Sepsis is defined as life-threatening organ dysfunction caused by a dysregulated host response to infection. ^, Survival after sepsis has improved over recent decades, resulting in a growing population of sepsis survivors. ^, In the United States, sepsis accounts for an estimated 1.7 million adult hospitalizations annually, of which approximately 350,000 result in death in hospital or discharge to hospice, underscoring the substantial burden of organ injury. Among intensive care unit (ICU) treated patients with sepsis, mortality remains high, and long-term recovery among survivors is variable. Survivorship after sepsis is frequently marked by significant long-term morbidity, including new functional dependence, cognitive impairment, anxiety, depression, post-traumatic stress symptoms, recurrent hospitalization, and excess long-term mortality. ^,,, In a recent prospective cohort of patients admitted to the ICU for sepsis, 38.7% of patients died before ICU discharge and 47.4% died before hospital discharge. Among ICU survivors, approximately 25% remained functionally dependent, 30% regained functional independence, and 45% had died by 3 years.

Historically, sepsis research has focused predominantly on short-term outcomes, and much of the literature has centered on improving survival through acute care processes, such as timely antimicrobial administration, source control, hemodynamic resuscitation, treatment duration, and antibiotic de-escalation, with success measured primarily in terms of survival. Consistent with this, landmark interventional trials in sepsis have typically used in-hospital 28 day or 90 day mortality as their primary endpoint rather than longer-term patient-important outcomes, such as functional recovery or quality of life. ^,,,,, However, over the past decade, critical care and sepsis research have progressively shifted to enhanced risk stratification, phenotypic enrichment, prognostication, and granular patient-centered outcomes, recognizing the need to address heterogeneity to offer more tailored care and that survival alone is an incomplete and wholly inadequate measure of recovery. ^,,,,, This broader view of risk ascertainment and recovery is now reflected in contemporary guideline recommendations, which emphasize individualized post-sepsis treatment as an integral component of sepsis care.

The aim of this article, therefore, is to present the current knowledge on the epidemiology of long-term outcomes among adult ICU sepsis survivors, with particular emphasis on patient-important outcomes, their prevalence, and contributing factors, and to discuss the emerging role of biomarkers, immunologic profiling, and artificial intelligence (AI) in prognostication, individualized prediction, and post-sepsis care pathways.

ICU sepsis survivors represent a specific subgroup within the broader population of sepsis survivors. Sepsis may occur across a broad spectrum of severity, from patients treated for short periods of time in the community and on hospital wards to those requiring prolonged ICU stays with multiple organ support. Patients who survive sepsis after an ICU stay are therefore not only survivors of infection, but also survivors of critical illness, often exposed to mechanical ventilation, corticosteroids, immobility and deep coma related to paralytics and sedation, delirium, and prolonged hospitalization, each of which may contribute to long-term morbidity. ^,,,

From the perspective of patients and their families, the outcomes that matter most extend beyond survival alone. Recovery is often understood in terms of physical function, cognitive abilities, emotional well-being, social participation, return to work or prior activities, and the ability to regain an acceptable quality of life. ^, Family members may also experience psychological distress, caregiver burden, and disruption in social or professional roles. For this reason, survivorship after sepsis should be approached through a patient- and family-centered lens rather than focused on mortality alone.

Within the ICU literature, the long-term consequences of critical illness are commonly conceptualized under the framework of post-intensive care syndrome (PICS), which encompasses new or worsening physical, cognitive, and mental health impairment arising after critical illness. ^,,, In the sepsis literature, the term post-sepsis syndrome (PSS) is often used to describe this multidimensional burden observed after sepsis. ^, Although PICS and PSS are not synonymous, they overlap substantially. PSS includes most of the physical, cognitive, and psychological sequelae captured by PICS, but further encompasses sepsis-specific features such as recurrent infection, rehospitalization, and persistent organ or immune dysfunction. In a multicenter prospective cohort study, mechanically ventilated ICU patients with sepsis had a higher risk of death than mechanically ventilated ICU patients admitted for nonsepsis diagnoses. However, among survivors, 6 month outcomes were similar between groups, including new disability, health-related quality of life, mental health outcomes, cognitive function, and return to work. These findings suggest that many long-term sequelae of sepsis overlap with those of critical illness survivorship more broadly.

