Model for Process Time Analysis in Magnetic Resonance Imaging

(1)

IN

DEGREE PROJECT MEDICAL TECHNOLOGY,

FIRST CYCLE, 15 CREDITS STOCKHOLM SWEDEN 2020,

Model for Process Time Analysis in Magnetic Resonance Imaging

Workflow Optimization to Reduce Access Time ELINA BROMAN

KTH ROYAL INSTITUTE OF TECHNOLOGY

SCHOOL OF ENGINEERING SCIENCES IN CHEMISTRY, BIOTECHNOLOGY AND HEALTH

(2)

(3)

i This project was performed in collaboration with

Philips Healthcare Transformation Services Supervisor at Philips: Elin Frid, Emelie Håkansson

Model for Process Time Analysis in Magnetic Resonance Imaging

Workflow Optimization to Reduce Access Time

Modell för processtidsanalys inom magnetresonanstomografi

Optimering av arbetsflöden för kortare väntetider

E L I N A B R O M A N

Degree project in medical engineering First level, 15 hp

Supervisor at KTH: Tobias Nyberg, Mattias Mårtensson Examiner: Mats Nilsson

KTH Royal Institute of Technology

School of Engineering Sciences in Chemistry, Biotechnology and Health SE-141 86 Flemingsberg, Sweden

http://www.kth.se/cbh 2020

(4)

ii

Abstract

Magnetic Resonance Imaging (MRI) can be used in many clinical situations, but they are limited by high costs and time-consuming properties. Most focus has been on improving the technical side of MRI, and not as much on process improvements. Long access times to MRI examinations can be the cause of inefficient workflows of the departments, which can cause adverse effects for the patients. A market research of the Swedish radiology departments that perform MRI examinations resulted in a wide range of median access times, ranging from approximately 21 days to 130 days. This indicates potential for improvement in their workflows. To improve workflows, they need to be analysed and measured. A model for process time efficiency analysis in MRI departments was created in this project which can assess 5 different metrics. These metrics are number of examinations, examination time, turnover time, scanner utility, and scheduling consistency.

Potential improvement strategies to reduce the access times associated with the metrics in the model is discussed. Examination time is mostly affected by the technique and the examination protocol but making the change to an abbreviated version of the protocol has the potential to significantly reduce examination time. This is especially useful for screening purposes. Reduction in turnover time can be achieved by analysing the process in between examinations and making suitable changes in preparation of patients and examinations, and in architecture for a more streamlined throughput. The scheduling process has a large impact on efficiency and reduction in access times and increasing utility rate. It is important for the scheduling process to be flexible to increase efficiency. As a result of this report, the

conclusion is that a benchmarking project could be conducted on Swedish radiology departments to determine best practices in workflows.

Keywords: MRI, market research Sweden, efficiency, access time

(5)

iii

Sammanfattning

Magnetresonanstomografi (MRT) kan användas i många kliniska situationer, men de är begränsade av höga kostnader och tidskrävande egenskaper. Störst fokus har tidigare varit på de tekniska aspekterna av MRT och inte lika mycket på processförbättring. Långa väntetider till MRT-undersökningar kan vara orsaken av ineffektiva arbetssätt på avdelningarna, vilket skapar negativa effekter för patienter. En marknadsundersökning av de svenska

radiologiavdelningarna som utför MRT-undersökningar resulterade i ett stort spann av väntetider, från ungefär 21 dagar till 130 dagar. Detta indikerar förbättringspotential i deras arbetssätt. För att förbättra arbetssätt måste de analyseras och mätas. En modell för att mäta effektivitet för processtder på MRT-avdelningar skapades i detta projekt med fem olika mätvärden. Dessa mätvärden är antal undersökningar, undersökningstid, tid mellan undersökningar, användningsgrad av maskiner och efterföljande av bokningstid.

Potentiella förbättringsstrategier för att minska väntetider länkade till mätvärderna i modellen diskuteras. Undersökningstiden påverkas mest av tekniken och protokollet för undersökningen, men genom att byta till ett förkortat protokoll kan undersökningstiden minskas avsevärt. Detta kan vara särskilt användningsbart vid screeningundersökningar.

Tiden mellan undersökningar kan minskas genom att analysera processen mellan undersökningarna och göra ändringar i förberedelser av patient och undersökning samt arkitektuella ändringar för en mer effektivserad genomströmning. Bokningsprocessen har en stor betydelse för effektiviteten, minskning av väntetider och ökad användningsgrad. Det är viktigt att bokningsprocessen är flexibel för att öka effektiviteten. Som resultat av denna rapport är slutsatsen att ett benchmarkning-projekt kan utföras på svenska

radiologiavdelningar för att hitta bra arbetssätt.

Nyckelord: MRT, marknadsundersökning Sverige, effektivitet, väntetider

(6)

iv

Acknowledgement

I would like to express my sincerest appreciation and gratitude to my supervisors Emelie Håkansson and Elin Frid at Philips for their support, guidance, and patience through this project. Their mentoring has provided me with a lot of valuable insights. Additionally, I would like to extend my deepest gratitude to Axel for always supporting and guiding me.

(7)

v

Introduction

The Swedish Association of Local Authorities and Regions (SALAR) compiled a report [1]

regarding the future supply of healthcare professionals in the Swedish healthcare system. In their report, they present factors that will have a negative impact on the future supply of healthcare professionals. Such factors include a larger amount of retirements that will occur as well as the structural changes to a more patient centered healthcare system. They predict that during a ten-year period, 8000 healthcare professionals are expected to retire per year, and an additional 4000 healthcare professionals will have to be employed every year. This is unless there are changes made to workflows. In addition to this, the National Board of Health and Welfare report [2] an expected increase of multimorbidity of approximately 74.4% over a 50 year period. They predict that this will lead to increased expenses for the regions in Sweden. More efficient workflows are demanded of the healthcare sector [1, 3], because the population is living increasingly longer, while at the same time not being able to work equally long [1].

Magnetic Resonance Imaging (MRI) is a modality that is widespread in more medical fields of application than before [4], but it is limited as a screening tool because of the high costs and time consuming properties [5]. Access time is defined as the time from referral to completion of examination [6]. For a quality healthcare system, it is important to adapt best practices that ensures efficiency [7], and long access times are often consequences, and indicators, of workflows that are not efficient [6, 8]. Long access times can also cause a delay in diagnosis of patients, which impacts the quality of care negatively [9]. An initial assessment of MRI examination access time provided insight of lengthy access times in Swedish hospitals. Most MRI optimization areas has previously been focused on the technical side, and not as much on the workflow optimization [10].

Philips Healthcare Transformation Services provide healthcare consulting and solutions to hospitals and regions in Sweden. Philips help radiology departments identify and implement improvement areas related to patient and staff experience, and operational efficiency. Philips want, by initiating this thesis, to further explore the radiology market in Sweden. This

includes finding new ways to optimize the MRI care flow at the radiology departments.

1.1 Aim

The first aim of this report was to evaluate the MRI access times for radiology departments in Sweden, to determine if there is potential for workflow optimization in Swedish MRI radiology departments. The second aim was to create a model that can analyse process time efficiency in MRI radiology and discuss strategies related to the model that can improve department workflow.

