Towards an Intelligent Triage Bracelet: A Conceptual Study of a

Semi-Automated Prehospital Triage Algorithm and the Integration of

Blood Pressure Measurement

Lorenz Gr

unerbel

, Ferdinand Heinrich

, Oliver Zett

, Kristjan Axelsson

and Maximilian Schumann

Fraunhofer EMFT, Hansastr. 27d, Munich, Germany

Keywords:

First Responder, Automated Triage System, Mass-Casualty Incident, Blood Pressure Monitoring, AI, Machine

Learning.

Abstract:

First responders often reach their limit when they have to ﬁnd and triage dozens of victims in a mass-casualty

incident. However, a delay in treatment directly affects the survival chances of seriously injured people. A

method to reduce the time for the prehospital triage could potentially save lives. Hence, this work outlines the

conceptual development of an intelligent bracelet that semi-automates the prehospital triage. This bracelet is

supposed to enable non-professional ﬁrst responders to help with the triage, which maximises the utilisation

of man power at a mass-casualty incident. The bracelet should automate the part of the triage that is based

on vital, position and movement data and it should guide through the necessary patient interactions. As a step

towards this goal, this work proposes a semi-automated triage algorithm that is based on mSTaRT. One of the

challenges to implement this concept is to measure the blood pressure with a small and easy to attach system.

Therefore, this work presents a wrist worn oscillometric blood pressure measurement prototype. Furthermore,

we discuss the use of machine learning methods to forecast triage level changes.

1 INTRODUCTION

After a mass-casualty incident, such as a train crash or

an environmental disaster, a large number of victims

immediately need emergency treatment. However, of-

ten only a few ﬁrst responders are quickly available.

Upon arrival, the ﬁrst responders, with the necessary

training, perform a comprehensive prehospital triage

of all victims to categorize them depending on the

severity of their injuries and consequently deﬁne the

order of treatment. It is impossible to immediately

treat everyone. Furthermore, a simultaneous monitor-

ing of all patients is not realizable, hence a sudden

deterioration of a patient’s health can remain unde-

tected.

Innovative technical assistance could accelerate

the triage process, which saves time for victim treat-

https://orcid.org/0000-0002-1932-481X

https://orcid.org/0000-0002-4626-2394

https://orcid.org/0000-0002-4074-9310

https://orcid.org/0000-0002-5044-2152

https://orcid.org/0000-0002-2528-2492

ment. Additionally, it makes continuous victim mon-

itoring possible. We hypothesize that the right tech-

nical assistance enables ﬁrst responders without dedi-

cated triage training to perform a semi-automated pre-

hospital triage. Consequently, this would maximize

the utilisation of man power at a mass-casualty inci-

dent.

To develop a semi-automated prehospital triage

concept we ﬁrst analysed which part of the modiﬁed

Simple Triage and Rapid Treatment (mSTaRT) triage

algorithm, as it is shown in the work of Paul et al.

(Paul et al., 2009), can be automated. We believe that

all necessary vital data recordings and the embedded

guidance for ﬁrst responders should be integrated in a

small, cost-efﬁcient and easy to attach sensor system.

Hence, our proposed solution is an Intelligent Triage

(ITRI) bracelet to be worn at the wrist. The continu-

ous non-invasive acquisition of relevant vital data in

this scenario is difﬁcult. This work focuses on the in-

tegration of wrist worn blood pressure measurement.

The acquired continuous vital data might not only be

used to semi-automate the triage decision, but it might

Grünerbel, L., Heinrich, F., Zett, O., Axelsson, K. and Schumann, M.

Towards an Intelligent Triage Bracelet: A Conceptual Study of a Semi-Automated Prehospital Triage Algorithm and the Integration of Blood Pressure Measurement.

DOI: 10.5220/0011746200003414

In Proceedings of the 16th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2023) - Volume 1: BIODEVICES, pages 169-176

ISBN: 978-989-758-631-6; ISSN: 2184-4305

 2023 by SCITEPRESS – Science and Technology Publications, Lda. Under CC license (CC BY-NC-ND 4.0)

169

be used to forecast deteriorating patient conditions.

