MTOS with study subset Variable MTOS Subset study data (n ¼ 1074) (n ¼ 80,544) Suitability for analysis Included (%) Excluded (%) 89 11 98.3 1.7 Type of injury Blunt (%) Penetrating (%) 79 21 90.4 9.6 Outcome Live (%) Dead (%) 91 9 79 21 Cause of injury Road traffic crashes (%) 50 Fall from heights (%) 17 39.2 26.7 Gender Male patients (%) Female patients (%) 84 16 Trauma scoring Revised Trauma Score (mean
S.D.) Injury Severity Score (mean
S.D.) 71 28 7.1
1.7 6.6
1.7 12.8
11.3 16.7
10.7 variation from the similar studies from the developed world.14 The observed mortality was 21.26%, whereas the expected mortality was 10.89%. The mortality in the MTOS group was 9%. The Z statistic was 14.1593 indicating a significant difference in outcome between the study subject and the MTOS group. The validity of a trauma scoring system is most accurately assessed by its impact on the mortality of severely injured patients who would be expected to die without rapid and aggressive treatment.17 We therefore, analysed the trauma outcome on the patients with severe and multiple injuries ðISS > 25Þ. The mean ISS was 16.7 and the mean RTS 6.61. The mean Ps was 89.73. The injury severity match between the study and the baseline patients can affect the Z statistic. The M statistic for the study group was 0.84, indicating a poor match of severity with the MTOS group (M value should be greater than 0.88). According to Boyd et al.,4 Z values associated with lower values of M should be viewed with some scepticism. Nevertheless, the TRISS method can be still be useful for year-to-year trend analysis of an individual institution. In comparison with the MTOS data, we can see that the average patient in the LTMG Trauma ICU is more severely injured (ISS) and more physiologically deranged (RTS). There are likely explanations for these findings, all associated with regional factors. In the first place, the patients’ physiological condition can easily deteriorate before admission to the emergency department because of the lack of prehospital care and triage. However, we found no correlation between pre-hospital delay and outcome (the average delay being 6 h). Very few victims were received at the hospital within the critical ‘‘golden hour’’. There is no formal Emergency Medical System in Mumbai, much like the rest of India. The police, the fire brigade, individual initiatives, private ambulances and non-government organisations provide some pre-hospital care to the injured. There is no central co-ordination for trauma patients. In some areas, ambulances owned by private hospitals operate on a fee-for-service basis. These are usually just transport vehicles that only have an oxygen cylinder. The ambulance staff is generally not trained in pre-hospital care. There is no common countrywide telephone number to access an ambulance or emergency health care services.1 The most common mode of transport of a patient to the casualty department is a personal or commercial vehicle and only 2% are transported in some kind of ambulance.16 The likely institutional factors for poor outcomes are economic constraints. The caregiver—patient ratio in ICUs is usually 1:1.2 However, in the LTM 390 Trauma ICU, it is 1:5. While it is acknowledged that informal care provided by relatives contributes positively to survival outcomes in India, the cost and contribution of this informal care needs to be studied objectively. Despite the government heavily subsidising the treatment, the trauma patient has to bear a significant financial load of the treatment. It has been shown that trauma mortality is inversely proportional to a country’s per capita gross national product.12,13 These are possible explanations for the higher expected mortality of the group under study when compared to the overall mortality in the MTOS group (21% versus 9%). The figures are similar to those from other developing countries.10,15 Both the TRISS methodology and the MTOS database find their origin in the United States. Therefore, their validity may be limited to specific conditions of the originating country or countries with patient populations that have similar characteristics as the United States.3 In literature, there are no studies testing the validity of the TRISS methodology in a developing country. The TRISS methodology20 being a western methodology may be an inappropriate comparison when applied to a developing country.10 This underlines the need for a regional trauma database and an injury scoring system that will account for geographical, economic and physical attributes of South Asia and the Indian subcontinent. References 1. Aggarwal P, Banga A, Kurukumbi M, Gupta M. Emergency physicians and emergency medicine: an imminent need in India. Natl Med J India 2001;14(5):257—9. 2. Bennett D, Bion J. Organization of intensive care. BMJ 1999;318:1468—70. 3. Bouillon B, Lefering R, Vorweg M, et al. Trauma score systems: the Cologne validation study. J Trauma 1997;42:652—8. V. Murlidhar, N. Roy 4. Boyd CR, Tolson MA, Copes SC. Evaluating trauma care: the TRISS method. J Trauma 1987;27:370—8. 5. Cales RH, Trunkey DD. Preventable trauma deaths: a review of trauma care system development. JAMA 1985; 254:1059—63. 6. Champion HR, Copes WS, Sacco WJ, et al. The Major Trauma Outcome Study: establishing national norms for trauma care. J Trauma 1990;30:1356—65. 7. Champion HR, Sacco WJ, Hunt TK. Trauma severity scoring. World J Surg 1983;7:4—11. 8. Gumber A. Burden of injury in India. Econ Pol Wkly 1997; (21 June):1478—91. 9. Hollis S, Yates DW, Woodford M, et al. Standardised comparison of performance indicators in trauma: a new approach to case-mix variation. J Trauma 1995;38:763—6. 10. Joosse P, Soedarmo S, Luitse JSK, Ponsen KJ. Trauma outcome analysis of a Jakarta University Hospital using the TRISS method. J Trauma 1990;30:1356—65. 11. McLellan BA. Trauma severity scoring: the language of trauma. In: McMutry RY, McLellan BA, editors. Management of blunt trauma. Baltimore: Williams & Wilkins; 1990. p. 11—9. 12. Mock CN, Adzotor KE, Conklin E, Denno DM, Jurkovich GJ. Trauma outcome in the rural developing world: comparison with an urban level 1 trauma center. J Trauma 1993;35:518. 13. Mock CN, Jurkovich GJ, nii-Amon-Kotei D, Arreola-Risa C, et al. Trauma mortality patterns in three nations at different economic levels: implications for global trauma system development. J Trauma 1998;44:804. 14. Mohan D, Roberts I. Global road safety and the contribution of big business. BMJ 2001;323:648. 15. Onwudike M, Olufemi AO, Olusola OAO. Teaching hospital perspective of the quality of trauma care in Lagos, Nigeria. World J Surg 2001;25:112—5. 16. PoSaw LL, Aggarwal P, Bernstein SL. Emergency medicine in the New Delhi area, India. Ann Emerg Med 1998;32: 609—15. 17. Shackford SR, Mackersie RC, Hoyt DB, Baxt WG, et al. Impact of a trauma system on outcome of severely injured patients. Arch Surg 1987;122:523—7. 18. Sidhu DS, Sodi GS, Bannerjee AK. Mortality profile in trauma victims. J Indian Med Assoc 1993;91:16—8. 19. Smith EJ, Ward AJ, Smith D. Trauma scoring methods. Br J Hosp Med 1990;44:114—8. 20. Yates DW. Scoring systems for trauma. BMJ 1990;301: 1090—4.