1.2 Limitation

The access times were evaluated for the year of 2019. The model is limited to the efficiency metrics examination time, turnover time, scanner utility, number of examinations performed, and scheduling consistency.

(10)

2

(11)

3 2

Background

This chapter covers the process of MRI image acquisition, why there is a time constraint, and what future technical applications in image acquisition techniques can make the MRI process faster and more efficient. It also covers process improvement projects and metrics used in efficiency analysis for MRI.

2.1 MRI Signals and Sequences

This section is summarized from Physics for Diagnostic Radiology by P.P Dendy [11]. MRI uses magnetism and radio waves to stimulate the hydrogen atom nuclei within the body and detect the resulting atomic effects. The MRI scanner consists of a large coil with current going through it, which creates a static magnetic field (B0). When performing an MRI examination, there are different sequences used that achieve different visual anatomical results. The simple spin echo sequence is covered briefly here. There are two parts of the MRI signal acquisition process: the excitation of the nuclei and the detection of the nuclei response to the excitation. To select different slices within the region to me imaged, a slice selection gradient is applied with varying magnetic field strengths. A radiofrequency coil (RF-coil), perpendicular to B0 sends excitation pulses and receives signals. The excitation pulses excite the nuclei and interrupts the equilibrium state causing M0 to flip 90° from the alignment with B0. At the exact moment when the 90° pulse is switched off, the nuclei are in phase precessing together in the transverse plane, creating net magnetization Mxy. This is a state that cannot be used for detection. The process is illustrated in Figure 1.

For the signal to be mapped to a specific region within the body we need differentiated signals from clusters of nuclei (referred to as voxels). This is achieved with two more magnetic field gradients, frequency, and phase encoding gradients, that produces different frequencies and phases for each voxel. Each voxel that has the same frequency within the section will have different phases, and the voxels with the same phase will have different frequencies. The result is that each voxel can be allocated to different regions. Next, a relaxation process begins, a response for the nuclei to release the excited energy acquired from the excitation pulse and return the magnetic vectors to the initial state before the

excitation. The signal is detected during this process. The relaxation times vary with tissue and with the field strength of the MRI machine. For example, the T1

relaxation times is 200 ms in fat, and 2060 ms in cerebrospinal fluid, for a 1,5 T MRI.

The sequence time parameters TE (echo time) and TR (repetition time) are varied in different sequences to produce different contrasts. TR is the repetition time of the excitation pulses. TE is the time between the RF pulse and receiving signal. Different tissues will have different contrasts in the resulting image, depending on how these parameters are chosen, because of the different relaxation times. For example, a long TR combined with a short TE will produce a proton-density image which visualises water content. Other sequences frequently used are gradient echo, which has one more sequence parameter, the flip angel, to better control the image contrast. Two often used sequences are short T1 inversion recovery Figure 1. Proton magnetization

vectors precess about B0 producing net magnetization M0. An excitation pulse flips M0 90°

creating net magnetization Mxy, and cause all proton vectors to sync, before they start to decay and dephase towards B0 again.

(12)

4 (STIR) and fluid attenuated inversion recovery (FLAIR), which are based on the spin echo sequence. Several different sequences are used during an MRI examination, which creates the protocol. The sampled signals from the MRI is stored in k-space, a two-dimensional array with frequency information. The information stored in k-space is transformed using Fourier transform to construct the final anatomical image.

2.2 MRI Time Constraints and Future Technical Advancement

MRI has time constraints because of both technical and physiological properties [12]. The technical constraints include the amplitude of the gradients, and the physiological include stimulations of nerves [12] that can occur when the magnetic fields of the gradients are switching on an off rapidly [13]. The process of creating an MRI image is also time consuming because of the many sequences required to obtain a diagnosable image. An example of a standard protocol for breast examinations can include 9 different sequences, each with 3 mm slice thickness, that together take approximately 30-40 minutes [14]. This time can also include processes necessary in between the imaging sequences, such as positioning or re-positioning of patient and injection of contrast agents [14].

There are technical advancements that have the potential to reduce the scan time in MRI.

Using a technique called Echo Planar Image Mix (EPIMix), A. F. Delgado [15] showed a reduction in scan time from approximately 750 seconds to 78 seconds for a full brain

examination. Each sequence had a reduced scan time. For example, a T1-FLAIR sequence in the standard MRI had a length of approximately 144 seconds and the same sequence with EPIMix had a length of 19 seconds. The sensitivity and specificity were high. The image quality, however, was rated lower than that of the standard MRI technique.

Reducing the sampling quantity is a technique that is used to reduce the scanning time by minimizing the quantity of the sampled signals [16]. It then produces a quality image through reconstruction algorithms [17]. There is potential for better reconstruction of undersampled k-spaces with Artificial Intelligence (AI). S. Z Dong et al. [18] highlight the potential of revolutionizing MRI imaging reconstruction in the future. Frameworks for deep learning are already being produced, such as those by Y. Han et al. [19], D. Lee et al. [20]

and S. Dar et al. [21]. In addition to the technical possibilities of AI, S. Z Dong et al. [18]

also mention possibilities for workflow improvement using AI. Such improvements include detecting abnormalities in images and classification of them, to decrease the reading time for radiologists.

2.3 Process Improvement Projects

Besides technical advancements in MRI, radiology departments can improve utilization of their resources by analysing and optimizing their processes [22]. E. P. Tamm et al. [23]

describe the procedure of process improvement in radiology departments. Their procedure is summarized in the following eight steps:

Step 1: Creating the aim. This first step includes identifying the needs of the customer. A customer in radiology is both the patient and clinicians. Collecting the information can be carried out by a control group or via questionnaires. After identifying the needs, they should be prioritized.

(13)

5 Step 2: Creating a team. The team should include people who are part of the relevant

processes to be improved. It should be a small and inclusive team with members that

understand all steps of the process. The team creates aim statements that are quantifiable and specific enough so that results of implementations can be measured.

Step 3: Choosing metrics and data sources. The metrics need to be a good representation of the process that the aim is focused on. It also needs to be reproductible to measure effects.

Eventual secondary metrics need to be chosen as well, to measure effects on other processes that changes to the main process will affect. Next, where the data can be collected needs to be identified. Such sources can include RIS, questionnaires or observations.

Step 4: Collecting data. An initial evaluation of the process current state is necessary to understand how the process behaves and to evaluate implementation strategies. The initial data needs to be over a large enough amount of time and in a stable state, not be affected by variations, to truly understand its behaviour.

Step 5: Process analysis. To understand what part of the process that is the source of the unwanted effects of the process that is aimed to improve, it needs to be identified. The analysis is commonly executed through flowcharts, tally sheets, and cause and effect diagrams to facilitate the analysis process. The analysis should result in process activities that have improvement opportunities.

Step 6: Creating an implementation plan. After deciding what strategies of change can be implemented in the process, the concrete ways to implement it needs to be outlined. This includes who should be responsible, what are the available resources and what steps are needed to accomplish this change.

Step 7: Implementing changes. Small incremental changes are recommended to minimise unexpected negative effects the changes might have. For example, initially implementing changes to only one MRI scanner. This also allows for sufficient training in the new strategies, if needed. New insights might come from this that required changes to the implementation plan.

