Trauma registry analysis

Setting up Chain-of-Survival Model trauma systems we should feel obliged to control the quality of treatment. Especially so when you break new ground and delegate life-saving skills to non-surgeons and non-doctors.

Quality control cannot be done by case studies; it takes a systematic effort where you analyze risk descriptors against outcome indicators for whole sets of patients treated. For that you have to gather the essential data on all patients, validate the data, and load them in a PC registry. This is a brief how-to-do guide. For details, please contact us.

Essential data: Risk descriptors (explanatory variables)

Let us list the explanatory variables that affect the probability of trauma survival/death.

Age: The physiological response to trauma is different in Infants and children. Also “old” people can take less. Consequently we have to register years-of-age as a continuous variable.

Time factors: The injured starts dying and the post-injury stress reactions develops by hours after injury. Consequently two time variables are important:

Time 1: minutes and hours from the time of injury to in-field life support starts.

Time 2: the total prehospital transit time from injury to hospital admission.

Physiological severity: Revised Trauma Score

The “golden standard” for physiological severity scoring is the Revised Trauma Score (RTS). It builds on three clinical signs of oxygen starvation: respiratory rate/RR, systolic blood pressure/BP, and central nervous system function as rated by Glasgow Coma Scale (GCS).

GCS score



Coded value

13 – 15

> 90

10 – 30


9 – 12

76 – 89

> 30


6 – 8

50 – 75

6 – 9


4 – 5

1 – 49

1 – 5



No carotic pulse

No breathing


The RTS for any patient can thus take on values from 0 (death) to 12 (no physiological derangement). 

Essential data: Physiological severity descriptors

Notice that the impact of risk descriptors depends on the local context. E.g. people get “old” at different years-of-age in North and South.  Also the physiological capacity of a trauma patient is obviously different in well-fed Westerners and malnourished people. The coded values set in the RTS scoring system is based on studies of urban trauma in the US and may not fit rural scenarios in the South. We therefore recommend registration also of the actual values of RR and BP.

Scientific studies document that the GCS is inaccurate due to inter-rater disagreement. We have therefore simplified scoring by substituting GCS values with plain ratings of consciousness level: awake =4, drowsy=3, coma reacts on sound=2, coma reacts on pain only=1, no reaction=0.

The physiological severity indicators in each and every patient vary by time, and should be registered at two points:

            When life support starts in-field                   On hospital (end-point) admission

RR1                                                               RR2

BP1                                                               BP2

Level of consciousness 1                                   Level of consciousness 2

RTS1 (coded)                                                RTS2 (coded)


Essential data: Anatomical severity descriptors

The RTS indicates the impact of the injury. But we also need a descriptor of the degree of tissue injury, how severe is the wound? For that, the Injury Severity Score/ISS is established as the standard international reference. The ISS score in the individual patient is defined from a comprehensive list of surgical diagnosis to be found in The Abbreviated Injury Scale (AIS) manual. If the actual patient has a penetrating chest injury and a compound femur fracture, you can find his ISS score from the list of diagnosis in the AIS manual:

  • “Fracture of 2 – 3 ribs with hemo/pneumothorax:      450220.3”
  • “Femur fracture open/displaced/comminuted:     851801.3”

The injury severity is indicated by the figure after the dot, the AIS code. In this case the AIS code is “.3” for both diagnosis. ISS is then calculated as the sum of the squared AIS codes for the three most severe injuries. For our patient the ISS = 3×3 + 3×3 = 18. By tradition we regard ISS values <9 as “moderate”; ISS 9 – 15 as “serious”; and ISS > 15 as “major trauma victim”.

ISS can thus be used to characterize the overall severity of a population of patients, and to stratify (divide into subsets) the population for statistical analysis.

Essential data: Other explanatory variables

Body temperature

At temperatures < 34°C blood coagulation fails and the probability of trauma death increases. Even in hot climate post-injury hypothermia is common in trauma victims (PEKER). Core temperature can be registered in the mouth by digital thermometers on hospital admission.


