In the article on Zero timestamps we have seen how timestamp problems can lead to faulty case durations. But faulty timestamps do not only influence the case durations. They also impact the variants and the process maps themselves, because the order of the activities is derived based on the timestamps.
For example, take a look at the following data set with just one faulty timestamp. There is one case with a 1970 timestamp (see screenshot below – click on the image to see a larger version). As a result, the ‘Create case’ activity is positioned before the ‘Import forms’ activity.
If we look at the process map, then you see that in all other 456 cases the process flows the other way. Clearly, the reverse sequence is caused by the 1970 timestamp.
And if we look at the average waiting times in the process map, then this one faulty timestamp creates further problems and shows a huge delay of 43 years.
As you can see, data quality problems due to timestamp issues can distort your process mining analysis in many different places. Therefore, it is important to carefully assess the process map and the variants, if possible together with a domain expert, to spot any suspicious orderings of activities.
If you have found a problem with the timestamps, then there can be different reasons for why this is happening. Zero timestamps are just one possible reason. Here is the next one: Wrong timestamp configuration during import.
Wrong Timestamp Pattern Configuration
When you import a CSV or Excel file into Disco, the timestamp pattern is normally detected automatically. You don’t have to do anything. If it is not automatically detected, Disco lets you specify how the timestamp pattern should be interpreted rather than forcing you to convert your source data into a fixed timestamp format. And you can even work with different timestamp patterns in your data set.
However, if you have found that activities show up in the wrong order, or if you find that your process map looks weird and does not really show the expected process, then it is worth verifying that the timestamps are correctly configured during import.
You can do that by going back to the import screen: Either click on the ‘Reload’ button from the project view or import your data again. Then, select the timestamp column and press the ‘Pattern…’ button in the top-right corner. You will see a few original timestamps as they are in your file (on the left side) and a preview of how Disco interprets them (in green, on the right side).
Check in the green column whether the timestamps are interpreted correctly. Pay attention to the lower and upper case of the letters in the pattern, because it makes a difference. For example, the lower case ‘m’ stands for minutes while the upper case ‘M’ stands for months.
How to fix: If you find that the preview does not pick up the timestamps correctly, configure the correct pattern for your timestamp column in the import screen. You can empty the ‘Pattern’ field and start typing the pattern that matches the timestamps in your data set (use the legend on the right, and for more advanced patterns see the Java date pattern reference for the precise notation and further examples). The green preview will be updated while you type, so that you can check whether the timestamps are now interpreted correctly. Then, press the ‘Use Pattern’ button
Wrong Timestamp Column Configuration
Another timestamp problem that can result from mistakes during the import step is that you may have accidentally configured some columns as a timestamp that are not actually a timestamp column in the sense of a process mining timestamp (but, for example, indicate the birthday of the customer).
In the customer service refund example below, the purchase date in the data has the form of a timestamp. However, this is a date that does not change over time and should actually be treated as an attribute. You can see that both the ‘Complete Timestamp’ as well as the ‘Purchase Date’ column have the title clock symbol in the header, which indicates that currently both are configured as a timestamp.
If columns are wrongly configured as a timestamp, Disco will use them to calculate the duration of the activity. As a consequence, activities can show up in parallel although the are in reality not happening at the same time.
How to fix: Make sure that only the right columns are configured as a timestamp: For each column, the current configuration is shown in the header. Look through all your columns and make sure only your actual timstamp columns are showing the little clock symbol that indicates the timestamp configuration. Then, press again the ‘Start import’ button.
For example, in the customer service data set, we would change the configuration of the ‘Purchase Date’ column to a normal attribute as shown below.
Data Scientists spend a large part of their day on exploratory analysis. In the 2015 Data Science Salary Survey, 46% of respondents said that they use one to three hours per day on the summarizing, visualization, and understanding of data, even more than on data cleansing and data preparation.
Process mining is focused on the analysis of processes, and it is an excellent tool in particular for the exploratory analysis of process-related data. If your data science project concerns business or IT processes, then you need to explore these processes and understand them first before you can train machine learning algorithms or run statistical analyses in any meaningful way.