Step 8: Repeating previous steps. The previous steps can be repeated until the aim of the project has been accomplished, or for new aims that arise during the project.

2.4 Metrics for Efficiency Measurement in MRI

M. Hu et al. [24] recommend metrics to analyse radiology department productivity and to use for benchmarking. Other studies have used similar metrics as efficiency metrics, of which some of them are covered in chapter 5. Discussion. The metrics of M. Hu et al. [24]

are illustrated in Figure 2 and described below.

- Examination time is defined as the time interval from the start of the first imaging sequence to the end of the last sequence for one patient, as recorded by a Radiology Information System (RIS). It is impacted by the chosen protocol for the examination, the interseries time, if images within a sequence has error and potential re-positioning of the patient. Things to consider with this parameter are the different kinds of protocols that exist for the same referral and the examination time can therefore differ in the analysis depending on which one was used.

(14)

6 - Turnover time is defined as the time between two patients, when the device is not being

utilized. For example, during preparation of the room, positioning of patient and other activities that are necessary in preparing for the examination. This time can be influenced by the health of the patient and placing a catheter for contrast or drug administration.

- Scheduled time is defined as the reserved time for the examination, from the scheduled start of one examination to the scheduled start of the next examination. It can also be expressed as the sum of the examination time and the turnover time.

- Scanner utility is defined as the time when the device is being used to perform examinations divided by the time it is available. It gives information about to what degree, in percentage, the scanner is being utilized.

- The interseries time is defined as the time between imaging sequences in a protocol, which is mainly impacted by the imaging equipment and the chosen protocol. Some activities that cause prolonged interseries time are a necessity. Such activities could include administration of contrast, repositioning of the patient, the placement of coils as well as the staff consulting with each other on protocol changes or other improvements.

Figure 2. The figure illustrates the efficiency metrics with two consecutive examinations. Scheduled time: start of examination for patient 1 to the start of the examination for patient 2. Examination time:

duration of the examination for one patient. Turnover time: end of examination for patient 1 to start of examination for patient 2. Interseries time: time between imaging sequence.

(15)

7 3

Method

This chapter covers the methods used to retrieve data on access times in Swedish hospitals in the market research. It will also cover the methods used to construct the model for process time efficiency analysis.

3.1 Market Research

An initial list of the existing hospitals in Sweden was created. An investigation was then conducted to determine which hospitals have radiology departments, and if they performed MRI examinations, by gathering information from their website and 1177 Vårdguiden [25].

Data for the year 2019, that included the monthly median access time for the hospitals, was collected from SALAR [26], to determine average access times for each hospital using equation 1.

𝐴𝑣𝑒𝑟𝑎𝑔𝑒 𝑚𝑜𝑛𝑡ℎ𝑙𝑦 𝑎𝑐𝑐𝑒𝑠𝑠 𝑡𝑖𝑚𝑒 = 𝛴 𝑚𝑜𝑛𝑡𝑙𝑦 𝑚𝑒𝑑𝑖𝑎𝑛 𝑎𝑐𝑐𝑒𝑠𝑠 𝑡𝑖𝑚𝑒

𝛴 𝑚𝑜𝑛𝑡ℎ𝑠 𝑜𝑓 𝑑𝑎𝑡𝑎 (1)

3.2 Model for Process Time Efficiency Analysis

The model was created using Microsoft Excel in Office 365, version 2006 (Microsoft Corporation, Redmond, WA, USA). The equations used in the model were constructed from definitions in section 2.3 Efficiency Metrics for MR, defined by M. Hu et al. [24]:

𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑒𝑥𝑎𝑚𝑖𝑛𝑎𝑡𝑖𝑜𝑛𝑠 = 𝛴 𝑒𝑥𝑎𝑚𝑖𝑛𝑎𝑡𝑖𝑜𝑛𝑠 (2)

𝐸𝑥𝑎𝑚𝑖𝑛𝑎𝑡𝑖𝑜𝑛 𝑡𝑖𝑚𝑒 = 𝑡𝑖𝑚𝑒𝑠𝑒𝑞𝑢𝑒𝑛𝑐𝑒𝑠𝑒𝑠 𝑒𝑛𝑑 𝑝𝑎𝑡𝑖𝑒𝑛𝑡 1− 𝑡𝑖𝑚𝑒𝑠𝑡𝑎𝑟𝑡 𝑠𝑒𝑞𝑢𝑒𝑛𝑐𝑒𝑠 𝑝𝑎𝑡ient 1 (3) 𝑇𝑢𝑟𝑛𝑜𝑣𝑒𝑟 𝑡𝑖𝑚𝑒 = 𝑡𝑖𝑚𝑒𝑠𝑒𝑞𝑢𝑒𝑛𝑐𝑒𝑠 𝑠𝑡𝑎𝑟𝑡 𝑝𝑎𝑡𝑖𝑒𝑛𝑡 2− 𝑡𝑖𝑚𝑒𝑠𝑒𝑞𝑢𝑒𝑛𝑐𝑒𝑠 𝑒𝑛𝑑 𝑝𝑎𝑡𝑖𝑒𝑛𝑡 1 (4)

𝑆𝑐𝑎𝑛𝑛𝑒𝑟 𝑢𝑡𝑖𝑙𝑖𝑡𝑦 𝑟𝑎𝑡𝑒 = ∑ 𝑒𝑥𝑎𝑚𝑖𝑛𝑎𝑡𝑖𝑜𝑛 𝑡𝑖𝑚𝑒

∑ 𝑡𝑖𝑚𝑒 𝑎𝑣𝑎𝑖𝑙𝑎𝑏𝑒 𝑓𝑜𝑟 𝑠𝑐𝑎𝑛𝑛𝑖𝑛𝑔 (5)

The scheduling metric, defined by M. Hu et al. [24], was modified to determine potential schedule consistency issues. The following equations were created:

𝑆𝑐ℎ𝑒𝑑𝑢𝑙𝑒 𝑐𝑜𝑛𝑠𝑖𝑠𝑡𝑒𝑛𝑐𝑦 𝑠𝑡𝑎𝑟𝑡 = 𝑡𝑖𝑚𝑒𝑠𝑐ℎ𝑒𝑑𝑢𝑙𝑒𝑑 𝑠𝑡𝑎𝑟𝑡 − 𝑡𝑖𝑚𝑒𝑒𝑥𝑎𝑚𝑖𝑛𝑎𝑡𝑖𝑜𝑛 𝑠𝑡𝑎𝑟𝑡 (6) 𝑆𝑐ℎ𝑒𝑑𝑢𝑙𝑒 𝑐𝑜𝑛𝑠𝑖𝑠𝑡𝑒𝑛𝑐𝑦 𝑒𝑛𝑑 = 𝑡𝑖𝑚𝑒𝑠𝑐ℎ𝑒𝑑𝑢𝑙𝑒𝑑 𝑒𝑛𝑑− 𝑡𝑖𝑚𝑒𝑒𝑥𝑎𝑚𝑖𝑛𝑎𝑡𝑖𝑜𝑛 𝑒𝑛𝑑 (7)

The model was created using three worksheets: Input, Calculations and Output. The input worksheet was structured with seven columns to accept input data extracted from Radiology Information Systems (RIS). The data fields include ID, examination start, examination end, scheduled start, scheduled end, info, modality, scanner availability in minutes per day, and scanner availability in days per week.

