International Journal of Medical Informatics 88 (2016) 34–43

Contents lists available at ScienceDirect

International Journal of Medical Informatics journal homepage: www.ijmijournal.com

Assessment of hospital processes using a process mining technique: Outpatient process analysis at a tertiary hospital Sooyoung Yoo a , Minsu Cho b , Eunhye Kim a , Seok Kim a , Yerim Sim b , Donghyun Yoo c , Hee Hwang a , Minseok Song d,∗ a

Center for Medical Informatics, Seoul National University Bundang Hospital, South Korea School of Business Administration, Ulsan National Institute of Science and Technology, South Korea c Patient’s Affairs, Seoul National University Bundang Hospital, South Korea d Department of Industrial & Management Engineering, POSTECH (Pohang University of Science & Technology), South Korea b

a r t i c l e

i n f o

Article history: Received 8 June 2015 Received in revised form 28 November 2015 Accepted 23 December 2015 Keywords: Process mining Outpatient care process Process analysis Process changes

a b s t r a c t Introduction: Many hospitals are increasing their efforts to improve processes because processes play an important role in enhancing work efficiency and reducing costs. However, to date, a quantitative tool has not been available to examine the before and after effects of processes and environmental changes, other than the use of indirect indicators, such as mortality rate and readmission rate. Methods: This study used process mining technology to analyze process changes based on changes in the hospital environment, such as the construction of a new building, and to measure the effects of environmental changes in terms of consultation wait time, time spent per task, and outpatient care processes. Using process mining technology, electronic health record (EHR) log data of outpatient care before and after constructing a new building were analyzed, and the effectiveness of the technology in terms of the process was evaluated. Results: Using the process mining technique, we found that the total time spent in outpatient care did not increase significantly compared to that before the construction of a new building, considering that the number of outpatients increased, and the consultation wait time decreased. These results suggest that the operation of the outpatient clinic was effective after changes were implemented in the hospital environment. We further identified improvements in processes using the process mining technique, thereby demonstrating the usefulness of this technique for analyzing complex hospital processes at a low cost. Conclusion: This study confirmed the effectiveness of process mining technology at an actual hospital site. In future studies, the use of process mining technology will be expanded by applying this approach to a larger variety of process change situations. © 2016 Published by Elsevier Ireland Ltd.

1. Introduction The processes of many organizations play an important role in advancing the performance and efficiency of work [1]. The organization must constantly improve processes to gain competency. First, the process must be understood, and one such tool for understanding the process is process mining technology [2]. Process mining refers to the analysis of the process log or data corresponding to the process of a business and extracting the required information [3,20]. Process mining technology consists of three main types—discovery, conformance, and enhancement

∗ Corresponding author at: 77 Cheongam-Ro, Nam-Gu, Pohang, Gyeongbuk 37673, South Korea. E-mail address: [email protected] (M. Song). http://dx.doi.org/10.1016/j.ijmedinf.2015.12.018 1386-5056/© 2016 Published by Elsevier Ireland Ltd.

[1,2,4,8,11,18]. That is, process mining is not only capable of discovering process models but also of identifying deviations between models and log, and of conducting performance analyses [1,16]. Process mining allows for the extraction of necessary information related to the process and allows one to derive, monitor, and improve the actual process [4]. Because process mining technology analyzes the log, which is already in the form of data, it can save time and reduce the costs of data collection. In addition, process mining technology only analyzes recorded data, which prevents distortion of information and ensures accuracy and objectiveness [1,5,6,15]. In hospital settings, clinical and administrative data have been computerized with the introduction of electronic health records (EHR) systems to improve the quality of medical care [7,21,22]. Such systems consist of a wide range of applications that record events [8], which further expand the opportunities for applying