2 CURRENT PREHOSPITAL

TRIAGE SYSTEMS

The procedure for a prehospital triage in the case of

a mass-casualty incident (MCI) is deﬁned by national

regulations or laws. In this work we will focus on

the mSTaRT triage algorithm, as it is shown in the

work of Paul et al. (Paul et al., 2009), because the

Bavarian directive about MCIs incorporates a ”Bavar-

ian Model” of the mSTaRT algorithm (Schm

oller and

Hagen, 2017).

The prehospital triage in the event of MCI is or-

ganised as follows: The ﬁrst arriving paramedic au-

tomatically becomes the operational leader. To es-

timate the number and severity of casualties every

available ﬁrst responder with dedicated triage train-

ing starts in a team of two with a ﬁrst comprehensive

triage. Patients with a chance of survival are grouped

in 3 categories: SK 1 (red) - life threatening - imme-

diate care/transport, SK 2 (yellow) - seriously injured

- transport based on urgency, SK 3 (green) - slightly

injured - transport when available. Every patient that

is able to walk is categorised in SK3 (green). If a life

ending injury is present a physician needs to conﬁrm

the death. Every patient that is not able to walk is

assumed to be SK 2 (yellow). If one condition from

a checklist of immediate life threatening conditions

is met the patient is classiﬁed as SK 1 (red). This

checklist includes checks for spontaneous breathing,

a reasonable breathing rate, spurting bleeding, radial

pulse and a check whether the patient can execute

simple commands. Patients are marked with paper

triage cards and barrier tape according to their triage

level. Classically the situation overview has to be ag-

gregated manually (Schm

oller and Hagen, 2017; Paul

et al., 2009).

Ideas for how to improve the prehospital triage

with technical aid were already published. In the fol-

lowing the use of vital sensors and the triage automa-

tion incorporated in these systems is recapitulated.

The commercially available system RescueWave

is an electronic replacement for the triage card that is

coupled with an application that tracks and visualizes

severity and position of each patient and also supports

transport organization. No vital data is recorded (Res-

cueWave, 2022).

In the literature multiple advanced electronic

triage tags that enable victim tracking, continuous vi-

tal data monitoring and medical data exchange over

information networks have been presented. The Wire-

less Internet Information System for Medical Re-

sponse in Disasters (WIISARD) limits the use of vi-

tal data to a pulse oximeter (Lenert et al., 2011). The

CodeBlue project focused on the sensor network for

emergency responses and demonstrates vital data ag-

gregation with an exemplary pulse oximeter and a

two-lead electrocardiogram (ECG). The authors ex-

pect that more vital data sensors will be available in

the future (Lorincz et al., 2004). For the Advanced

Health and Disaster Aid Network (AID-N) an elec-

tronic triage tag that integrates a pulse oximeter, blood

pressure arm cuffs and a two-lead ECG was devel-

oped. Vital data is continuously monitored and an

alarm is raised on patient speciﬁc thresholds (Gao

et al., 2007). The authors of the AID-N network point

out that an arm cuff to measure non-invasive blood

pressure has to be attached properly to deliver accu-

rate sensor reading which might be challenging with

clothed patients. Furthermore, the cuff adds extra ma-

terial cost and makes the system much larger (Gao

et al., 2006).

Automation of the triage decision based on elec-

tronic triage tags has also been studied in the litera-

ture. The eTriage tag comes in a version that mea-

sures blood oxygen saturation (SpO2) and heart rate

with a pulse oximeter and an extended version that

additionally measures the breath rate via a nasal can-

nula. The triage category can either be manually set

by the triage ofﬁcer or in a simple triage mode the

priority is deduced from heart rate and SpO2 thresh-

old values (Sakanushi et al., 2013). A Korean elec-

tronic triage system measures the heart rate, the SpO2

and the respiratory rate with a pulse oximeter. Ad-

ditional information including the body temperature,

the blood pressure, the consciousness of the patient

and the walking ability can be set by a ﬁrst responder

via a mobile app. The triage category is automatically

determined based on a checklist of combined patient

conditions (Park, 2021). One German triage tag in-

troduced the concept of pre triage levels as a general

and simple reference. To determine these pre triage

levels accelerometer-based activity classiﬁcation dif-

ferentiates walking patients, i.e. minor affected, and

lying or sitting patients, i.e. major effected. Addition-

ally, a pulse oximeter records the SpO2 as well as the

heart rate and thresholds for triage levels are deﬁned.