The calculation worksheet was created to include three different calculation tables. The first table, column A through E, calculates the weekday, examination time (equation 3), turnover time (equation 4) and consistency of scheduled start and end (equations 6 and 7) for each examination. The second table, column G and F, calculates the number of entered weeks of data from the input worksheet. The third table, column J through R, includes averages per

(16)

8 week, and averages per weekday, for each metric. The metrics are number of examinations performed (equation 2), turnover time, examination duration, scanner utility (equation 5), early and late start of examination, and early and late endings of examination.

The output worksheet was created with a header consisting of the model name, a text field to include information about the model, and six 2-D clustered column charts. The charts

display the results for number of examinations, turnover time, examination time, scanner utility, examination start compared to scheduled start, and examination end compared to scheduled end. The charts display the calculated metrics in average per week and average per weekday.

(17)

9 4

Results

This chapter presents the market research with access times in Swedish MRI radiology departments. It also describes the model created to measure process time efficiency.

Figure 3 represents average access times using equation 1 presented in the method chapter, for 18 (out of 21) regions in Sweden that reported data to SALAR [26]. It includes 51 out of the 75 surveyed hospitals that perform MRI examinations. The regions that did not report data was Stockholm, Västerbotten and Jämtland-Härjedalen. Not all hospitals reported data for every month of 2019, they varied between 12 months to 3 months, as presented in Appendix 1. The access times presented here are the averages of the monthly median access time for the hospitals that reported data and is defined as days between referral and

examination completion.

The Swedish hospitals had an access time with median value of 42.6 days per month, as illustrated in figure 4. The Swedish regions had a median value of 39.5 days per month, as illustrated in figure 3. Motala, in the region of Östergötland, had the shortest access time of 20.7 days. Karlstad, in the region of Värmland, had the longest access time of 129.7 days.

Motala reported data for 12 months and Karlstad only reported data for 3 months. The access time information for each hospital, along with number of examinations and reported months, is presented in Appendix 1.

Figure 3. The figure illustrated the average monthly access time for 18 regions in Sweden, sorted by increasing access time. Access time is defined as days from referral to examination completion. The

data is for the year 2019.

25 31 32 33 36 37 38 38 39 40 41 47 49 49 51

61 70

85

0 10 20 30 40 50 60 70 80 90

Days

Region

Average Monthly Access Time per Region

(18)

10 Figure 4. The figure illustrated the average value of access time in 2019, calculated from their monthly

median values. Access time is defined as days from referral to examination completion.

20,7 24,3 24,9 27,9 28,0 28,6 30,4 30,9 31,2 32,0 32,7 34,7

36,1 36,2 36,5 36,5 36,7 37,4 37,5 38,1 39,0 39,3 39,3 40,4 41,5

42,6 42,7 43,8 44,4 44,7 45,4 46,0 46,2 46,3 46,6 47,4 49,0

49,8 52,9 53,5 55,1

56,4 58,5 58,6 59,3 60,1 61,8

65,5 70,7

81,8

129,7

0,0 50,0 100,0 150,0

Lasarettet i Motala Skaraborgs sjukhus…

Hallands sjukhus Halmstad/Varberg/Kungsbacka Södra Älvsborgs sjukhus Mälarsjukhuset Eskilstuna NU-sjukvården…

Gävle sjukhus Lasarettet i Enköping Örebro univeritetssjukhus Örnsköldsvik sjukhus Helsingborg lasarett Vrinnevisjukhuset Norrköping Piteå sjukhus Värnamo sjukhus Lasarettet i Ystad Hudiksvall sjukhus Länssjukhuset Ryhov Jönköping Kungälvs sjukhus Visby Lasarett Gällivare sjukhus Alingsås Lasarett Universitetsjukhuset Linköping Länssjukhuset i Kalmar Oskarshamn sjukhus Ängelholm sjukhus Sunderby sjukhus Västerviks sjukhus Akademiska sjukhuset Lasarettet i Trelleborg Lasarettet i Landskrona Sahlgrenska univeritetssjukhuset Skånes Universitetsjukhus Malmö Skånes Universitetsjukhus Lund Höglandssjukhuset Eksjö Västmanlands sjukhus Köping Nyköpings sjukhus Blekingesjukhuset Karlshamn och Karlskrona Hässleholm sjukhus Angereds närsjukhus Sjukhuset i Torsby Länssjukhuset Sundsvall Härnösand Västmanlands sjukhus Västerås Sollefteå sjukhus Ljungby Lasarett Mora Lasarett Centralsjukhuset Kristianstad Falu Lasarett MRAB Medicinsk röntgen, Ellenbogen Praktikertjänst Sjukhuset i Arvika Centrallasarettet Växjö Centralsjukhuset i Karlstad

Days

Hospital

Average Median Access Time for Hospitals in Sweden

(19)

11 4.2 Model for Process Time Efficiency Analysis

The model calculates efficiency metrics using data extracted from RIS. Blue text indicates user input data fields, and black text indicates calculation fields. The input worksheet is illustrated in figure 5. The following data is accepted in the input worksheet, where required data is listed in bold text.

- ID: Identification number for each examination,

- Examination start: date and time for when each examination started, in the format yyyy-mm-dd hh:mm:ss,

- Examination end: date and time for when each examination ended, in the format yyyy-mm-dd hh:mm:ss,

- Scheduled start: date and time for the scheduled start of each examination, in the format yyyy-mm-dd hh:mm:ss,

- Scheduled end: date and time for the scheduled end of each examination, in the format yyyy-mm-dd hh:mm:ss,

- Info: information regarding the examination. For example, if it was terminated or had an error,

- Modality: the specific scanner that the data extracted from RIS is from,

- Available minutes per day: number of minutes the scanner can be used each day, - Available days per week: number of days per week the scanner is available to be

used.

The calculation worksheet is illustrated in figure 6. The calculations fields are formatted for 12 000 examination inputs with possibilities for extension by the user. The logic of the calculation fields is the following:

- Weekday: retrieves the weekday for each examination,

- Duration: calculates the difference between examination start and examination end - Turnover: calculates difference between end of one examination and the start of the

next one. Returns 0 if the examinations are on different dates.

- Start on time: calculates the difference between scheduled start and examination start. Returns negative value for late starts and positive for early starts.

- End on time: calculates difference between scheduled end and examination end.

Returns negative value for late endings and positive for early endings.

- Weeks of input data: calculates the number of weeks of input data from input worksheet.

- Table with averages per week: sums up each metric value for each weekday, and in total, and divides it by number of weeks of input data. Scanner utility sums the examination times and divides by the amount of time the scanner is available for usage.

The output sheet displays the results in weeks in 2-D clustered column charts and is illustrated in figure 7. The scheduled hours graph illustrates late and early starts of

examinations in minutes. Green indicates an early examination start, and red indicates late examination start.