S. Yoo et al. / International Journal of Medical Informatics 88 (2016) 34–43

process mining technology in the field of medicine. Hospitals with these data apply a care system that is based on the process, which allows for the selection of medical services based on patients’ conditions [9]. This approach improves the patient care and quality of care, thus underscoring the importance of the role of the process [10,19,23]. However, standardizing such processes in hospitals is not easy, and discovering problems and implementing improvements remain challenging. Limitations include a lack of new research in this area, an inability to apply a formal process flow due to large discrepancies across individual patients and hospital organizations, and difficulties in identifying the types of health care processes that occur even in a group of patients with the same diagnosis due to numerous diverging factors [11]. In addition, a data-based, automated tool for evaluating hospital process changes after improvements are made is currently lacking [12], and therefore the quality of health care delivery can only be evaluated based on indirect clinical indicators, such as mortality and readmission rates [9,10]. One previous study that analyzed the treatment process of stroke patients [8] and another that assessed the treatment process of female oncology patients [11] used process mining technology to analyze a complex process, such as that of a hospital, and confirmed the applicability of the technology. However, these previous studies have not devised the strategies to apply and assess the results of the process mining in the actual hospital sites. In our previous study, we compared an expert-driven outpatient care process with a machine-driven process using a process mining technique and measured task-based performance [13]. However, we did not evaluate the performance in relation to changes in the hospital environment or the achieved process. In this study, we used process mining technology to analyze process changes based on changes in the environment and the effects of the changes in terms of the consultation wait time, the amount of time required for each step, and the total outpatient care process. This approach was used to assess the applicability of process mining technology as an objective tool for hospital process evaluation and study the effectiveness of the technology at a real medical site. 2. Methods 2.1. Materials The study site, Seoul National University Bundang Hospital, is a tertiary hospital in South Korea that includes 1400 beds and 38 operating rooms. The hospital was established in May of 2003 and

35

has been operating as a fully digital hospital since its establishment. In April of 2013, the hospital opened a new building that was annexed to the original building and updated the emergency room, intensive care unit, cancer center, and clinical neuroscience center to accommodate the new space. The new building houses the emergency room, intensive care unit, comprehensive cancer center, and clinical neuroscience center to increase collaboration among departments and the efficiency of treating patients with severe, rare, or incurable diseases. After the inception of the new building, the cancer center moved from the second floor of the original building to the second floor of the new building, and the clinical neuroscience center moved from the first floor of the original building to the second floor of the new building. Moreover, additional administrative and register counters were installed in the new building. A room dedicated to anti-carcinogenic serum injection was established on the fifth floor of the new building, and outpatients of the cancer center were able to receive the injection service in the new building. To evaluate the effectiveness of the changes in the hospital facility environment before and after the establishment of the new building in terms of the process, we selected the EHR log data of the outpatient clinics at the new building, such as the cancer and clinical neuroscience centers, as the analysis object, where there have been changes in the outpatient care environment due to the establishment of the new building. We collected one month of data prior to the establishment of the new building in July of 2012 and one month of data after the establishment of the new building in July of 2013. Since most of indexes regarding the evaluation of care processes are measured per month or day, data for a month is used. 2.2. Event log collection and pre-processing We defined the event log of the outpatient care process as shown in Table 1. We collected the activity completion time, the corresponding department, and the related detail information for all the outpatient care activities that are performed when patients visit our hospital. However, regarding treatment events, we collected the treatment start date rather than the treatment completion date because the system did not have information regarding the activity completion time. We assumed that the treatment event would be inserted in the last stage of all processes. Since we have to identify case ID in the analysis to separate events according to patients, the anonymized patient ID is generated for each patient. Instead of the anonymized ID, any personal information such as social security

Table 1 Types and attributes of event logs for analyzing outpatient care processes. Event type (activity)

Attribute

Sign on selective medical service Referral registration Outside image registration Payment Test registration Test Consultation registration

Case ID, Activity completion time, Resource ID, Resource department code Case ID, Activity completion time, Resource ID, Resource department code Case ID, Activity completion time, Resource ID, Resource department code Case ID, Activity completion time, Resource ID, Resource department code Case ID, Activity completion time, Resource ID, Resource department code, Test code, Type of test, Scheduled test date Case ID, Activity completion time, Resource ID, Resource department code, Test code, Type of test, Scheduled test date Case ID, Activity completion time, Resource ID, Resource department code, Patient type, Department code, Appointment method, Appointment Date Case ID, Activity completion time, Resource ID, Resource Department code, Patient type, Department code, Appointment method, Appointment Date Case ID, Activity completion time, Resource ID, Resource department code, Patient type, Practitioner ID, Scheduled department code, Scheduled consultation date Case ID, Activity completion time, Resource ID, Resource department code, Test code, Type of test, Scheduled test date Case ID, Activity completion time, Resource ID, Resource department code Case ID, Activity completion time, Resource ID Case ID, Activity completion time, Resource ID, Resource department code Case ID, Activity completion time, Resource ID, Resource department code Case ID, Treatment start date, Resource ID, Resource department code, Treatment code

Consultation Consultation scheduling Test scheduling Admission scheduling Outside-hospital prescription printing In-hospital prescription receiving Certificate issuing Treatment

Case ID: a unique ID for identification of outpatients of the day, Resource ID: a unique ID for identification of someone or something that performed the specific activity, Resource department code: a unique code for identification of the resource ID’s departments.