Importantly, recovery after ICU should be viewed as a trajectory rather than a fixed outcome assessed at a single time point. ^,, Patients do not all follow the same path: some recover rapidly, some remain persistently impaired, and others develop new morbidity or die over subsequent months and years. This dynamic nature of survivorship is reflected in studies showing heterogeneous trajectories, with disability and excess mortality persisting for up to 5 years or beyond after an ICU stay. ^,,,

The long-term sequelae of sepsis represent a systemic health care burden requiring longitudinal care models beyond the traditional critical care boundaries. ^, Sepsis short-term mortality rates have significantly improved, and this has created an expanding population of survivors living with multidimensional long-term morbidity. The mechanisms underlying these chronic sequelae are multifactorial and may include persistent inflammation, prolonged immune dysregulation, persistent organ injury, and the progression of pre-existing comorbid disease. ^,,

Functional disability represents one of the most prevalent domains of long-term morbidity among sepsis survivors. Recent meta-analytic data indicate that the prevalence of physical impairments exhibits significant heterogeneity, ranging from 25% to 80% across diverse patient populations. A primary contributor to this profound functional decline is ICU-acquired weakness, defined as weakness that cannot be explained by a cause other than critical illness. ICU-acquired weakness has been reported in approximately 55% of assessable long-stay ICU patients. Sepsis is itself an important risk factor for ICU-acquired weakness, alongside other reported risk factors including greater severity of illness, multiple organ failure, prolonged immobilization, prolonged mechanical ventilation, hyperglycemia, and older age. ^, The pathophysiology encompasses critical illness polyneuropathy, myosin-depletion myopathy, disuse atrophy, and evident as muscle wasting, which may appear as early as the first week of an ICU stay. Frailty, defined as a state of reduced physiologic reserve and increased vulnerability to stressors, is another important dimension of post-ICU functional disability. Baseline frailty is associated with worse ICU and post-ICU outcomes, including lower quality of life (QoL) and greater functional dependency, and frailty severity may further worsen during ICU hospitalization, with higher discharge frailty associated with poorer subsequent recovery. ^, Functional improvement is often greatest in the first 6 to 12 months after hospital discharge, but many patients do not fully regain their pre-illness functional baseline and may remain persistently disabled for years or indefinitely. ^, Longitudinal data suggest that severe sepsis survivors acquire approximately 1.5 new functional limitations after hospitalization, with these impairments persisting for years after the acute episode. Moreover, the persistence of muscular weakness at the time of ICU discharge is independently associated with significantly elevated 1 year mortality and increased health care costs. Lastly, survivors of sepsis may experience an increased long-term risk of cardiovascular events for up to 5 years after hospital discharge. In a recent meta-analysis, the pooled cumulative incidence of myocardial infarction, stroke, and heart failure in sepsis survivors was as high as 9%, potentially contributing to impaired functional recovery.

Cognitive impairment is an important and potentially disabling long-term sequela after critical illness, with up to 34% of patients presenting persistent cognitive impairment 1 year after an ICU stay. The severity of this acquired dysfunction is considerable: one-quarter of post-ICU patients develop cognitive profiles comparable to early-stage Alzheimer’s dementia, while 33% demonstrate severe neurocognitive profiles equivalent to those observed following traumatic brain injury. Importantly, survivors of severe sepsis experience a greater increase in moderate-to-severe cognitive impairment than patients admitted for general, nonseptic hospitalizations, with an absolute increase of 10.6% points. Implicated pathophysiological pathways include blood-brain barrier disruption, neuroinflammation, microvascular injury, neurotransmitter dysregulation, and hippocampal neuronal loss. In the BRAIN-ICU study, which included patients admitted to the ICU for any reason, a longer duration of delirium in the hospital was independently associated with worse global cognition and executive function at 3 and 12 months after discharge. In sepsis-specific studies, risk factors for poor cognitive outcomes included depressive symptoms, central nervous system infection, length of hospitalization due to infection, temporal proximity to the last period of infection, the time to first antibiotic dose, cumulative dose of haloperidol during the ICU stay, mean glycemia, severity scores on ICU admission scores, neuron-specific enolase, and IFN-γ levels at ICU discharge. ^, Although partial improvement may occur over the first year, these impairments often represent a major change, resulting in reduced autonomy and substantial caregiver dependence. ^,

