MEASURING PATIENT PREFERENCES THROUGH THE TIME TRADE-OFF METHOD FOR ORTHOPEDIC CONDITIONS ON LARGE SAMPLES

Background: In cost-effectiveness analyses, Quality-Adjusted Life Years (QALY) remains one of the most widely used health effect measure. Among the various methods of estimating utility values, time trade-off (TTO) has traditionally been one of the dominant methods for eliciting utilities, however it has been presenting several practical impediments to provide a high and fast collecting process.Objective: To test a method of collecting TTO-derived utilities using a platform called Amazon’s Mechanical Turk (MTurk) that provides reliable, fast and inexpensive data.Methods: A pre-programmed interactive questionnaire was design to simulate a live TTO interview using Qualtrics. To validate the results members of the Research on Research (RoR) Group not aware of the research agreed to answer the same questions on a videoconference live interview. We determined feasibility through assessment quality and cost/benefit relation indicators. In addition, this paper followed the framework for reproducible research reports proposed by our group.Results: Results: Our results showed that the MTurk population is representative of the US population (based on 2012 census) and there were no differences on the willingness to live when comparing the MTurk sample and the live interview sample, and also no differences of the WTL when comparing the different questionnaire designs developed. Preference results showed differences only for race (between others and African-Americans, and other and white), and overall median values of 0.83 (Q1=0.83;Q3=0.90).Conclusions: MTurk is a reliable web place to collect large sample using the TTO method, and should be used to collect utility data for CEA.


QUEST I O N N A I R E D E V E L O P M E N T A N D D E S I G N
).

TECHNICAL FEATURES
We a priori determined that the jumping questions questionnaire required approximately 7.5 minutes to be answered  question.
In addition, we excluded any respondents whose time to completion of the questionnaire was less than 3.5 minutes, given that the video alo ne lasted 3.5 minutes. However, we did include these last two items for validatio n analysis.

Data Analisys
We initially performed an exploratory data analysis. Descriptiv e statistics were presented as relative frequencies, median We defined the lev el of significance as p=0.05, and all statistical procedures and graphs were perform ed using R language software20.

REPRODUCIBLE RESEARCH FRAMEWORK
This paper followed the framework for

SAMPLE DESCRIPT I O N
The total amount of respondents in the  ( Figure 2 ).

FEASIBILIT Y A N D VALIDIT Y O F T H E METHOD
Regarding feasibility, 352 (90.95%) of the original sample completed the instrument within a priori estimated acceptable levels.

Figure 4 -Comparison of the Willingness to Live variables between total sample and samples controlled for attention, duration and inconsistency
Validity testing showed no statistical differences between the evaluatio n methods   Figure  6).

DISCUSSION
To the best of our kno wledge this is the first study to validate a TTO instrument using the MTurk platform. Previous study suggested that it would be feasible to conduct quality of life research in patients via the Internet25, and other internet based Rev i sta El etr ôn ic a Ge st ão & Soc ied ad e v. 12 , n . 31 , p . 2 17 3 -2 19 3 | Jan e iro /Ab ri l -20 17 IS SN 1 9 80 -57 56 | D OI: 1 0. 21 17 1/ g e s.

EXPLAINING W H Y U S E T H I S M E T H O D
The design and report of this questionnaire comes from the idea that collecting utilities using the TTO method through MTurk provides a series of solutio n to several limitatio ns of utilities collection in particular the traditio nal TTO method, such as: Compariso n with P rior Work

EXPLAINING W H Y U S E T H I S M E T H O D
The design and report of this questionnaire comes from the idea that collecting utilities

FUTURE R E S E A R C H A N D I M P L I C A T I O N S F O R P R A C T I C E
For future research, we have found this tool to be extremely efficient as a metho d to collect preferences for decisio n analyses and CEA; use of MTurk could conceivably be applied to build a utility database in a fast and inexpensive manner. More specifically, we are planning to integrate the specific values we estimated within the context of a knee OA Markov Model. In addition, the reproducible research framework in which our study was conducted is specifically designed to allow the use of the sam e methodology in sim ilar utility collecting projects involving other disease processes and health states.