36

S. Yoo et al. / International Journal of Medical Informatics 88 (2016) 34–43

Table 2 Process mining technique. Analysis items

Purpose of items

Process mining techniques

Analysis method and measured matrix

Data overview analysis

To provide general information

Analyzing basic data statistics

Frequency analysis

To check out what is important of each task in process To figure out how long patients stay in the hospital To figure out how long patients receive each service To investigate how the construction has affected the process in depth To figure out how events are distributed within

Basic performance analysis [1,11] Basic performance analysis [1,11] Basic performance analysis [1,11] Basic performance analysis [1,11] Basic performance analysis [1,11] Dotted chart analysis [17]

Performance analysis for outpatient care process Performance analysis for each task Detailed comparative analysis of processes before and after construction Analysis of event distribution

Analysis of relationship between precedence and following task Comparative analysis using standard process

To identify flows and order relationship among tasks To evaluate deviation of the discovered process model from the standard one

Discovery of process model Process pattern analysis

To develop process model To identify the most frequent care flows

number, address, phone number, etc. is not used. In this way, we tried to protect the privacy of patient records. When pre-processing the collected data, duplicate data were deleted when the same activity occurred consecutively within three minutes or less. That is, when an activity was repeated consecutively due to the staff’s computational task, for instance, payment → payment → payment, the duplicated data were deleted. Computational data that were not derived from actual patient visits were deleted, cleaned, and analyzed. 3. Process mining technique The process mining analysis technique used in this study is shown in Table 2. This method uses basic performance analysis, which is a technology that analyzes accomplishments by measuring the frequency of case, activity, and originator from the event log data, execution and wait time, and total time spent [1,11]. We analyzed the data according to a frequency analysis, the hourly distribution of patients, the total time for an outpatient care process, the time spent per task, the patient-to-task ratio, and a comparative analysis of effects before and after the task. To analyze the task event distribution, we used a dotted chart analysis and derived a two-dimensional graph by indicating task events using dots based on time or frequency [17]. Process discovery is a technology that includes heuristic mining and handover mining [14,16]. Heuristic mining is used to derive a process model based on an algorithm, whereas handover mining schematizes a standard process model using the main flow derived from the process model that is based on the task’s precedence relationship and frequency. We used the process discovery technology to execute specific tasks, such as the analysis of a task’s precedence relationship, derivation of the process model, and comparison of the standard process. The derived process and compared standard process included the mega pro-

Handover mining [16] Handover mining [16]

Heuristic mining [14] Pattern analysis [16]

Computing frequency and rate of occurrence for each task Calculating total time for outpatient care process Conducting comparative analysis of time per task Conducting fine-grained comparative analysis to find out detailed differences Analyzing task event distribution per patient, range that is completed in given amount of time, and number of task events per patient Diagnosing relationship and frequency of precedence and following task Conducting delta analysis and computing matching rate between the discovered and the standard model Deriving process model Discovering care flows and identifying major ones

cesses derived from a previous study [13]. Finally, we used pattern analysis technology [16] to analyze the process pattern, which is a technology that derives process patterns and executes performance analyses for each pattern, such as the frequency, rate, and time.

4. The process changes assessment framework Summing up all the previously proposed contents, we provided a comprehensive framework which represents the overall flow of the assessment of process changes using data analysis. Fig. 1 represents the process mining analysis framework for assessing process changes based on changes in the environment. In the data preparation step, we extracted event logs from EHR log data and performed pre-processing for improving the quality of data. Considering the period of environmental changes, we prepared two event logs in July of 2012 and 2013 from the pre-processed log data. Based on extracted event logs, we performed process mining analyses including the following 10 analyses items—data overview analysis, frequency analysis, performance analysis for outpatient care process, performance analysis for each task, detailed comparative analysis of processes, analysis of event distribution, precedence and following relationship analysis, comparative analysis using standard process, discovery of process model, and process pattern analysis. In addition, each item was analyzed using five process mining techniques which are basic performance analysis, dotted chart analysis, handover mining, heuristic mining, and pattern analysis. After that, we compared two results on before and after of the construction of the new building based on three standards: wait time, process time, and matching rate of the process model. By following the suggested steps in the framework, we have performed process mining data analysis and detailed results are provided in the Section 5.

