A Hybrid Visual Analytics Framework to Evaluate Trauma

Incidences and Enhance Patient Care

Waqar Haque

1 a

, Jordan Oliver

2 b

, Sonal Bajaj

2 c

and Navjot Kaur

1 d

Department of Computer Science, University of Northern British Columbia, Prince George, Canada

Northern Health, Prince George, Canada

Keywords: Healthcare Analytics, Trauma Care, Visual Analytics, Online Analytical Processing.

Abstract: Large volumes of data has been collected on unintentional injuries and mortality. Besides years of life lost,

traumatic injuries account for a significant portion of healthcare expenditure. With intelligent visual analytics,

the collected data can be used for informed decision making and resource allocation. A multi-dimensional

online analytical processing (OLAP) cube has been developed using data from BC Trauma Registry (BCTR)

and Discharge Abstract Database (DAD). We propose a comprehensive framework which uses the OLAP

cube, a web-based data entry platform populating a standalone database, filters consistent with Accreditation

Canada (AC) measures, AC inclusion/exclusion criteria, and tools which render reports from historical and

operational perspectives. These reports are grouped in higher level categories with intuitive drill-down

capabilities for navigating at finer granularity along multiple dimensions. Access control for data entry is

enabled with provisions for nested groups.

1 INTRODUCTION

Trauma is a global health concern and traumatic

injuries can have an impact on individuals leading to

long term disability and even deaths in some cases

(Provincial Health Services Authority, n.d.).

Depending upon the severity of injury, trauma can be

classified as minor, moderate or major. “Major

trauma is the principal cause of death for people under

age 45 in Canada. Over 700,000 people are injured

each year in BC and more than half of them seek

medical attention. Of these 700,000 British

Columbians injured each year, about 1,800 die, 9,000

suffer permanent disability, 27,000 are hospitalized

and an estimated 41,000 potential years of life are

lost. The yearly cost of injury exceeds $2.8 billion in

BC and is close to $19.8 billion nationally. Falls and

motor vehicle-related injuries are the main causes of

injury-related hospitalization in BC.” (Provincial

Health Services Authority, n.d.) This raises a

compelling question whether any of these injuries or

https://orcid.org/0000-0002-6921-8097

https://orcid.org/0000-0002-7031-0528

https://orcid.org/0000-0002-9323-9453

https://orcid.org/0000-0001-5542-7501

deaths could have been prevented by providing better

trauma care and timely assistance.

To analyze the factors associated with each

specific trauma case and to identify gaps in trauma

care, the aid of analytical tools is vital. Business

Intelligence (BI) is defined as “broad category of

applications and technologies for gathering, storing,

analyzing, sharing, and providing access to data to

help enterprise users make better business decisions”

(Ali -Özkan, Crvenkovski, & Johnson, 2016). BI can

integrate data from a wide variety of internal and

external data sources and provide an effective

information platform for decision makers in

healthcare (Foshay & Kuziemsky, 2014). The

extracted information can then be displayed in the

form of an intuitive dashboard providing a visual

analytics platform for decision makers to take

appropriate actions and enable enhanced trauma care.

The metrics of interest include injury severity score

(ISS), location of incidence, time taken for trauma

care to be provided, type of trauma care provided,

transport information, risk groups, and mortality rate.

Haque, W., Oliver, J., Bajaj, S. and Kaur, N.

A Hybrid Visual Analytics Framework to Evaluate Trauma Incidences and Enhance Patient Care.

DOI: 10.5220/0008894500230033

In Proceedings of the 13th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2020) - Volume 5: HEALTHINF, pages 23-33

ISBN: 978-989-758-398-8; ISSN: 2184-4305

The data is collected by various Health Service

Delivery Areas (HSDA) in several formats and sent

to the Provincial Health Services Authority (PHSA)

for standardization and use in an integrated provincial

trauma system. Over the last few years, the Trauma

Services of BC wing of PHSA has developed a data

warehouse and built a very stable online analytical

processing cube. Several measures and dimensions

have also been identified (Microsoft). We have

integrated this cube with a database thus creating a

hybrid platform to provide the integrated reporting

layer together with ASP.NET based web forms to

enter data related to Trauma Team Activations (TTA)

and other local data sources.