With process mining you can get a process view of the data. The specific process view results from the following three parameters:
Case ID: The selected case ID determines the scope of the process and connects the individual steps of a process instance from the beginning to the end (for example, a customer number, order number or patient ID)
Activity: The activity name determines the steps that are shown in the process view (such as “order received” or “X-ray examination completed”).
Timestamp: One or more timestamps per step (for example for the beginning and the end of an X-ray examination) are used to calculate the process sequence and to derive parallel process steps.
When you analyze a data set with process mining, then you determine at the beginning of the analysis, which columns in the data correspond to the Case ID, activity name, and timestamps. You can set these parameters in the configuration when importing the data into the process mining tool.
When importing a CSV file into the process mining software Disco, you can specify for each column in your data set how it should be interpreted.1
In the following example of a purchasing process, the Case ID column (the purchase order number) is configured as Case ID, the start and complete timestamps as Timestamp, and the Activity column as Activity. As a result, the process mining software automatically produces a graphical representation of the actual purchasing process based on historical data. The process can now be further analyzed based on facts.
Usually, the first process view– and the import configuration derived from it–follows from the process understanding and task at hand.
However, many process mining newcomers are not yet aware of the fact that a major strength of process mining, as an exploratory analysis tool, is that you can rapidly and flexibly take different perspectives on your process. The above parameters function as a lens with which you can adjust process views from different angles.
Here are three examples:
1. Focus on Another Activity
For the above purchasing process, we can change the focus on the organizational process flow by setting the Role column (the function or department of the employee) as Activity.
This way, the same process (and even the same data set) can now be analyzed from an organizational perspective. Ping-pong behavior and increased transfer times when passing on operations between organizational units can be made visible and addressed.
2. Combined Activity
Instead of changing the focus, you can also combine different dimensions in order to get a more detailed picture of the process.
If you look at the following call center process, you would probably first set the column “Operation” as activity name. As a result, the process mining tool derives a process map with six different process steps, which represent the accepting of incoming customer calls (“Inbound Call”), the handling of emails, and internal activities (“Handle Case”).
Now, imagine that you would like to analyze the process in more detail. You would like to see how many first-level support calls are passed on to the specialists in the back office of the call center. This information is actually present in the data. The attribute “Agent Position” indicates whether the activity was handled in the first-level support (marked as FL) or in the back office (marked as BL).
To include the “Agent Position” in the activity view, you can set both the column “Operation” and the column “Agent Position” as activity name during the data import step. The contents of the two columns are now grouped together (concatenated).
As a result, we get a more detailed view of the process. We see for example that calls accepted at the first-level support were transferred 152 times to the back office specialists for further processing. Furthermore, no email-related activities took place in the back office.
3. Alternative Case Focus
Finally, we could question whether the service request ID of the CRM system, which was selected as the case ID, provides the desired process view for the call center process. After all, there is also a customer ID column and there are at least three different service requests noted for “Customer 3” (Case 3, Case 12 and Case 14).
What if these three requests are related and the call center agents just have not bothered to find the existing case in the system and re-open it? The result would be a reduced customer satisfaction because “Customer 3” has had to repeatedly explain the problem with every call.
The result would also be an embellished “First Call Resolution Rate.” The “First Call Resolution Rate” is a typical performance metric for call centers, which measures the number of times a customer problem could be solved with the first call.
That is exactly what happened in the customer service process of an Internet company. In a process mining project, initially the customer contact process (via telephone, Internet, e-mail or chat) was analyzed with the Service ID column chosen as the case ID. This view produced an impressive “First Contact Resolution Rate” of 98%. Of 21,304 incoming calls, apparently only 540 were repeat calls.
Then the analysts noticed that all service requests were closed fairly quickly and almost never re-opened again. To analyze the process from the customer’s perspective, the Customer ID column was chosen as a case ID. This way, all calls of a specific customer in the analyzed time period were summarized into one process instance and repeating calls became visible.
The “First Contact Resolution Rate” in reality amounted to only 82%. Only 17,065 cases were actually started by an incoming call. More than 3,000 were repeat calls, but were counted as new service requests in the system (and on the performance report!).