The prevalence of psychological sequelae, including symptoms of anxiety, depression, and posttraumatic stress disorder (PTSD), is high after an ICU stay. In the year following critical illness, survivors have a median point prevalence of clinically significant depression of 28% to 33%, alongside substantial rates of PTSD and anxiety. ^, In a nation-wide population-based study, among 21,980 sepsis survivors, 54.8% were diagnosed with a new mental health issue in the 12-month postdischarge. Reported risk factors for post-ICU psychological sequelae include pre-existing psychiatric illness, longer duration of mechanical ventilation or ICU stay, delirium, benzodiazepine exposure, and distressing ICU memories. ^, An elevated longitudinal risk of both self-harm and completed suicide has also been documented among survivors. These psychological sequelae are often persistent, with up to one-third of patients still symptomatic at 12 months. They are associated with long-term deficits in both mental and physical health-related quality of life. Importantly, a recent study focused on PTSD symptoms found that a quarter of septic patients reported few symptoms at discharge, but an increase in symptoms in the subsequent 2 years, highlighting the importance of long-term follow-up.

QoL is consistently and significantly reduced among sepsis survivors. It has been reported that approximately one-third of septic survivors admitted to the ICU have not returned to independent living by 6 months, with frequent impairments in mobility, usual activities, and self-care. Successful return to work often serves in the literature as a general surrogate for broader functional recovery, and approximately two-fifths of ICU survivors remain unemployed 1 year after an ICU stay. Sepsis survivors frequently experience significant occupational disruption, manifesting as reduced working hours, involuntary career changes, and severe financial strain. A recent nationwide Norwegian registry study of 12,260 previously employed survivors highlighted this important socioeconomic morbidity; 2 years following discharge, successful return to work was achieved by only 64.2% of ward-managed sepsis patients and only 52.2% of those requiring ICU admission. The multidimensional disruption that persists after critical illness is particularly relevant for younger working-age survivors.

The sequelae of sepsis after critical illness are not isolated to the individual patient but also significantly affect their family members and caregivers. This has been described in the ICU literature as PICS-family. In most cases, caregiver support is provided by an informal family member, which may result in considerable psychological, social, and financial strain. In the Canadian RECOVER cohort, 67% of the caregivers reported depressive symptoms at the time of the discharge and 43% remained symptomatic at 1 year. In a French study from the FAMIREA group, PTSD symptoms were reported 3 months after an ICU stay in 33.1% of caregivers, reaching 81% among those who were involved in end-of-life decisions. New mental illness has been reported in approximately one-quarter of spouses or partners of patients with sepsis or septic shock. ^, Importantly, worse patient outcomes have been shown to be associated with greater caregiver burden and with reduced mental health-related QoL. Other reported risk factors for caregiver distress include younger patient age, history of mental illness, being a female, patient death and patient physical impairment. ^, Given this substantial and sustained burden, caregivers should be recognized as an essential component of post-sepsis care and integrated into the ICU care continuum. Moreover, as many patients with sepsis die during their hospital stay, caregivers and relatives of these patients should be considered. New mental health disorders, such as depression or generalized anxiety, have been reported in 34% of relatives of patients who died in the ICU, and complicated grief in up to 52% at 6 months after the event. ^,

Children of critically ill adults are also a vulnerable and often forgotten group. They are exposed to a disruption of family routine, altered parental relationship, and to the consequences of financial strain. Although direct data are limited, they may also be at risk of internalizing symptoms, school underperformance, and psychological distress. With an increase in patient comorbidity and complexity, a greater proportion of patients will sustain longer ICU stays and children may need specific additional psychological, social, and school support to navigate new physical and cognitive disability in a parent. This reality has not yet been broadly integrated into the design of current ICU follow-up care.

Health care workers are also affected by the care of these patients. Critically ill sepsis patients often require prolonged ICU stays and undergo numerous invasive interventions and procedures. When care appears futile, when the treatment plan becomes uncertain, or when a patient is later seen after discharge with major disabilities and questions arise about the appropriateness of the care provided, health care workers may experience distress. ^, This can manifest as care-related regret, for which perceived futility has been highlighted as a major driver, and is associated with burnout and intention to change profession. Although a direct link with patient long-term outcomes has not been clearly established, health care workers’ well-being is relevant to the ICU care continuum because the delivery of sustained, patient- and family-centered care depends on a workforce that is adequately supported. This is especially relevant in the post-COVID-19 context, marked by substantial staff attrition, workforce shortages, and persistently high burnout rates.