2 RELATED WORK

The application of BI concepts in healthcare is

growing rapidly. Many healthcare institutes (Institute

for Health Metrics and Evaluation (IHME), n.d.)

(World Health Organization (WHO), n.d.)

(University of Utah, n.d.) (Fojut, 2016) (The Trauma

Audit & Research Network (TARN), n.d.) (UMC

Utrecht, n.d.) have used their own data repositories to

develop online analytics and visualization tools for

various diseases (such as cardiovascular, diabetes,

cancer, etc.) and injuries. The Institute of Health

Metrics and Evaluation (IHME) provides a tool which

analyzes data (1990-2016) related to premature

deaths, disabilities, and injury from over 130

countries and can be visualized along several

dimensions including demographics, mortality,

disease causes and risk factors. The visualization is

available in different formats such as map, treemap,

line chart, patterns bar chart, pyramid chart, arrow

chart and heat map. The dashboard can be drilled

down to specific countries and states (Institute for

Health Metrics and Evaluation (IHME), n.d.). The

WHO’s GHO (Global Health Observatory) is a web-

based analysis and visualization tool for global health

priorities (World Health Organization (WHO), n.d.).

The data presented is in form of 35 different theme

pages where each theme presents summary and

reports on global health conditions such as

HIV/AIDS, child health, tuberculosis and road safety.

These dashboards are not completely trauma data

centric and do not provide any information about the

causes of trauma related incidence and deaths, or

trauma patient movement.

The Intermountain Injury Control Research Center

(IICRC) at University of Utah is the central data

repository for trauma-related data. IICRC compiles

and analyzes trauma data making it available for

research (Bureau of Emergency Medical Services &

Preparedness, Utah Department of Health, 2015). A

Trauma Audit dashboard is developed by University

of Utah School of Medicine and is available at state,

hospital and region level. The key performance

indicators (KPIs) used in the dashboard are age,

population, ISS, mortality, field triage, level of care,

transfers (inter or intra-hospital) and mode of

transport amongst others (University of Utah, n.d.).

Centura Health developed a trauma BI application for

monitoring hospital’s trauma system performance

using data from trauma registry and other facilities

(Fojut, 2016). The dashboard is made of pivot tables

allowing users to analyze and report with drill-down

capabilities. The reports can be filtered by injury type

using ICD codes and can be further sliced by month,

fiscal period, quarter, facility and regional groups.

The reports provide insights into financial measures

like cost per case or contribution margin.

The Trauma Audit and Research Network

(TARN) (The Trauma Audit & Research Network

(TARN), n.d.) has one of the largest trauma database

in Europe, receiving data from 80% of trauma centers

across the United Kingdom. TARN measures and

monitors trauma care process and provides local,

regional and national information on trauma patient

outcome. The Major Trauma Dashboard provides

quarterly and ad-hoc reports to the clinicians and also

published population statistics on the epidemiology

of trauma. Jedox (Jedox, n.d.) and Celcus B.V.

(Celcus, n.d.), Netherlands based companies, have

developed a trauma BI application for University

Medical Centre Utrecht’s (UMC Utrecht) (UMC

Utrecht, n.d.) trauma centre. This application

integrates data collection, reporting and analysis

process. Users can enter data manually and the

application validates and transforms the data. The

data is then stored in relational data sources and is

analyzed and visualized by using dashboards and

dynamic reports.

3 METHODOLOGY

The purpose of our proposed framework is to provide

healthcare professionals and decision-makers with

real-time aggregated information on regional trauma

cases, trauma team activations and other trauma-

related information. The primary source of data is the

OLAP cubes maintained by the Provincial Health

Services Authority (PHSA) using the Discharge

Abstract Database (DAD) and the British Columbia

Trauma Registry (BCTR). The BCTR data includes

moderate and severe trauma whereas the DAD cube

HEALTHINF 2020 - 13th International Conference on Health Informatics

also includes cases categorized as minor trauma. Both

of these cubes are coupled with a relational database

for comprehensive reporting and regional data entry

through web forms. Microsoft’s BI tool stack

consisting of SQL Server, various business

intelligence components and ASP.net framework has

been used for data integration, reporting and analysis

[14]. A challenge in designing the framework was the

integration of disparate data files from different data

sources. An abstract level interaction diagram of the

interrelated components of the proposed framework

is shown in Figure 1.