Process mining allows you to get a process perspective on your data. Moreover, it is worthwhile to consider different views on the process. Look out for other activity perspectives, possible combinations of fields, and new perspectives on what constitutes the case in the process.
You can take different views to answer different questions. Often, multiple views are necessary to obtain an overall picture of the process.
Do you want to explore the perspective changes presented in this article yourself in more detail? You can download the used example files here and analyze them directly with the freely available demo version of our process mining software Disco.
Note: For the open-source software ProM (http://www.promtools.org/) you often use XML formats such as XES or MXML, which contain this configuration. ↩
Have you completed a successful process mining project in the past months that you are really proud of? A project that went so well, or produced such amazing results, that you cannot stop telling anyone around you about it? You know, the one that propelled process mining to a whole new level in your organization? We are pretty sure that a lot of you are thinking of your favorite project right now, and that you can’t wait to share it.
We want to help you showcase your best work and share it with the process mining community. This is why we are introducing the Process Miner of the Year awards. The best submission will receive this award at this year’s Process Mining Camp, on 10 June in Eindhoven.
What we are looking for
We want to highlight process mining initiatives that are inspiring, captivating, and interesting. Projects that demonstrate the power of process mining, and the transformative impact it can have on the way organizations go about their work and get things done.
There are a lot of ways in which a process mining project can tell an inspiring story. To name just a few:
Process mining has transformed your organization, and the way you work, in an essential way.
There has been a huge impact with a big ROI, for example through cost savings or efficiency gains.
You found an unexpected way to apply process mining, for example in a domain that nobody approached before you.
You were faced with enormous challenges in your project, but you found creative ways to overcome them.
You developed a new methodology to make process mining work in your organization, or you successfully integrated process mining into your existing way of working.
Of course, maybe your favorite project is inspiring and amazing in ways that can’t be captured by the above examples. That’s perfectly fine! If you are convinced that you have done some great work, don’t hesitate: Write it up, and submit it, and take your chance to be the Process Miner of the Year 2016!
How to enter the contest
You can either send us an existing write-up of your project, or you can write about your project from scratch. It is probably better to start from a white page, since we are not looking for a white paper, but rather an inspiring story, in your own words.
In any case, you should download this Word document, which contains some more information on how to get started. You can use it either as a guide, or as a template for writing down your story.
When you are finished, send your submission to email@example.com later than 30 April 2016.
We can’t wait to read about your amazing projects!
Eindhoven can be reached conveniently through a direct train connection from Amsterdam’s Schiphol airport. Mark the day in your calendar, and start making plans for your trip to the birthplace of process mining! You should also sign up for the camp mailing list to receive updates about this year’s camp, and to be the first to know when ticket sales open.
Share your story
We are currently busy putting together the program of this year’s camp, and we have already secured a number of speakers with great stories to tell. A lot of you have been doing great work lately, and some of the best process mining stories that we are aware of have already made their way onto this year’s camp program.
Before we finalize the program, we wanted to give all of you the opportunity to help us shape this year’s camp. Would you like to point us to interesting stories or topics that may not be on our radar yet? Do you have a great process mining story you would like to share at this year’s camp, or do you know someone who might? Send Christian an email at firstname.lastname@example.org and let us know!
See you on 10 June!
Process mining camp is our annual practitioner conference for process miners all over the world. It is not only a place to hear interesting and inspiring talks from other process miners, but also the annual family meeting of the global process mining community. Over the past four years, process mining enthusiasts from more than 17 different countries (including Australia, Korea, Brazil, South Africa and the United States) have come together to exchange their experiences and meet their peers.
In 2012, more than 70 smart and driven people joined us for the first Process Mining Camp. In 2013, we moved Process Mining Camp to the Zwarte Doos and added workshops, and we had a great day with more than 100 process mining enthusiasts from all over the world. In 2014, camp tickets sold out very quickly, and process mining enthusiasts from more than 16 countries came for a varied program including workshops, keynotes, and a panel discussion. In 2015, we moved to the auditorium to make more room, and 173 people from 17 different countries joined us at camp.
This year will be the greatest camp ever, and we cannot wait to meet you in Eindhoven!
