I really enjoyed Vera Tilson’s presentation yesterday on how she used a Wall Street Journal article on the OVA1 test as a case study for her class. I usually attend sessions where we see the results of someone else’s research, but it was a nice change of pace to see a presentation on teaching methods. As a new faculty member, I appreciated it.
Prof. Tilson had some great suggestions about how to turn this article into a case for MBA students. It can be used to get students to understand how to structure a decision analysis problem and how to apply statistical concepts in the real world. She was able to cover a lot of concepts all based on a short WSJ article.
95% of the sessions I attended were focused on research, but I’m glad we can also share teaching methods at this conference.
Today I focused primarily on talks on healthcare operations management. I was really pleased with the attendance at my presentation, interest shown, and questions asked. The sessions offered a number of well-thought out papers that can lead to improving our ability to schedule doctors and procedures. I also found a paper on remote electronic monitoring for home health to be particularly relevant to some of my own research.
The weather was great and I decided to duck out late afternoon for a short workout and dip in the pool. We had dinner at Dunn’s Deli, which I was told is the :second best” deli in Montreal. All in all, an informative, interesting, day with a bit of down time added as well.
I just came from a talk that I believe was OR at its best. Ngai-Hang Leung of MIT presented his work with Jeremie Gallien and a host of others on their work in Zambia to improve the distribution of essential medicines, such as anti-malarial drugs. They have already achieved impressive results with a redesign of the distribution system, e.g, by using cross-docking.
Now they are working to improve the inventory policies used by hundreds of remote clinics. Perhaps surprisingly, the problem has many similarities with the problem faced by fashion retailers, including extreme seasonality of demand (malaria hits during the rainy season when the mosquitoes are active). The team formulated a dynamic min/max reorder point model that is similar to previous models developed by Jeremie and others and implemented by the retailer Zara. In simulation experiments, the model achieves impressive results in avoiding stockouts and saving lives. Here’s hoping that it achieves a similar performance as it is implemented by a consortium of aid organizations.
This is my last blog post. I’m heading down to New York for the INFORMS Revenue Management and Pricing Conference at the end of the week. I hope to see some of you there. Thanks to the other blog posters, and thanks again to the organizers of this excellent conference.
So far today I saw a talk on empirical OM used to look at workload, a talk integrating a network of queues with self-routing behavior, and heard a discussion about someone working in mental health and addiction.
Other quick thoughts:
–chronic disease is becoming the biggest problem in our health system. How can we prevent and manage effectively to reduce long-term effects?
–obesity is a major contributor to costs, diabetes, various comorbidities. What can OR offer this area?
–electronic medical records and their associated information exchanges will become ubiquitous in the next few years. What will the role of personalized medicine be, and how can OR help inform it?
–medical decisions (at many levels) are driven by numerous factors including incentives (of clinics, companies, physicians, patients). What do we need to do to incorporate in our research models?
–the health system in the US has increasing and significant costs. What are the main contributors to this, and thus what are the most important things for us to focus on as researchers?
signing off now to catch a plane.
How many of you have personal health stories, that sometimes drive your interest in health or your research in the area? I have heard anecdotes from my colleagues so I know some of you do.
Today I will make a note about a personal health story. It’s not related to my research, although indirectly it’s related to my experience at the conference. My co-author, Dr. Nicoleta Serban, was scheduled to present our joint work yesterday. However, this year her sister was diagnosed with leukemia. My age, and given a prognosis of a few months without a transplant. They searched for a match and couldn’t find one that was 100%, but decided to go forward with a half-match from Nicoleta. She was scheduled to make the bone marrow transplant for this Monday, thus I came to the conference earlier and made the presentation. You can get more information about Nicoleta’s sister Mihaela on a website that they update http://www.match4mihaela.org/.
You want to do something positive in healthcare? Register to be a bone marrow donor. It’s quick and painless to register (no needles!). I think they can mail you a kit and you mail it back. I have been on the list for years and never called, so I infer that the probability they will need you is small, but if they do you save someone’s life. Still not willing? you can donate to help support the cost of typing and identifying matches.
don’t wait until your death to provide a(n organ) transplant to someone, you can be a match today.
In case you are wanting to explore Montreal and take a break from the conference…The Montreal Tourism Department publicly published these fun facts:
Here is another list of Free For All activities in Montreal and the surrounding area.
Enjoy your stay..
Many attendees have stopped by the INFORMS table and told me the program is excellent. Congratulations to all the organizations involved and their affiliates — INFORMS, CORS, EURO, POMS, and IIE. The mere existence of this conference speaks volumes about the spirit of partnership among the “sister” societies for the betterment of the profession. Congratulations, too, to Vedat Verter and the organizing committee for their hard work and committment.
