Monday, 14 December 2015

Week 5 [14-20.12.2015] Process Mining Methodology for Health Process Tracking Using Real-Time Indoor Location Systems



Hi all,
I would like to present an article, which is part of my research activities. It is an area of process analysis and its performance optimization. Process mining gives us a possibility to take a look on our process in action (in production environment).

Process mining [27] is a paradigm that comes from the business process management research field,  which  allows  the  discovery  and  graphical  representation  of  human-understandable  models that  represent  the  real  execution  of  a  process.   Process  mining  discovery  algorithms  use  logs  for creating workflows, which represent the processes’ possible paths and their associated statistics in a  graphical  way.

It is based on logs generated during process activities. This method shows process in real view.

Process discovery is a process mining technique that is based on analyzing existing logs in order to represent the model of the process.

Hospital surgery case will show you how powerful and usable tool a process mining is.
Based on data collected during research we can discover new solutions of existing issues. It presents real process flow in our organisation. This methodology gives us tools to understand and improve it.

My questions are:

What do you think about process mining as a method of process understanding?

What are disadvantages of this approach?

What do you do to improve process? What is the best way to analyze flow and improve it?

28 comments:


  1. I think that the key for process mining is to monitor real life situations, not process model designed by software engineers. For this kind of monitoring logs should be very detailed to properly identify process loops and branches - treatment process varies between patients even if disease is the same.
    I think that process mining (specially in medical treatment) should not be treated as a method for understanding process. I really hope that hospital staff knows what they're supposed to do. I consider it a good way to improve the process iteratively - analyze, improve, get feedback, repeat.
    I don't have experience in modelling and optimizing medical processes, but I did design some in telco sector. To improve a process one has to identify business needs, define what data need to be processed and via which endpoints, find bottlenecks (most common is human manual tasks or long asynchronous batch tasks) and eliminate or at least reduce them.

    ReplyDelete
    Replies
    1. Thank you for your opinion.

      I agree. Flow monitoring is a target of process mining (my goal is real time process monitoring). This approach gives you a possibility to compare process model with reality described by logs. Data quality (logs) is one of more important attributes in this method. Process mining is dependent upon inbound data.

      In this point I am a little bit confused, because I partially agree with you ... but to ensure proper level of service quality we develop procedures. Medical staff knows what they have to do, but we are talking about various situations. We want to minimize staff effort and human errors risk. When I was writing about process understanding I thought about taking a look an process in real life. We have to deal with situations, which are interfering our process.

      I am agree, and I want to highlight point identify business needs. Companies sometimes create a processes without any business purpose.

      Delete
  2. Process mining can improve our understanding about organisation. It can help you in optimising existing processes in your organisation. I think the key benefit of such an approach is that, that almost everybody can understand it and it can eliminate bottleneck and show how you can improve many things. Especially business, if they see a picture they feels like home. Main disadvantages (i think that it touches Poland) of this approach is that, it's stays only on paper and that is sad.

    ReplyDelete
    Replies
    1. Hi. It stays on paper - it is truly sad if our work lands on paper and case is closed. If your process has bottlenecks then you should improve it, otherwise you are losing money. We should answer on the question: does process have a sense? Process mining should give us a picture - how our process looks like or how it changes itself?

      Delete
  3. As a teacher I prefer to visualize activities than pure command lines codes to present some results.
    I think every method to simplify is acceptable. Of course provided this methods do not distort final results. I know that a deep analysis of logs, testing variety of situations, constant tracing In real life, collecting data, iterative tests to optimize the results can give us the whole view of the flow of the process. But all this sophisticated issues can be shown to the audience more easily with process mining and for me this is very useful.
    One of the disadvantages is that it is really hard to compare single cases of the treated people. But engineers should of course keep on trying to improve the system no matter how long it can take because it can be very useful .

    ReplyDelete
    Replies
    1. This comment has been removed by the author.

      Delete
    2. Thank you. As I wrote it is very important to take care of input data quality. It bases on input and incorrect logs can distort process visualization. We need valuable input in this method.

      Another disadvantage is that you have to collect input before you start any activities.