This week, we are moving to the timestamp problems. Timestamps are really the Achilles heel of data quality in process mining. Everything is based on the timestamps: Not just the performance measurements but also the process flows and variant sequences themselves. So, over the next weeks we will look at the most typical timestamp-related issues.
Zero timestamps (or future timestamps)
One data problem that you will most certainly encounter at some point in time are so-called zero timestamps, or other kind of default timestamps that are given by the system. Often, zero timestamps were initially set as an empty value by the programmer of the information system. They can either be a mistake or indicate that the real timestamp has not yet been provided (for example, because an expected process step has not happened yet). Another reason can be typos in manually entered data.
These Zero timestamps typically take the form of 1 January 1900, the Unix epoch timestamp 1 January 1970, or some future timestamp (like 2100).
To find out whether you have Zero timestamps in your data, you can best go to the Overview statistics and take a look at the earliest and the latest timestamps in the data set. For example, in the screenshot below we can see that there is at least one 1900 timestamp in the imported data (click on the screenshot to see a larger version).
You should know what timeframe you are expecting for your data set and then verify that the earliest and latest timestamp confirm the expected time period. Be aware that if you do not address a problem like the 1900 timestamp in the picture above, you may end up with case durations of more than 100 years!
How to fix: You can remove Zero timestamps using the Timeframe filter in Disco (see instructions below).
You may also want to communicate your findings back to the system administrator to find out how these Zero timestamps can be avoided in the future.
To understand the impact of the Zero timestamps, you first need to investigate in more detail what is going on.
You want to find out whether just a few cases are affected by the Zero timestamps, or whether this is a wide-spread problem. For example, if Zero timestamps are recorded in the system for all activities that have not happened yet, you will see them in all open cases.
To investigate the cases that have Zero timestamps, add a Timeframe filter and use the ‘Intersecting timeframe’ mode while focusing on the problematic time period. This will keep all those cases that contain at least one Zero timestamp. Then use the ‘Copy and filter’ button to create a new data set focusing on the Zero timestamp cases (see screenshot below).
As a result, you will see just the cases that have Zero timestamps in them. You can see how many there are. Furthermore, you can inspect a few example cases to see whether the problem is always in the same place or whether multiple activities are affected. In our example, just two cases contain Zero timestamps (see below).
Now, let’s move on to fix the Zero timestamp problem in the data set.
Then: Remove cases or Zero timestamps only
Depending on whether Zero timestamps are a wide-spread problem or not you can take two different actions:
If only a few cases are affected, you can best remove these cases altogether. This way, they will not disturb your analysis. At the same time you will not be left with partial cases that miss some activities because of data issues.
If many cases are affected, like in the situation that Zero timestamps were recorded for activities that have not happened yet, you can better remove just the events that have Zero timestamps and keep the rest of these cases for your analysis.
In our example, just two cases are affected and we will remove these cases altogether. To do this, add a Timeframe filter and choose the ‘Contained in timeframe’ option while focusing your selection on the expected timeframe. This will remove all cases that have any events outside the chosen timeframe (see screenshot below).
If you just want to remove the activities that have Zero timestamps, choose the ‘Trim to timeframe’ option instead. This will “cut off” all events outside of the chosen timeframe and keep the rest of these cases in your data (see below)
Note that if your Zero timestamps indicate that certain activities have not happened yet, it would be better to keep the timestamp cells in the source data empty, rather than filling in a 1900 or 1970 timestamp value (see example below).
Events with empty timestamps will not be imported in Disco, because they cannot be placed in the sequence of activities for the case. So, keeping the timestamp cell empty for activities that have not occurred yet will save you this extra clean-up step in the future.
Finally: Make a clean copy
Once you have cleaned up the Zero timestamps from your data, you can best make a new copy using the ‘Apply filters permanently’ option to get a fresh start (see screenshot below). The result will be a new (cleaned) data set, which can now serve as the starting point for your analysis.
That’s it! You have successfully removed your Zero timestamps and any new filters that you add from now an will be based on your cleaned data.
Even if your data imported without any errors, there may still be problems with the data. For example, one typical problem is missing data. Keep reading to learn more about some of the most common types of missing data in process mining.
