Dear Soheil

I don't know if some people used AHP for determining the location of urban health care centers, but if not, I do know why:

Simply because the method works with preferences and this scenario needs to consider real data and issues, not data from intuition as in AHP. Can you imaging building health centers based on the opinion of a DM or DMs?

Determining health locations in a city is not an easy matter and certainly not for AHP, because you need to consider data and criteria that are no subject to preferences, and also because you need to work with GPS data, which are

are objective. For instance, suppose that you have to find out the best location for four health facilities in a city, and that the four of them have to be built.:

* Population of each area or district in a city where a facility could be built

* Minimum and maximum population to be served, in each area

* Density of each area

* Minimum and maximum capacity of each center, in order not to have a center too small or too big.

* Determining the center of gravity or barycentre of the district according to the above requirements

* Determining the average distance to be traveled by patients

* Determining the average distance from the facility to high complexity centers or main hospitals to derive emergencies.

* Available plots of land to built the facility

Can this problem be solved? Definitely, but not with AHP or ANP

I know that MCDM like AHP is commonly used in healthcare sectors. Many articles you can found it at my research items USED MCDM methods. Also you can see it in my scholar

https://scholar.google.com/citations?user=VOveIgkAAAAJ&hl=en

Regards

Soheil,

I have worked on a similar area. Checked my work titled "Multi-Criteria Allocation Configuration Problem for Mobile Health Clinics". https://scholar.google.com/citations?user=C9T1VCgAAAAJ&hl=en

Dear A.s and Egbe:

Soheil question was very specific when inquiring:

'Why Multi criteria decision making techniques, like AHP, are not common for allocation of healthcare facilities?'

Following references from A.s and Egbe, I was reading the titles of their 26 articles, and none of them refers to allocation of healthcare facilities. However, there is an article from Egbe on Mobile Health Clinics. Obviously, without reading the article, since the author, does not show it, it is impossible to make any comment. In addition, he does not say which method employed.

Most probably there are many articles written by doctors where they, based on their experience, use preferences in some aspects, but of course, I am not qualified to comment about them.

I maintain that problems such as determining locations for health centers, are scenarios that can't be solved by AHP.

By the way, if anybody has solved the problem of health centers allocation with AHP I would be glad to know about it , and thus to have the opportunity to read the paper.

Dear Nolberto Munier

Thank you so much for your great reply. This is true that expert opinions is not enough for allocating a healthcare facility but is it possible to integrate these opinions with model that made from real data?

For example this is my idea: The output of real data model could be one of the inputs of AHP. Other factors like cost and type of land use, proximity to schools, main streets, public places and so on could be overlay with the output model to improve the preferences with policymakers criteria.

In my case, I'm working on a project for allocation/relocation of EMS ambulances, and I think if I just consider calls-data for my analysis the acceptance of the results wouldn't be much for policymakers.

I would be thankful to here more from you and others.

Dear Soheil

Well, obviously I misunderstood your original question, since I assumed that you were talking about physical centres for health, when in reality you are talking about locations of ambulances.

I don't understand how you can input real data in AHP, when the method precisely ignore it.

I don't understand either how you can overlay another real data

Why do you need preferences from the DM? About what? If a hospital receives a call from an injured or sick person, it is not matter of preferences but a matter of action. Probably the dispatcher will send the most nearby ambulance, and the decision is based on distances and ambulance availability. I am completely ignorant about how a hospital works in these calls; therefore, I am only expressing what common sense tells me.

If you talk about say the best treatment for a patient I can understand it because it needs the experience of a physician, and probably in a complicated case, the decision is taken by a an agreement between several doctors. There is no MCDM method that can replace it, but for real data, I don't see the advantage.

Regarding your project, again, it can't be based on preferences. It must be based in frequency of calls, average seriousness of the call, number of ambulances, number of para medics, equipment on ambulances, etc., you know it better than me.

In my opinion you have to base your study on statistics to extract average values, and it appears to me that you don't need any MCDM method.

Dear Soheil,

Yes expert opinions can be integrated into real data on the allocation of healthcare facilities and other resources.

Dear Egbe

Can you pls. explain how you combine real data with experts' opinions using AHP?

I am not saying that you are wrong, I am just trying to learn, and an example will be very useful

Thank you again dear Nolberto Munier

This is just a raw idea. A significant part of our research is about allocation of physical facilities which are EMS stations. The policy makers want to allocate new stations and the question is where?

I have data of EMS callers with their locations for one year and I can find out which districts are more in need of station with different computational models. but the question is that in reality for some reasons like cost, the policy maker prefer to allocate station in a less optimum location. this is where i think AHP can balance these aspects.

The output of computational model can be a polygon or raster layer which contains a score based on priority for new resource. This layer and other GIS layers like land-use, distance form main streets, cost of lands and so on can enter to AHP to be integrate and demonstrate preferred areas based on both actual need and policymakers criteria.

I don't know about the possibility of this idea and I wish to hear more.

Egbe-Etu Emmanuel Etu Thank you for your response.

Do you know any example of this in literature?

Nolberto Munier This is one of the rare samples of using AHP for finding priority areas for establishing new healthcare facility

Article The application of geographic information systems (GIS) in i...

For allocation problems, I'd say that using multicriteria methods is not a frequent approach . But you still can find some.

Regarding the health context, the health experts are not aware of multicriteria problems. They generally do not cover these types of management concepts in school, for example. So, this might constraints the conexion between the two areas, although there is a huge field of exploration. Added to this, health applications are sensitive because of the specifics and caractheristcs of the problems, say biological or chemical behaviors. In order to develop a management approach like multicriteria you need to gather a multidisciplinary group.

