First Advisor

Wayne W. Wakeland

Term of Graduation

Fall 2007

Date of Publication


Document Type


Degree Name

Doctor of Philosophy (Ph.D.) in Systems Science


Systems Science




Diagnosis, Chinese Medicine, Temporomandibular joint -- Diseases, Recreational therapy -- United States



Physical Description

1 online resource (2, xv, 219 pages)


Astonishingly, most Traditional Chinese Medicine (CM) research in the West proceeds without CM diagnoses, perhaps because CM diagnosis is time consuming or not considered relevant. One way to improve the feasibility of incorporating CM diagnosis would be to prescreen participants using questionnaires. This would allow cost savings in recruitment, and balancing of treatments by CM diagnosis. Consequently, the hypothesis of this research was that pre-treatment questionnaires can predict CM diagnosis.

Baseline questionnaires from 195 participants with temporomandibular joint disorder (TMD) were examined to test the hypothesis. Two methods, logistic regression (LR) and reconstructability analysis (RA), were used in conjunction to test the hypothesis. Models were created that predicted CM diagnosis from pretreatment questionnaires. First, LR models were prepared to predict the diagnosis for each subject using direct effects only. Then variable-based (VBRA) and statebased RA (SBRA) were used to select potentially important interaction terms. These terms were then introduced into the original LR model and assessed for clinical relevance, model simplification, and improved diagnosis prediction.

Prediction equations were identified for each of the four CM diagnoses that increased prediction accuracy by 3.9% to 22.2% compared to a naive model that simply used population prevalence. VBRA successfully identified one clinically significant interaction term that improved classification by 1.8% and simplified the model. SBRA identified a state that improved classification by 3.2% and further simplified the original model. The classification errors were reduced between 23.9% and 56.0% when adjusting for baseline prevalence. When assessing the trade-off of specificity and sensitivity, the model efficiencies for the diagnoses ranged from 0.61 to 0.83 where 0.50 indicates a model with no improvement over a model that simply assumes population prevalence and 1.00 indicates a perfect model.

Since the study identified successful prediction algorithms for three of the most common diagnoses, the research successfully demonstrated the use of questionnaires to help select patients with specific CM diagnoses. The use of VBRA and SBRA with LR to find interaction terms was also demonstrated. Future studies of CM diagnoses should build upon the current study.


In Copyright. URI:

This Item is protected by copyright and/or related rights. You are free to use this Item in any way that is permitted by the copyright and related rights legislation that applies to your use. For other uses you need to obtain permission from the rights-holder(s).


If you are the rightful copyright holder of this dissertation or thesis and wish to have it removed from the Open Access Collection, please submit a request to and include clear identification of the work, preferably with URL.

Persistent Identifier