ArticlePDF Available

Abstract

Patients frequently complain of long waiting times in phlebotomy units. Patients try to predict how long they will stay in the phlebotomy unit according to the number of patients in front of them. If it is not known how fast the queue is progressing, it is not possible to predict how long a patient will wait. The number of prior patients who will come to the phlebotomy unit is another important factor that changes the waiting time prediction. We developed an artificial intelligence (AI)-based system that predicts patient waiting time in the phlebotomy unit. The system can predict the waiting time with high accuracy by considering all the variables that may affect the waiting time. In this study, the blood collection performance of phlebotomists, the duration of the phlebotomy in front of the patient, and the number of prior patients who could come to the phlebotomy unit was determined as the main parameters affecting the waiting time. For two months, actual wait times and predicted wait times were compared. The wait time for 95 percent of the patients was predicted with a variance of ± 2 minutes. An AI-based system helps patients make predictions with high accuracy, and patient satisfaction can be increased.
© 2020. Dilek Orbatu, Oktay Yıldırım, Eminullah Yaşar, Ali Rıza Şişman & Süleyman Sevinç. This is a research/review paper,
distributed under the terms of the Creative Commons Attribution-Noncommercial 3.0 Unported License http://creativecommons.
org/licenses/by-nc/3.0/), permitting all non-commercial use, distribution, and reproduction in any medium, provided the original
work is properly cited.
Patient Wait Times in the Phlebotomy Unit
By Dilek Orbatu, Oktay Yıldırım, Eminullah Yaşar, Ali Rıza Şişman & Süleyman Sevinç
Izmir University of Health Sciences
Abstract-
Patients frequently complain of long waiting times in phlebotomy units. Patients try to
predict how long they will stay in the phlebotomy unit according to the number of patients in front
of them. If it is not known how fast the queue is progressing, it is not possible to predict how long
a patient will wait. The number of prior patients who will come to the phlebotomy unit is another
important factor that changes the waiting time prediction.
We developed an artificial intelligence (AI)-based system that predicts patient waiting time
in the phlebotomy unit. The system can predict the waiting time with high accuracy by
considering all the variables that may affect the waiting time. In this study, the blood collection
performance of phlebotomists, the duration of the phlebotomy in front of the patient, and the
number of prior patients who could come to the phlebotomy unit was determined as the main
parameters affecting the waiting time. For two months, actual wait times and predicted wait times
were compared. The wait time for 95 percent of the patients was predicted with a variance of ± 2
minutes. An AI-based system helps patients make predictions with high accuracy, and patient
satisfaction can be increased.
Keywords: phlebotomy, wait time, prediction, artificial intelligence.
GJMR-K Classification: NLMC Code: WG 108
PatientWaitTimesinthePhlebotomyUnit
Strictly as per the compliance and regulations of:
Global Journal of Medical Research: K
Interdisciplinary
Type: Double Blind Peer Reviewed International Research Journal
Publisher: Global Journals
Online ISSN: 2249-4618 & Print ISSN: 0975-5888
Volume 20 Issue 1Version 1.0 Year 2020
Patient Wait Times in the Phlebotomy Unit
Dilek Orbatu α, Oktay Yıldırım σ, Eminullah Yaşar ρ, Ali Rıza Şişman Ѡ & Süleyman Sevinç ¥
Abstract- Patients frequently complain of long waiting times in
phlebotomy units. Patients try to predict how long they will stay
in the phlebotomy unit according to the number of patients in
front of them. If it is not known how fast the queue is
progressing, it is not possible to predict how long a patient will
wait. The number of prior patients who will come to the
phlebotomy unit is another important factor that changes the
waiting time prediction.
We developed an artificial intelligence (AI)-based
system that predicts patient waiting time in the phlebotomy
unit. The system can predict the waiting time with high
accuracy by considering all the variables that may affect the
waiting time. In this study, the blood collection performance of
phlebotomists, the duration of the phlebotomy in front of the
patient, and the number of prior patients who could come to
the phlebotomy unit was determined as the main parameters
affecting the waiting time. For two months, actual wait times
and predicted wait times were compared. The wait time for 95
percent of the patients was predicted with a variance of ± 2
minutes. An AI-based system helps patients make predictions
with high accuracy, and patient satisfaction can be increased.
Keywords: phlebotomy, wait time, prediction, artificial
intelligence.
he ability to predict patient flow and wait times has
become one of the essential clinical management
tools. "How long will it take?" is one o the question
most frequently asked at hospital reception desks [1].
