AB0931 Characteristics associated with the perception of high-impact disease (PsAID ≥4) in patients with recent-onset psoriatic arthritis. Model based on machine learning

Abstract

BackgroundThe Psoriatic Arthritis Impact of Disease (PsAID) questionnaire is the standard tool for evaluating the impact of psoriatic arthritis (PsA) on quality of life [1]. Variables associated with high disease impact were studied in patients with long-standing established disease. The characteristics associated with high-impact PsAID in recent-onset PsA remain unknown.ObjectivesTo evaluate which patient and disease characteristics are associated with the perception of high-impact disease (PsAID ≥4) in recent-onset PsA.All patients gave their informed consent. The study was approved by the Clinical Research Ethics Committee of the Principality of Asturias.We conducted a cross-sectional analysis. The dataset was generated using data for the independent variables at the 3 visits (baseline, first year, and second year of follow-up) matched with the PsAID values at each of the 3 visits. PsAID was categorized into two groups, namely, <4 and ≥4 [1]. We trained logistic regression models and a random forest–type machine learning algorithm to analyze the association between the outcome measure and the variables selected in the bivariate analysis (statistical significance defined as p value <0.05). We used a confusion matrix to visualize the performance of the final model. This matrix shows the real class of the data items, together with the class predicted by the machine learning algorithm, and records the number of hits and misses.ResultsThe sample comprised 158 patients. 20.9% were lost to follow-up. Of the patients who attended the clinic, 45.8% scored PsAID ≥4 at baseline; 27.1%, at the first follow-up visit, and 23.0%, at the second follow-up visit. The variables associated with PsAID ≥4 selected in the logistic regression analysis were HAQ, patient global pain during the previous week, educational level, and level of physical activity in the previous week. The association was positive for the first 2 variables and for level of physical activity and negative for educational level. When physical activity was introduced as a categorical variable, a possible negative association was observed for a moderate level (although this was not statistically significant) and a positive association was observed for a high level (Table 1).Table 1.Variables associated with PsAID ≥4: Logistic regression analysis.VariableRegression coefficient95% CIp value (Wald test)HAQ10.394[7.777, 13.011]<0.001Patient global pain in the previous week5.668[4.016, 7.320]<0.001Educational level-2.064[-3.515, -0.613]0.005Moderate level of physical activity in the previous week-0.341[-1.255, 0.573]0.465High level of physical activity in the previous week1.221[0.158, 2.283]0.024When the random forest–type machine learning algorithm was trained with these 4 variables, the order of importance (from more to less) attributed by the model was: patient global pain, HAQ, educational level, and physical activity. The percentage of hits in the confusion matrix was 86.14%.ConclusionPain control and control of the disease as a whole, preventing patients from suffering a decrease in their functional capacity, are first-order treatment objectives. PsA patients should take regular physical exercise, but with a moderate or low impact on their joints and entheses.