During real emergencies it was found that especially

minor affected patients tend to temporally stop mov-

ing without clinical relevant reasons, which hindered

the detection of minor affected patients based on ac-

tivity classiﬁcation. The vital data based triage clas-

siﬁcation was better at identifying the minor affected

patients but struggled to clearly identify the major af-

fected ones (Rodriguez et al., 2014).

BIODEVICES 2023 - 16th International Conference on Biomedical Electronics and Devices

170

3 PROPOSED PREHOSPITAL

TRIAGE SYSTEM

We envision an intelligent triage system that utilizes

the available man power at a mass-casualty incident

by enabling ﬁrst responders without explicit triage

training to perform the prehospital triage. Therefore,

we introduce the concept of an easily attachable Intel-

ligent Triage (ITRI) bracelet that records vital, posi-

tion as well as movement data and features embedded

triage guidance. With this system patients could be

monitored continuously to detect condition changes.

Additionally, machine learning methods could fore-

cast triage category changes. The location of all pa-

tients could be shown on a central device with an

overview map to give the operations manager a clear

situation picture. Figure 1 displays a schematic rep-

resentation of this system. It would allow the ﬁrst

responders to monitor the whole mass-casualty event

well-arranged and continuously over the total ﬁrst aid

intervention period.

AI methods:

condition forecasting

vital data

slightly injured

SK III

seriously injured

SK II

life-threatening

SK I

semi automated triage:

ITRI mSTaRT

position and movement

Figure 1: ITRI-system vision: The ITRI bracelet records

vital data as well as position and movement data. Based

on this data the triage levels are determined with a semi-

automated algorithm, which is based on mSTaRT. AI meth-

ods are used to forecast condition changes. On a central

device the health state of all patients during a mass-casualty

incident can be monitored.

3.1 Prehospital Triage Algorithm

As argued before, the semi-automated prehospital

triage algorithm of the ITRI bracelet will be based on

the mSTaRT algorithm (Paul et al., 2009). We iden-

tiﬁed the parts of the mSTaRT scheme which might

be automated with vital sensor or position and move-

ment data and show the resulting triage classiﬁcation

scheme in Figure 2. After attaching the ITRI bracelet,

the ﬁrst triage decision depends on the patient’s abil-

ity to walk to a collection point. This can be su-

pervised by GPS in combination with accelerometer-

based movement detection. At the collection point a

critical ﬁnding checklist is assessed. Besides checks

that have to be conducted by a medical expert, the

list also includes threshold values for the respiratory

rate, the SpO2 and the systolic blood pressure. Pa-

tients that are not able to walk are tested for sponta-

neous breathing. The check for spontaneous breath-

ing as well as the respiratory rate can be deduced from

a pulse oximeter. If no spontaneous breathing is de-

tected, the ﬁrst responder should clear the airways.

When the life-saving measure is not successful a med-

ical expert has to be called to the site to determine the

chances of survival with the help of an electrocardio-

gram. Next, the respiratory rate is checked. Detec-

tion of squirting bleeding is not recognizable with vi-

tal sensor data. Thus, the ITRI system needs a user

interface, which asks the ﬁrst responder to check for

squirting bleeding and an option to send for a med-

ical expert if necessary. Next the presence of radial

pulse is assessed. Lastly the system needs to inter-

act with the ﬁrst responder to test the consciousness

of the patient, for example with a simple movement

instruction.