Gaps in the timeline
Check the timeline in the ‘Events over time’ statistics to see whether there are any unusual gaps in the amount of data over your log timeframe.
The picture above shows an example, where I had concatenated three separate files into one file before importing it in Disco. Clearly, something went wrong and apparently the whole data from the second file is missing.
How to fix:
If you made a mistake in the data pre-processing step, you can go back and make sure you include all the data there.
If you have received the data from someone else, you need to go back to that person and ask them to fix it.
If you have no way of obtaining new data, it is best to focus on an uninterrupted part of the data set (in the example above, that would be just the first or just the third part of the data). You can do that using the Timeframe filter in Disco.
Unexpected amount of data
You should have an idea about (roughly) how many rows or cases of data you are importing. Take a look at the overview statistics to see whether they match up.
For example, the picture below shows a screenshot of the overview statistics of the BPI Challenge 2013 data set. Can you see anything wrong with it?
In fact, the total number of events is suspiciously close to the old Excel limit of 65,000 rows. And this is what happened: In one of the data preparation steps the data (which had several hundred thousand rows) was opened with an old Excel version and saved again.
Of course, this is a bit more subtle than an obvious gap in the timeline but missing data can have all kinds of reasons. For some systems or databases, a large data extract is aborted half-way without anyone noticing. That’s why it is a very good idea to have a sense of how much data you are expecting before you start with the import (ask the person that gives you the data how they structured their query).
How to fix:
If you miss data, you must find out whether you lost it in a data pre-processing step or in the data extraction phase.
If you have received the data from someone else, you need to go back to that person and ask them to fix it.
If you have no way of obtaining new data, try to get a good overview about which part of the data you got. Is it random? Was the data sorted and you got the first X rows? How does this impact your analysis possibilities? Some of the BPI Challenge submissions noticed that something was strange and analyzed the data pattern to better understand what was missing.
Unexpected distribution or empty attribute values
Similarly, you should have an idea of the kind of attributes that you expect in your data. Did you request the data for all call center service requests for the Netherlands, Germany, and France from one month, but the volumes suggest that the data you got is mostly from the Netherlands?
Another example to watch out for are empty values in your attributes. For example, the resource attribute statistics in the screenshot below show that 23% of the steps have no resource attached at all.
Empty values can also be normal. Talk to a process domain expert and someone who knows the information system to understand the meaning of the missing values in your situation.
How to fix:
If you have unexpected distributions, this could be a hint that you are missing data and you should go back to the pre-processing and extraction steps to find out why.
If you have empty attribute values, often these values are really missing and were never recorded in the first place. Make sure you understand how these missing (or unexpectedly distributed) attribute values impact your analysis possibilities. You may come to the conclusion that you cannot use a particular attribute for your analysis because of these quality problems.
It is not uncommon to discover data quality issues in your original data source during the process mining analysis, because nobody may have looked at that data the way you do. By showing the potential benefits of analyzing the data, you are creating an incentive for improving the data quality (and, therefore, increasing the analysis possibilities) over time.
Cases with unexpected number of steps
As a next check, you should look out for cases with a very high number of steps (see below). In the shown example, the callcenter data from the Disco demo logs was imported with the Customer ID configured as the case ID.
What you find is that while a total of 3231 customer cases had up to a maximum of 30 steps, there is this one case, (Customer 3) that had a total of 583 steps in total over a timeframe of two months. That cannot be quite right, can it?
To investigate this further, you can right-click the case ID in the table and select the “Show case details” option (see below).
This will bring up the Cases view with that particular case shown (see below). It turns out that there were a lot of short inbound calls coming in rapid intervals. The consultation with a domain expert confirms that this is not a real customer, but some kind of default customer ID that is assigned by the Siebel CRM system if no customer was created or associated by the callcenter agent (for example, because it was not necessary, or because the customer hung up before the agent could capture their contact information).
Although in this data set there is technically a case ID associated, this is really an example of missing data. The real cases (the actual customers that called) are not captured. This will have an impact on your analysis. For example, analyzing the average number of steps per customer with this dummy customer in it will give you wrong results. You will encounter similar problems if the case ID field is empty for some of your events (they will all be grouped into one case with the ID “empty”).
