1,720,967 research outputs found
Impulsive Compulsive Behaviour-related Harm in Older Adults: Rethinking Our Approach to Assessment
No full textImpulsive compulsive behaviours (ICB) are prevalent amongst older adults diagnosed with Parkinson’s, typically reported by clinicians and researchers as problematic gambling, compulsive eating, compulsive shopping and hypersexuality. ICBs are characterized as overwhelming urges to repeatedly participate in activities. Acquiescing to the impulse or compulsion provides temporary relief from the pressure of resisting the urge, but over the longer term, the consequences of the behaviour cause harm or distress to the individual and or others. The overarching objective of this study is to identify predictors of ICB-harm using path analysis and machine learning models.
We are examining two main hypotheses in this study.
Hypothesis 1
The current method of assessing ICBs only provides patients with a limited list of seven phenotypes, which may result in an incomplete understanding of the range and burden of ICBs in Parkinson’s.
The following three outcome measures will examine the prediction that the full range of ICB phenotypes is hidden, and expanding the list of phenotypes from 7 to 35 will reveal a much greater range of presentations than has hitherto been reported.
i. The Questionnaire of Impulsive Compulsive Behaviours in Parkinson’s (QUIP-rs);
ii. The Impulsive Compulsive Behaviours Checklist (34 phenotypes listed);
iii. The Internet Addiction test.
Hypothesis 2
The psychological signature of ICB will differ depending on the choice of the construct (symptom severity or harm), but the predictors will overlap to a greater or less extent.
If defined in terms of symptoms (e.g., urges, frequency, attempts to cut-down), cue sensitivity and self-reported distress will be critical when not engaged in the activity. On the other hand, critical correlates of harm will include denial and disengagement; if insight is present, then guilt and shame will be core features.
We suspect a wide range of predictors have only selected a subset here: personality traits (sensation-seeking, e.g., low boredom threshold for example, and pleasure-seeking), thinking styles (a tendency obsessive-compulsive), negative beliefs about the self (I am weak) traits, state negative affect may act as nonspecific predictors. Mediators may include a range of ‘constructs’ linked to impaired self-directedness (deceased premeditation, decreased perseverance, decreased ability to learn from mistakes and motivation to instigate change, self-distraction).
i. The QUIP-rs
ii. The Short (Gambling) Harm screen (adapted for ICB by the researchers).
iii. The Experience of Compensation Gratification
iv. The Temptation and Restraint Inventory
v. Difficulties in Emotional Regulation Scale
vi. The Self-regulation Scale
vii. Coping Orientation to Problems Experienced Inventory
viii. Apathy Motivation Index
ix. Urgency, Perseverance, Premeditation and Sensation Seeking Scale
x. The Yale-Brown Obsessive Compulsive Scale.
Design
This is a multicentre, single-cohort survey of older adults with or without a Parkinson’s diagnosis.
Participants
People from the general and clinical populations with a higher likelihood of developing ICB, including Parkinson's, will receive an invitation to participate. Individuals from different countries are welcome to participate, but NHS sites in the United Kingdom will not be included.
Inclusion criteria
To ensure a diverse sample, we have kept the inclusion criteria as broad as possible. We only ask that participants are 40 years old or older and have internet access to complete the survey.
Sample size
To conduct path analysis, Kline suggests a sample size of 20 times the number of psychological variables. Since our study involves 11 variables, our target sample size is 220 participants.
Data management
We will be using multiple imputations to deal with any missing values. To protect the participant’s confidentiality, we will store their information in a secure database with password protection to protect the participants' privacy; every participant will receive a unique code different from Qualtrics’ IP identifier. This method maintains the study’s accuracy and reduces privacy risks by making it challenging to identify individuals from whom the data was gathered. However, it is essential to note that this method does not guarantee the complete anonymity of the dataset.
Analytic strategy
The questionnaire data will be normalised within each sample to minimise bias. Normalisation involves subtracting the mean of all data points from each data point.
First, we will estimate associations between the data from each questionnaire through bivariate Pearson correlation coefficients (r). |r| > 0.10–0.24 will be considered a small effect size, |r| > 0.24–0.37 a moderate effect size and |r| > 0.37 a large effect size (71).
Second, direct and indirect relations between predictors (X), and outcome variables (Y), including mediation links= (M), will be evaluated through path analysis.
We will use the maximum-likelihood estimation method of parameter estimation and evaluating goodness-of-fit through standard statistical measures [including the root mean square error of approximation (RMSEA), Bentler’s Comparative Fit Index (CFI), the Tucker-Lewis Index (TLI), and the standardised root mean square residual (SRMR)] (73). Good model fit will be considered non-significant by chi-square (χ2) tests if the following criteria were met (73): RMSEA <0.08, TLI > 0.9, CFI > 0.9, and SRMR <0.1. We will measure the model’s global predictive capacity using the coefficient of determination (CD).
Network plots will be used for visualisation purposes. Strongly correlated variables, either positively or negatively, will be linked more closely and with more robust paths. Any paths that are not significant will be eliminated. Paths will also be coloured by their sign (red for positive and blue for negative). Finally, ellipses will be drawn around the sub-scores included in a factor. We will only include variables with rho coefficients >0.3 will be included in the exploratory factor analysis.
Finally, linear regression models will estimate the amount of variation explained/captured by the developed model. The Coefficient of Determination or R-Squared (R2) always ranges between 0 & 1. Overall, the higher the R-squared value, the better the model fits the data. To determine the accuracy of the model fit, we will analyse F values and significance levels and the Root Mean Squared Error. The RMSE specifies the absolute fit of the model to the data, i.e. how close the observed data points are to the predicted values. Additionally, we will use standardised betas to evaluate the predictive value of each independent variable
- …