      Delete
  4. What do you think about process mining as a method of process understanding?

    First of all, I need to make a disclaimer (or spoiler to your research activity?) that there are pretty advanced business process discovery solutions already available (in most of the cases, successfully deployed to IT environments in the biggest companies).

    It is definitely one of means to understand the process and to be able to identify any bottlenecks (as pointed in the the article).

    Disadvantages? How to improve?

    This kind of approach usually adds much of processing overhead (collecting all the data, not only the relevant and significant data), which immediately impacts the overall performance of the process. This drawback is notoriously observed in IT environments where such capability is deployed. Because of this, it is advisable to use it in non-production environments to verify the process and identify any inefficiencies.
    To me, it is a total misunderstanding to deploy this approach in health services.
    It is even more bizarre trying to speed up the surgeon during any procedure just because ‘it takes too long’- where is the individual approach to each patient? Everyone is different. Statistical approach may be ok to evaluate the cost of procedure, not to the duration of the procedure and not to speed a surgeon up.

    ReplyDelete
    Replies
    1. Regarding to disclaimer, I know about different methods of process discovery, but I shown you one of most interesting for me. Process understanding in this case was: take a picture of implemented solution. It is hard to find disadvantages in a process designed by yourself. You can catch interesting insights and observations only when someone tests your process or after implementation in production environment. Your process can be smooth and lean, but process mining gives you another input to continuous improvement.

      Publication describes extreme use case. In my opinion, authors were looking for difficult use case with "spaghetti syndrome". Health care environment is familiar for everyone. Process mining gives you additional output. How to use this data depends on you.

      Delete
  5. Well, generally speaking I don’t know what to say. I agree with the author that, this kind of process may bring us many advantages, for example in medical science, because medicine is developing very dynamically. However, we can’t say for sure that this process is ok, because in next two or three years it may be out of date.

    ReplyDelete
    Replies
    1. This solution can help us with understanding some events. Why it happened? Where is the problem?

      Delete
  6. What do you think about process mining as a method of process understanding?
    In my opinion, if you have enough logs from the different patients you actually improve certain procedures. However, it's hard to find patterns, every single case is a bit different and so on. Everything depends on algorithms and so on.
    What are disadvantages of this approach?
    There is a high chance that you actually won't get anything new. Wasting a lot of resources and so on.
    What do you do to improve process? What is the best way to analyze flow and improve it?
    I don't have enough knowledge to actually answer to this question. It's hard to code process mining to get anything from it. We can try to improve algorithms, train it on different cases and so on.

    ReplyDelete
    Replies
    1. You can not exclude lost time issue. Process mining needs time and space - we have to generate logs (time) and store them (space). That is why I am repeating - the most difficult is valuable data collection.

      Sometimes I am trying to draw a process. When I am implementing solution based on paper draw - I have to make a lot of changes during this process to fit it to reality.

      Delete
  7. Hi,

    Thanks for sharing and for your thoughts on the article. I think that anything that simplifies understanding of any problem (be it processes here) is a good thing. Since it involves graphical representations its probably more understandable for people. One of the best examples i.e. is Einstein's famous e=mc^2 (something of great complexity brought into a simple mathematical formulae).

    Sometimes oversimplifying things leads to worse results or loss of accuracy on certain aspects. I am afraid that since the article focuses on the medical use cases it would be risky to use.

    ReplyDelete
    Replies
    1. Hi, I agree with oversimplifying sentence but process mining should protect us before that. It shows process in production environment. We have a chance to compare our vision to reality.

      Delete
  8. What do you think about process mining as a method of process understanding?

    This is good idea, you usually analyse logs when something bad happens, we don't we use it very often i.e. for conclusions for the to improvements. There should be such a practice..

    What are disadvantages of this approach?
    Time and resource consumption.

    What do you do to improve process? What is the best way to analyze flow and improve it?

    I agree that debugging and logs which are form of debugging are most precise, but you can also watch the results, outputs or metadata and conclude what is happening.

    ReplyDelete
  9. What do you think about process mining as a method of process understanding?