How to fix:
You can simply remove the cases with such a large number of steps in Disco (see below). Make sure you keep track of how many events you are removing from the data and how representative your remaining dataset still is after doing that.
To remove the “Customer 3” case from the callcenter data above, you can right-click the case in the overview statistics and select the Filter for case ‘Customer 3’ option.1
In the filter, you can then invert the selection (see the little Yin Yang button in the upper right corner) to exclude Customer 3. To create a new reference point for your cleaned data, you can tick the ‘Apply filters permanently’ option after pressing the ‘Copy and filter’ button:
The result will be a new log with the very long case removed and the filter permanently applied (you have a clean start).
Alternatively, you could also use a Performance filter with the ‘Number of events’ metric to remove cases that are overly long. ↩
However, most of that data was not originally collected for process mining purposes. And especially data that has been manually entered can always contain errors. How do you make sure that errors in the data will not jeopardize your analysis results?
Data quality is an important topic for any data analysis technique: If you base your analysis results on data, then you have to make sure that the data is sound and correct. Otherwise, your results will be wrong! If you show your analysis results to a business user and they turn out to be incorrect due to some data problems, then you can lose their trust into process mining forever.
For example, the delimiting character (“,” “;” “I” etc.) cannot be used in the content of a field without proper escaping. If you look at the example snippet below then you can see that the “,” delimiter has been used to separate the columns. However, in the last row the activity name itself contains a comma:
Another problem might be that your file has less columns in some rows compared to others (see example below).
Other typical problems are invalid characters, quotes that open but do not close, and there are many more.
If Disco encounters a formatting problem, it gives you the following error message with the sad triangle and also tries to indicate in which line the problem occurs (see below).
In most cases, Disco will still import your data and you can take a first look at it, but make sure to go back and investigate the problem before you continue with any serious analysis.
We recommend to open the file in a text editor and look around the indicated line number (a bit before and afterwards, too) to see whether you can identify the root cause.
How to fix: Occasionally, the formatting problems have no impact on your data (for example, an extra comma at the end of some of the lines in your file). Or the number of lines impacted are so few that you choose to ignore it. But in most cases you do need to fix it.
Sometimes, it is enough to use “Find and Replace” in Excel to replace a delimiting character from the content of your cells and export a new, cleaned CSV that you then import.
However, in most cases it will be the easiest to point out the problem that you found to the person who extracted the data for you and ask them to give you a new file that avoids the problem.
Figure 1: Process map of ED #1 – Cumulative time (click to enlarge)
This is a guest article by Matthew H. Loxton, a senior analyst for healthcare at WBB. You can request an extended version of this case study with detailed recommendations from Matthew directly. An overview paper about process mining for quality improvement in healthcare environments can be found here.
Historically, Quality Improvement (QI) projects have used a combination of received workflow and observational studies to derive the as-is process model. The process model is used to target interventions to reduce waste and risk, and to improve processes that lead to gains in the target performance indicators. Process mining enables QI efforts to more rapidly discover areas for improvement, and to apply a perspective that was historically not available to QI teams.
Since process mining is algorithmic and uses electronic health record (EHR) data, it can be deployed at scale, and can be used to find process improvement opportunities across an entire healthcare system without undue resource requirements or disruption to clinical operations.
The case studies involved two of the busiest Emergency Departments (ED) in the U.S., and give the reader a picture of how process mining can be used as part of a long-term process improvement regime.
The WBB team used the process mining software Disco to mine ED and EHR data for two EDs for the period 06/04/2015 to 08/02/2015. Data included 2,628 cases for ED #1 and 2,447 cases for ED #2. Each case represents a unique patient transitioning through the ED to arrive at a disposition.
The WBB team also conducted interviews and facilitated sessions with various flow management application stakeholders to identify benefits and challenges, and to provide recommendations for future improvements. Interview participants related to EHR included ED directors, ED physicians, chiefs of staff, chiefs of medicine, and members of the EHR program office.
The discovered process models showed a high degree of variation (see Figure 1 at the top of this article), and the team used filters to manage the process model complexity to a point where the models were useful in identifying and contrasting paths and their performance. The team obtained concurrence from the point of contact at each of the two facilities that the process model was a fair depiction of how their ED operated.