See find a paper that work with both discipline.

Chapter FITradeoff Method for the Location of Healthcare Facilities ...

Conference Paper Multicriteria Decision Making for Healthcare Facilities Loca...

Combining experts opinion is a group decision problem, you can find some studies using fuzzy and/or multicriteria concepts. Also, it depends on the type of method you are using. I'm not an expert in AHP, but it might be possible since AHP uses and agregation method. So, in a part of the process, to make de calculations possible, weights are elicitated from experts which is combined with the real data (decision matrix) of the problem.

Dear Soheil Hashtarkhani,

My research was on allocating mobile health clinics (MHCs) to areas in Michigan that do not currently have MHCs. We utilized AHP, a K-means clustering algorithm and set covering to optimize the coverage area. Although, this work did not capture the direct input of policymakers, it captured the inputs of patients who utilize MHCs for treatment. Parameters such as cost, distance, type of service, population was utilized to capture data from the patients regarding allocation of MHCs. The result obtained from the research enabled policymakers to make decisions on how which area is best to allocate MHCs.

If you want to find out more please read my paper titled: "Multi-Criteria Allocation Configuration Problem for Mobile Health Clinics". You can also perform a journal backtracking in the reference section to get ideas from other papers cited in the work.

Nolberto Munier I believe that if the data containing patients preferences on where to allocate resources is captured and analyzed using tools such as clustering and AHP, data containing preferences of policymakers can also be captured and merged with data of patients and a much better model will be developed. My research has captured the preferences of patients as that was what the policymakers in my case where looking for in allocating MHCs.

Dear Soheil

SH - This is just a raw idea. A significant part of our research is about allocation of physical facilities which are EMS stations. The policy makers want to allocate new stations and the question is where?

NM - That is precisely the job for MCDM. If in a city you have it divided in several areas and also you have several hospitals and health centers, the MCDM method can tell you where the ambulances should be parked as well as the number of them for each center, just for Emergency Medical Services, but as you can realize this is a mathematical problem and then I don’t see where the Decision-Maker (DM) fits.

However, the DM is the most important component in decision-making, because this activity usually has a large subjective component. How then can I explain this dichotomy?

The answer is related to the time when the DM has to act, and it is not certainly at the beginning of the process by inventing data. His most important role is post result, not ante result. It means that the DM has to design the initial decision matrix considering all alternatives and criteria, and most important, he has to consider those aspects that are hidden in the scenario. For instance, that certain routes may be temporarily closed such as drawbridges that are periodically open for maritime or river traffic, or interruption provoked by low rail barriers, or traffic bottlenecks, or heavy traffic, etc.

Once he introduces all this information in the decision matrix, and that reflect reality, he must run a MCDM model and get results. After a solution is reached, he must analyze it and can make changes, or even reject the solution, based on his experience. In this way he can make whatever changes or modifications he wishes, but now based on a solid result, not on conjectures. There can be various several DMs and each one with his ideas, but all of them start from the same base, and therefore, possibly a final solution can be attained among them.

SH – I have data of EMS callers with their locations for one year and I can find out which districts are more in need of station with different computational models. but the question is that in reality for some reasons like cost, the policy maker prefer to allocate station in a less optimum location. this is where i think AHP can balance these aspects.

NM- Yes, it is quite possible, but as I said, it is a DM decision after a result is found; it is not a matter of preferences. The DM may certainly allocate some ambulances in less optimal locations, but this decision must be based on solid reasons. If for instance, he decides to allocate in another location, he can modify the MCDM initial matrix by adding that new location and running again the chosen MCDM method. If the general objective is to have a better service, he needs to have some yardstick in order to judge if, in general, the new location is better that the old one.

As you can see, there is no room here to use AHP. Just consider that real-world is in general not transitive, therefore, why to force it to be using AHP?

SH - The output of computational model can be a polygon or raster layer which contains a score based on priority for new resource.

NM. I agree, but remember that AHP does not consider resources, let alone their availability. Since you have to have a certain number of paramedics and ambulances, that is resources, how are you going to introduce them in an AHP model?

SH -This layer and other GIS layers like land-use, distance form main streets, cost of lands and so on can enter to AHP to be integrate and demonstrate preferred areas based on both actual need and policy makers criteria.

NM- Yes, all GPS layers used alone or combined give you a very good information. My question is, how are you going to introduce it into AHP? If you think in feeding the DM with areas, potential routes, bottlenecks, speeds, statistics of calls, etc., how can a policy maker handle all of them simultaneously?

SH - I don't know about the possibility of this idea and I wish to hear more.

NM- My friend, I understand that decision-making is not just using preferences, which of course can change if another DM simply because he sees reality differently. Reality is only one, and perhaps there could be many different solutions, and that is also real. Then, how do you know that out of say three solutions, which is the best if you don’t have any yardstick to compare to?

Don’t you think that getting a solution, obtained by reflecting as close as possible a scenario and then processed without any human interference, is better that one obtained by intuition?

In so doing the DM can work on solid basis not on wobbly terrain.

Well, this is only my opinion

Nolberto Munier I think your current response answers it all.

Francisco Viana Thank you so much for your great answer. The article you suggested seems really great for my question. I asked the author for full-text to consider it carefully.

Thanks

Soheil Hashtarkhani You are welcome. Feel free to make others questions when you'll be working with it.

Greetings.