The use of automated reception and turnaround time
(TAT) management system was associated with a
decrease in overall TAT and improved workflow at the
phlebotomy room. Despite efforts at improving the
workflow, the inconvenience to patients has not been
completely mitigated. Repetitive waiting steps increased
the overall zaten pasif TAT of the phlebotomy service
and lead to patient dissatisfaction with the hospital
service [2]. Long wait times among important sources of
dissatisfaction with medical özgün care delivery and a
barrier to the further use of healthcare wiley metni
facilities by affected patients [3]. Studies on customer
değilmiş, psychology in waiting situations reported that
real- time prediction of waiting time improved patient
satisfaction and service quality, especially when queues
were unclear to customers. Patient satisfaction is
positively correlated with patients receiving more
Corresponding Author α: Izmir University of Health Sciences Tepecik
Training and Research Hospital Medical Director, Izmir, Turkey.
e-mail: drdilekorbatu@gmail.com
Author σ ρ ¥: Department of Computer Engineering, Dokuz Eylul
University, Izmir, Turkey.
Author Ѡ: Medicine Faculty, Department of Clinic Biochemistry, Dokuz
Eylul University, Izmir, Turkey.
information, and when patients' actual waiting time is
shorter than the expected waiting time [4]. In another
study, the authors studied the applicability of machine
learning models to predict waiting times. Models built
with ML can reflect sophisticated trends that are hard to
capture with conventional regression approaches [5]. In
our previous özgün ack.nas research, we have
developed a system (Phlerobo) to plan metni değilmiş
and manage all the processes, and resources to
minimize the errors that can occur during the
phlebotomy phase and allow the patients to comfortably
wait for the shortest time [6].
The study was conducted on the waiting times
of 25 thousand patients who applied to the blood
collection unit of SBU TEAH on 42 working days
between 01.09.2019-31.10.2019. At the SBU TEAH
blood collection center, the phlebotomy processes are
being managed by using the Phlerobo system, which
we mentioned in our previous research [6]. The
Phlerobo system is capable of storing all phases of the
phlebotomy process in real- time. Thus, the actual
waiting time is known. Also, other information required
for waiting time prediction is also obtained from the data
stored by the Phlerobo. The Phlerobo stores the
patient's arrival to the phlebotomy unit, the
announcement of the patient's name, the start of the
blood collection process, the completion of the blood
collection process, in real- time. In this process,
demographic information such as age, gender,
outpatient clinic, diagnosis, requested tests, the number
of tubes to be taken, which phlebotomist sample was
taken are also known by Phlerobo. In addition, the
number of phlebotomists working, the number of
patients waiting in the waiting room can be obtained
from the Phlerobo. Because of these features, the
Phlerobo provides a reliable data set for the inputs of
the study and increases the reliability of the results
obtained. For high accuracy prediction of the waiting
time, it is significant to correctly determine the factors
affecting the waiting time. However, it is not easy to
identify all factors in advance. Parameters that affect the
waiting time but are not taken into account when
calculating the waiting time determine the error rate of
the prediction results. All factors that affect the waiting
time should be identified to ensure that the waiting time
estimate is within the acceptable error rate. The high
importance of some parameters on the waiting time is
known, but it is not always possible to predict their
T
1
Year 2020
Global Journal of Medical Research Volume XX Issue I Version I
(DDD
D)
K
©20
20 Global Journals
variable forms. For these reasons, predicting the waiting
time is a much more difficult problem than thought.
Problems caused by errors in waiting time prediction
should be well analyzed. The effect of prediction errors
on patients' needs to be carefully examined. It is almost
impossible for the predicted waiting time to be the same
as the actual waiting time. It is necessary to determine
the acceptable error rate for the performance
measurement of the algorithm. When the patient waits
longer than the estimated waiting time, the patient may
complain that the queue has not yet turned to him / her,
although the estimated time has elapsed. The actual
waiting time is shorter than the predicted waiting time
may also cause dissatisfaction. The patient may have
moved away from the waiting room, assuming that it is
time for him / her to arrive, or may not be ready for
phlebotomy when it is his / her turn. Taking these
considerations into account, the effects of the prediction
error rate on the patient should be well evaluated. These
factors also appear in predictive systems used in
different areas. For example, operating systems predict
the time remaining during file transfer. The predicted
remaining time changes depending on various reasons,
such as the instant workload of the computer and the
file size transferred. Sometimes we see that the file
transfer is completed much longer than or shorter than
the initial estimate. Considering the estimated time, we
may have canceled the file transfer, which may be
completed in a much shorter time. Sometimes we had
to wait for much longer than the estimated remaining
time for the file transfer. Another example of a predictive
system is the prediction of how long the car will go with
the remaining fuel. The estimate of how long the vehicle
can go with the fuel in the tank is given, taking into
account its fuel consumption over the last trip. However,
many parameters such as road conditions, speed of the
car, its load-carrying capacity have an impact on how
long can be driven with the remaining fuel, and these
factors should be considered in the estimation. To
eliminate the variability in the assessments, the
estimates are usually presented in multiples of 10, such
as 110km, 240km. We don't commonly see estimates
like 117 or 243 on vehicle trip computers. Many vehicles
do not show the estimated trip information after a
specified amount of fuel and warn that fuel needs to be
refueled. Because an incorrect estimate can cause the
vehicle to run out of fuel on the road. As can be seen in
these two examples, thept.scribd.com/document/36090
8384/… özgün metni değilmiş effective presentation of
information is crucial to the success of the prediction. As
mentioned above, the most outstandingstage of the
study is the correct determination of the parameters that
will affect the waiting time. and the accurate estimation
of the impact of each parameter is the most crucial step
for the success of the study. The parameters
determined in this step are explained in detail below.
