DS0411 - Recherche translationnelle en santé

Minimal clinically Important DIfference determination for Patient reported outcomes in presence of REsponse Shift – MIDIPRES

Minimal clinically Important DIfference determination for Patient reported outcomes in presence of REsponse Shift

A méthodological approach to solve the problem of the Minimal clinically Important DIfference determination

Context

Patients Reported Outcomes (PRO) are more and more used in clinical research or in medical practice for measuring quality of life (QoL), pain, anxiety or fatigue for instance. Nevertheless, these measures are often difficult to interpret for physicians and statisticians. The Minimal Clinically Important difference (MCID) is a useful tool for the interpretation of such data. Unfortunately, few studies focus on the determination of the MCID of a PRO. Indeed, such studies are expensive and can be difficult to perform and there is no consensus to date on the best method for the MCID determination.<br /><br />When PRO are used to estimate a phenomenon, the unit of measurement is classically the score (Classical Test Theory – CTT), but a first difficulty for the MCID determination on the score is the lack of interval property of the scores (property implying that a score unit always represents the same quantity of the studied phenomenon). Without this property, the interpretation of the MCID value can depends of the initial value of the score. Another potential difficulty is the presence of Response-Shift (RS): this phenomenon appears when the perception of the measured concept by the patients changes over time. In presence of RS, and without adjustment on it, an important risk of bias is possible regarding the MCID determination. Actually no method has been found in the literature to determine the MCID in presence of RS.

-Rasch Measurement Theory (RMT) is an interesting alternative to the use of scores, because these models allow obtaining a measure of the PRO, the latent trait, having by definition the interval property. More, RMT allow an easy handling of missing data. Indeed, in these models obtained estimations of the latent trait independent of the observed items, and consequently these estimations are unbiased even if missing data are informative (missing not at random and possibly linked to the patient's health state).

The main objective of this project consists in comparing methods for the MCID determination in terms of bias on the MCID, sensitivity and specificity to detect individuals with true change, and robustness to the presence of missing data and response-Shift.

In a first time, the model characterizing the relationship between the MCID and the latent trait (eg: quality of life) will be built. The procedures allowing the MCID determination will be then defined for CTT and RMT. In a second time, a simulation study will be performed in order to compare different method of MCID determination within these two measurement paradigms.

Last, the procedures will be used on real data chosen for their diversity: a study regarding the link between satisfaction of care and quality of life (SATISQOL), a study evaluating the Health-related Quality of Life for transplanted patients (ReTransQol), a study on the adaptation of patients with cancer to their illness (ELCCA) and a study measuring the cognitive distortions of problematic gamblers (GABS+GRCS).

An extension of the conceptual model is under process. New simulations studies have been schedulded for 2019.

The procedures will be implemented under a Stata module and under the PRO-online software (http://pro-online.univ-nantes.fr), in order to make them freely available for researchers.

Several papers have been submited to scientific journals.

Patients Reported Outcomes (PRO) are more and more used in clinical research or in medical practice for measuring quality of life (QoL), pain, anxiety or fatigue for instance. Nevertheless, these measures are often difficult to interpret for physicians and statisticians. The Minimal Clinically Important difference (MCID) is a useful tool for the interpretation of such data. Unfortunately, few studies focus on the determination of the MCID of a PRO. Indeed, such studies are expensive and can be difficult to perform and there is no consensus to date on the best method for the MCID determination.

When PRO are used to estimate a phenomenon, the unit of measurement is classically the score (Classical Test Theory – CTT), but a first difficulty for the MCID determination on the score is the lack of interval property of the scores (property implying that a score unit always represents the same quantity of the studied phenomenon). Without this property, the interpretation of the MCID value can depends of the initial value of the score. Another potential difficulty is the presence of Response-Shift (RS): this phenomenon appears when the perception of the measured concept by the patients changes over time. In presence of RS, and without adjustment on it, an important risk of bias is possible regarding the MCID determination. Actually no method has been found in the literature to determine the MCID in presence of RS.

Rasch Measurement Theory (RMT) is an interesting alternative to the use of scores, because these models allow obtaining a measure of the PRO, the latent trait, having by definition the interval property. More, RMT allow an easy handling of missing data. Indeed, in these models obtained estimations of the latent trait independent of the observed items, and consequently these estimations are unbiased even if missing data are informative (missing not at random and possibly linked to the patient's health state).

The main objective of this project consists in comparing methods for the MCID determination in terms of bias on the MCID, sensitivity and specificity to detect individuals with true change, and robustness to the presence of missing data and response-Shift.

In a first time, the model characterizing the relationship between the MCID and the latent trait (eg: quality of life) will be built. The procedures allowing the MCID determination will be then defined for CTT and RMT. In a second time, a simulation study will be performed in order to compare different method of MCID determination within these two measurement paradigms.

In a third time, the procedures will be implemented under a Stata module and under the PRO-online software (http://pro-online.univ-nantes.fr), in order to make them freely available for researchers.

Last, the procedures will be used on real data chosen for their diversity : a study regarding the link between satisfaction of care and quality of life (SATISQOL), a study evaluating the Health-related Quality of Life for transplanted patients (ReTransQol), a study on the adaptation of patients with cancer to their illness (ELCCA) and a study measuring the cognitive distortions of problematic gamblers (GABS+GRCS).

Project coordination

jean-benoit hardouin (Université de Nantes - EA4275-SPHERE)

The author of this summary is the project coordinator, who is responsible for the content of this summary. The ANR declines any responsibility as for its contents.

Partner

SPHERE Université de Nantes - EA4275-SPHERE

Help of the ANR 276,567 euros
Beginning and duration of the scientific project: January 2016 - 48 Months

Useful links

Explorez notre base de projets financés

 

 

ANR makes available its datasets on funded projects, click here to find more.

Sign up for the latest news:
Subscribe to our newsletter