Table 3 Changes before and after the construction of the new building according to two different center. Cancer center

Total number of outpatients Total time for outpatient care Consultation wait time Test wait time Number of tests per patient

Clinical neuroscience center

Before

After

Growth (%)

1000 116.97 23.24 11.94 0.68

2546 127.33 22.18 19.43 0.77

154.6 8.9 −4.6 62.7 4.4

P-value

Before

After

Growth (%)

P-value

0.023 0.271 0.001 0.027

1337 88.23 27.08 7.17 0.75

2243 90.73 23.72 6.61 0.68

67.8 2.8 −12.4 −7.8 −9.3

0.520 0.005 0.112 0.177

S. Yoo et al. / International Journal of Medical Informatics 88 (2016) 34–43

37

Fig. 1. The process changes assessment framework using process mining techniques.

5. Results 5.1. Analysis of wait time To evaluate the efficiency of operating the outpatient clinic after the establishment of the new building, we analyzed various key performance indices, such as the total time of the outpatient care

process, the consultation wait time, and the test wait time, while considering changes in the number of patients. The total time of the outpatient care process indicates the time from when the process was logged after a patient’s visit to the hospital to the completion of the final process stage. The consultation wait time refers to the duration from the time after consultation is registered until the

38

S. Yoo et al. / International Journal of Medical Informatics 88 (2016) 34–43

Fig. 2. Dotted chart ((a) cancer center; (b) clinical neuroscience center) before and after the construction.

time when the consultation begins. The test wait time refers to the time after the test is registered until the time when the test begins.

Fig. 2 shows dotted charts of the task events over time, where each dot represents a task event, and Table 3 shows the result of these indices. In the figure, the y-axis and the x-axis are configured

S. Yoo et al. / International Journal of Medical Informatics 88 (2016) 34–43

39

Fig. 3. Time required per task for outpatient care processes ((a) Cancer center, (b) Clinical neuroscience center).

as cases and duration, respectively, and all the cases are sorted by duration. From the figure, we can identify the spread of the total time for outpatient care. For example, as shown in Fig. 2(a), 40% of patients in the cancer center received all medical services within one hour both before and after of the construction of the new building. Overall, there is no big difference of the distributions of the total time for outpatient care. We found that the number of outpatients increased by 154.6% (approximately 2.5 times) in the cancer center and 67.8% (approximately 1.7 times) in the clinical neuroscience center compared to the number of outpatients before the new building was established. However, the total time required for outpatient care increased by 8.9% (10.36 min) in the cancer center and by 2.8% (2.5 min) in the clinical neuroscience center. The total time required for outpatient care did not increase significantly considering the growth rate of the number of patients. Rather, the consultation wait time decreased by 4.6% (1.06 min) in the cancer center and by 12.4% (3.36 min) in the clinical neuroscience center compared to the consultation wait times before opening the new building. The test wait time increased by 62.7% (7.49 min) in the cancer center but decreased by 7.8% (0.56 min) in the clinical neuroscience center. Moreover, the number of tests per patient increased by 4.4% in the cancer cen-

ter and decreased by 9.3% in the clinical neuroscience center. For a detailed analysis of the wait time for each test, we categorized the tests by considering the characteristics of each test. The tests were divided into five groups, namely, a specimen test, medical imaging test, special test, departmental test, and miscellaneous tests. For the specimen tests, we included tests such as tests by laboratory departments, including the department of nuclear medicine and the department of pathology. For the medical imaging tests, we included tests from the departments of radiology and medical imaging. A special inspections category included tests conducted by the department of special inspections, such as endoscopy and spirometry tests. The departmental inspections included various specifics tests that were performed in each clinical department. For the miscellaneous section, we included test data logs remaining in groups other than the four aforementioned categories, such as the medical support services team, education training support team, nursing unit, and office of radiation safety management. Based on a comparison of the changes in the test wait time and the number of tests in the cancer center and the clinical neuroscience center by the types of tests, the number of specimen tests for the cancer center increased (Table 4). The number of specimen tests conducted by the laboratory department in the cancer cen-

40

S. Yoo et al. / International Journal of Medical Informatics 88 (2016) 34–43

Table 4 Changes in the test wait time and the number of tests. Test wait time growth (%)

Specimen test Medical imaging test Special test Departmental test Misc.