Figure 1: Framework Components.

Users interact with the framework through data entry

components and reporting interfaces. The data entry

web forms are accessible from remote sites and can

be used to populate regional information such as

Trauma Team Activations (TTA), ground and air

ambulance services, and staff credentials. This

information is currently outside the scope of OLAP

cubes maintained by PHSA and thus is captured

directly at designated sites. The web interface gives

the user the ability to easily edit, delete and validate

the underlying data. The information entered is stored

in a relational database which is independent of the

OLAP cubes, but integrated in the reporting module.

The relational tables include information such as

mechanism/motivation of injury, admitting services,

methods of transport, mortalities, sites, and trauma

team activations. This database has logical

relationships between tables so that data can be

accessed and re-assembled by the SQL Server

Reporting Services (SSRS) to extract meaningful

information. Likewise, the reporting component

securely connects users to the aggregated information

and drill downs of the underlying data. It renders

different dashboards grouped by specific sets of

reports such as Incidence, Mortality, Transportation,

and TTA. These reports pull information from the

database and cubes to provide a visual representation

of various KPIs in the form of intuitive graphs, pie

charts, tables, and bar/line graphs.

A server-based reporting platform, SQL Server

Reporting Services (SSRS), has been used to create

the reporting dashboard. Besides providing a

visualization platform, SSRS has provision to export

reports to applications such as MS Excel, MS Word,

PDF, etc. In addition, SSRS provides the ability to

create reports with drill through/drill down actions,

expand/collapse toggles, tooltips and report

parameters for interactive navigation. The SSRS

reports can also be filtered, grouped and sorted

according to user’s requirements. Once deployed, the

reports can be viewed by multiple users via a standard

browser through a web-based connection and does

not require any configuration of the client machine.

Users can open data entry web-form and reporting

dashboard in different windows and can use both

components at the same time. Access control is

accomplished through an Active Directory. Data

entry can only be done by designated sites/staff

whereas the reports can be viewed broadly upon

authentication.

4 DATA VISUALIZATION

The reporting module consists of over forty reports

representing data from the sources mentioned earlier.

The top level dashboards display aggregated results

in visually appealing formats and allow intuitive

navigation to finer granularity via drill down and drill

through reports. The framework allows tracking of

several metrics including transportation times, injury

severity, most responsible diagnosis, inter-facility

transfers for higher level of care,

mechanism/motivation of injury, trends, all broken

down by gender, demographics and other specified

parameters. These dashboards not only provide

access to useful information but also enable

administrators to make more informed decisions.

Where applicable, historical data is used to show

trends. The reports are highly optimized and render

within 2-7 seconds.

4.1 Landing Page

The landing page (Figure 2) provides an overview of

the dashboard together with the ability to navigate

through various web pages. The page is divided into

two sections for reporting and data entry. The

reporting section is further divided into three sub-

sections based on the data sources: BCTR, DAD and

Pre Registry Database. As stated earlier, BCTR

reports only include moderate and major trauma cases

A Hybrid Visual Analytics Framework to Evaluate Trauma Incidences and Enhance Patient Care

Figure 2: Landing Page

whereas DAD also includes trauma cases which are

categorized as ‘Minor Trauma’. The icons on the

left side navigate to different dashboards representing

the BCTR data (Incidence, Mortality, and Transport).

In addition the PHSA Reports icon allows viewing of

static annual reports which are periodically uploaded

through the data entry section. The next three icons

produce similar reports for the DAD data. In

addition, a Comparison icon allows selection of up to

three facilities for comparison purposes. Finally, the

TTA icon generates a comprehensive report

displaying various metrics related to trauma team

activations. The data entry section presents different

web forms for entering data which is not supplied by

PHSA cubes. This includes Early Fixed Wing

Launch, TTA, and uploading of Annual Reports

compiled by PHSA.