Of course with all due respect to the other organizations, I am here reprsenting INFORMS and all we have to offer. If you are not a member, you can sign up for the rest of 2011 at half off. Many O.R. professionals have pointed out to me that this is not really a discount when you take time to think about it. (I really wish you wouldn’t do that, but you are right!) However, if you are planning on attending the 2011 INFORMS Annual Meeting in Charlotte — which, by the way, is planning to have a robust healthcare track — it is in your own best interest to sign up now as a half year member at $74 for regular members. Why? Because when you sign up for the Annual Meeting as a member, you will save over $175 that non-members pay. This is called personal revenue management, I believe! Сайт знакомств
I won’t even get into the quality of the information you’ll receve from INFORMS for the rest of the year. Well, maybe just one item — our journals. By way of introducing you to our journals, INFORMS staff with the help of members Ritu Agarwal and Brian Denton have put together a volume called “Focus on Healthcare.” This volume features 16 healthcare-related articles drawn from many of our 11 highly respected journals. If you did not get a free thumb drive containing this volume, I’m sorry, but I’ve run out. However, we did place this special volume on the web for your perusal. Please visit it at http://www.informs.org/Sites/IPOL/Focus-on-Healthcare. I’ll be closing up shop on Wednesday to head back to INFORMS HQ in Maryland, so please stop by the INFORMS table soon.
I arrived a little later than I expected but still in time to get to a session before noon. I was impressed with the number of attendees — healthcare is clearly becoming a major app area for OR.
I was charged with spurring conversation over lunch, and things went well once we got rolling. The conversation was interesting partly because we had a couple of practitioners at our table and so us academics were exposed to another point of view.
I sometimes get frustrated with the overmodeling that occurs. One predictive modeling paper had a 1% improvement over a fairly basic logit model and it was touted as a major improvement. The proposed model was incredibly complex. I guess more effort should be directed toward finding a parsimonious model that does the job even if it’s not esoteric. The proof of the pudding is in the utility of the model, not its complexity.
The reception was surprising, in terms of the amount of beef I was offered — it was piled high on my plate (I didn’t ask for seconds!). Apparently there was a high variance across carving stations, I was told.
Looking forward to some great sessions tomorrow.
Some of my favorite things about the healthcare conference so far:
1. Networking lunch. This is a great idea to get us all together. Maybe this is only possible at a smaller conference like this where everyone has similar interests (instead of the full-blown INFORMS conference). We all have to eat, so this was a nice opportunity to meet some new people.
2. Presentation quality. I know it’s only the first day, but I think the quality of presentations is slightly better than at the regular INFORMS conference. Perhaps the more comprehensive abstract submission and the 30 minute talk format has given us more fully-developed talks.
3. Attendance. I think attendance is better: I was surprised to see almost 25 attendees at an 8:30 am session on the first day.
What do you think? Was I wrong? Do you have other suggestions?
The three talks in this morning’s session “Capacity Management in Hospitals” led me to question three basic assumptions about queueing systems: that arrival rates, service times, and capacities are fixed and exogenous. Our traditional queueing framework takes these quantities as given, outside of the boundaries of the model.
Take the G/G/s system (please). The first talk, by Jonathan Helm and Mark Van Oyen, discussed how arrival rates into hospital wards may be regulated to balance census levels. Therefore, when we look at historical data on patient arrivals, we should not take these at face value (exogenous) but instead consider how the arrival rate might be endogenous – may depend upon real-time system loading.
For the next ‘G’ in G/G/s, Don Eisenstein and a team from the University of Chicago described data that strongly linked length of stay (LOS) in hospital beds with both bed utilization and the allocation of beds among specialties. Higher utilization leads to shorter LOS, and focusing groups of beds towards particular specialties lowers LOS. Therefore, service time may be endogenous as well.
Finally, Wen-Ya Wang and Diwakar Gupta discussed how the ‘s’ in G/G/s might be endogenous: in some empirical studies (but not all), nurses who anticipate higher workloads during their shifts may be less likely to show up. Yikes! We can’t even take the number of servers for granted?
That is, should we build Supermodels, in which everything is endogenous? But such models, encrusted with so many variables and feedback loops, will be analytically intractable. And if we want to use simulation models to make real health care decisions, by the time we collect the data necessary to populate such complex simulation models, it will probably be time for us to retire (or the hospital we work for will have folded). It’s an old tension in modeling, between capturing the complexity of the real world and making appropriate approximations for the sake of efficiency and even accuracy. Therefore, we want to ask not only whether certain parameters are endogenous, but also whether it matters when making real decisions. That’s a tough question, but it is one we must answer when implementing our ideas.
* Thanks to Armann Ingolfsson for suggesting this title.