Type I Error Inflation in the Presence of a Ceiling Effect

Austin, Peter C.; Brunner, Lawrence J.

doi:10.1198/0003130031450

Cited by 113 publications

(95 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…9,10 It is heavily influenced by culture and institution and has a stronger correlation with patient expectations, subsequent patient experience, strong therapeutic relationships, 14,15 work activity, and procedural outcome rather than with professionalism or anesthetic outcome. 16,17 The discriminate utility of satisfaction as an indirect measure of quality of recovery is also limited by its lack of a uniform definition or assessment tool, rapid early recovery, 18,19 and an inconsistent relationship with traditional recovery markers (nociceptive complications). Thus, a disconnect remains between an institution's unidimensional performance indicators and patients' perceived multidimensional recovery.…”

Section: Patient Satisfactionmentioning

confidence: 99%

The importance of postoperative quality of recovery: influences, assessment, and clinical and prognostic implications

Bowyer

Royse

2015

Can J Anesth/J Can Anesth

View full text Add to dashboard Cite

Quality of recovery is a complex construct whose definition is influenced heavily by the opinions and biases of the individual patient, clinician, or institution. As a result, recovery assessment tools differ in their fundamental definitions of recovery, breadth, and assessment time frame. Accurate assessment of recovery is essential as suboptimal recovery has both economic and prognostic implications. Quality of care is often substituted as a surrogate at the institutional level for quality of recovery, but it is ideologically distinct from patients' perceived quality of care, recovery, and satisfaction. Recovery tools also differ in their assessment of recovery as a continuous vs dichotomous variable and in their focus at the group vs individual level. Ideally, recovery measures should assess outcomes in a simple dichotomous fashion and maintain relevancy by assessing in multiple domains at various time points. Assessment of recovery in a dichotomous fashion also has both clinical and research applications. It allows identification of suboptimal recovery at both individual and group levels, respectively, and when performed in real time, it allows the opportunity for timely targeted intervention specific to individual patients as well as for resource rationalization.Résumé La qualité de récupération est un construit complexe dont la définition est fortement influencée par les opinions et les biais du patient, du clinicien, voire de l'établissement. Dès lors, les outils d'évaluation de la récupération diffèrent de par la définition-même de la récupération, leur envergure, et le cadre temporel de l'évaluation. Une évaluation précise de la récupération est essentielle; en effet, une récupération sous-optimale a des implications tant économiques que pronostiques. Au niveau institutionnel, on utilise souvent la qualité des soins comme substitut de la qualité de récupération mais, d'un point de vue idéologique, ce concept se distingue de la qualité des soins, de la récupération et de la satisfaction telles que perçues par les patients. Les outils de récupération diffèrent également dans leur évaluation de la récupération en tant que variable continue ou dichotomique et dans leur emphase sur le groupe ou l'individu. Dans l'idéal, les mesures de la récupération devraient évaluer les pronostics de manière dichotomique simple et maintenir leur pertinence en évaluant plusieurs domaines à différents moments dans le temps. L'évaluation dichotomique de la récupération a également des applications cliniques et en recherche. Elle permet d'identifier une récupération sous-optimale tant au niveau du groupe que de l'individu. D'autre part, lorsque ce type d'évaluation est réalisé en temps réel, elle nous donne la possibilité d'intervenir d'une manière ciblée et opportune spécifique à chaque patient tout en rationnalisant les ressources.Author contributions Andrea Bowyer and Colin Royse contributed substantially to the conception and design of the manuscript and drafted the version to be published.

show abstract

Section: Patient Satisfactionmentioning

confidence: 99%

The importance of postoperative quality of recovery: influences, assessment, and clinical and prognostic implications

Bowyer

Royse

2015

Can J Anesth/J Can Anesth

View full text Add to dashboard Cite

show abstract

“…the pre-intervention measurement) may be restricted by the ceiling effect. Austin & Brunner 20 and Austin & Hoch 21 point out that most research into the impact of ceiling effects or censored data deal with the impact of censoring the dependent variable. Their research identifies the possibility that a censored independent variable may increase the likelihood of a Type I error (i.e.…”

Section: Literature Reviewmentioning

confidence: 99%

“…While these may be effective at minimizing the impact of the ceiling effect -like the maximum likelihood method does for an increase in Type 1 error -the statistical power is reduced by the technique, and it is recommended that data be collected that is not subject to ceiling effects for best results 20 .…”

Section: Literature Reviewmentioning

confidence: 99%

“…Various analytical techniques are available for dealing with ceiling effects, including Tobit models for censored dependent variables 14,20 , or a simple discard method and the maximum likelihood method for independent variables 20,21 . While these may be effective at minimizing the impact of the ceiling effect -like the maximum likelihood method does for an increase in Type 1 error -the statistical power is reduced by the technique, and it is recommended that data be collected that is not subject to ceiling effects for best results 20 .…”

Section: Literature Reviewmentioning

confidence: 99%

See 1 more Smart Citation

Spatial Ability Instrument Ceiling Effect and Implications

Call¹,

Goodridge²,

Sweeten³

2016 ASEE Annual Conference &Amp; Exposition Proceedings

View full text Add to dashboard Cite

This research paper investigates the potential existence of and implications for a ceiling effect observed in sophomore engineering students' spatial ability scores when using a common spatial ability instrument. Repeated use of the Purdue Spatial Visualization Test: Visualization of Rotations (PSVT:R) -shown herein over the course of two semesters -has revealed potential limitations when using the assessment with undergraduate engineering students during their sophomore year. The correlation between spatial ability and academic performance in engineering education has been thoroughly established. The PSVT:R and its revision are commonly used in academic spatial ability research. However, with the observed high average performance typical of engineering students on the PSVT:R, a ceiling effect may pose limitations to its utility. Sophomore engineering students in a Statics class -the first class in the Engineering Mechanics series -were each given the PSVT:R and Mental Cutting Test (MCT) assessments twice per semester. Results showing that the MCT may be more capable of differentiating student abilities, despite having a lower possible maximum score, are presented. Scores from similarly aged students in an Anatomy class are provided for comparison. The impact of ceiling effects for the education of high-performing populations, such as Engineering Mechanics students, will be discussed and actions for improvements in spatial ability measurement will be proposed. An argument is also put forth to understand how these tests relate to students' engineering capabilities.

show abstract

“…Using a variable suffering from either one as a covariate in ordinary (and generalized) linear regression model can results in bias and false findings if the covariate is correlated with another 'independent' (predictor) variable (Austin & Brunner 2003;Brunner & Austin 2009;Carroll et al 2006). These problems could be addressed, along with those arising from more familiar overfitting problem (Babyak 2004), from the point of view of combined cross-validation and partial least squares regression (Abdi 2010;Rosipal & Kramer 2006).…”

Section: Statistical Problemsmentioning

confidence: 99%