    This is good idea, you usually analyse logs when something bad happens, we don't we use it very often i.e. for conclusions for the to improvements. There should be such a practice..

    What are disadvantages of this approach?
    Time and resource consumption.

    What do you do to improve process? What is the best way to analyze flow and improve it?

    I agree that debugging and logs which are form of debugging are most precise, but you can also watch the results, outputs or metadata and conclude what is happening.

    ReplyDelete

  10. What do you think about process mining as a method of process understanding?
    I think that this is a very good way to understand the basics of a process. Why basics ? Because we are humans, we often need to feel "the thing", the small detail to perfectly understand something. Medicine is a very good example, every surgeon know the alghoritm, he knows what and how he should perform the operation. But the best of them can feel it, their actions are not only based on knowlage, but on something inside them. I think same thing is visible even in programming. We all know the basics alghoritms, patterns etc, we know when to use it but sometimes we just know that something recognized as the best just isn't in our case and we should do it in different way.

    What are disadvantages of this approach?
    I think that we can miss some important details in this approach. Don't get me wrong, I don't think that this approach is bad BUT it can't be the only one.

    ReplyDelete
    Replies
    1. Exactly! This approach must be one of tools, which help us improve our process. Advantage of this tool is that it show us process in production environment. We can miss something and it is important to remember about that like in any other approach.

      Delete

  11. What do you think about process mining as a method of process understanding?
    I think that it is a good idea in order to understand the process. There is a need for a sufficient number of logs which will include different states and cases. They definitely need to be representative.
    What are disadvantages of this approach?
    One of the disadvantages is that we will not get satisfying results despite the fact that lots of resources are spent.
    What do you do to improve process? What is the best way to analyze flow and improve it?
    Maybe improvement of the algorithms? Inclusion of some kind of predictions?

    ReplyDelete
    Replies
    1. Thank you for answer. I agree - costs can overwhelm the profit. It is a risk.

      Delete
  12. The article is very interesting. I believe that the visuals are helpful in understanding the problem, such as selecting only the relevant prerequisites and rejection not related to the study subject, the problem are the methods of selection, too weak not give a clear picture and still have it look too strong deprive us of the knowledge of the relevant premises. I agree or medical use may be the least risky.

    ReplyDelete
    Replies
    1. I think it is good point. We have to find borders where our process ends and another one starts. We should try to analyze process flow like it is. If you reject something then you can receive incomplete picture of your process or results might be murky. Regarding to medical case - me too.

      Delete
  13. What do you think about process mining as a method of process understanding?

    I think that it can simplify and organize some processes, which can help to understand them better. It also uses a great idea that is visualisation. Most people better understand visualised actions.

    What are disadvantages of this approach?

    I think sometimes the model of the process might not be accurate to specific case, or just strictly following the process may cause too much generalizations, that are not always good in a fields such as medicine.

    What do you do to improve process? What is the best way to analyze flow and improve it?

    I think that some kind of control on the process is needed, but I don't have enought knowledge to suggest particular solutions.

    ReplyDelete
  14. > What do you think about process mining as a method of process understanding?

    Sounds right. Guess most applications today use this approach.

    > What are disadvantages of this approach?

    Can't tell. Can you elaborate on this more ?

    ReplyDelete
  15. What do you think about process mining as a method of process understanding?
    Using the graphs it's easy to show in fast way how process are connected to each other and how they are correlated. That's why that is good method of process understanding. Graphical process representation allows us to discover new sense of human-understandable models.

    What are disadvantages of this approach?
    The graph can only represent an action, not intuition or motivation. This mechanical approach can disqualified human intuition.

    What do you do to improve process? What is the best way to analyze flow and improve it?
    To analyze flow you must understand the connection between the action. The good understanding of context is necessary in my opinion.

    ReplyDelete
  16. I like process mining as a method, how ever I find it applicable only to a limited number of processes - mainly the explicit organizational processes that are covered by a vast amount of data. It is extremely hard to mine social processes, because people are afraid to disclose their personal information. This is a big pity of course, doctoral researches would be so much more interesting in such a case.

    ReplyDelete