Waiting patients count: When the patient arrives at the
phlebotomy unit, the number of patients waiting in the
hall is the predominant factor affecting the patient wait
time. However, contrary to expectations, the number of
patients waiting in the phlebotomy unit cannot be used
directly in the waiting time calculation. Not all patients
who are physically staying in the phlebotomy unit are
waiting for phlebotomy. Patient accompanists or
patients from other hospital units may be waiting in the
waiting room of the phlebotomy unit. A group of patients
may be waiting to re- sample at a given hour, such as
OGTT, postprandial blood glucose tests. Therefore, the
actual number of patients waiting in the hall should be
determined. The 80 number of patients still waiting for
their turn should be retrieved from the phlebotomy unit
management system. In this study, the number of
waiting patients stored in real- time by the Phlerobo
system was used. Priority information of waiting patients
should be known along with the number of waiting
patients, because the patient who has recently arrived at
the phlebotomy unit will wait for the patients who have
higher priority. Waiting patients who have low priority do
not effect the patient's wait time. For example, routine
priority patients waiting in the hall will not be influential
90 in calculating the waiting time of the patient with the
elderly priority, since a patient with elderly priority will be
served before the routine patients waiting in the hall.
Patient priority: Triage is the main parameter affecting
the process in the management of the phlebotomy unit.
The management of patients who arrive at the
phlebotomy unit is determined according to the priority
rights defined by legal rules. When an emergency
priority patient comes to the blood collection unit, the
blood collection process is completed before the non-
emergency priority patients waiting in the hall. The
priority of the patient is compared with the priority of the
other patients waiting in the waiting room, and the
number of patients with higher priority is determined.
Potential patient count: Just like the number of patients
waiting for in the phlebotomy waiting hall, the number of
patients coming to the phlebotomy unit is an important
parameter on the waiting time. If the priority of the
patients who will arrive at the phlebotomy unit is higher
than the waiting patients, the wait time of the waiting
patients will increase. At this stage, the number of new
patients arriving at the phlebotomy unit, and their
priorities should be known, and the effect of these
patients should be added to the waiting time. For
example, for a patient who is calculated to wait 5
minutes according to the number of patients waiting in
the waiting hall, it is added that patient will have to wait
an extra 2 minutes because of the high-priority patients
who will arrive at the phlebotomy unit during the
5-minute waiting period and the patient is predicted to
wait 7 minutes. However, it is not easy to know the
number of new patients and their priorities. Different
2
Year 2020
Global Journal of Medical Research Volume XX Issue I Version I
(DDD
D)
K
©20
20 Global Journals
estimation methods can be developed to determine this
number. One of these methods is the calculation using
historical data, taking into account the day and time of
the week. With this method, the number of patients
arriving at the phlebotomy unit in the same time interval
on the same day in the past weeks is estimated. The
parameters that affect the performance of this
estimation method can be highly variable. For example,
due to factors such as seasonal changes, days before
or after long holidays, a high variance will occur in this
statistical study. As another method, it can be used to
estimate the number of future patients in a near time
interval according to the patient’s habit of arriving at the
blood collection unit. Although the prediction
performance of this method is high, it is not sufficiently
successful in crowds that occur suddenly. As another
method, a computational algorithm was developed and
used in this study to determine the accumulation of
patients that occur suddenly, taking into account the
laboratory test orders made by the polyclinics. Patients
for whom who laboratory tests are ordered during their
outpatient assessments a come to the phlebotomy unit
after the examination. Laboratory test order information
is taken from the hospital information management
system in real- time and the number of potential patients
arriving at the phlebotomy unit is calculated. However, it
takes time for the patients to come to the phlebotomy
department from the outpatient clinic. This time is also
taken into account when calculating the estimated
arrival time of the patients to the phlebotomy unit.
However, during the study period, it was arxiv.org/pdf/
1905.07271v1 özgün olmayan sözcükler found that not
all patients who have test orders in the outpatient clinic
arrived at the blood collection unit that day. The majority
of patients who receive a test order in the morning arrive
at the phlebotomy unit, but this rate decreases as time
goes on. When the cause of this condition was
examined, it was found that especially patients with tests
requiring fasting left the hospital to give blood the next
day because they did not meet the fasting prerequisite.
The arrival trend to the blood collection unit is observed,
and it is estimated that how many of the patients who
have test orders will come to the phlebotomy unit. Thus,
the arrival trend to the phlebotomy unit, the number of
patients to who have test order and the priority
information of the patients are used together to
determine the number and priority of the potential
patients.