The necessary user interface between the ﬁrst re-

sponder and the ITRI system, which might be a com-

bination of speech commands and buttons for user

feedback, is outside the scope of this work. As dis-

cussed in section 2 the necessary hard- and software

for victim tracking has been addressed in the liter-

ature. Furthermore, the feasibility to use a pulse

oximeter to obtain SpO2, the heart rate as well as

the respiratory rate have been shown. A two-lead

electrocardiogram has also been integrated in elec-

tronic triage tags already. In the classic mSTaRT the

use of an electrocardiogram is limited to the deter-

mination of death. ECG readings might be helpful

for continuous patient monitoring, but the attachment

of an ECG also takes a relatively long time. The

systolic blood pressure is a vital parameter used in

the critical ﬁnding check list, but an automated mea-

surement with an arm cuff has practical limitations.

Therefore, this work focuses on a concept for a wrist

worn non-invasive intermittent blood pressure mea-

surement. Additionally, such a system would also be

able to detect and quantify the radial pulse.

3.1.1 Potential Triage Forecasting

As outlined on the right side of Figure 2 the con-

tinuous recording of vital sensor data in combination

Towards an Intelligent Triage Bracelet: A Conceptual Study of a Semi-Automated Prehospital Triage Algorithm and the Integration of

Blood Pressure Measurement

171

AI methods:

condition forecasting

now

+10 min

ITRI mSTaRT

mass-casualty incident

attach ITRI bracelet

send to collection point

slightly injured

SK III

seriously injured

SK II

life-threatening

SK I

walks to

collection point

SK III

SK II

SK I

spontaneous

breathing

first responder interaction

legend:

ITRI automated decision

respiratory rate

RR > 30/min or <10/min

spurting bleeding

radial pulse

simple movement

instruction

clear airways

critical finding checklist

(vital data):

- RR > 30/min or <10/min

- SPo2 < 90%

- syst. blood pressure < 90mmHg

yes

medical expert interaction

stop bleeding

movement not recognized

movement recognized

spontaneous

breathing

no spontaneous

breathing

determination of death

Other critical findings

ECG

Figure 2: Proposed semi-automated prehospital triage scheme which is based on data from the ITRI-bracelet and is supposed

to be performed by non professional ﬁrst responders.

with available static patient data could be used to fore-

cast the triage status with machine learning methods.

This can improve the continuous monitoring of pa-

tients, because the medical experts can focus on pa-

tients with worsening conditions. A challenge for the

development of machine learning algorithms is the

scarcity of available data: Data from mass-casualty

incidents is difﬁcult to obtain, because these incidents

happen relatively rare and complete documentation is

not guaranteed. Additionally, no continuous monitor-

ing of vital signs is currently used.

In the literature, studies with static data from a

pre-hospital setting or clinical daily routine suggest

that the triage could be improved with machine learn-

ing models. One retrospective study with data from

two emergency departments took vital signs, the chief

medical complaint and active medical history to pre-

dict likelihood of acute outcomes. Compared with

the emergency severity index the proposed random

forest model showed superior results. (Levin et al.,

2018) Another study analysed data in two regions in

the Netherlands to improve pre-hospital triage. The

study identiﬁed eight signiﬁcant predictors based on

clinical reasoning and built a regression model to pre-

dict whether a patient is severely injured and thus

needs to be transported to a higher level trauma cen-

ter. The authors also present a mobile app that advises

to which trauma center a patient should be transported

(van Rein et al., 2019).

When continuous vital data is available, machine

learning methods for times series data are applica-

ble. The PhysioNet/Computing in Cardiology Chal-

lenge 2012 challenged participants to predict the in-

hospital mortality of intensive care unit patients based

on ﬁve general static descriptors and 36 time series of

vital signs and laboratory results. Multiple methods

of competitors have shown to obtain signiﬁcantly bet-

ter scores compared to a classic acuity score baseline

algorithm. Beside the conventional logistic regres-

sion other model architectures like support vector ma-

chines, neural networks, random forests and ensem-

ble learning methods were used by participants (Silva

et al., 2012). This challenge was based on the MIMIC

data set. This data set is recorded at critical care units

of the Beth Israel Deaconess Medical Center and the

newest version is organised in modules that reﬂect the

provenance of the data. For potential triage forecast-

ing algorithms the emergency department and inten-

sive care unit modules are the most relevant (MIMIC,

; Johnson et al., 2022).