4.2 TTA Web Form

The Trauma Team Activation web form (Figure 3)

provides an interactive mechanism to enter

information/records of all admitted cases where the

trauma team was activated and those where the

admitted patient met TTA criteria, but an activation

did not occur. The data captured includes different

sites, demographics, pre-registry number, emergent

care (TTA Activated, Mechanism of injury,

Motivation of injury, etc.), disposition (Admitting

services, OR arrival time Transferred, etc.) and

outcome information (Length of Stay, ISS, Vital

Status). This web-form is highly interactive with

ability to search data, and enforces data integrity via

validation constraints in real time. Existing records

are displayed in a grid view under the form which

allows easy editing and deletion of records. Any

changes made through the web form are reflected

immediately in the associated reports.

4.3 Incidence (BCTR)

The Incidence dashboard provides a snapshot of

trauma incidence related statistics (Figure 4). The

report header displays the total number of major

trauma cases for the selected fiscal year and facility.

A trend chart shows the number of moderate and

severe trauma incidences over the last five years. The

trauma incidence count is broken down by KPIs such

as age-range, gender, Injury Severity Score (ISS),

motivation/mechanism of injury, risk groups, falls,

and alcohol related injuries. The mechanism of injury

section provides transport and non-transport related,

and work related trauma incidence statistics. These

charts can be further drilled down for information on

the specific KPI at finer granularity. In addition, there

are several buttons on the dashboard which allow

navigation to other drill-down reports. For example,

Figure 5 shows the trend in Emergency Department

HEALTHINF 2020 - 13th International Conference on Health Informatics

Figure 3: Trauma Team Activation Data Entry Web Form.

Figure 4: Trauma Incidences (BCTR).

A Hybrid Visual Analytics Framework to Evaluate Trauma Incidences and Enhance Patient Care

Length of Stay (ED LOS) by months over the selected

fiscal year. The blue bars show the total number of

cases while the orange bars show the total number of

ED LOS days for each month. The monthly trend of

average patient length of stay is also displayed. For

the fiscal year 2016/17, the total number of trauma

cases admitted to the emergency department was 78

with a total of 177 emergency department days. Based

on this, it is observed that a trauma patient spends on

average two days in ED.

Figure 5: Length of Stay drill-down.

4.4 Mortality (BCTR)

The mortality report (Figure 6) uses BCTR data to

report on the number of trauma related deaths by

facility and fiscal year. A breakdown of this number

by regions or local health authorities is shown on an

interactive map. The mortality trend chart shows the

number of deaths that occurred in the last five years

with respect to the total trauma incidence count and

severity level of the injury. Additional indicators

displayed on the dashboard allow the user to view

other parameters related to trauma mortality. For

instance, the numbers are broken down by gender and

age-group which illustrates that more male trauma

patients died than female in the selected year.

The monthly bar graph shows that more deaths

were reported in November and December in

comparison to other months. The mortality by

location pie chart shows that three deaths occurred in

the Intensive Care Unit (ICU), two in the Nursing

Unit and one in the Operating Room. Clicking on this

pie chart renders a detailed drill down that displays

the location of the deaths by ISS level. Similarly, the

other charts (tables, graphs) explain various trends

and changes in mortality rates by falls grouped by

four main ISS levels. The three tables at the end of

the report allow the user to easily identify the

mechanisms, types and motivations of injury that

contribute to the patients’ deaths.

4.5 Transport (BCTR)

Transportation times are a key factor in the outcome

from a traumatic injury. The ultimate goal is to

identify bottlenecks and minimize transfer times.

Figure 7 shows the average time it took for a trauma

patient to arrive at a definitive care facility either from

scene of injury, or as a transfer. The transfer times are

also shown by the injury severity level. For instance,

it can be seen that all severe trauma patients (ISS>12)

arrived at the definitive care facility within four

hours. A breakdown of the number of cases

transported from scene and/or first facility to

definitive care is an important metric to analyze

patient’s journey and identify bottlenecks, if any.