Phlebotomist Count: The number of phlebotomists is the
main parameter that determines the patient flow speed
of the phlebotomy unit. The more phlebotomists are
working; the more blood samples are taken in parallel.
This speeds up the patient flow. Although the number of
patients waiting in the hall is too high, the phlebotomy of
waiting patients can be completed in a very limited time
özgün olmayan metin if enough phlebotomists are
working. Likewise, even if a few patients waiting in the
hall, if the number of phlebotomists is not sufficient, long
waiting times occur. The number of phlebotomists
working at the time the patient arrives at the blood
collection unit is known, allocation status of the
phlebotomists is unknown until the patient is served. A
definite pattern of increases and decreases in the
number of phlebotomists could not be determined.
Normally all phlebotomists are expected to work during
working hours, but occasionally phlebotomists need a
break, which leads to a reduction and increase in the
number of phlebotomists during the day. This variability
could not be included in the prediction of waiting time,
as it is not possible to determine precisely when these
decreases and increases will occur.
Phlebotomy rate: Although the number of phlebotomists
determines the patient flow speed in the phlebotomy
unit, the phlebotomist's phlebotomy performance is also
influential in determining the patient flow speed. When
the mean blood collection time of phlebotomists was
examined, it was observed that there were significant
differences between phlebotomists. Moreover, the same
phlebotomist showed different performances on
different days. The number of patients served by
phlebotomists varies between 15 and 32 per hour. This
fact showed that the number of phlebotomists was not
sufficient in the waiting time calculation and that the
number of patients served during a certain period
should be taken into account. Another variable affecting
the phlebotomy time is how many tubes should be taken
from the patient. Also, the demographic information of
the patient affects the preparation of the patient for
phlebotomy and the finding of blood vessels, thus
causing variability in blood collection time. However,
these factors do not cause marked perturbations on
average when calculating the patients taken at a given
time. Taking into account the variability between
phlebotomists and patients, the average patient flow
performance of the entire phlebotomist team was used
to estimate the waiting time.
After estimating the values of the parameters
described in detail above, the waiting time is predicted
by a calculation. First of all, the number of patients
waiting in the hall and having the potential to come to
the phlebotomy unit, whose priority is higher or equal, is
calculated. Depending on the number of working
phlebotomists, and the blood collection performance of
the phlebotomists, the waiting time of the patient is
predicted by calculating the total time it takes to
complete the blood collection process. The most
challenging difficulty in predicting is the individual
calculation of the parameters of the algorithm which is
directly related to the quality of data stored in the
process. Reliable real-time information from the
information system enables us to make high accuracy
3
Year 2020
Global Journal of Medical Research Volume XX Issue I Version I
(DDD
D)
K
©20
20 Global Journals
estimates. The support of the Phlerobo system cannot
be denied in the success of our study.
The wait time returned from the algorithm is a
numeric value in minutes. However, presenting the
number returned from the algorithm to the end- user will
not be a correct representation. When we say that your
estimated waiting time is 1.4 minutes, we talked about
the negativities that will occur in patients when waiting
longer or shorter than 1.4 minutes. Instead, it would be
more functional for patients to provide estimates to
provide sufficient information to the patient. For
example, categorizing and presenting the waiting time
as short, medium, and long can be an easy-to
understand presentation style by the end- user.
However, this easy-to-understand presentation may not
render enough information. The prediction in the form of
a time interval will provide both sufficient information and
increase the likelihood of making an accurate prediction.
For example, your estimated waiting time is between 1-2
minutes instead of your estimated waiting time is 1.4
minutes, which means that it will provide sufficient
information to the patient and increase prediction
success. The range of the minute interval to be
presented needs to be well defined. As the predictive
success of the wider ranges, increases, the value of the
information provided will decrease, whereas in the
narrow predictive ranges, the significance of information
increases, the predictive success decreases.
Determining the optimum range will be essential for the
success of the study. Closer intervals can be provided
for shorter waiting times and wide ranges for longer
waiting times. For shorter waiting times, the estimated
waiting time may be presented to the patient in a narrow
range, such as from 1 to 3 minutes, while for longer
waiting times, the calculated waiting time may be
provided in longer intervals, such as from 10-15
minutes. The presentation of predetermined waiting time
to patients may vary according to the preferred method
of each hospital. We aimed to deviate ± 2 minutes as
the success target of our study. All display modes
including ± 2 minute deviation can be selected.
We predicted the wait times of patients who
applied to the SBU TEAH phlebotomy unit between 2
September 2019 and 31 October 2019 for two months
using the estimation method, which we developed. We
analyzed the predicted wait time for 44 different working
days. To use the prediction method, the parameters
required for prediction must be available. Especially in
the first minutes of the day and in the absence of
sufficient data after a lunch break, waiting time cannot
be predicted. The wait time for 94.6% of all patients were
predicted. We examined whether the estimates were
shorter or longer than the actual waiting time. 14.1% of
the predictions fell short than the actual wait time and
85.9% were longer than the actual wait time. It is
preferred that the deviation is mostly positive. There will
be no dissatisfaction as the patient will receive service
before the predicted time. If the estimated time was
shorter than the actual waiting time, the patient's
complaint might increase after the time elapsed after the
estimated time. The mean waiting time with ± 1 minute
deviation was predicted for 42.4% of the patients.