Growth in the number of tests (%)

Cancer center

P-value

Clinical neuroscience center

P-value

Cancer center

Clinical neuroscience center

53.6 43.8 0.0 −35.5 −61.7

0.000 0.206 0.999 0.692 0.252

−23.1 25.9 −58.3 297.4 −60.7

0.024 0.310 0.267 0.054 0.028

172.9 120.4 146.4 881.8 87.9

53.4 52.5 48.4 55.3 40.7

ter increased approximately 2.5-fold (380 to 960 tests). However, tests conducted by this department have limitations. For example, patients’ test times are not absolute because these tests are not scheduled, which consequently affects the number of patients tested. In addition, tests conducted by the laboratory department did not accurately reflect the patients’ wait time because these tests measured the time that elapsed from the printing of the specimen test label to the time when the specimen was recorded as being taken at the testing site. In contrast, for the scheduled tests, the time was recorded from when a patient was registered for a test to when the test was initiated.

5.2. Process time The process time is shown in Fig. 3. We analyzed the processes from two perspectives: processes that exhibited similar changes at both the cancer center and the clinical neuroscience center and processes that exhibited contrasting changes at the two centers. First, the time for consultation and the issuance of prescriptions

decreased at both centers. The processes that showed an increase in time included activities related to medical support services, such as in-hospital receipt of prescriptions, payment, certificate issuing, consultation referral registration, signed agreement for selective medical service, and registration for video resources of other hospitals. Processes that showed an increase in time at the cancer center but a decrease in time at the clinical neuroscience center included consultation registration, tests, test registration, and consultation scheduling. In contrast, test scheduling, admission scheduling, and severe, rare, or incurable disease registration showed a decrease in time at the cancer center and an increase in time at the clinical neuroscience center. The decrease in the consultation wait time at both centers can be explained by the fact that when considering the time spent for consultation as a combination of the consultation wait time and actual consultation time, the consultation wait time decreased by 1.06 min at the cancer center and by 3.36 min at the clinical neuroscience center. The time required for in-hospital prescriptions also decreased at both centers, which may be the result of an increase in the number of self-checkout machines,

Fig. 4. The most frequent outpatient care processes in the cancer center and clinical neuroscience center ((a) before establishment of the new building, (b) after establishment of the new building).

S. Yoo et al. / International Journal of Medical Informatics 88 (2016) 34–43

41

Table 5 The ten most frequent outpatient care process patterns in the cancer and clinical neuroscience centers (a: the cancer center, b: the clinical neuroscience center). (A) Cancer center Before

After

Process

Frequency

Percentage (%)

Process

Certificate issuing > payment > treatment Consultation registration > consultation > consultation scheduling > payment > inhospital prescription receiving > treatment Consultation registration > consultation > consultation scheduling > payment > treatment Consultation registration > consultation > payment > treatment

36

3.6

5.37

35

3.5

137 Sign on selective medical service > consultation > payment > treatment 115 Certificate issuing > payment > treatment

34

3.4

106

4.16

33

3.3

65

2.55

20

2

61

2.39

18

1.8

35

1.37

Consultation registration > consultation > payment > inhospital prescription receiving > end Consultation registration > consultation > Certificate issuing > payment > treatment Certificate issuing > Certificate issuing > payment > treatment

10

1

34

1.33

10

1

1.10

9

0.9

Consultation registration > consultation > consultation scheduling > test scheduling > payment > in-hospital prescription receiving > treatment

9

0.9

28 Consultation registration > consultation > payment > certificate issuing > payment > treatment Consultation registra25 tion > consultation > payment > prescription printing > treatment 23 Consultation registration > consultation > test scheduling > consultation scheduling > payment > treatment

Process

Frequency

Percentage (%)

Process

Frequency

Percentage (%)

1

Certificate issuing > payment > treatment

114

8.53

172

7.67

2

Test scheduling

88

6.58

149

6.64

3

57 Consultation registration > consultation > payment > treatment Test registration > test 49

4.26

101

4.5

71

3.17

48

3.59

Consultation registration > consultation > consultation scheduling > payment > prescription printing > treatment Certificate issuing > payment > treatment Consultation registration > consultation > payment > treatment Payment > test registration > test Test scheduling

44

1.96

22

1.65

Test registration > test

44

1.96

22

1.65

43

1.92

18 18

1.35 1.35

39 37

1.74 1.65

10

0.75

Consultation registration > consultation > consultation scheduling > payment > inhospital prescription receiving > treatment Payment > treatment Consultation registration > consultation > payment > issue prescription > treatment Consultation registration > consultation > payment