Sepsis is one of the most costly conditions treated in hospital, with a substantial economic burden reported in Europe and in the United States. ^, In a systematic review of 26 studies, the median sepsis costs per stay were €36,191. However, the economic burden of sepsis extends beyond the index admission and is driven by readmissions, recurrent hospitalization, ongoing rehabilitation needs related to persistent disabilities, and prolonged use of health care resources. ^, Recent data in the United States demonstrate that in the first year after discharge, the median total health care expenditure for sepsis survivors reached 28,719 USD, and these substantial costs continued into the second postdischarge year. Notably, the median cost associated with a single hospital readmission for this cohort was 20,320 USD. Resource utilization remains elevated for years, with sepsis survivors spending roughly 3 times more time in hospital or care facilities after the index episode than before their sepsis episode.

Long-term outcomes are the cumulative product of the interplay of multiple factors occurring across the trajectory of illness and recovery. These determinants can be broadly grouped into pre-existing patient vulnerability, severity, and management of the critical illness, and the quality of postdischarge recovery and follow-up.

Pre-existing health status is one of the strongest determinants of long-term outcomes. Older age, frailty, and burden of comorbid disease are consistently associated with mortality, poorer functional recovery, and lower health-related QoL after sepsis and critical illness. ^,, Among these factors, frailty and sarcopenia are particularly relevant, as they independently predict both short- and long-term mortality, persistent functional impairment, and reduced quality of life. ^,,,,

The severity of the acute sepsis episode is another main determinant. Critical illness severity, often captured by ICU severity scores, complicating shock, number of organ failures, and the need for invasive organ support including mechanical ventilation, renal replacement therapy, and vasopressors are all well-established risk factors for worse long-term outcome. Among sepsis-specific factors, timely recognition, early, appropriate antimicrobial therapy adhering to antimicrobial stewardship tenets, adequate hemodynamic resuscitation, and prompt source control remain the most actionable in-ICU determinants of prognosis. ^,, However, their association with longer-term patient-centered outcomes remains poorly studied. At the same time, prolonged or unnecessary broad-spectrum antimicrobial exposure may contribute to other complications, including the selection of multidrug-resistant organisms, gut microbiota dysbiosis and Clostridium difficile infections. ^, These downstream complications, in turn, may increase the likelihood of rehospitalization and greater long-term morbidity.

Critical illness due to sepsis does not end at ICU or hospital discharge; for many patients, recovery extends over months to years. This supports the concept of an ICU and sepsis care continuum, in which discharge represents a transition point rather than the end of care. Contemporary guidelines reflect this perspective. In the 2026 Surviving Sepsis Campaign, clinicians are advised to provide survivors and their families with written and verbal information about sepsis, treatments received, and common post-sepsis impairments, while health systems are advised to facilitate postdischarge assessment and follow-up for physical, cognitive, and emotional problems. The same guideline also suggests postcritical illness follow-up services for survivors of hospitalization with sepsis or septic shock, but acknowledges that the optimal format, timing, and intensity of follow-up remain uncertain. The transition from hospital to community is a vulnerable period during which multiple frequently neglected factors may shape long-term recovery. Medication reconciliation failures, underrecognized new impairments, the absence of structured follow-up, and fragmented communication with primary care represent the main challenges. ^, Access to rehabilitation might represent an important outcome modifiers: sepsis and septic shock survivors discharged to a rehabilitation facility had the best 5 year survival across severity strata, suggesting that structured postdischarge care may improve long-term recovery.

A first practical implication is to identify patients at highest risk of poor long-term outcomes as early as possible. As discussed earlier, older age, baseline frailty, comorbidity burden, prolonged ICU stay, multiorgan failure, prolonged mechanical ventilation, delirium, and severe physical impairment at discharge are consistently associated with worse recovery trajectories after critical illness and sepsis. ^,, These factors may help clinicians tailor information provided to patients and families, guide early discussions about likely recovery and potential sequelae, discuss goals of care, and flag high-risk survivors to primary care and rehabilitation services at the time of discharge.

A second implication is that some sequelae may be preventable, at least in part, through improvements in acute ICU care. Early mobilization has been associated with better physical recovery and, in a recent randomized trial, with lower long-term cognitive impairment 1 year after critical illness. Delirium—common in patients with sepsis and a known risk factor for subsequent cognitive dysfunction—may be reduced through multicomponent approaches such as the ABCDEF bundle and through careful minimization of deep sedation and benzodiazepine exposure. ^, More recently, higher ABCDEF bundle compliance has also been associated with less disability in instrumental activities of daily living and better health-related QoL at 12 months after an ICU stay. For mental health outcomes, family engagement, structured communication, and flexible visitation may improve the ICU experience and reduce distress, although evidence for durable long-term psychological benefit remains inconsistent. ICU diaries may also help reduce posttraumatic stress symptoms in patients and family members after critical illness, although their effects have not been consistent across studies. ^,