To potentially integrate machine learning models

trained on clinical data into the ITRI bracelet one

has to use transfer learning techniques. For this use

case differences in vital data sensors and sample rates

as well as inhomogeneous patient populations pro-

pose challenges. Lastly, a validation with real mass-

casualty incident data is indispensable.

BIODEVICES 2023 - 16th International Conference on Biomedical Electronics and Devices

172

3.2 Blood Pressure Measurement

The blood pressure (BP) readings have to be non-

invasive and enable automated intermittent monitor-

ing during the whole duration of an emergency re-

sponse. The selected approach is the oscillometric

measurement method. This method is in widespread

clinical and ambulatory use for automated cuff based

BP measurement systems, that either measure at the

arm or at the wrist.

3.2.1 Oscillometric Measurement Principle

The automated oscillometric BP measurement with

electronic readout via a pressure sensor is based on

the work of Ramsey (Ramsey, 1979). The typical

measurement principle is well described by Sharman

et al. (Sharman et al., 2022) and consists of the fol-

lowing steps: A pneumatic cuff with an integrated

pressure sensor is placed around the arm and inﬂated

to a pressure greater than the systolic blood pres-

sure. Then the cuff is deﬂated and the pressure sensor

records the pressure decrease overlayed with a pul-

satile component, i.e. the pulse waves. These pulse

waves result from the pulsatile heartbeat and can be

analysed separately after high pass ﬁltering. During

the cuff pressure decrease the pulse wave amplitudes

ﬁrst increase sharply when the cuff pressure is near

the systolic BP, then the amplitudes reach a maximum

when the cuff pressure is equivalent to the mean arte-

rial BP and for even lower cuff pressure the amplitude

of the pressure waves decrease again. The recorded

pulse waves during the cuff pressure decrease is called

the oscillometric waveform and the approximation of

their amplitude change is called the oscillometric en-

velope. The cuff pressure at which the oscillometric

envelop is maximal is equivalent to the mean arte-

rial BP. The systolic BP is typically estimated to be

the cuff pressure at the rising site of the oscillometric

envelope at about 50% (range 45–73%) of the maxi-

mal amplitude. Equivalently the diastolic BP is typ-

ically estimated to be the cuff pressure at the falling

site of the oscillometric envelope at about 70% (range

69–83%) of the maximal amplitude (Sharman et al.,

2022).

The oscillometric measurement method has some

known drawbacks. The placement of the device rela-

tive to the heart can alter the BP measurement due to

hydrostatic pressure effects. During the measurement

the patients should not move. Conventional oscillo-

metric BP measurements are known to underestimate

systolic BP (Sharman et al., 2022). Differences of ar-

terial stiffness, which for example can be observed in

elderly or diabetic patients, also inﬂuence the accu-

racy of the measurements (van Montfrans, 2001). It

remains to be evaluated if these drawbacks prevent the

usage of the oscillometric method.

3.2.2 System Implementation

The BP measurement component should be a small

and lightweight system that is comfortable to wear

and easy to use.

Generating a varying cuff pressure with a pre-

cise resolution on a small footprint is a challenging

task that can be solved with micropumps. We pro-

pose the use of piezoelectric micropumps to combine

high pressure generation (25 kPa), large enough ﬂow

rates (50 ml/ min), a relative low power consump-

tion (<300 mW) as well as a small system size (Ø

20 mm x 2.1 mm) (Bußmann et al., 2021b; EMFT

Steel Pumps, ). The operating principle is based on

the indirect piezoelectric effect that describes the me-

chanical deformation of a material due to an applied

electric ﬁeld. The actuator is a piezoceramic, be-

cause this material exhibits a strong piezoelectric ef-

fect. An alternating voltage at the piezoceramic ac-

tuator leads to its extraction and contraction, which

bends the metal diaphragm it is glued on. The bending

displaces the volume inside the pump chamber and

moves the ﬂuid according to the two integrated pas-

sive check valves (Bußmann et al., 2021a). Figure 3

presents the schematic layer-stack of a metal microp-

ump that can be used to generate the cuff pressure.