28

1.25

1 2

3

4

5

6

7

8

9

10

Consultation registration > consultation > prescription printing > consultation scheduling > payment > treatment Payment > treatment

Consultation registration > consultation > consultation scheduling > payment > treatment Consultation registration > consultation > consultation scheduling > payment > inhospital prescription receiving > treatment Consultation registration > consultation > consultation scheduling > payment > prescription printing > treatment Consultation registration > consultation > certificate issuing > payment > treatment Payment > treatment

Frequency

Percentage (%)

4.51

0.98

0.90

(B) Clinical neuroscience center Before

4 5 6 7

8 9

10

Payment > test registration > test Certificate issuing > Certificate issuing > payment > treatment Consultation registration > consultation > prescription printing > consultation scheduling > payment > treatment Payment > treatment Consultation registration > consultation > Certificate issuing > payment > treatment Test registration > test > test registration > test

After

3.66

42

S. Yoo et al. / International Journal of Medical Informatics 88 (2016) 34–43

which automatically issue a prescription after receiving payment. The number of self-checkout machines increased from 14.58% to 22.90% at the cancer center and from 17.10% to 22.35% at the clinical neuroscience center. Consultation registration, tests, and test registration increased only at the cancer center, likely because the time for the tests prior to consultation increased, which consequently increased the time needed for test registration, the actual test, and consultation registration. Although both centers use the same administrative and register counter, the time for scheduling tests, admission scheduling, and severe, rare, or incurable disease registration increased only at the clinical neuroscience center and decreased at the cancer center. This difference likely occurs because patients at the cancer staff seemed to be guided well. In-hospital prescriptions increased at both centers because the outpatient pharmacy is located in the original building, resulting in a greater amount of time for outpatients using the new building to receive inhospital prescriptions. Tasks such as payment, consultation referral registration, signed agreement for selective medical service, and registration for video resources of other hospitals also occur at the same administrative and register counter. However, we hypothesize that the increase in the time of these tasks was due to an insufficient number of administrative and register counters in the new building relative to the number of patients. 5.3. Matching rate of the process model Using process mining technology, we analyzed the patterns of patients in the outpatient care clinics of the cancer and clinical neuroscience centers. As shown in Fig. 4 and Table 5, we derived the most frequent outpatient care processes before and after the establishment of the cancer and clinical neuroscience centers at the new building and confirmed the frequency of each process. We found that there were no marked changes in the most frequent outpatient care processes before and after the establishment of the new building. Based on a comparison of the processes used at the cancer and clinical neuroscience centers, the cancer center appeared to have more patients with severe, rare, or incurable disease due to a higher number of in-hospital prescriptions issued and a higher rate of self-injection prescriptions. The clinical neuroscience center was hypothesized to have more patients en route from other hospitals due to the higher rate of consultation referral registrations and registrations for video resources of other hospitals compared to the cancer center. The matching rate, which is the ratio of matches with the expertdriven model in the total flow frequency, increased from 87.0% before the establishment of the new building to 88.9% after the establishment of the new building at the cancer center. However, the matching rate decreased from 86.8% to 85.2% after the establishment of the new building at the clinical neuroscience center.

checkout machines in the new building seemed to contribute to decrease the time of outside-hospital prescription printing. After the establishment of the new building, outpatient found to spend more time on the process of administrative activities such as in-hospital receipt of prescriptions, payment, certificate issuing, consultation referral registration, signed agreement for selective medical service, and registration for video resources of other hospitals. Especially, regarding in-hospital prescriptions, because the outpatient pharmacy was located in the original building, it took longer for outpatients from the new building to receive in-hospital prescriptions. Thus, we believe it is necessary to further evaluate the discomfort experienced by outpatients by considering the labor requirements and cost of operating the current outpatient pharmacy in the original building. Moreover, we found that an additional location should be installed in the new building where patients can receive in-hospital prescriptions. Previous studies that applied process mining technology include a study of the treatment process of stroke patients [8] and a study of the process of female oncology patients [11] that analyzed complex processes, such as those of hospitals, and confirmed the applicability of this technology. However, these previous studies did not utilize the results to seek strategies that could be applied at hospital sites or compare the results to the flowchart that the hospitals expected. In this study, we evaluated the effects of the process on changes in the hospital’s facility environment and identified improvements, thus confirming the effectiveness of process mining technology in hospitals. The limitation of our work would be that the treatment time could not be analyzed since our system records only the start time of the treatment. After the system is modified to collect both the start and completion time of the treatment, it should be analyzed in the future.

