From reh0303@doh.wa.gov Thu May 6 11:57:24 1999 Received: from mxu3.u.washington.edu (mxu3.u.washington.edu [140.142.33.7]) by lists.u.washington.edu (8.9.3+UW99.02/8.9.3+UW99.01) with ESMTP id LAA21584 for ; Thu, 6 May 1999 11:57:23 -0700 Received: from smtp.doh.wa.gov (hub.doh.wa.gov [192.230.1.5]) by mxu3.u.washington.edu (8.9.3+UW99.02/8.9.3+UW99.01) with SMTP id LAA19930 for ; Thu, 6 May 1999 11:57:22 -0700 Received: from Connect2 Message Router by smtp.doh.wa.gov via Connect2-SMTP 4.32; Thu, 6 May 1999 11:59:38 -0700 Message-ID: Date: Thu, 6 May 1999 11:57:48 -0700 From: "Hoskins, Richard E." Sender: "Hoskins, Richard E." To: waphgis@u.washington.edu Subject: Fwd: RE: Hospital catchment area X-SMF-Hop-Count: 2 X-Mailer: Connect2-SMTP 4.32 MHS/SMF to SMTP Gateway Concerning the hospital catchment area question from Dr Zucchi, Dr Zucchi: What GIS software do you have available? What are trying to calculate or determine? Some ideas 1. Try building Theissien polygons around each hospital location. These "areas of influence" are the areas that are closest to each hospital, also called Thiessen Polygons or a Voronoi Diagram. Caliper's (www.caliper.com) Maptitude can automatically create areas of influence around any number of points in a layer. Likely the other desktop packages can as well. >From Maptitude's HELP: "To create areas of influence, you simply choose the set of point features you want to use. Maptitude creates the areas of influence, and saves them in a geographic file. As an option, Maptitude will use an overlay to estimate the attributes for each area based on data stored in some other layer in the map. "When Maptitude creates areas of influence, it starts by building a large circle containing all the points of interest. Then it divides the circle into areas, with one area for every input point. As a result, the areas that are built around the points on the edge can be much larger than those in the middle. " If you use an overlay to estimate the attributes of areas of influence, remember that the ones around the edge are larger than the others. To build areas around the points so that every area is the exact same size, use the Tools-Bands command." When you are finished you have the Thiesssen Polygons and you have an estimate of whatever variable you need from the polygon coverage (say, census tract) below it - if you need this information. Once the Thiessen polygon coverage is made then a point -in- polygon query over a point patient residence layer will produce a table with an indication of which patient is in which polygon which can then be exported for calculation. 2. However my experience is that you might be surprised what the catchment area of a hospital really is. It is not necessarily a simple geographic relationship as of course you know. One would expect this in referral hospitals, but not in local ones. We have found that often this is not the case, that is, there is not a simple homogeneous geographic relationship of a hospital's location to the origin of its patients. It is age/sex/ & disease/condition dependent. 3. Building a distance matrix is tough in the desktop packages, programming is required. It may be simpler to write a SAS (or SPSS) program to do it for you. One would get a table of patients and distance to each hospital. From that you could divide up patients who are closest to hospital 1, then hospital 2, etc. I have SAS code you can have if it suits your need to set up a distance matrix. You will need your coordinates in UTM for convenience. Long/lat can be handled but one needs conversion routines to linear coordinates. One then could look at the hospital a patient actually went to vs the one they are closest to. 4. To get the hospital catchment area for each patient using a buffer - you can end up with 200,000 buffers, etc. Building buffers around the hospitals will not give the result that I think you need. I f I have missed the point of what you are trying to accomplish, send me a note "off-line" and I'll try again. Richard E. Hoskins reh0303@doh.wa.gov ********************* Forwarding Message ********************* -----Original Message----- From: WAPHGIS-owner@u.washington.edu [mailto:WAPHGIS-owner@u.washington.edu]On Behalf Of Alberto Zucchi Sent: Tuesday, May 04, 1999 5:33 AM To: waphgis@u.washington.edu Subject: Hospitals catchment area Dear Listers, I'd like to receive some suggestion on the following problem. I have a data-set containing all paediatric admissions (say, 200,000) in all the hospitals in a region (say, 30). I have two tables. The first contains hospitals' localization (with related coordinates). The second contains each admissions these hospitals did in one year, with each patient's city of residence (say approx. 200,000 patients for 2,000 cities/villages). I'd like to create a "catchment area" for each hospital, in relation to their patients' residence. I thought to build a "distance matrix" in kilometres, and the calculate the mean or median distance for each city/hospital, and finally superimpose a "buffer" (or maybe drawing a simple circle...to simplify reality). Does anyone have some suggestion/trick on the way to perform this? Thank you in advance. Alberto Zucchi, MD Epidemiology Office ASL Bergamo .