In addition to producing visual depictions of the underlying workflow and performance, a number of “special cases” were observed in which patient travel through the process model were unexpected and revealed opportunities for improved use of EHR, data governance, and monitoring of unusual patient transactions. For example, some processes are incomplete and do not follow the “should-be” process by omitting the Discharged status.
Among others, the team found opportunities for improvement related to data governance risks, functionality of EHR and inconsistent use of EHR status and disposition in the following areas:
1. Cases of unedited EHR labels existed in the data.
One benefit of process mining is that unknown or unexpected transitions can be identified. The activity items in the data are a combination of national terms and locally configured terms. Locally configured terms are used to describe a location or status that is required to suit local needs such as specialty wards or services unique to the local patient population or facility specialties.
When a locally configured term is created, the default name is “new#”, where # is the next available sequence number. The name is manually edited and renamed to be meaningful to the facility (e.g. “admit to psychiatry”). The process model revealed two transition states in the live data, “new2”, and “new3”. Since “new2” and “new3” have 9 and 28 cases respectively, it proved worthwhile to examine the cases.
The event labels stemmed from unfinished additions of new labels that had been inadvertently left in the EHR data. The discovery of these labels led to a process improvement exercise in data cleanup, and discussions regarding processes for adding or editing fields.
2. Loops in the process model due to incorrect sequence entry.
Process loops are expected in some process models, and may indicate normal functioning of the process. However, in processes that are expected to be linear and branching, such as many care flows in the ED, a process loop can indicate either clerical or clinical error, or a process issue.
Figure 2: ED #1 Process loops
In this case, the data revealed that the loops were the result of some events being entered in reverse order due to functionality in the EHR (see Figure 2 for an example).
The EHR grid view contains all the editable fields, and a user can select the disposition and status in any order. The choice and availability is not constrained or guided by business rules within EHR. As a result, the elapsed times in reports that use a formula for elapsed time based on the status timestamps may be negative, and skew EHR and productivity reports.
This discovery initiated a discussion on enhancement of the EHR and policies regarding use of the grid view. Furthermore, a review of the current reporting algorithms will be performed to ensure that negative values are not skewing or biasing data.
3. “Pinball Patients” with high event counts.
The distribution curve of events per case is an indicator of one dimension of complexity in a process model. Although the ED-1 distribution shows that most cases have four events, it can also be seen that a small number of variants have far more events per case (see Figure 3).
Figure 3: ED #1 Events per case
To help identify opportunity for process improvement, it is useful to examine cases that have fewer or more events than chance would predict. For ED-1, the team examined cases that had less than two events, and cases that had more than eight events.
Cases with abnormally low or high event counts may reveal clerical errors, or process gaps that do not adequately address some patient situations.
The ED-1 process model showed three variants in which there were only two events (none that had fewer than two):
44 cases in which the patient was entered in error
31 cases in which the patient was sent to a clinic
7 cases in which the patient eloped
Cases in which patients are entered in error should be evaluated for potential training, EHR functionality, or process issues. Patient elopement is also a situation that deserves examination to see if there are delays or process issues resulting in patient dissatisfaction.
In some cases, there were an unexpectedly high number of status changes. The ED-1 process model showed 24 variants in which there were eight or more events, and two in which there were 10 events.
The following graphic shows the process model for a single case in which the patient had 10 events (see Figure 4).
Figure 4: ED #1 “Pinball patient”
Cases with both more than two standard deviations of events per variant above or below the mean merit further scrutiny to understand the causes. These cases were examined by the senior ED physician to determine root causes and any evidence of patient safety risks.
This case study illustrates how process mining can reveal questions and potential risks and issues that might not have been otherwise visible. The program office can examine facility processes and formulate specific and targeted questions without unnecessarily interrupting or burdening the facility staff.
Discretion must be used when evaluating elapsed time between transitions; since short times may be due to administrative bundling of tasks and long times may indicate administration being carried out after the fact. For example, short transition times such as from “Admitted” to “Admitted to ICU”, “Operating Room,” “Admitted to Telemetry,” and “Admitted to Ward,” showed that the events were administrative actions in the EHR, and are not due to patient movements.
