A Platform for Assessing Cancer Patients’ Quality of Life - New Contributions in Information Systems and Technologies - page 51

analysis of the test does not require strictly parallel tests to verify the trust; the
statistics of the items and the ability of the examinees are both reported on the same
scale. IRT models have potential advantages over the CTM method in the evaluation
of responses of self-reported health. These models provide estimates for the items and
the latent trait that does not vary with the characteristics of the population. Meaning
that measures can be obtained and the variation does not depend on the used
instrument or individuals [18]. IRT models typically use a logistic function to link the
observed variables to the latent characteristic, Georges Rasch, besides this
relationship in their model established a relationship between two other parameters. In
fact, the Rasch model is a mathematical expression (1) for the relationship between
the probability of success ( P ) and the difference between the ability of the individual
( θ ) and the difficulty of the item ( β ).

ʸ − ʲ

e

P =

(1)

ʸ − ʲ

1 + e

From (1) it is possible to obtain the relation in (2):

P

⊛ ⊞

ln = ʸ − ʲ

⊜ ⊝ ⊟ ⊠ (2)

1 − P

Therefore, this model assigns a probability of success on the answer to a question
in terms of only two parameters to be estimated: Efficiency θ s of the subject s and
Difficulty β i of the item i . In the Rasch model, the scores of the subjects in the test or
questionnaire are sufficient to estimate the ability of people. All people have the same
score and have the same ability regardless of the responded positively items.
Furthermore, item parameters can be estimated without the need to estimate the
abilities of individuals [19]. This joint measurement on the scale, the skills of the
individuals and item difficulty allows the analysis of interactions between individuals
and items. These characteristics fit perfectly the needs of a questionnaire for the
assessment of the Quality of Live (QoL), whereas the ability of the subject is their
QoL and that the items are the questions of the questionnaires. In case the answer is
not located within the expected value, it is possible to mark it as deviant, comparing
the probability of correct response and depending on their QoL (ability) with the
response of fact.

4 Methodology

The QoLis platform uses the concepts of knowledge discovery designed to exploit large
amounts of data. The objective was to try to replace the questionnaires to measure and
quickly indicate the quality of life. Next the methods, procedures to collect data and
analysis are explained.
The target population is composed by patients with cancer from Otorhinolaryngology
and Head and Neck services of an oncologic institution. The sample was composed of
3013 patients. A space named as Office of Evaluation of Quality of Life became the