Figure 3: Diaphragm micropump consisting of four layers:

Body plate with holes, two valve sheets in the middle, and

an actuator with a piezoelectric element on top.

In Figure 4 a schematic representation of the com-

plete ﬂuidic setup is shown. The piezoelectric mi-

cropump ﬁlls the cuff with environmental air to gen-

erate the required cuff pressure. An integrated pres-

sure sensor is used to measure the cuff pressure and

the pulse waves. A ﬂow restriction with a high ﬂu-

idic resistance in parallel to the micropump acts as

a over-pressure protection and pressure relief. Since

the cuff pressure is directly controlled by the actua-

tion frequency of the micropump the pressure build

up can be precisely controlled. This enables to mea-

sure the oscillometric waveform during the pressure

build up.

Towards an Intelligent Triage Bracelet: A Conceptual Study of a Semi-Automated Prehospital Triage Algorithm and the Integration of

Blood Pressure Measurement

173

2. pressure cuff inflated

flow restriction

1. pressure cuff deflated

environment

∆p

micropump

∆p

environment

pressure

time

pressure

time

Figure 4: Fluidic principle of proposed oscillometric blood

pressure measurement method (ﬁgure taken from (Gruener-

bel, 2022)).

Our design does not use a complete cuff, but

a pressurised area that is ﬁxated to the wrist by a

bracelet. During system design a proper placement

of the pressurised area with its center above an artery

is ensured as suggested for the centre of an arm cuff

by Petrie et al. (Petrie et al., 1986). Furthermore,

we apply a pressurised area that is large enough to

ensure good signal quality but still comes with a rea-

sonable ﬁlling time. Figure 5 shows the current state

of the housing and the integrated electronics of the

BP measurement bracelet. The Housing consists of

three 3D-printed parts, which are manufactured us-

ing stereo lithography. The 100 µm thick thermoplas-

tic polyurethane foil at the bottom of the housing is

pressed against the housing base with a ring. Addi-

tionally the foil is glued to the housing base. The ring

ﬁxating the elastic foil and the upper housing lid are

pulled together with four screws. The rubber seals be-

tween the micropump and the housing base are com-

pressed when the upper housing lid presses down on

the printed circuit board, pushing the micropump in

the housing base. To make silicon sealing easier, a

chamber is positioned at the pressure sensor opening.

The ﬂuidic channels are printed directly into the hous-

ing base. The air inlet is positioned opposite the wrist

strap mounting arms so that it will be hidden behind

the wrist strap to prevent unintentional blocking of the

inlet. A rubber wrist strap that is commercially avail-

able is used.

3.2.3 Feasibility Study - Results

The introduced BP monitoring system is tested in

an early feasibility study. Therefore, a test person

wears the bracelet. As soon as the applied pressure

is high enough, the pulse waves become visible. An

exemplary pulse wave recording is shown in Figure

6. The y-axis represent the measured pressure inside

the pressurised area. The pressure pulses can be dis-

tinguished clearly. It should be feasible to calculate

Figure 5: Cross section of the blood pressure monitoring

bracelet with: 1) housing lid, 2) PCB with electronics, pres-

sure sensor and contact pad for micropump, 3) the metal

micropump, 4) rubber sealings, 5) a pressure sensor, 6) the

outer housing with integrated ﬂuid channels, 7) a ﬁxture for

the elastic foil, 8) the elastic foil, i.e. the pressurised area.

the mean arterial blood pressure and estimate the sys-

tolic and diastolic blood pressure using the algorithm

for the oscillometric method described in section 3.2.

Additionally, the sensor readings can be used to quan-

tify the radial pulse and the heart rate.

Figure 6: Exemplary measurement of blood pressure pulses

acquired with the presented demonstrator.

4 CONCLUSION

It was analysed which part of the mSTaRT algorithm

can be automated to potentially enable all ﬁrst respon-

ders to perform a semi-automated prehospital triage.