Protracted pain is one main trigger of the devastating post-injury cascade reaction of complications and thus affects mortality probability. The level of pain can be registered by the life support provider at first contact in-field (point 1), and again on hospital admission (point 2). For this we use Visual Analogue Scales/VAS. 

If the pain-2 is less than pain-1, prehospital treatment has been effective and we expect it to have increased the probability of survival. List VAS values at point1 and point 2 as continuous variables.

Total operation time

Surgery – even well performed – is read by the body as a physiological trauma. Extensive primary surgery is one potent trigger of the post-injury stress response by enhancing immuno-depression. List total time used for primary surgery (minutes) as a continuous variable.

Malaria Falciparum

Unless given early antimalarial treatment, 1/3 of patients with asymptomatic malaria Falciparum develop post-injury symptomatic malaria within 48 hours after injury. The complication increases the risk of bacterial wound infection and protracts recovery. In malaria-endemic areas we should therefore identify the parasite carriers. This can easily be done by rapid test dip-sticks. The test result is registered as dichotome variable: malaria yes/no.

Laboratory data

If you have access to laboratory service, base excess on hospital admission is a good indicator of post-injury physiological derangement.

The explanatory variables should be registered there and then on a simple chart. This is the Field Chart used by Trauma Care Foundation and partners. Notice that essential information is registered at two points: first in-field encounter (point 1), and again on hospital admission (point 2).


































Essential data: Outcome indicators (result variables)


Trauma mortality

Our aim is to save lives, and to find out what makes a survivor. Trauma mortality is thus the main result variable. Register death: yes/no. Also register:

Time of death (hours after injury)

Death on-site (before prehospital life support started): yes/no

Death during prehospital treatment: yes/no

In-hospital death: yes/no


Effect of prehospital life support

The aim of prehospital life support is temporary damage control, to stop physiological deterioration and even improve the physiological condition of victims. For this we compare physiological indicators on-site and again on hospital admission. RTS values are already registered at two points  as explanatory variables (PEKER). From this you can construct a new result variable:

RTS2 – RTS1.  

RTS2 ≥ RTS1 is considered good prehospital outcome.

RTS2 ≤ RTS1 can also be used to select the subset of patients who deteriorated despite life support provided. The management of every one of these patients should be scrutinized.


Postoperative sepsis

The probability of  postoperative bacterial wound infection mainly depends on the quality of primary surgical care. Still we should expect that good prehospital life support should helps reduce the rate of post-injury sepsis: bacterial wound infection and/or abscess formation and/or pneumonia and/or bacteremia. Register sepsis: yes/no


Post-injury malaria

In areas with endemic malaria Falciparum, the rate of symptomatic post-injury malaria is an indicator of treatment quality. Symptomatic malaria is defined as clinical signs of malaria (fever > 38.5 °C plus chills and sweating) and positive rapid test or microscopic smear examination. Register: post-injury malaria yes/no.


In-hospital days

The duration of the hospital treatment may be an indicator of post-injury recovery. However, this variable may be inaccurate because other factors may affect in-hospital time (actions of war, security measures, in-hospital physical rehabilitation etc.).



Essential data: Gathering end-point variables


The result variables are registered at the end-point, which is the hospital providing definitive trauma surgery. Fort his we use a simple form to register the essential variables and factors:




































The data base

Patient ID, essential variables, and other factors are gathered from the Field and Hospital, carefully validated, and then loaded in a computer registry. The simple and safe way is to  make the registry in Excel spread sheats. From Excel the data can later be imported to any statistics soft-ware for analysis. Find enclosed an Excel registry template you can adapt to your setting. Other options are to build a relations data base can be made (Microsoft Access etc.),  or to load the data directly in a program for medical statistics (JMP etc. PEKER).


