7. Conclusion In this study, we confirmed the effectiveness of process mining technology after evaluating changes in the facility environment of a hospital before and after the establishment of a new building. We also assessed the applicability of process mining technology as an automated tool based on an analysis of processes in the hospital. In future studies, process mining technology will be applied to a wider range of situations, such as inpatient care processes and emergency room processes, and will be used to evaluate the performance before and after various changes to processes are implemented, such as process addition, deletion, and changes in the order of operations. We believe that these future studies will expand the effectiveness of process mining technology.

6. Discussion Author contributions Using a process mining technique, we found that the total time spent in outpatient care did not increase markedly considering that the number of patients also increased and that the consultation wait time decreased after the establishment of a new building. Before the establishment of the new building, it appeared that outpatient spent more time on the consultation and outside-hospital prescription printing. These results suggest that the operation of the cancer and clinical neuroscience centers was more efficient after changes in facility environment and processes were implemented. Discrepancies in consultation registration, tests, and test registration may have occurred due to the characteristics of each clinic, such as the occurrence of a test prior to consultation or the types of tests performed. In addition, the increased number of self-

S. Yoo, E. Kim, and S. Kim designed the study and drafted the manuscript. M. Cho, Y. Sim, and M. Song analyzed the data and D. Yoo and H. Hwang contributed to the discussion of data and reviewed the manuscript. M. Song reviewed and revised the manuscript.

Conflicts of interest There are no conflicts of interests that could inappropriately influence the authors’ findings.

S. Yoo et al. / International Journal of Medical Informatics 88 (2016) 34–43

Summary points What was already known • Process mining analyzes recorded log data to derive, monitor, and improve business processes and there are some research works to use it to analyze treatment process in hospitals. • In the medical industry, it has been utilized to analyze the treatment process of stroke patients and female oncology patient. • However, there have been no studies to evaluate various key performance measures, such as consultation waiting time, consultation time, etc. under the changes in care processes/facilities in a hospital. What this study added • In a hospital setting, process mining can be utilized to analyze process changes (e.g., changes in the environment). We suggested a way to investigate process changes with process mining techniques in which several measures such as consultation waiting time, the amount of time required for each step, the total duration of outpatient care processes, etc. can be calculated from EHR data. • With process mining techniques, we analyzed processes in a tertiary hospital and compared the analysis results before expanding a hospital with the results after the expansion. We confirmed the efficiency of the operation of the new building and found some rooms for improvement on the process for outpatient pharmacy and administrative department. • Analysis of the most frequent outpatient care processes of two different clinical departments was useful to find common processes and different processes, reflecting characteristics of disease.

Acknowledgements This work was supported by the Industrial Strategic Technology Development Program (10040142) funded by the Ministry of Trade, Industry and Energy. References [1] W.M.P. van der Aalst, A.J.M.M. Weijters, B.F. van Dongen, A.K. Alves de Medeiros, M. Song, H.M.W. Verbeek, Business process mining: an industrial application, Inf. Syst. 32 (2007) 713–732. [2] M. Song, S.H. Siadat, M. Pechenizkiy, A comparative study of dimensionality reduction techniques to enhance trace clustering performances, Exp. Syst. Appl. 40 (2013) 3722–3737. [3] W.M.P. van der Aalst, J. Herbst, L. Maruster, G. Schimm, A.J.M.M. Weijters, Workflow mining: a survey of issues and approaches, Data Knowl. Eng. 47 (2003) 237–267.