(20)

12 Figure 7 shows the output of example data input consisting of the information presented in figure 6. The example data is consisting of four examinations performed on two consecutive Mondays. The scanner information is 540 available hours per day, and 5 available hours per week listed in the calculation worksheet. This example data has no real significance and is chosen at random solely for the purpose of illustrating the output worksheet. Enlarged versions image 5,6 and 7 is presented in Appendix 2

Figure 5. Screenshot of example data from input worksheet, used to illustrate output worksheet in figure 7.

Figure 6. Screenshot of the calculation worksheet using example data presented in figure 5 and graphically illustrated in output worksheet in figure 7.

Figure 7. Screenshot of output worksheet using example data presented in figure 5.

(21)

13 5

Discussion

This chapter discusses the market research results for Swedish MRI radiology departments.

It also discusses the model in terms of usability, what improvements can be made to

workflows in practical settings and potential for further improvements of the model. Lastly, recommendations for future work is presented.

The collected data in the market research is from 2019. It can be argued that this is a vital data set. Because of the Covid-19 pandemic data from 2020 might be a misleading

representation caused by the additional pressure put on the healthcare system, at least during the first half of the year.

Karlstad, with the longest access times, only reported data for 3 months, which could be an indication of an inaccurately high representation of their access time. Three Swedish regions were not evaluated for access times: Stockholm, Jämtland-Härjedalen and Västerbotten. This is because they did not report data to SALAR [26], and the information was therefore not accessible with the data collection method used in this project. There was an attempt to collect the data by contacting a coordinator at the Unit for Care Analysis and Statistics in Region Stockholm. The response was that they do not routinely collect access times.

Because of the epidemic situation, we did not want to contact the hospitals for the data and left it therefore for future assessments. Other methods used to gather the missing data could also serve as a complement to collect more comprehensive data from the regions that did not report access times for all 12 months of 2019. Not all hospitals in the presented regions in figure 3 reported data. This means that the regional averages for some regions have missing data and could therefore have a different real average value. It might for that reason be better to study Figure 5 or Appendix 1 with access times for each hospital. The data collected does not disclose how many MRI machines the hospitals have, which could be an important consideration for future studies.

The access times have a large span between the regions, as illustrated in figure 3, which could indicate inefficient workflows with potential to optimize. The access times also vary with hospitals in the same region and with number of patients examined, as illustrated in image 4. These findings are in line with a study by Y. Jiang et al [27] where they analysed 74 hospitals, including 3.7 million data records. They found that hospitals with many patient examinations is not guaranteed to have longer access times, and hospitals with a small number of patients is not guaranteed shorter access times. They also found that the access times can vary largely even with hospitals that are regionally close to each other. This is something that the market research in this report also shows. It should be noted that the collected data is the median value for access times for each hospital, which means that the access times were both shorter and longer for a lot of the patients than the ones reported here. Y. Jiang et al. [27] argue that the access time metric is meaningless on its own if the hospital has different targets in access times depending on patient category. They mention that to these hospitals, it is of greater importance for a priority one patient to receive care within its patient category target access time and not to exceed this target, than it is for a lower priority category. The time the patient exceeds the target defines the exceeding time.

They reported a significant difference in overflow patients, the patients who exceed their patient category access times. Priority level one, two, three and four had overflow rates of 8.8%, 13.4%, 62% and 71%, respectively. Their results showed that most of the average exceeding time falls on patient 4 categories and very little on patient one categories, which

(22)

14 they conclude is because of the severity of the patient one categories health. As a result, they recommend using both exceeding time and overflow rate as additional metrics along with access time.

Ultimately, because the healthcare system is moving towards a patient-centred organisation the focus needs to be on the patient experience and health. Long access times can negatively affect the patient and their families because of increased stress levels [28]. Long access times can also cause a delay in treatment of diseases, such as operations of breast cancer. A study in Canada by C. Baliski et al. [28] could show a delay in treatment with surgery for breast cancer patients of four weeks, compared to the patients who did not need MRI examination prior to their surgery. The waiting aspect for women with breast cancer is the most stress inducing events, apart from getting the diagnosis [29]. This further highlights the need to reduce access times as much as possible.

Waiting times processes during hospital visits is beyond the scope of this project but is something that radiology departments additionally can analyse to increase patient

satisfaction [30] and decrease patient anxiety [31]. H. Thu et al. [31] showed a correlation between increased anxiety levels while waiting for the MRI examination for certain people.

They stated that for individuals who experience anxiety in general, particularly women, the added anxiety in the waiting room becomes much larger. The perception of the waiting time for patients can increase their satisfaction and make them experience a shorter wait time than the real wait time [32]. The perception can be improved by increasing the comfort in waiting rooms through better aesthetics and educational or entertaining activities [32].

5.2 Model for Process Time Efficiency Analysis

The model in this project was initially intended to be used to analyse and benchmark MRI radiology department workflow in Swedish hospitals. Because of the current SARS-CoV-2 pandemic, the intended collaboration with the hospitals could not be carried out. The model has, for that reason, been validated with test data and should be verified in real settings as well. Excel was chosen as the tool for the project because it is easily accessible, can handle a large amount of structured data sets and can illustrate the results graphically.

The model does not distinguish between different examination types, different scanners (if the extracted data includes several scanners) or errors or cancellations that occur during examination that will be displayed in the info field of the input worksheet. However, the data field for information and modality in the input worksheet can be used to filter and remove out data the user deems inappropriate. It is expected to get numerical and visual

representations of efficiencies in workflow from the model. The individual workflows, however, must be analysed on an individual level to see the reasons for the results, and investigate if and how they can be optimized.

5.3 Applications of Model and Improvement Strategies

This section discusses the applications of the model, and what strategies to consider if prolonged times and low utility rate are revealed from the model. These improvement strategies, along with the model, could serve as tools and strategies in further projects by Philips when data can be collected from hospitals. The processes that are discussed are examination time, turnover time, the scheduling processes and to some degree the utilization rate. Visser and Beech state in their book, Health Operations Management, that “… the design and planning of a process is influenced by its predictability.” [7, p.46]. To design a well performing organization it is important to have the tools to measure, understand and

(23)

15 plan for improvement strategies [33].

When analysing MRI examination time, the specific protocol that is used needs be taken into consideration optimising examination time efficiency [24]. Another possibility is to explore implementation of abbreviated protocols. The implementation of MRI in more areas of imaging that has occurred for the past decades has led to new image sequences being added to already existing protocols, causing added time to the imaging process [4]. An abbreviated version of the MRI protocol has the potential of obtaining a diagnosable image without including all of the imaging sequences of a standard protocol, therefore reducing the scanning time [4]. The aim of an abbreviated protocol is to reduce the scanning and

interpretation time [4, 5], reduce costs associated with MRI and increase the comfort of the patient [4]. There are several studies that show a significant reduction in scanning time when comparing standard protocols to abbreviated versions. V.L Mango et al. [14] demonstrated a reduction from approximately 30-40 minutes to 15 minutes for breast cancer detection. A. B Ross et. Al. [34] presented a reduction from 23 minutes to 6.5 minutes for fractures. S. C Harvey et al. showed a reduction from 23 minutes to 4.4 minutes in breast cancer screening.

Reducing the scanning time also allows for reduction of scheduled time slots and in return increasing the number of patients that can be examined [35].