Analyzing trauma system quality: basic comparisons

Two strategies should be used: Analyze outcome indicators by year to see if rates of mortality and complications come down by system improvement. Also compare trauma system outcome between different districts/branches. We use confidence interval analysis to compare difference of proportions and means (PEKER).

Trauma mortality by year/period

Trauma System X

Patients treated


Year 1



Year 2




Load the proportions in a statistical calculator, e.g. the CIA soft-ware, and find that the trauma mortality rate was reduced from 19.4% in year 1 to 13.0% in year 2. The reduction is statistically significant at the 95% level, the confidence interval (CI) for the difference being 1.1% to 11.7%.

Notice: If the 95% CI for difference does not contain zero, the difference is considered statistically significant at that given level of confidence (equivalent to p < 0.05).


You can compare two separate systems in the same way:


Patients treated


Trauma System X



Trauma System Y




This year System Y had mortality rate at 15.5% as compared to 19.4% in System X. It seems that System Y has the better quality of performance. However, the samples are small in numbers, therefore the difference is not statistically significant; 95% CI ranges from – 4.1% to 11.9%.


Effect of prehospital life support

We will examine the prehospital treatment effect by year using RTS differences (PEKER):

Trauma System X


Mean         SD

Year 1

1.1           0.65

Year 2

1.55           0.8


The difference is significant, 95%CI for difference 0.34 – 0.56. System X seems to have improved the prehospital performance during the period.




Analyzing trauma system quality: are the samples really comparable?

Is it fair to compare Year 1 and Year 2? Or System X with System Y? It is fair enough on the single condition that the distribution of the main risk factors is the same in both samples. For this you need a computer program for medical statistics (PEKER) to compare the mean values and the spread (standard deviation, SD) of the observations. Let us say that System X in Year 1 had 330 patients with a mean ISS of 12.5 (SD 10.3) and the system in Year 2 had 440 patients with mean ISS of 9.8 (SD 11,1). Let us also assume that the local scene of injury changed so that total prehospital transport time (Time 2 PEKER) was 5.5 hours ( SD 3.1) in Year 1 and 3,0 hours (SD 4,1) in Year 2. This is the result of confidence interval comparisons:

System X

95%CI for ISS difference

95%CI for time difference

Year 1 versus Year 2

1.2 – 4.2 ISS points

2.0 – 3.0 hours


So, both anatomical severity and prehospital transit times are clearly significantly different in Year 1 and in Year 2. That makes the two patient populations not statistically comparable regarding outcome indicators. So, our statistical comparisons on the previous page are not valid; we have to stratify the two populations before doing outcome comparisons.



Take out slices/subsets of the two populations that have (approximately) the same distribution of the main risk factors. E.g.:


Subset 1

Subset 2

Subset 3


Moderate: < 9

Serious: 9 – 15

Major Trauma: > 15

Time 2

< 2 hours

2 – 4 hours

> 4 hours


> 10

8 – 10

< 8


Now we are ready to compare outcome indicators (mortality, sepsis etc.) by year for each subset separately. Notice that stratification criteria should be defined after careful studies of the distribution and the impact of the actual explanatory variables (scatter plots and ROC analysis PEKER).



Quality comparisons between trauma systems are invalid if the social context is significantly different. E.g. it is not fair to compare fatality rates of high-tech urban Western systems with grassroots Chain-of-Survival Model systems in the rural South – even after stratification. Also differences in distribution of statistics make inter-system comparisons invalid (PEKER).




Severity scoring: TRISS

Existing systems for trauma severity scoring are based on studies of urban trauma in the North and are inaccurate in severity classification of rural trauma in the South. Accurate classifications are important for two reasons:

  1. To triage and monitor the treatment of individual patients.
  2. To monitor the quality of performance in trauma systems and identify patients with unexpected outcome, e.g. patients dying despite survival risk ratio < 0.5.