As a step towards integrating all necessary vital data

sensors a small, cost-efﬁcient and easy to attach pro-

totype of an oscillometric blood pressure monitoring

bracelet was designed and build. Measurements with

the demonstrator show promising raw sensor value

readings with clearly identiﬁable pressure pulses. The

exact algorithm to determine the oscillometric enve-

lope and the calculation of the systolic and diastolic

blood pressure values remains to be deﬁned. The

blood pressure reading must than be veriﬁed with, ei-

ther an intra-arterial measurement or the classic aus-

cultatory method with a sphygmomanometer.

The integration of all other vital data sensors, the

BIODEVICES 2023 - 16th International Conference on Biomedical Electronics and Devices

174

position determination, the movement detection, the

embedded triage guidance together with the user in-

terface of the bracelet as well as the central situation

overview still have to be designed.

The integration of machine learning to forecast

possible triage status changes is an interesting re-

search topic. In both prehospital and clinical settings

data based triage models show promising results. A

prehospital triage forecasting model based on time

series data can only be developed after it was deter-

mined which vital data can be measured continuously

during a prehospital triage.

The effectiveness of an electronic triage system

can only be validated with emergency drills or even

better with real emergency situations. Especially

our assumption that ﬁrst responders without spe-

ciﬁc triage training are enabled to perform a semi-

automated prehospital triage has to be veriﬁed this

way. Additionally, in the same manner, the validity

of the proposed blood pressure measurement method

must be conﬁrmed.

REFERENCES

Bußmann, A. B., Durasiewicz, C. P., Kibler, S. H. A., and

Wald, C. K. (2021a). Piezoelectric titanium based

microﬂuidic pump and valves for implantable medi-

cal applications. Sensors and Actuators A: Physical,

323:112649.

Bußmann, A. B., Gr

unerbel, L. M., Durasiewicz, C. P.,

Thalhofer, T. A., Wille, A., and Richter, M. (2021b).

Microdosing for drug delivery application—a review.

Sensors and Actuators A: Physical, 330:112820.

EMFT Steel Pumps. Stainless steel micropumps for

medical-, laboratory and analysis technology.

https://www.emft.fraunhofer.de/content/dam/emft/

emft-neu/deutsch/dokumente/infoblatt-stainless-

steel-micropumps-medical-laboratory-analysis-

technology-fraunhofer-emft.pdf. Accessed: 2023-01-

02.

Gao, T., Massey, T., Bishop, W., Bernstein, D., Selavo,

L., Alm, A., White, D., and Sarrafzadeh, M. (2006).

Integration of triage and biomedical devices for con-

tinuous, real-time, automated patient monitoring. In

2006 3rd IEEE/EMBS International Summer School

on Medical Devices and Biosensors, pages 34–39.

Gao, T., Massey, T., Selavo, L., Crawford, D., Chen, B.-

r., Lorincz, K., Shnayder, V., Hauenstein, L., Dabiri,

F., Jeng, J., Chanmugam, A., White, D., Sarrafzadeh,

M., and Welsh, M. (2007). The advanced health and

disaster aid network: A light-weight wireless medical

system for triage. IEEE Transactions on Biomedical

Circuits and Systems, 1(3):203–216.

Gruenerbel, L. (2022). Experimental and theoretical

methodologies for high-ﬂow optimization of microp-

umps in medical applications. Dissertation (not pub-

lished yet), page 165.

Johnson, A., Bulgarelli, L., Pollard, T., Horng, S., Celi,

L. A., and Mark, R. (2022). Mimic-iv. https://

physionet.org/content/mimiciv/2.0/. Accessed: 2022-

11-16.

Lenert, L. A., Kirsh, D., Griswold, W. G., Buono, C., Lyon,

J., Rao, R., and Chan, T. C. (2011). Design and eval-

uation of a wireless electronic health records system

for ﬁeld care in mass casualty settings. Journal of the

American Medical Informatics Association : JAMIA,

18(6):842–852.

