Interpretation | Risk prediction models for post-intensive care syndrome of ICU discharged patients: A systematic review.

Post Intensive Care Syndrome (PICS) is a new or worsening cognitive, physical and/or mental health disorder that occurs after a critical care survivor is discharged from the ICU. According to research, more than half of ICU survivors continue to experience at least one domain of PICS symptoms or dysfunction within two years of discharge. These impairments not only severely impact the health-related quality of life of patients and their caregivers, but may also lead to recurrent hospitalizations and death. Therefore, early identification of those at risk for PICS and targeted interventions are critical.

Based on this, Fang Fang et al. from Shanghai Jiao Tong University School of Medicine published an article titled “Risk prediction models for post-intensive care syndrome of ICU discharged patients: A systematic review” in the International Journal of Nursing Sciences, conducting a systematic evaluation of risk prediction models for adult critical care survivors with PICS. The aim was to assess the characteristics and feasibility of existing models and provide guidance for clinical practice.

The study followed the PRISMA guidelines and systematically searched 10 Chinese and English databases, including Cochrane Library, PubMed, Embase, and CNKI, by November 1, 2023, combining Mesh terms with keywords and manually tracing references to the included literature. After screening, 16 studies were finally included, covering cognitive, psychological, physiological and multi-domain prediction models of PICS outcomes. The studies were assessed for methodological quality using the Predictive Model Risk of Bias Assessment Tool (PROBAST), and the extracts included sample source, definition of PICS outcomes, and predictors, etc. The 16 studies were published in the last decade, and the geographic distribution covered Europe, China, the United States, Australia, Japan, and Iran. The studies were designed to be prospective, most of them were multicenter studies, and the patients were mainly from comprehensive ICUs.Except for one study that aimed to improve an existing model, all of them focused on the development and validation of new models.

PICS outcome assessment tools in the study were varied, with the Montreal Cognitive Assessment and the Brief Mental State Examination commonly used for cognitive impairment, the Hospital Anxiety and Depression Scale and the Revised Impact of Events Scale for psychological impairment, and the Barthel Index and the Katz Activities of Daily Living Index for physical impairment.The prevalence of PICS ranged from 19% to 58.9%, with a follow-up time spanning 3 to 72 months. Assessment time points ranged from 1 day to 6 months after hospital discharge. Model discriminatory power ranged from 0.68 to 0.90 in the derivation phase and 0.68 to 0.88 in the validation phase, with the majority of models demonstrating good to excellent predictive accuracy. For calibration assessment, seven studies used the Hosmer-Lemeshow test, with some models showing good agreement between predicted and actual outcomes, but overall calibration information was not adequately reported.

The predictors mainly covered demographic characteristics, pre-ICU functional status, ICU experience and early onset of new symptoms. Among them, delirium and age were the core predictors in the cognitive impairment model; pre-ICU psychological problems and intra-ICU depressive symptoms were mostly seen in the psychological impairment model; and age and motor dysfunction at discharge were the main predictors in the physical impairment model. Regarding methodological quality, only 1 study had a low risk of bias and 14 were at high risk.

Existing models, although excellent in discriminatory power, generally suffer from methodological shortcomings. In addition, patient self-reported subjective bias may interfere with outcome accuracy and needs to be optimized in conjunction with objective assessment tools. Predictor analyses showed that age and baseline functional status, as immutable factors, could be used for early risk stratification; whereas adverse experiences in the ICU and early onset of new symptoms were intervenable, suggesting that clinical management of delirium, early activity, and psychosocial support during the ICU should be strengthened to reduce the risk of developing PICS.