Prediction model for sleep disorders in patients transferred out of ICU: Development and validation

Abstract Background Sleep disturbance is one of the clinical manifestations of post‐intensive care syndrome (PICS) in ICU. Sleep disorders can cause changes in patients' emotional, cognitive, memory, immune, and motor systems, resulting in delayed wound healing, increased infection rate, readmission rate, mortality rate and complication rate. However, developing prediction models for sleep disorders in post‐ICU patients is an understudied problem. Aim To explore the risk factors of sleep disorders, establish the prediction model, and verify its prediction efficiency internally and externally, providing a scientific basis for clinical staff to prevent the occurrence of sleep disorder in patients transferred out of ICU in the early stages. Study Design A total of 405 patients transferred from the intensive care department of the Affiliated Hospital of Nantong University in China from May 2022 to December 2022 were selected as the study subjects by convenience sampling method and were divided into a modelling group of 270 patients and an internal verification group of 135 patients. A total of 67 ICU patients admitted to the same hospital from July 2023 to September 2023 were selected as the external validation group. General data and clinical data questionnaires were used to collect information on the influencing factors. The Pittsburgh Sleep Quality Index (PSQI) was used for follow‐up 2 weeks after ICU transfer. According to the follow‐up results, the patients were divided into a non‐sleep disorder group and a sleep disorder group. Univariate analysis was used to analyse the risk factors of sleep disorders in ICU patients. To avoid multicollinearity, LASSO regression was used to filter variables. Through binary logistic regression, the forward step method and likelihood ratio test were selected to further screen the variables. R language was used to establish a riskprediction model and draw a column graph. Receiver operating characteristic (ROC) and Hosmer‐Lemeshow (H‐L) tests were used to verify the prediction effect of the model. Results The influencing factors for sleep disorders in patients transferred out of ICU were pre‐hospital sleep disturbance [Odds Ratio: 4.467, 95% CI (1.191–16.749), p = .026], APACHE II score ≥15 [Odds Ratio: 6.452, 95% CI (1.777–23.434), p = .005], moderate comorbidities [Odds Ratio: 18.045, 95% CI (1.568–66.731), p = .015], severe comorbidities [Odds Ratio:12.083, 95% CI (2.785–116.911), p = .002], remifentanil use [Odds Ratio: 12.083, 95% CI (2.716–53.756), p = .001], RCSQ total score <45 [Odds Ratio: 18.037, 95% CI (4.907–66.300), p < .001], moderate depression [Odds Ratio: 70.659, 95% CI (8.195–609.219), p < .001] and severe depression [Odds Ratio: 8.563, 95% CI (1.165–62.936), p = .035]. The prediction model was as follows: Logit (P) = −10.529 + 1.497* (pre‐hospital sleep disorder) +1.864* (APACHE II score ≥15) +2.325* (moderate complication) +2.893* (severe complication) +2.492* (remifentanil use) +2.892* (RCSQ) Total score <45 +0.574* (mild depression) +4.258* (moderate depression) +2.147* (major depression). The area under the ROC curve of the prediction model was 0.916, the sensitivity was 81.9%, and the specificity was 96.0%. The H‐L test showed that χ 2 = 4.301, p = .829 ( p > .05). The area under the internal verified ROC curve (AUC) was 0.896, and the H‐L test revealed that χ 2 = 3.683 and p = .885 ( p > .05). The area under the external verified ROC curve was 0.739, the sensitivity was 72.7%, and the specificity was 64.7%. The H‐L test results showed that χ 2 = 4.683, p = .699 ( p > .05), indicating that the model had a good prediction effect. Conclusions The risk histogram of sleep disorders in ICU patients can predict the risk of sleep disorders in ICU patients, and can be used to assess the high risk of sleep disorders in ICU patients, and can help nurses to formulate corresponding intervention measures. Relevance to Clinical Practice The dynamic nomogram can be used to systematically monitor various factors associated with sleep disorders, and the prevention of sleep disorders can improve outcomes and quality of life for patients discharged from the ICU. Furthermore, nurses need to develop and accurately apply nursing interventions, taking into account all relevant variables, thereby reducing the occurrence of sleep disorders.