Levin, S., Toerper, M., Hamrock, E., Hinson, J. S., Barnes,

S., Gardner, H., Dugas, A., Linton, B., Kirsch, T., and

Kelen, G. (2018). Machine-learning-based electronic

triage more accurately differentiates patients with re-

spect to clinical outcomes compared with the emer-

gency severity index. Annals of Emergency Medicine,

71(5):565–574.e2.

Lorincz, K., Malan, D. J., Fulford-Jones, T., Nawoj, A.,

Clavel, A., Shnayder, V., Mainland, G., Welsh, M.,

and Moulton, S. (2004). Sensor networks for emer-

gency response: Challenges and opportunities. IEEE

Pervasive Computing, 3(4):16–23.

MIMIC. The medical information mart for intensive care.

https://mimic.mit.edu/. Accessed: 2022-11-16.

Park, J. Y. (2021). Real-time monitoring electronic

triage tag system for improving survival rate in

disaster-induced mass casualty incidents. Healthcare,

9(7):877.

Paul, A. O., Kay, M. V., Huppertz, T., Mair, F., Dierking,

Y., Hornburger, P., Mutschler, W., and Kanz, K.-G.

(2009). Validierung der vorsichtung nach dem mstart-

algorithmus: Pilotstudie zur entwicklung einer multi-

zentrischen evaluation. Der Unfallchirurg, 112(1):23–

30, 32.

Petrie, J. C., O’Brien, E. T., Littler, W. A., and de Swiet,

M. (1986). Recommendations on blood pressure mea-

surement. Br Med J (Clin Res Ed), 293(6547):611–

615.

Ramsey, M. (1979). Noninvasive automatic determination

of mean arterial pressure. Medical & biological engi-

neering & computing, 17(1):11–18.

RescueWave (2022). Digitale sichtungskarte f

ur manv.

https://rescuewave.de/. Accessed: 2022-06-29.

Rodriguez, D., Heuer, S., Guerra, A., Stork, W., Weber, B.,

and Eichler, M. (2014). Towards automatic sensor-

based triage for individual remote monitoring during

mass casualty incidents. undeﬁned.

Sakanushi, K., Hieda, T., Shiraishi, T., Ode, Y., Takeuchi,

Y., Imai, M., Higashino, T., and Tanaka, H. (2013).

Electronic triage system for continuously monitoring

casualties at disaster scenes. Journal of Ambient Intel-

ligence and Humanized Computing, 4(5):547–558.

Schm

oller, G. and Hagen, F. (2017). Sichtung und vorsich-

tung bei massenanfall von verletzten und erkrankten.

retten!, 6(05):372–381.

Sharman, J. E., Tan, I., Stergiou, G. S., Lombardi, C., Sala-

dini, F., Butlin, M., Padwal, R., Asayama, K., Avolio,

A., Brady, T. M., Murray, A., and Parati, G. (2022).

Automated ’oscillometric’ blood pressure measuring

Towards an Intelligent Triage Bracelet: A Conceptual Study of a Semi-Automated Prehospital Triage Algorithm and the Integration of

Blood Pressure Measurement

175

devices: how they work and what they measure. Jour-

nal of Human Hypertension, pages 1–8.

Silva, I., Moody, G., Scott, D. J., Celi, L. A., and Mark,

R. G. (2012). Predicting in-hospital mortality of icu

patients: The physionet/computing in cardiology chal-

lenge 2012. Computing in cardiology, 39:245–248.

van Montfrans, G. A. (2001). Oscillometric blood pressure

measurement: progress and problems. Blood pressure

monitoring, 6(6):287–290.

van Rein, E. A. J., van der Sluijs, R., Voskens, F. J., Lansink,

K. W. W., Houwert, R. M., Lichtveld, R. A., de Jongh,

M. A., Dijkgraaf, M. G. W., Champion, H. R., Beeres,

F. J. P., Leenen, L. P. H., and van Heijl, M. (2019).

Development and validation of a prediction model for

prehospital triage of trauma patients. JAMA Surgery,

154(5):421–429.

BIODEVICES 2023 - 16th International Conference on Biomedical Electronics and Devices

176