Reducing the scan time is not all that needs to be considered, the image must also be of good quality. The detection rate of the abbreviated protocol versus the full protocol seems to be, in general, very high. V. L Mango et al. [14] showed a breast cancer detection sensitivity of 100% for one of the interpreters and 92% from three other, with a reading time of 44 seconds from what they mention as a standard time of just under 5 minutes. The

interpretation time reduces because the interpreter will have fewer sequences of the scan to review [4], which also can enable double reading of the images using a second interpreter or the same interpreter in a different session [35]. Although switching to abbreviated protocols show promising results, the protocols that are suggested in research literature is not

supported by all experts in the field or included in professional guidelines yet [4] and the implementation strategy should be specific for the department to best suit their workflow, patients and needs [35]. The economic benefits could be an incitement for MRI departments to investigate the use of abbreviated protocols in certain situations. C. Besta et al. [36]

reported that potential cost reductions of the MRI process when using abbreviated protocols could be as high as 30.7-49%.

Because of recent advances on the technical aspects of MRI, the total time that the patient is in the scanning room has been decreased [10]. As a result, the turnover time takes up a larger portion of the workflow of the radiology department, therefore making it an important metric to consider, but it is rarely assessed [10]. Dividing activities necessary to perform the

examination appropriately, such as preparatory tasks and transportation, are good practices for efficiency [37]. A study conducted by M. P. Recht et al. [10] compared the turnover time in their old MRI facility with their newly constructed facility. When creating a new MRI center, they took the opportunity to redesign their workflow to reduce the turnover time.

They assembled a team with many different stakeholders to design the workflow and the architecture of the new facility. The new architecture of the facility included optimal placement of doors, with one for entering and one for exiting, a room dedicated for preparations of patients, optimal angle placements between the scanners and the pathway, dockable tables for the scanners, and duplicates of the most used coils. Their new workflow was focused on making sure the patient was ready ahead of the examination, in their new preparation room. The preparations included coil placement, instructing the patient of the

(24)

16 examination, earplug placement, and IV catheter to be put in place. K. L. S K Tukour et al.

[9] found that patients arriving to their scheduled MRI examination needing contrast, or other drug administration, without an IV catheter in place have a significant impact on the total MRI examination process time. If there already is a policy for IV placement in place ahead of time, they emphasized ensuring compliance to the policy and reinforcing it if needed.

M. P. Recht et al. [10] evaluated the turnover times for both prepared and not prepared patients. The old facility has turnover times of 430 respectively 481 seconds. The new facility had turnover times of 115 respectively 141 seconds. The patients that were not ready on time were late mostly because of late arrivals or not arriving at all to their scheduled appointment. Their initial approximated turnover time, at the old MRI centre, was 10 minutes and they reduced it by a mean of 5 minutes and 28 seconds at the new centre. They concluded that these decreases in turnover time would allow for more patients to be

examined and it simplified their scheduling process by only needing 30-minute time slots.

K. Beker et al. [22] used the Lean methodology to evaluate utilization rate at a hospital in the United States. The aim of using Lean as an analysis, they describe, tool is to divide process time into value-adding time, business-value-adding time, and non-value-adding time. By doing so, they could identify and correct the areas that did not bring any real value to the MRI process. Their results were that over 29% of the total patient stay time in their

department consisted of non-value-added time, approximately 32% of business-value-added time, and about 38% value-added time. Their study further highlights the importance of preparing the patient ahead of time, as problems occurring with the IV catheter placement had the biggest impact on patient stay.

Different programming models have been proposed to improve the scheduling process [38].

However, the present mathematical models in research literature are difficult for departments to implement themselves. They are not sufficiently enough based on real settings, and are focused on specific problems and departments [33]. Examples of such problems are solving the problem of examination times with uncertain duration [38] and planning strategies for present and future examination demands [39]. V. A. Loving et al. [32] recommend using simulations instead of mathematical formulas because of their ability to handle more

complex variables within an operation than a mathematical model can. Simulations can also be used to evaluate a certain improvement strategy or model, and can be more cost effective, as well as remove negative impact on patient care, than a real time experiment will [40]. One common factor is the importance of a flexible scheduling process which has been shown to increase scanner utility rate [6, 9, 39]. Patient no-shows has also been shown to increase with access time, which if eliminated often results in better decisions within the operation

management [41].

Access time is often used as a metric when dealing with scheduling issues [27]. J. R. van Sambeek [6] conducted a study in the Netherlands focused solely on optimizing the

scheduling process. Their new strategy made changes to the block arrangement for different imaging categories after analysing the existing process, simulating different new strategies, and implementing the best strategy. It resulted in an increase in utility rate by 10% and a reduction in access time between 6 and 29 days, depending on patient category and scanner.

In the study conducted by Y. Jiang et al. [27], as discussed in section 5.1 Market Research, they created two scheduling policies which are aimed to reduce the prolonged access times compared to the target access times. Their policies were simulated using the real data they

(25)

17 had collected. The simulation resulted in a significant decrease in access time. Some patients had a decrease of over 200 days in access time. All patients, however, had an access time of less than 54 days. Furthermore, the Swedish region of Kronoberg [42] reports success in reducing its access time after implementing a more flexible scheduling process which resulted in a decrease of access time from 21 to 8 weeks. Its strategy also included adding evening and weekend shifts and using consultants to perform Computed Tomography (CT) examinations.

The model could, and was intended to, measure efficiency and to benchmark hospitals in Sweden. Benchmarking is a comparison of performance between similar organizations using objective data to determine best practices [43]. The goal with benchmarking is to find areas within an organisation that have worse results than their peers, to potentially improve processes within the organisation [43]. Getting the objective data from different hospitals, while simultaneously studying the workflow processes, could provide information as to what practices within radiology departments are more efficient and time saving than others. If there are new strategies being implemented, the model can also be used to measure the performance of the new strategies to determine if they were successful or not, as a so-called Key Performance Indicator (KPI). The turnover time, for example, was a KPI in the project by M. P. Recht et al. [10].

Furthermore, when analysing workflow areas for improvement and implementing new workflow strategies, it can be beneficial to include an outsider part to remove eventual biases resulting from expectations. J. R. Van Sambeek et al. [9] found indications of biases in the analysis process when only including the radiology team. Their conclusion was that biases can come from the fact that the radiology team is more likely to look outside of their own workflow, and control, for the cause of undesired effects in their organisation. They recommend creating a team that includes both insiders and outsiders of the department to create more objective base for creating new workflows and strategies. This creates an incitement for Philips, as an outsider part, to assist in workflow improvement projects.

5.4 Potential to Expand the Model

The model measures efficiency with 5 different metrics. There is potential to expand the model with more metrics to include a more comprehensive and specific analysis of workflow. The model could be extended to be able to display results for more modalities than one and different examination types. This would make the model more user friendly and need less preparation of input data. Using pivot tables and charts could possibly allow for better sorting and filtering options, for example the exclusion of certain dates. Including analysis capability for monthly output could also be beneficial.

M. Mayer and R. Sebro [44] conducted a study investigating the impact on MRI process turnaround times for weekend examinations in musculoskeletal radiology. They define turnaround time measurements for both the clinician (cTAT) and radiologist (rTAT), as well as the sum of them both (pTAT). Their findings were that reports of examinations performed and reported by radiologist during weekends were not as likely to be opened and read by the clinician during the weekend. The cTAT time increased with a median of 1.6 days when performed during weekends compared to weekdays. Evaluating such turnaround times when implementing strategical changes in workflow, or to evaluate current workflows, could be a possible extension to the analysis model.