Across health systems, survivorship care pathways are highly heterogeneous. National initiatives in countries such as Australia and Switzerland have emphasized discharge planning, patient and family education, and improved transition from hospital to community care. ^, However, interventional studies of post-sepsis follow-up have produced mixed results. In the SMOOTH randomized trial, a structured primary care–focused aftercare program did not improve mental health–related QoL at 6 months after ICU discharge, although it informed subsequent work on sepsis aftercare. In the recent ENCOMPASS trial, a navigator-led telehealth Sepsis Transition and Recovery program did not reduce the composite of 90 day readmission or death, despite a lower mortality signal in a secondary analysis. A recent systematic review found that postacute interventions for sepsis survivors varied substantially in content, timing, and intensity, with inconsistent effects across outcomes. Rehabilitation-oriented strategies showed some promise, whereas evidence for bundled follow-up, care coordination, and mental health interventions remained limited.

Accordingly, future survivorship programs should move beyond a one-size-fits-all clinic model and instead use risk-stratified approaches that integrate discharge information, premorbid vulnerability, ICU exposures, and post-sepsis symptoms to guide referral toward rehabilitation, mental health support, medication review, or coordinated primary care follow-up. This approach is consistent with the 2026 Surviving Sepsis Campaign, which recommends physical rehabilitation services for selected survivors, mental health support after discharge, medication reconciliation at transitions of care, and improved communication with primary care and community services. Fig. 1 illustrates the trajectory of sepsis survivors and proposes a longitudinal, multidisciplinary model of post-sepsis survivorship care, integrating pre-existing vulnerability, acute critical illness exposures, postdischarge outcomes, and coordinated follow-up needs across the recovery trajectory.

Longitudinal determinants, consequences, and care model for post-sepsis recovery. The care continuum extends from pre-ICU status through the ICU stay, hospital ward, and rehabilitation, to community reintegration and return to previous social roles. Patient trajectories may be risk stratified and modified by factors arising at each stage, and further refined through biomarkers, immunophenotyping, and AI. The figure highlights outcomes at 3 levels: patients, family members/caregivers, and health care workers. Post-sepsis care includes family and peer support, specialist referral, cognitive and psychological support, patient and family education, rehabilitation, medication reconciliation, and coordination with primary care.

(Developed by the authors based on concepts described by Herridge and Azoulay. )

Many studies of sepsis still focus on short-term outcomes, most commonly 28 day or 90 day mortality, without adequately assessing long-term outcomes. Although 28 day mortality was long considered a pragmatic endpoint for evaluating acute treatment efficacy, it does not reflect meaningful recovery as studies have described persistent disabilities and excess mortality up to 5 years after an ICU stay. ^, As a result, studies restricted to short-term survival may substantially overestimate treatment success by failing to capture persistent disability and late death.

Another limitation is the inadequate characterization of patients before sepsis. Many studies summarize premorbid status using comorbidity counts or indices. While these are important predictors of poor long-term outcomes, they do not fully capture frailty, baseline functional status, or preadmission quality of life, which have all been highlighted as important determinants of long-term patient-centered outcomes. ^,,, Without this information, it can be difficult to determine whether impairments observed after sepsis are new or simply reflect pre-existing vulnerability. ^, Retrospective patient-reported or proxy-based assessments may partly address this issue, although both approaches have recognized limitations. ^,,

Long-term follow-up studies are also vulnerable to several major sources of bias. First, attrition bias, resulting from differential loss to follow-up, is common because follow-up is often incomplete and nonrandom, with patients who are either most impaired or fully recovered being less likely to remain in the study. ^, Second, death is a competing risk for many patient-centered outcomes, and treating death as simple censoring may lead to overestimation of recovery. ^, Third, analyses restricted to survivors are vulnerable to collider bias, a form of selection bias that can occur when analyses condition on survival, potentially creating misleading associations between exposures and long-term outcomes. ^,

These limitations have important implications for study design and analysis. Follow-up should describe the nonsurvivors and extend beyond the early postdischarge period and ideally include repeated assessments over time, since recovery trajectories differ across physical, cognitive, and psychological domains. Outcome standardization is also needed, as the use of numerous different instruments has limited comparability across studies. ^,, On the analytical side, methods that account for competing risks, nonrandom attrition, and heterogeneous recovery trajectories are likely to provide more valid estimates of long-term outcome than conventional complete-case or survivor-only approaches.