Predictions were made for 77% of patients with ± 2 min
deviation. The mean waiting time with a lapse of ± 3
minutes was predicted for 91.5% of the patients. The
wait time was predicted for 95.5% of patients with ± 4
min elapse and 97.3% of patients with ± 5 min
deviation. When we analyzed the results for the priority
patients, it was predicted for 47.1% of the patients with a
deviation of ± 1 minute. 86.2% of patients were
predicted to have an accuracy of ± 2 minutes. For
98.5%, the wait time with a deviation of ± 3 minutes was
predicted. It is seen that the prediction success of the
priority patients is higher because of the short waiting
time average. The targeted deviation of ± 2 minutes
was achieved for 77% of the patients. However, for 42%
of patients, the predictive success was accurate to ± 1
minute. When the differences between the days were
examined, it was observed that the deviations in the
estimates were high when the number of phlebotomists
suddenly decreased and increased during the day and
that the divergences in the estimation were low when the
number of phlebotomists was constant [Table 1].
Table 1: Prediction Error Rate
±1 min ±2 min ±3 min ±4 min ±5 min
Prior 47 % 86 % 98 % 100 % 100 %
Routine 40 % 72 % 88 % 93 % 96 %
Total 42 % 77 % 92 % 96 % 97 %
It is almost impossible to know the wait time
exactly. However, it is possible to predict the wait time
with approximate accuracy. All variables affecting the
waiting time must be recognized for the waiting time
prediction to be successful. However, it is not possible
to know the exact values of some variables in advance.
With the developed algorithm, it is tried to determine the
possible significance of unknown parameters for the
moment by predicting the behaviors that patients and
phlebotomists will exhibit in the future from their past
experiences. Thus, the closest waiting time to the actual
waiting time will be predetermined. Factors such as
patients who do not come to the blood collection unit
because they are not hungry despite the doctor's
request for the test, patients who arrive at the blood
collection unit early or late than expected, and
phlebotomist have more breaks than expected cause
the difference of the estimated waiting time to be higher.
New innovative solutions for these problems are to be
4
Year 2020
Global Journal of Medical Research Volume XX Issue I Version I
(DDD
D)
K
©20
20 Global Journals
expected to increase predictive success. It will be an
appropriate study to examine the effects of waiting time
prediction on patients. If the retrospective estimation
success is scrutinized, and the system makes
acceptable estimates, the positive and negative effects
may be examined by presenting the predictions to the
patients for a certain period. In this study, different
methods of presentation of waiting time prediction can
be tried to determine which presentation method is
more successful. Thus, the wait time prediction study
will be utilized in the most effective way.
We would like to thank the managers of İzmir
University of Health Sciences Tepecik Training and
Research Hospital and the staff of the blood collection
unit for supporting the field practices of the study and
providing the blood collection unit data. We also would
like to thank LABENKO BİLİŞİM AŞ, the developer of the
artificial İntelligence- based management system of
blood collection unit (Phlerobo), for providing reliable
data with high-tech infrastructures and for the support of
database and and application server hardware for
performing the algorithm tests of the study.
Oleg S. Pianykh, PhD and Daniel I. Rosental, MD
“Can We Predict Patient Wait Time?Journal of the
American College of Radiology, 2015-10-01, Volume
12, Issue 10, Pages 1058-1066.
2. Soon Gyu Yun, M.T., Jeong Won Shin, M.D., Eun Su
Park, M.D., Hae In Bang, M.D., and Jung Gu Kang
M.T. “Efficiency of an Automated Reception and
Turnaround Time Management System for the
Phleotomy Room” Annals of Laoratory Medicine.
3. Yo Seop Woo, Young Dae Kwon, Mi-Kyung Lee,
Young Joo Cha and ye Ryoun Kim “A Real-Time
Computer imulation Program for Reducing
Outpatients Pleotomy Wait Time” ournal of Clinical
Laboratory Analysis 29:255-258, 2015.
4. Yan Sun, PhD, Kiok Liang Teow MSc, Bee Hoon
Heng, MBBS, MSc, Chee Kheong Ooi, MBBS, MSc,
Seow Yian Tay, MBBS, MSc “Real-Time Prediction of
Waiting Time in the Emergency Department, Using
Quantile Regression” Annals of Emergency
Medicine, Volume 60, No 3, September 2012.
5. Caterine Curtis, MS, Chang Liu MS, Thomas J.
Bollerman, MS and Oleg S. Pianykh PhD “Machine
Learning for Predicting Patient Wait Times and
Appointment Delays” Journal of American College of
Radiology, 2018-09-01, Volume 15, Issue 9, Pages
1310-1316.