(26)

18 Interseries time, a metric defined by M. Hu et al. [24] and described in 2.3 Metrics for

Efficiency Measurement in MRI, was not included in the model but it could be extended to include it. It was not included in this model because of uncertainty if and how the data would be extracted from RIS. When using interseries times as a metric, M. Hu et al. [24]

emphasize that a simultaneous analysis of the chosen protocols and the interseries time would be beneficial. Because of the discussion about abbreviated protocol in the previous section, both the interseries time and implementation of abbreviated protocols could be worth investigating.

5.5 Future Work

There is a wide range of access times to MRI examinations between hospitals and regions in Sweden, which indicates potential for improvements to be made to workflows. It is therefore my recommendation that the intended benchmarking project is carried out by Philips when data from hospitals are accessible. This will enable Philips to potentially find best practice approaches in radiology workflow. They will also be able to give recommendations to radiology departments that can make their workflow more efficient, and potentially reduce access times to MRI examinations in Sweden. The process time analysis model presented in this report can be used to benchmark the MRI processes.

(27)

19 6

Conclusion

The Swedish hospitals have a large span of different access times for MRI examinations.

Long access times indicates potential to improve workflow. The model created in this project can be used to measure six different process efficiency metrics in MRI radiology.

Future analysis and improvements of processes within MRI departments and their workflows can reduce the access times for MRI in Sweden.

(28)

20 7

References

[1] SKL, "Hälso- och sjukvårdsrapporten 2019," 2019. [Online]. Available:

https://webbutik.skr.se/bilder/artiklar/pdf/7585-729-9.pdf?issuusl=ignore

[2] Socialstyrelsen, "Grund och struktur för lägesrapportering om kroniska sjukdomar,"

2014. [Online]. Available: https://www.socialstyrelsen.se/globalassets/sharepoint- dokument/artikelkatalog/ovrigt/2014-9-40.pdf

[3] Socialstyrelsen and Universitetskanslersämbetet, "Framtidens vårdkompetens," 2019.

[Online]. Available: https://www.socialstyrelsen.se/globalassets/sharepoint- dokument/artikelkatalog/ovrigt/2019-8-6244.pdf.

[4] R. Canellas et al., "Abbreviated MRI Protocols for the Abdomen," Radiographics, vol.

39, no. 3, pp. 744-758, May-Jun 2019, doi: 10.1148/rg.2019180123.

[5] R. M. Mann, J. C. M. van Zelst, S. Vreemann, and R. D. M. Mus, "Is Ultrafast or Abbreviated Breast MRI Ready for Prime Time?," Current Breast Cancer Reports, vol. 11, no. 1, pp. 9-16, 2019, doi: 10.1007/s12609-019-0300-8.

[6] J. R. van Sambeek, P. E. Joustra, S. F. Das, P. J. Bakker, and M. Maas, "Reducing MRI access times by tackling the appointment-scheduling strategy," BMJ Qual Saf, vol. 20, no. 12, pp. 1075-80, Dec 2011, doi: 10.1136/bmjqs.2010.049643.

[7] J. B. Vissers, Rogers, Health Operations Management - Patient flow logistics in health care. New York: Routledge, 2005.

[8] Y. Y. Cheung, E. M. Goodman, and T. O. Osunkoya, "No More Waits and Delays:

Streamlining Workflow to Decrease Patient Time of Stay for Image-guided Musculoskeletal Procedures," Radiographics, vol. 36, no. 3, pp. 856-871, 2016.

[9] K. L. S. TOKUR, D.D. TERRIS1 M.N. JARCZOK, S. BENDER, S.O.

SCHOENBERG and A. G. WEISSER, "Process analysis to reduce MRI access time at German University Hospital," International Journal for Quality in Health Care 2012, vol. 24, pp. 95-99, 2011.

[10] M. P. Recht et al., "Optimization of MRI Turnaround Times Through the Use of Dockable Tables and Innovative Architectural Design Strategies," American Journal of Roentgenology, vol. 212, no. 4, pp. 855-858, 2019, doi: 10.2214/ajr.18.20459.

[11] B. H. P. P Dendy, Physics for diagnostic radiology, 3 ed. (Series in Medical Physics and Biomedical Engineering). Taylor & Francis Group, 2012.

[12] M. Lustig, D. Donoho, and J. M. Pauly, "Sparse MRI: The application of compressed sensing for rapid MR imaging," Magnetic Resonance in Medicine, vol. 58, no. 6, pp.

1182-1195, 2007, doi: 10.1002/mrm.21391.

[13] M. Davids, B. Guerin, A. Vom Endt, L. R. Schad, and L. L. Wald, "Prediction of peripheral nerve stimulation thresholds of MRI gradient coils using coupled

electromagnetic and neurodynamic simulations," Magn Reson Med, vol. 81, no. 1, pp.

686-701, Jan 2019, doi: 10.1002/mrm.27382.

[14] V. L. Mango et al., "Abbreviated protocol for breast MRI: are multiple sequences needed for cancer detection?," Eur J Radiol, vol. 84, no. 1, pp. 65-70, Jan 2015, doi:

10.1016/j.ejrad.2014.10.004.

[15] A. F. Delgado et al., "Diagnostic performance of a new multicontrast one‐minute full brain exam (EPIMix) in neuroradiology: A prospective study," Journal of Magnetic Resonance Imaging, vol. 50, no. 6, pp. 1824-1833, 2019, doi: 10.1002/jmri.26742.

[16] J. Tsao and S. Kozerke, "MRI temporal acceleration techniques," J Magn Reson Imaging, vol. 36, no. 3, pp. 543-60, Sep 2012, doi: 10.1002/jmri.23640.

[17] D. Jizhong, L. Yu, and Z. Liyi, "Bregman Iteration Based Efficient Algorithm for MR Image Reconstruction From Undersampled K-Space Data," IEEE Signal Processing Letters, vol. 20, no. 8, pp. 831-834, 2013, doi: 10.1109/lsp.2013.2268206.

(29)

21 [18] S. Z. Dong, M. Zhu, and D. Bulas, "Techniques for minimizing sedation in pediatric

MRI," Journal of Magnetic Resonance Imaging, vol. 50, no. 4, pp. 1047-1054, 2019, doi: 10.1002/jmri.26703.

[19] Y. Han, L. Sunwoo, and J. C. Ye, "k -Space Deep Learning for Accelerated MRI,"

IEEE Trans Med Imaging, vol. 39, no. 2, pp. 377-386, Feb 2020, doi:

10.1109/TMI.2019.2927101.

[20] D. Lee, J. Yoo, S. Tak, and J. C. Ye, "Deep Residual Learning for Accelerated MRI Using Magnitude and Phase Networks," IEEE Trans Biomed Eng, vol. 65, no. 9, pp.

1985-1995, Sep 2018, doi: 10.1109/TBME.2018.2821699.