43

[4] W.M.P. van der Aalst, M.H. Schonenberg, M. Song, Time prediction based on process mining, Inf. Syst. 36 (2011) 450–475. [5] W.M.P. Van der Aalst, A.J.M.M. Weijters, L. Maruster, Workflow mining: discovering process models from event logs, IEEE Trans. Knowl. Data Eng. 16 (2004) 1128–1142. [6] C.W. Gunther, W.M.P. Van der Aalst, Fuzzy mining: adaptive process simplification based on multi-perspective metrics, Proceedings of the 5th International Conference on Business Process Management (2007) 328–343. [7] J.A. Linder, J. Ma, D.W. Bates, B. Middleton, R.S. Stafford, Electronic health record use and the quality of ambulatory care in the United States, Arch. Intern. Med. 167 (2007) 1400–1405. [8] R. Mans, H. Schonenberg, G. Leonardi, S. Panzarasa, A. Cavallini, W.M.P. van der Aalst, Process mining techniques: an application to stroke care, Stud. Health Technol. Inform. 136 (2008) 573–578. [9] E.D. Peterson, M.T. Roe, J. Mulgund, E.R. DeLong, B.L. Lytie, R.G. Brindis, S.C. Smith Jr., C.V. Pollack Jr., L.K. Newby, R.A. Harrington, W.B. Gibler, E.M. Ohman, Association between hospital process performance and outcomes among patients with acute coronary syndromes, JAMA 295 (2006) 1912–1920. [10] R.M. Werner, E.T. Bradlow, Public reporting on hospital process improvements is linked to better patient outcomes, Health Aff. 29 (2010) 1319–1324. [11] R. Mans, M.H. Schonenberg, M. Song, W.M.P. van der Aalst, P.J. Bakker, Application of process mining in healthcare: a case study in a Dutch hospital, in: A. Fred, J. Filipe, H. Gamboa (Eds.), Biomedical Engineering Systems and Technologies, Springer, Heidelberg, 2009, pp. 425–438. [12] P. Gemmel, D. Vandaele, W. Tambeur, Hospital Process Orientation (HPO): the development of a measurement tool, Total Qual. Manage. 19 (2008) 1207–1217. [13] E. Kim, S. Kim, M. Song, S. Kim, D. Yoo, H. Hwang, S. Yoo, Discovery of outpatient care process of a tertiary university hospital using process mining, Healthcare Inform. Res. 19 (2013) 42–49. [14] W.M.P. van der Aalst, A.J.M.M. Weijters, A.K. Alves de Medeiros, Process mining with the heuristics miner-algorithm, Technische Universiteit Eindhoven, Tech. Rep. WP 166 (2006) 1–34. [15] M. Song, W.M.P. van der Aalst, Towards comprehensive support for organizaional mining, Decis. Support Syst. 46 (2008) 300–317. [16] M. Cho, M. Song, S. Yoo, A systematic methodology for outpatient process analysis based on process mining, Int. J. Ind. Eng. 22 (2015) 480–493. [17] M. Song, W.M.P. van der Aalst, Supporting process mining by showing events at a glance, in: K. Chari, A. Kumar (Eds.), Proceedings of 17th Annual Workshop on Information Technoogies and Systems (WITS 2007), Montreal, Canada, 2007, pp. 139–145. [18] R. Mans, W.M.P. van der Aalst, R. Vanwersch, Process Mining in Healthcare: Evaluating and Exploiting Operational Healthcare Processes, Springer International Publishing, 2015, pp. 17–26. [19] S.J. Boas, T.F. Bishop, A.M. Ryan, S.C. Shih, L.P. Casalino, Electronic health records and technical assistance to improve quality of primary care: lessons for reginal extension centers, Healthcare 2 (2014) 103–106. [20] R.P. Bose, W.M.P. van der Aalst, Dealing with concept drifts in process mining, Neural Netw. Learn. Syst. IEEE Trans. 25 (2014) 154–171. [21] H. Hypponen, K. Saranto, R. Vuokko, P. Makela-Bengs, P. Doupi, M. Lindqvist, M. Makela, Impacts of structuring the electronic health record: a systematic review protocol and results of previous reviews, Int. J. Med. Inform. 83 (2014) 159–169. [22] P. Carayon, T.B. Wetterneck, B. Alyousef, R.L. Brown, R.S. Cartmill, K. McGuire, P.L.T. Hoonakker, J. Slagle, K.S. Van Roy, J.M. Walker, M.B. Weinger, A. Xie, K.E. Wood, Impact of electronic health record technology on the work and workflow of physicians in the intensive care unit, Int. J. Med. Inform. 84 (2015) 578–594. [23] B.W. Pickering, Y. Dong, A. Ahmed, J. Giri, O. Kilickaya, A. Gupta, O. Gajic, V. Herasevich, The implementation of clinician designed, human-centered electronic medical record viewer in the intensive care unit: a pilot step-wedge cluster randomized trial, Int. J. Med. Inform. 84 (2015) 299–307.

Assessment of hospital processes using a process mining technique: Outpatient process analysis at a tertiary hospital.

Many hospitals are increasing their efforts to improve processes because processes play an important role in enhancing work efficiency and reducing co...
3MB Sizes 0 Downloads 9 Views