6. Orbatu D., Yıldırım O, “Use of Artificial Intelligence in
Phlebotomy Unit” Turkish Journal of Biochemistry
(2018), 43(Supplement), pp. 22-29. Retrieved 15
Oct. 2019, from doi:10.1515/tjb-2018-43s144
5
Year 2020
Global Journal of Medical Research Volume XX Issue I Version I
(DDD
D)
K
©20
20 Global Journals
1.
... It has been suggested that the maintenance of social distancing could reduce the spread of disease, however, the phlebotomy area is almost always crowded in a hospital, especially in peak times. The latest study proposed an artificial intelligence-based system to predict patient WT in the phlebotomy unit [30]. Subjects may have blood collection completed according to the predicted time to avoid gathering in the waiting area. ...
Article
Full-text available
Background: The waiting time (WT) for a phlebotomy is directly related to patient satisfaction with a health service. However, the processing time varies widely depending on the type of patients. Monitoring of the WT alone may not enable an effective evaluation of the lean performance of the medical staff for patients with different characteristics. The objective of this study was to use process cycle efficiency (PCE) to assess the performance of an intelligent tube preparation system (ITPS) which automatically labeled test tubes and conducted patient rerouting for phlebotomy services, and to interpret the WT during peak hours. Methods: Three time periods were used. The baseline period was from 1 July to 31 July 2014. Phase 1 was after the establishment of the ITPS, with patients ≥80 years old being rerouted. In phase 2, patients ≥78 years old were rerouted. Those data were recorded with a calling system and ITPS, respectively. Results: PCE was significantly improved from 12.9% at baseline to 51.1% (p < 0.001) in phase 1 and 53.0% (p < 0.001) in phase 2. The WT of 16.9 min at baseline was reduced to 3.8 min in phase 1 (p < 0.001), and 3.6 min in phase 2 (p < 0.001). Moreover, the results showed that a WT < 10 min was consistent with a PCE ≥ 25%. Conclusions: Establishing an ITPS for phlebotomy can significantly increase PCE and shorten the WT. Furthermore, the PCE ≥ 25% could be a good assessment reference for the management of appropriate human resources for phlebotomy services, although it is a complex parameter.
... However, there are efforts to improve the preanalytic, analytic, and post-analytic processes of the clinical laboratories with the AI-support. Examples of such studies are an AI-supported system that predicts patient waiting time in the phlebotomy unit and organizes the entire blood collection process or the autoverification of test results using an ML approach [17,18]. ML approach can also implement for further processes, such as predicting out-of-control events in internal quality control studies, detecting instrument failures before even they occur, or determining compatibility between analyzers in central laboratories where several instruments running the same parameters are tested. ...
Article
Full-text available
A new trend in healthcare The current trend in health care is a shift from a disease-centered model to a patient-centered model, from a pater-nalistic physician-patient relationship to an egalitarian partnership , and from empirical to data-based evidence [1, 2]. In parallel with the increasing number and variety of medical tests, it is estimated that the number of data generated in the evaluation of a patient will reach 10.000 in 2020. The output obtained from molecular biology studies and the widespread use of electronic health records (EHR) combined with the growing capabilities and capacity of computers has attention to "big data. " It is once again clear how important it is to process big data in the fight against the disease and in the development of vaccines or drugs during these days when we fight the Covid-19 outbreak as humanity [3]. Processing and extracting meaningful interpretations from such vast and complex data, as seen in the Covid-19 outbreak, require the use of various computational tools and methods. In parallel with the increasing number and variety of medical tests and the widespread use of electronic health records combined with the growing capabilities and capacity of computers have attention to "big data. " Processing and extracting meaningful interpretations from such vast and complex data require artificial intelligence (AI) that refers to complex software systems that enable computers to augment and even imitate human intelligence and decision-making. Machine learning (ML) is a subfield of AI that uses algorithms to parse and learn data and then apply this new learning to make predictions and informed recommendations. In recent years, the effects that the digitalization of healthcare services will have on medicine, especially laboratory medicine as seen in the industry, the economy, and social life. The abundance of health data will lead to a shift from analytical competence in diagnostic tests to the ability to integrate data and simultaneously interpret them within the clinical context. Therefore, "computational laboratory medicine" units should be established and integrated into resident and undergraduate education curricula. Using the computational approach, the promise of improved medical interpretation will further increase the effectiveness of laboratory diagnostics in the process of intensive dialogue/ consultation and clinical decision-making. Medical laboratories may play an active role in the future as a "nerve center of diagnostics" and joining the patient and physician to form a "Diagnostics 4.0" triangle. As the big data continue to grow in healthcare, the need for implementing AI and ML techniques into laboratory medicine is inevitable. In this new AI-supported era, clinical laboratories will move towards a more specialized role in translational medicine, advanced technology, management of clinical information, and quality control of results generated outside the laboratory. The field of laboratory medicine should consider such a development sooner rather than later. Abstract
... However, there are efforts to improve the preanalytic, analytic, and post-analytic processes of the clinical laboratories with the AI-support. Examples of such studies are an AI-supported system that predicts patient waiting time in the phlebotomy unit and organizes the entire blood collection process or the autoverification of test results using an ML approach [17,18]. ML approach can also implement for further processes, such as predicting out-of-control events in internal quality control studies, detecting instrument failures before even they occur, or determining compatibility between analyzers in central laboratories where several instruments running the same parameters are tested. ...