These charts also show the distribution of patient

transport modes such as land ambulance, private

transport, and fixed-wing or helicopter ambulance.

The trend line shows the average time to definitive

care for the previous five years which allows one to

assess the impact of resource allocation, if any.

Drill-down reports provide a monthly breakdown of

this information. Often, patients are transferred to

facilities where a higher level of care can be provided.

This information is also displayed on the dashboard.

In cases where such transfer occurs directly from

scene, a report on regional transfers provides relevant

information for the selected facility. Finally, the

dashboard also allows drilling down to another KPI

which shows the number of cases that required a

higher level of care by different risk groups (i.e.

penetrating, truncal, elderly, and polytrauma).

A similar set of reports is also available for the

Discharge Abstract Database (DAD). In addition to

the metrics defined earlier, this cube also contains

data for minor trauma and provides a finer granularity

comprising of all locations whether or not those are

designated as trauma care facilities. Thus, the number

of cases in BCTR is a subset of those included in the

DAD cube. However, the ISS scores are not captured

in the latter.

HEALTHINF 2020 - 13th International Conference on Health Informatics

Figure 6: Trauma Mortality (BCTR).

A Hybrid Visual Analytics Framework to Evaluate Trauma Incidences and Enhance Patient Care

Figure 7: Transport Report.

4.6 Comparison

Trauma statistics at facilities which are similar in size

and resources can be compared to obtain useful

information. This is accomplished by a comparison

report (Figure 8) which allows selection of up to three

facilities for side-by-side comparison. The top section

of this report provides a comparison of general

trauma statistics such as incidence, mortality count,

mortality percentage, complications, comorbidities,

and age range. The trauma statistics section provides

a more comparison of other metrics such as injury

type and motivation/mechanism of injury. The

transfers section of the report shows the discharged

and transferred cases by mode of transport for each of

the selected facilities. With the ability to compare

facilities side-by-side, executives/users have the

ability to determine how incidences and mortalities

differ across similar facilities.

HEALTHINF 2020 - 13th International Conference on Health Informatics

Figure 8: Facilities Comparison.

4.7 TTA

Trauma Team Activation (TTA) is the pre-hospital

notification by Emergency Medical Services (EMS)

based on scene triage criteria that is provided to the

emergency department. This ensures that the trauma

team is immediately available upon patient arrival to

address the clinical needs of seriously injured

patients. This information is entered through web

forms and can be simultaneously seen in the reporting

module. The report header displays the total number

of trauma team activations (split by gender) for the

selected fiscal year and facility (Figure 9). An

aggregated overview of the data entered is displayed

in top portion of this report. Some of the key statistics

include inpatient trauma consults, number of TTAs

cohorted, number of early fixed wing launches,

average length of stay and peak periods of arrival.

Average times and total number of cases for various

departments (CT, Hospital, OR, Ward, SCU and ICU)

are also presented.

The report further groups the TTA cases by ISS,

month, age range, motivation/mechanism of injury,

admitting services, mode of transport and transfers.

For instance, Figure 8 shows a total of 91 TTAs, out

of which 6 died in ER, ICU and Ward. For 2017/18,

A Hybrid Visual Analytics Framework to Evaluate Trauma Incidences and Enhance Patient Care

July was the peak month with twelve trauma team

activations. Additionally, the highest number of TTA

injuries had a reported Injury Severity Score (ISS)

less than 9. A total of 31 cases were transferred into

UHNBC, out of which 15 arrived directly from the

incidence scene. The highest number of cases were

transferred out to Vancouver General Hospital.

Each chart in the TTA report can be further drilled

down to compare the TTA data of the selected

Facility with other Trauma sites. Figure 10 is an

example of the drill down from “Cases by Age-Range

& Gender” chart. The drill down report shows a

higher number of overall TTA cases for males when

compared to females. Across all three TTA sites, the

highest number of cases are recorded for the age-

rnge 15-24, followed by 25-34 and 35-44.