[21] S. U. H. Dar, M. Ozbey, A. B. Catli, and T. Cukur, "A Transfer-Learning Approach for Accelerated MRI Using Deep Neural Networks," Magn Reson Med, vol. 84, no. 2, pp. 663-685, Aug 2020, doi: 10.1002/mrm.28148.

[22] K. Beker, A. Garces-Descovich, J. Mangosing, I. Cabral-Goncalves, D. Hallett, and K.

J. Mortele, "Optimizing MRI Logistics: Prospective Analysis of Performance, Efficiency, and Patient Throughput," AJR Am J Roentgenol, vol. 209, no. 4, pp. 836- 844, Oct 2017, doi: 10.2214/AJR.16.17698.

[23] E. P. Tamm, J. Szklaruk, L. Puthooran, D. Stone, B. L. Stevens, and C. Modaro,

"Quality Initiatives: Planning, Setting Up, and Carrying Out Radiology Process

Improvement Projects," Radiographics, no. September-October, pp. 1529-1543, 2012.

[24] M. Hu et al., "Informatics in Radiology Efficiency Metrics for Imaging Device Productivity," RadioGraphics, vol. 31, no. 2, pp. 603-617, 2011.

[25] Vårdguiden. "1177 Vårdguiden." www.1177.se (accessed 2020-04-07.

[26] SKL. Undersökningar [Online] Available:

https://www.vantetider.se/Kontaktkort/Sveriges/Undersokningar

[27] Y. Jiang, H. Abouee-Mehrizi, and Y. Diao, "Data-driven analytics to support scheduling of multi-priority multi-class patients with wait time targets," European Journal of Operational Research, vol. 281, no. 3, pp. 597-611, 2020, doi:

10.1016/j.ejor.2018.05.017.

[28] C. E. M. Christopher Baliski, Caitlyn M. Liberto, Sandra Broughton, Susan Ellard, Marianne Taylor, Janet Bates, Anky Lai, "Influence of nurse navigation on wait times for breast cancer care in a Canadian regional cancer center," The American Journal of Surgery, vol. 207, no. 5, pp. 686-692, 2014, doi:

https://doi.org/10.1016/j.amjsurg.2014.01.002.

[29] B. L. Green et al., "Prevalence of Posttraumatic Stress Disorder in Women With Breast Cancer," Psychosomatics, vol. 39, no. 2, pp. 102-111, 1998, doi:

10.1016/s0033-3182(98)71356-8.

[30] F. Bielen and N. Demoulin, "Waiting time influence on the satisfaction‐loyalty

relationship in services," Managing Service Quality: An International Journal, vol. 17, no. 2, pp. 174-193, 2007, doi: 10.1108/09604520710735182.

[31] H. Thu, S. E. Stutzman, C. Supnet, and D. M. Olson, "Factors Associated With

Increased Anxiety in the MRI Waiting Room," Journal of Radiology Nursing, vol. 34, no. 3, pp. 170-174, 2015, doi: 10.1016/j.jradnu.2015.04.009.

[32] V. A. Loving, R. L. Ellis, R. Rippee, J. R. Steele, D. F. Schomer, and S. Shoemaker,

"Time Is Not on Our Side: How Radiology Practices Should Manage Customer Queues," Journal of the American College of Radiology, vol. 14, no. 11, pp. 1481- 1488, 2017, doi: 10.1016/j.jacr.2017.06.006.

[33] R. K. Jha, B. S. Sahay, and P. Charan, "Healthcare operations management: a structured literature review," Decision, vol. 43, no. 3, pp. 259-279, 2016, doi:

10.1007/s40622-016-0132-6.

[34] A. B. Ross, B. Y. Chan, P. H. Yi, M. D. Repplinger, D. J. Vanness, and K. S. Lee,

(30)

22

"Diagnostic accuracy of an abbreviated MRI protocol for detecting radiographically occult hip and pelvis fractures in the elderly," Skeletal Radiology, vol. 48, no. 1, pp.

103-108, 2018, doi: 10.1007/s00256-018-3004-7.

[35] S. C. Harvey, P. A. Di Carlo, B. Lee, E. Obadina, D. Sippo, and L. Mullen, "An Abbreviated Protocol for High-Risk Screening Breast MRI Saves Time and Resources," Journal of the American College of Radiology, vol. 13, no. 4, pp. 374- 380, 2016, doi: 10.1016/j.jacr.2015.08.015.

[36] C. Besa et al., "Hepatocellular carcinoma detection: diagnostic performance of a simulated abbreviated MRI protocol combining diffusion-weighted and T1-weighted imaging at the delayed phase post gadoxetic acid," Abdominal Radiology, vol. 42, no.

1, pp. 179-190, 2016, doi: 10.1007/s00261-016-0841-5.

[37] Y. A. Ozcan and J. S. Legg, "Performance measurement for radiology providers: a national study," Int. J. Healthcare Technology and Management, vol. 14, no. 3, 2014.

[38] H. Qiu, D. Wang, Y. Wang, and Y. Yin, "MRI appointment scheduling with uncertain examination time," Journal of Combinatorial Optimization, vol. 37, no. 1, pp. 62-82, 2017, doi: 10.1007/s10878-017-0210-5.

[39] A. P. Carpenter, L. M. Leemis, A. S. Papir, D. J. Phillips, and G. S. Phillips,

"Managing magnetic resonance imaging machines: support tools for scheduling and planning," Health Care Manag Sci, vol. 14, no. 2, pp. 158-73, Jun 2011, doi:

10.1007/s10729-011-9153-z.

[40] B. V. Wessman, A. K. Moriarity, a. Ametlli, and D. J. Kastan, "Reducing Barriers to Timely MR Imaging Scheduling," Radiographics, vol. 34, no. 7, pp. 2064-2071, 2014.

[41] N. Liu and S. Ziya, "Panel Size and Overbooking Decisions for Appointment-Based Services under Patient No-Shows," Production and Operations Management, vol. 23, no. 12, pp. 2209-2223, 2014, doi: 10.1111/poms.12200.

[42] RegionKronoberg. "Kortare väntetider till röntgen."

http://www.regionkronoberg.se/nyhetsarkiv/2019/minskade-vantetider-till-rontgen/

(accessed 2020-04-09.

[43] A. Wind and W. H. van Harten, "Benchmarking specialty hospitals, a scoping review on theory and practice," BMC Health Serv Res, vol. 17, no. 1, p. 245, Apr 4 2017, doi:

10.1186/s12913-017-2154-y.

[44] M. Mayer and R. Sebro, "An Important and Often Ignored Turnaround Time in Radiology - Clinician Turnaround Time: Implications for Musculoskeletal Radiology," J Belg Soc Radiol, vol. 103, no. 1, p. 49, Aug 14 2019, doi:

10.5334/jbsr.1834.

Model for Process Time Analysis in Magnetic Resonance Imaging

Model for Process Time Analysis in Magnetic Resonance Imaging

Workflow Optimization to Reduce Access Time ELINA BROMAN

Model for Process Time Analysis in Magnetic Resonance Imaging

Modell för processtidsanalys inom magnetresonanstomografi

Abstract

Sammanfattning

Acknowledgement

Contents

Introduction

Background

Method

Results

Average Median Access Time for Hospitals in Sweden

Discussion

Conclusion

References