Article
Full-text available
In parallel with the increasing number and variety of medical tests and the widespread use of electronic health records combined with the growing capabilities and capacity of computers have attention to “big data.” Processing and extracting meaningful interpretations from such vast and complex data require artificial intelligence (AI) that refers to complex software systems that enable computers to augment and even imitate human intelligence and decision-making. Machine learning (ML) is a subfield of AI that uses algorithms to parse and learn data and then apply this new learning to make predictions and informed recommendations. In recent years, the effects that the digitalization of healthcare services will have on medicine, especially laboratory medicine as seen in the industry, the economy, and social life. The abundance of health data will lead to a shift from analytical competence in diagnostic tests to the ability to integrate data and simultaneously interpret them within the clinical context. Therefore, “computational laboratory medicine” units should be established and integrated into resident and undergraduate education curricula. Using the computational approach, the promise of improved medical interpretation will further increase the effectiveness of laboratory diagnostics in the process of intensive dialogue/ consultation and clinical decision-making. Medical laboratories may play an active role in the future as a "nerve center of diagnostics" and joining the patient and physician to form a "Diagnostics 4.0" triangle. As the big data continue to grow in healthcare, the need for implementing AI and ML techniques into laboratory medicine is inevitable. In this new AI-supported era, clinical laboratories will move towards a more specialized role in translational medicine, advanced technology, management of clinical information, and quality control of results generated outside the laboratory. The field of laboratory medicine should consider such a development sooner rather than later. Keywords: Artificial intelligence, big data, black box, diagnostic 4.0, computational laboratory medicine, explainable artificial intelligence, machine learning
Article
Objectives Laboratory testing, crucial for medical diagnosis, has 3 phases: preanalytical, analytical, and postanalytical. This study set out to demonstrate whether automating tube labeling through artificial intelligence (AI) support enhances efficiency, reduces errors, and improves outpatient phlebotomy services. Methods The NESLI tube-labeling robot (Labenko Informatics), which uses AI models for tube selection and handling, was used for the experiments. The study evaluated the NESLI robot’s operational performance, including labelling time, technical problems, tube handling success, and critical stock alerts. The robot’s label readability was also tested on various laboratory devices. This research will contribute to the field’s understanding of the potential impact of automated tube-labeling systems on laboratory processes in the preanalytical phase. Results NESLI demonstrated high performance in labeling processes, achieving a success rate of 99.2% in labeling parameters and a success rate of 100% in other areas. For nonlabeling parameters, the average labeling time per tube was measured at 8.96 seconds, with a 100% success rate in tube handling and critical stock warnings. Technical issues were promptly resolved, affirming the NESLI robot’s effectiveness and reliability in automating the tube-labeling processes. Conclusions Robotic systems using AI, such as NESLI, have the potential to increase process efficiency and reduce errors in the preanalytical phase of laboratory testing. Integration of such systems into comprehensive information systems is crucial for optimizing phlebotomy services and ensuring timely and accurate diagnostics.
Article
Full-text available
Background Recent advances in laboratory information systems have largely been focused on automation. However, the phlebotomy services have not been completely automated. To address this issue, we introduced an automated reception and turnaround time (TAT) management system, for the first time in Korea, whereby the patient's information is transmitted directly to the actual phlebotomy site and the TAT for each phlebotomy step can be monitored at a glance. Methods The GNT5 system (Energium Co., Ltd., Korea) was installed in June 2013. The automated reception and TAT management system has been in operation since February 2014. Integration of the automated reception machine with the GNT5 allowed for direct transmission of laboratory order information to the GNT5 without involving any manual reception step. We used the mean TAT from reception to actual phlebotomy as the parameter for evaluating the efficiency of our system. Results Mean TAT decreased from 5:45 min to 2:42 min after operationalization of the system. The mean number of patients in queue decreased from 2.9 to 1.0. Further, the number of cases taking more than five minutes from reception to phlebotomy, defined as the defect rate, decreased from 20.1% to 9.7%. Conclusions The use of automated reception and TAT management system was associated with a decrease of overall TAT and an improved workflow at the phlebotomy room.