Figure 10: TTA Cases by Age and Gender.

Figure 9: Trauma Team Activations (TTA).

HEALTHINF 2020 - 13th International Conference on Health Informatics

5 CONCLUSION

A large amount of data related to trauma incidences

has been residing in multiple repositories in legacy

formats. A multi-dimensional OLAP cube was built

by PHSA, but the data still remained in a format

which could not be used for decision making. We

have proposed a framework which allows

visualization of this data along several KPIs. The

dashboard provides a platform which can be easily

navigated to reports at finer granularity. The reports

are grouped on the basis of incidences, mortality,

transportation and trauma levels, and are tightly

coupled with web forms which capture data not

included in the OLAP cubes. This creates a hybrid

environment where data is extracted from both cubes

and relational databases in a transparent manner.

Simultaneous users can access the interface through

standard web browsers while access control is

implemented using Active Directory.

The interactive platform allows decision makers

to view trends, bottlenecks and performance along

multiple dimensions to identify areas which can result

in improved trauma care. The platform is extendable

to seamlessly include other health service delivery

areas and add other metrics of interest.

ACKNOWLEDGEMENTS

This work was supported by the research

collaborative grant RC15-3180 received from

Northern Health British Columbia, Canada. Other

students who contributed to this work include Devin

Calado, Giridhar Krishnan, Helal Lutfi and Lila

Mansour.

REFERENCES

Ali -Özkan, O., Crvenkovski, P., & Johnson, H. (2016).

Using a business intelligence data analytics solution in

healthcare. IEEE 7

Annual Information Technology,

Electronics and Mobile Communication Conference

(IEMCON), (pp.1-6) doi:10.1109/IEMCON.2016.77463

Bureau of Emergency Medical Services & Preparedness,

Utah Department of Health. (2015). Traumatic Injury

Outcomes in Utah's Trauma System, 2001-2013. Salt

Lake City: Division of Family Health and Pre-paredness.

Retrieved 09 10, 2018, from https://bem sp.utah.gov/wp-

content/uploads/sites/34/2016/05/utr_2013.pdf

Celcus. (n.d.). Retrieved September 3, 2018, from

http://www.celcus.nl/en

Fojut, R. (2016, February 22). Trauma System News.

Retrieved August 9, 2018, from http://www.trauma-

news.com/2016/02/how-a-trauma-system-uses-

business-intelligence-to-evaluate-protocols-track-

costs-monitor-quality-optimize-billing-and-more/

Foshay, N., & Kuziemsky, C. (2014). Towards an

implementation framework for business intelligence in

healthcare. International Journal of Information

Management, 34(1), 20-27.

Institute for Health Metrics and Evaluation (IHME). (n.d.).

GBD Data. Retrieved September 3, 2018, from

http://www.healthdata.org/gbd/data

Jedox. (n.d.). Retrieved September 3, 2018, from https://

www.jedox.com/en/

Microsoft. (n.d.). Retrieved June 29, 2018, from https://ww

w.microsoft.com/en-us/sql-server/business-intelligence

Microsoft. (n.d.). What Are Measures and Dimensions?

Retrieved July 1, 2018, from https://docs.micro

soft.com/en-us/biztalk/core/what-are-measures-and-

dimensions

Provincial Health Services Authority. (n.d.). Trauma

Services BC. Retrieved July 18, 2018, from http://www

.phsa.ca/our-services/programs-services/trauma-servic

es-bc

The Trauma Audit & Research Network (TARN). (n.d.).

Retrieved July 15, 2018, from https://www.tarn.ac.uk/

UMC Utrecht. (n.d.). Retrieved July 15, 2018, from

https://www.umcutrecht.nl/en/1

University of Utah. (n.d.). Trauma Audit Dashboard.

Retrieved August 3, 2018, from http://www.utahtr

auma.org/registryMembers/documents/TraumaAuditD

ashboardUserManual.pdf

World Health Organization (WHO). (n.d.). Global Health

Observatory. Retrieved September 3, 2018

A Hybrid Visual Analytics Framework to Evaluate Trauma Incidences and Enhance Patient Care