Massage Medical Applications Project (MMAP)

The MMAP Data Process and Markup Language (MMAML)

A major goal of MMAP is to characterize the connections between treatment goals, conditions being treated, secondary conditions requiring protocol modifications, and sources of information. The sources of information are considered to be direct contributions by and discussion among members of the massage professional community, massage applications in the medical literature, and other benchmark references. The MMAP projects thus has four overall tasks:

The first task implies that the sources under consideration, including discussion and comments, should be recorded as a primary reference and historical documentation. This becomes explicit in the fourth task, providing a public context and benchmark for any conclusions drawn from the output of this project and a reference for later revisions.

The second and third tasks imply a means of recording the information itself and the interconnections between entries. From an information management or knowledge domain standpoint, this is a process of defining a graph with nodes of several types and the edges between nodes. For this purpose, we are in the process of defining (in use) the Massage Medical Applications Markup Language (MMAML). MMAML is an application of XML, the eXtensible Markup Language defined by the W3C (World Wide Web Consortium). MMAML draws on the knowledge of several existing graph definition and communications languages, including GraphML (Graph Markup Language), GraphXML, GXL (Graph Exchange Language), and XGMML (eXtensible Graph Markup and Modeling Language). The main conceptual difference between MMAML and the existing markup languages is that MMAML starts with collects of heterogeneous nodes applicable to massage medical applications. Thus it appeared, at least in the onset, to define a language to match the data rather than to force the data into the general concept of “nodes”. At the same time, the edge notation for MMAML is based on that for GraphML. This should help in later mapping MMAML nodes and edges to formats used by existing graph display tools to look for data clustering.

We are developing a human friendly viewer for MMAML. At present the viewer provides a multiple outline perspective of the raw data in MMAML. The MMAP database in MMAML format is also human readable.

While the data constitute a graph, if a starting point is chosen and the links broken at a given depth, the data can be displayed as a tree. The MMAML viewer thus can display outlines of the MMAP data from different starting perspectives such as treatment goals, conditions, and contributor input. The viewer has also made use of concepts developed for OPML and ActiveRenderer to manage interactive outline displays. Thus sections of the outline display can be expanded or collapsed by clicking on the triangle symbols on the left edge.

The current MMAML node types are:

Tag Data Content
goals Treatment Goals
conditions Condition prompting Treatment
techniques Massage Techniques
complications Complicating secondary conditions affecting massage protocols
references Links to information on external reference documents
contributors Input contributed to MMAP by individual practitioners and educators

The additional "edge" tag defines connections between nodes, each node entry having a unique ID.

References

To handle archiving references, we have settled on the Library of Congress's MODS–XML format. PubMed provides an ability to save search results in PubMed XML , but that contains a lot of extraneous material for use by PubMed itself. The PubMed XML can be transformed to MODS–XML using Chris Putnam's bibutils. Initially, we had used a further transformation into BibTexML to get the references up, but we dropped that now that we have an XSL stylesheet for MODS. An added feature is the ability to highlight (light-yellow) review articles likely to be of high interest. That's been done for a couple of records. Reference identifier strings (e.g., Ernst2003) are now displayed at the beginning of each reference. The HTML generated by the XSL transformation includes named anchors (internal links) based on these identifier keys, giving a quick way to get to an individual reference if the key is known (once some editing to insure that these are unique is complete). This will also allow references to be listed by identifier key in the MMAML database, with a link from there to the actual reference, similar to how contributor input is now included.

The initial cut at finding medical articles referencing massage applications resulted in 375 viewable references. These were citations from 1995 through 2005 having abstracts, having massage in the title, and not containing cardiac or carotid as keywords. Note that the references are currently unedited by human hand and not otherwise included into MMAP data structures. The XML file containing them is about one MB in size. The journal titles link to a Google search, which can be useful. The doi (digital object identifier) field and uri/url fields, where present, will in most instances successfully link to the actual article source. No claims are made at this point that the splinters have been sanded off or that it's even been varnished, let alone that the varnish has dried.

Settling on a reference archive format will facilitate editing of the existing file and the addition of other article and book references. There are reviewed articles, yet to be merged in, from the Journal of Bodywork and Movement Therapies. These are not indexed in PubMed but listed in Elsevier's Science Direct. For the moment, a separate preliminary file of these JBMT citations is available. Most of the longish URL's still have to be replaced by more succinct DOI's. There are also some relevant articles on outcome-based medical education and determination of competency to include. These come from several sources. There is now, also, a separate file of selected references on professionalism, outcome-based education, competence, and physician liability from use of CAMS (New!).

In addition to article selection on content, there are some character encoding problems, mostly in author's names, that will be corrected in the near future. Some of the record identifier strings also have to be corrected to insure that they are unique and informative. The references will, in the not two distant future, be broken up into smaller files based on year of publication. What follows are files that are the start of breaking references up by time and integrating various sources together.

Reference files