Article
Being able to accurately predict waiting times and scheduled appointment delays can increase patient satisfaction and enable staff members to more accurately assess and respond to patient flow. In this work, the authors studied the applicability of machine learning models to predict waiting times at a walk-in radiology facility (radiography) and delay times at scheduled radiology facilities (CT, MRI, and ultrasound). In the proposed models, a variety of predictors derived from data available in the radiology information system were used to predict waiting or delay times. Several machine-learning algorithms, such as neural network, random forest, support vector machine, elastic net, multivariate adaptive regression splines, k-th nearest neighbor, gradient boosting machine, bagging, classification and regression tree, and linear regression, were evaluated to find the most accurate method. The elastic net model performed best among the 10 proposed models for predicting waiting times or delay times across all four modalities. The most important predictors were also identified.
Article
Purpose The importance of patient wait-time management and predictability can hardly be overestimated: For most hospitals, it is the patient queues that drive and define every bit of clinical workflow. The objective of this work was to study the predictability of patient wait time and identify its most influential predictors. Methods To solve this problem, we developed a comprehensive list of 25 wait-related parameters, suggested in earlier work and observed in our own experiments. All parameters were chosen as derivable from a typical Hospital Information System dataset. The parameters were fed into several time-predicting models, and the best parameter subsets, discovered through exhaustive model search, were applied to a large sample of actual patient wait data. Results We were able to discover the most efficient wait-time prediction factors and models, such as the line-size models introduced in this work. Moreover, these models proved to be equally accurate and computationally efficient. Finally, the selected models were implemented in our patient waiting areas, displaying predicted wait times on the monitors located at the front desks. The limitations of these models are also discussed. Conclusions Optimal regression models based on wait-line sizes can provide accurate and efficient predictions for patient wait time.
Article
Background: Our goal was to enhance the productivity of phlebotomists and reduce outpatient wait times. Here, we aimed to develop a computer simulation program in which resources would be shifted from the laboratory to assist with phlebotomy. We evaluated the efficacy of computer simulation approaches for phlebotomy wait time and provide phlebotomy help time. Methods: We evaluated the performance of the following three approaches: no helping system (NHS), the conventional assistance system (CAS), and the computer simulated helping system (CSHS). The CSHS predicted the phlebotomy waiting times based on computer simulation approaches, decided the assist times, and sounded an alarm. Results: The wait time for the NHS was significantly longer than that of the CAS and the CSHS (P < 0.05). We divided the wait time into the three parts: <5 min, 5-10 min, and >10 min. Significant differences between the three systems were detected (P < 0.05). The phlebotomy computer simulation system significantly decreased the help time of the phlebotomists (CAS was 93.3 ± 19.7 min vs. CSHS was 79.5 ± 17.7 min, P = 0.03). Conclusion: We designed a computer-based predicted alarm system. This system could effectively decrease help time for phlebotomists and outpatients phlebotomy wait times.
Article
Emergency department (ED) waiting times can affect patient satisfaction and quality of care. We develop and validate a model that predicts an individual patient's median and 95th percentile waiting time by using only data available at triage. From the existing ED information system, we extracted date and time of triage completion, start time of emergency physician consultation, and patient acuity category (1=most urgent, 3=least urgent). Quantile regression was applied for model development and parameter estimation by using visits from January 2011. We assessed absolute prediction error, defined as the median difference between the 50th percentile (median) predicted waiting time and actual waiting time, and the proportion of underestimated prediction, defined as the percentage of patients whose actual waiting time exceeded the 95th percentile prediction. The model was validated retrospectively with June 2010 data and prospectively with data from April to June 2011 after integration with the existing ED information system. The derivation set included 13,200 ED visits; 903 (6.8%) were patient acuity category 1, 5,530 (41.9%) were patient acuity category 2, and 6,767 (51.3%) were patient acuity category 3. The median and 95th percentile waiting times were 17 and 57 minutes for patient acuity category 2 and 21 and 89 minutes for patient acuity category 3, respectively. The final model used predictors of patient acuity category, patient queue sizes, and flow rates only. In the retrospective validation, 5.9% of patient acuity category 2 and 5.4% of category 3 waiting times were underestimated. The median absolute prediction error was 11.9 minutes (interquantile range [IQR] 5.9 to 22.1 minutes) for patient acuity category 2 and 15.7 minutes (IQR 7.5 to 30.1 minutes) for category 3. In prospective validation, 4.3% of patient acuity category 2 and 5.8% of category 3 waiting times were underestimated. The median absolute prediction error was 9.2 minutes (IQR 4.4 to 15.1 minutes) for patient acuity category 2 and 12.9 minutes (IQR 6.5 to 22.5 minutes) for category 3. Using only a few data elements available at triage, the model predicts individual patients' waiting time with good accuracy.
Use of Artificial Intelligence in Phlebotomy Unit
  • D Orbatu
  • O Yıldırım
Orbatu D., Yıldırım O, "Use of Artificial Intelligence in Phlebotomy Unit" Turkish Journal of Biochemistry (2018), 43(Supplement), pp. 22-29. Retrieved 15