Sepsis is a dominating cause of mortality with 11 million global deaths annually [1]. Notably, sepsis is closely linked with a multi-faceted health-related complication known as post-intensive care syndrome (PICS) [2,3], distinguished by physical impairment, a decline in quality of life, and posttraumatic stress syndrome [[4], [5], [6]]. Despite the provision of multidisciplinary interventions to aid patients with sepsis in regaining their prior functional status [[7], [8], [9], [10]], the incidence of physical impairment at hospital discharge remains high up to 61.4% [4,[11], [12], [13]]. Physical impairment at hospital discharge emerges as a composite outcome of sepsis sequalae and a critical determinant of discharge disposition; thus, emphasizing the urgency to reduce the risk of post-sepsis physical impairment.

Previous research has revealed that ICU-acquired weakness is a primary contributor to physical impairment at hospital discharge among critically ill patients [12,13]. ICU-acquired weakness is significantly associated with disease severity and the administration of aminoglycoside antibiotics and steroids [[14], [15], [16]]. While these earlier studies have identified the risk factors for physical impairment, unexplored risk factors for physical impairment may exist, and the methodologies employed in these studies may not effectively measure the association between pre-existing comorbidities and post-sepsis physical impairment.

In cases of sepsis, comorbidities prior to ICU admission and ICU-acquired weakness are associated with a higher risk of in-hospital mortality [[17], [18], [19]]; however, no preceding study has examined the association between pre-existing comorbidities and physical impairment at hospital discharge. Moreover, traditional studies investigating physical impairment in critically ill patients have excluded those with pre-existing physical impairment or those who died in the hospital [4,11,13,20]. These exclusion criteria induce a selection bias, underestimating the association between pre-existing comorbidities and physical impairment at hospital discharge. An evaluation of pre-existing comorbidities, while adjusting for these selection biases, would offer a more accurate depiction of the relationship between pre-existing comorbidities and post-sepsis physical impairment at hospital discharge. Such an analysis could aid in populations vulnerable to physical impairment following sepsis, thereby enhancing rehabilitation interventions in critical care.

Inverse Probability Attrition Weighting (IPAW) is extensively used to adjust the selection bias to estimate the association to the outcome by balancing the excluded populations using propensity score (PS) [21]. IPAW is relevant to adjust for selection bias when an association between physical impairment and death following sepsis are not ordinal. To estimate PS, it has been suggested that non-linear machine learning algorithms showed higher performance in mitigating selection bias compared with a conventional linear model, such as logistic regression model [[22], [23], [24], [25]].

In this context, we explored the hypothesis that pre-existing comorbidities prior to ICU admission were associated with physical impairment at hospital discharge among adult sepsis patient after adjusting for potential bias factors, such as pre-existing physical impairment before ICU admission and in-hospital mortality. We further evaluated the relationship between specific comorbidity subgroups and physical impairment upon hospital discharge.