Process discovery is a critical component of QI. The ability to compare accurate depictions of what was intended with what is actually being done is a central part of being able to identify variances, and to correctly target and monitor QI interventions. Traditional methods of process discovery have proven very effective, but have significant disadvantages in terms of accuracy, timeliness, and cost. Process mining enables QI practitioners to more rapidly discover as-is process maps, and thereby to identify deviations, delays, and bottlenecks. Rapid discovery of actual workflow enables faster and more targeted interventions that can increase efficiency, reduce risk, and reduce cost.
When people talk about the ‘First Time Right’ principle, they typically refer to the goal of running through a business process without the need to redo certain steps (because they were not right the first time). You also do not want to do unnecessary extra steps (referred to as ‘Waste’ in Lean) that ideally should not be there.
So, when you analyze your process with process mining you often want to focus on these repetitions, the extra steps and other kind of rework, to understand where and why these inefficiencies are happening.
But one of the goals of your process mining analysis might be to find out how many cases follow the ‘First Time Right’ process in the first place. Is 80-90% of the process going through the ‘First Time Right’ process flow? Or is it more like 30%?
In the above video, we show you how you can perform such a ‘First Time Right’ analysis with Disco very quickly.
In a nutshell, the steps are as follows:
1. Prepare your data
If you still have to clean or otherwise prepare your data, do this first. For example, you might want to remove incomplete cases from your data set using the Endpoints filter.
2. Make a permanent copy of your data set
The cleaned data set will be your new reference point. For example, if your data only contains 80% completed cases, then you want these 80% to be “the new 100%” in terms of your ‘First Time Right’ analysis.
To do this, press the ‘Copy’ button in the lower right corner and enable the ‘Apply filters permanently’ option.
3. Remove unwanted steps and paths
You could simply determine and filter the variant that corresponds to the ‘First Time Right’ process, but often there are more than one and the total number of variants can grow very quickly. An easier way is to work yourself towards the ‘First Time Right’ process in a visual way directly from the process map.
You start by clicking on the unwanted steps and paths and use the filter shortcuts from the process map, in an iterative way. Before applying each filter, you invert the configuration so that you do not keep all cases that perform the step (or follow the path) that you clicked on, but precisely the ones that do not.
4. Read off the remaining percentage of cases
When you are finished, you can simply look at the percentage indicator for the cases that remain in the lower left corner. This will be the portion of process instances that follow the ‘First Time Right’ process (out of all completed cases in your data set).
You can of course also look at the number of cases and performance statistics, as well as inspect the remaining variants in the ‘Cases’ tab.
If you have not done this before, try it! Process mining can not only help you to focus on the parts that go wrong but also quickly show you the portion of the process that goes right. Make sure to keep copies of your different analyses, so that you can compare them.
We are happy to announce the immediate release of Disco 1.9.1!
Disco 1.9.1 is a maintenance update with no user-facing changes, so you should feel right at home if you are used to Disco 1.9.0. However, we have improved a number of core components of Disco under the hood, greatly improved the performance, and fixed a number of annoying bugs in this release. As such, we recommend that all users of Disco update to 1.9.1 at their earliest convenience.
Disco will automatically download and install this update the next time you run it, if you are connected to the internet. You can of course also download and install the updated installer packages manually from fluxicon.com/disco.
What is new in this version
Overdrive: Greatly improved performance for repeated mining of the same data set.
Airlift: Support for servers providing multiple data catalogs.
CSV Import: Improved header auto-detection.
CSV Import: Improved accuracy of import settings auto-detection.
Log Import: Fixed a bug where some data files containing illegal characters failed to load properly.
Process Map: Fixed a bug where setting the detail percentages explicitly could fail to work for some setups.
Export: Improved and extended audit report filter summary.
Log Filter: Fixed a bug that could prevent display of the filter view in exceedingly rare circumstances.
Bug Fixes: This update fixes several minor issues and user interface inconsistencies.
We hope that you like this update, and that it makes getting your work done with Disco an even better experience. Thank you for using Disco!