The international “golden standard” for severity scoring and probability of survival (Ps) calculation is the Trauma and Injury Severity Score/TRISS (PEKER). TRISS is a composite calculator based on anatomical (ISS PEKER) and physiological (RTS PEKER) severity indicators. For TRISS operation, the RTS parameters (RR, BP, GCS) are vected, vectors being deducted by logistic regression analysis on large US trauma databases:

RTS (vected) =  0.9386(GCS code) + 0.7326(BP code) + 0.2908(RR code). RTS can thus take on values from 0 to 7.848.

                                                               – b

TRISS is based on a probability distribution: Ps =  1 / (1 + e)

The value of b is set by the regression equation: b = b0 + b1(RTS) + b2(ISS) + b3(AGE). AGE is defined as a dichotome variable: (AGE ≤55 years) = 0; (AGE >55 years) = 1.The value of b is set separately for blunt and penetrating injuries.


Problems with TRISS

  1. ISS is context dependent: The AIS codes are based on what is “moderate” or “critical” in high-tech US urban trauma systems.
  2. RTS is context dependent: The vectors are derived from studies of well-fed, (mainly) healthy Westerners.
  3. AGE is context dependent: “Old” being defined as age >55 years does not apply to all trauma settings.
  4. The time factor is not included: Being briefly hypotensive (BP = 70, RTS code = 2) and being hypotensive for 3 hours (BP = 70, RTS code = 2) makes a big difference (PEKER).
  5. Methodological failure – 1:  The GCS parameter in RTS is inaccurate. The GCS scores are skewed toward motor response impact, patients with the same score may have significantly different Ps. There are also rater failures, particularly for inexperienced raters and for scoring at intermediate levels of consciousness. The rate of GCS scoring failures may be as high as 50% (PEKER).
  6. Methodological failure – 2: In logistic regression analysis there should not be co-linearity between the predictors. But RR, BP, and GCS are tied, all three variables being indicators of oxygen starvation, which may give false high TRISS prediction.
  7. Methodological failure – 3: The TRISS goodness-of-fit is low in your study population if the distribution of predictors differs significantly from the distribution in the US reference population.

Conclusion: Make your own “TRISS” !



Severity scoring: Make your own severity calculator

In the TCF trauma systems we use a modified and simpler version of the RTS. We have replaced the Glasgow Coma Scale with a 5-ranked grading of consciousness level. Also the intervals for ranking of respiratory rate/RR and systolic blood pressure/BP is set differently from the US version of RTS:


Consciousness level



Coded value


> 90

10 – 24



70 – 90

25 – 35


Coma, responds to sound

50 – 69

> 35


Coma, responds to pain only

< 50

< 10


No response

No carotic pulse

No breathing



Test calculator accuracy

We tested the accuracy of this Physiological Severity Score/PSS using Receiver Operating Characteristics/ROC analysis in 700 mine- and war injured patients in Iraq and Cambodia, a patient population that had a mean prehospital transit time of 5.5 hours. The study question was: how accurate is PSS in predicting risk of trauma death? We found that the accuracy of PSS in that actual population was high (area under ROC curve 0.93). Interestingly we found that one single risk indicator, respiratory rate after in-field pain relief/RR2, predicted trauma death as well as the comprehensive PSS calculator.



Where the ROC curve comes closest to the upper-left corner is the optimal cut-off for that variable. The graph demonstrates that respiratory rate remaining > 25/minute after pain relief is a critical sign.


Scrutinize patients with unexpected outcome

With a solid physiological severity calculator and the ISS you can study patients with unexpected outcome in a scatter plot. The diagonal on the graph is the line of

50% probability of survival/Ps. Patients under the line (lower-left)

had  had injuries of low severity, Ps >50,

but they died; they are the unexpected negatives.

The patients to the upper-right had Ps <50 and still

survived, they are the unexpected positives. From

these two categories of patients we have a lot to

learn. The trauma team should do “file autopsy” on all

these cases to identify failure and success factors.

Decision-tree analysis may be a useful method to study

this subset.

Leave a Reply