In septic ICU patients, research on biomarkers and immunophenotyping is growing; however, this has mainly been studied in relation to short-term mortality. Examples include inflammatory markers such as IL-6 and procalcitonin, endothelial markers such as proadrenomedullin, and immune markers such as monocytic HLA-DR, all of which have been explored as predictors of early severity and death. ^,, A useful biomarker for long-term outcomes would ideally serve 3 main functions.

First, early risk stratification: identifying which patients are most likely to develop poor long-term outcomes while there is still a window for intervention. NT-proBNP elevation within the first 24 hours, for example, has been associated with reduced physical function and muscle strength at 6 and 12 months. NSE and IFN-γ levels at ICU discharge have been associated with poor cognitive outcomes in septic patients. ^, Dynamic profiling of angiogenic, metabolic, and immune markers may also help identify patients at risk of developing PICS. ^,

Second, mechanistic bridging: linking acute-phase biological disturbances with the sequelae that emerge months or years later. Some studies have moved beyond short-term survival and suggest that persistent biological dysregulation may contribute to adverse long-term trajectories. A study found that abnormalities in the host immune response persisted for up to 1 year after sepsis and were associated with a higher risk of readmission and long-term mortality. Another study similarly reported prolonged immune disturbances 1 year after ICU discharge in septic shock survivors, while the FROG-ICU cohort linked sepsis-survivor heterogeneity to distinct biological profiles at ICU discharge and late adverse events. ^,, Together, these findings suggest that failure to restore biological homeostasis may contribute to later morbidity and may help identify therapeutic targets capable of altering recovery trajectories.

Third, therapeutic targeting: biomarkers may enable predictive enrichment by matching the right intervention to the right patient. This approach is already emerging, with biomarkers being explored to identify biologically distinct subgroups that may benefit from more targeted interventions. In a recent systematic review, biomarkers associated with successful enrichment strategies included suPAR, antithrombin, mHLA-DR, IL-6, and endotoxin activity assay.

Importantly, most studies have evaluated the predictive utility of biomarkers on short-term endpoints, typically 28 day mortality. Few studies have investigated the association between biomarkers and patient-important outcomes and further studies are needed.

Recent guidelines recommend postdischarge assessment and follow-up after sepsis or septic shock, but the evidence supporting specific follow-up models remains limited. One major challenge is that sepsis survivors are heterogeneous, with substantial variation in premorbid vulnerability, ICU exposures, organ dysfunction, and postdischarge trajectories. As a result, patients are unlikely to derive equal benefit from the same intervention delivered at the same time point. A second challenge is that existing survivorship interventions differ widely in timing, intensity, and content, ranging from global transition-of-care programs to rehabilitation-focused strategies and ICU follow-up clinics. Together, these observations argue against a one-size-fits-all model of post-sepsis care and support the need for individualized risk stratification.

AI may offer clinically useful tools to identify patients at highest risk of poor long-term outcomes and to guide more personalized follow-up. Several recent studies have developed prediction models for post-ICU or post-sepsis sequelae. A recent study using routinely collected data developed an interpretable machine-learning model to predict new functional impairment after critical illness, showing good discriminative performance with area under the ROC curve (AUROC) values up to 0.90. This illustrates the feasibility of early prediction of functional decline using easily accessed clinical data. Similarly, a recent study developed and temporally validated an electronic health record–based machine-learning model to predict stressor-related psychiatric disorders within 1 year after hospitalization with sepsis. In an independent test sample, patients in the highest-risk decile accounted for nearly one-third of subsequent psychiatric cases, suggesting that routinely available inpatient data may help identify survivors at risk for later psychological morbidity. For cognitive outcomes, a recently developed interpretable machine-learning model predicted post-sepsis cognitive impairment with good discrimination. In this study, key predictors included age, education level, septic shock, benzodiazepine exposure, illness severity, and IL-10. Similar work has extended prediction to family outcomes: a machine-learning model to predict PTSD in relatives of ICU patients has been developed, highlighting the possibility of stratifying risk not only for patients but also for family members.

At present, most evidence in this field stops at prediction. There remains little evidence that implementing AI-based risk models to guide post-sepsis follow-up improves long-term patient-centered outcomes. The next step will therefore be to improve model performance and external validation, and in a second step to test whether AI-informed survivorship pathways can meaningfully alter recovery trajectories after septic critical illness.

Survival after sepsis has improved significantly over time. Yet recovery extends far beyond survival alone. Survivors of sepsis-associated critical illness, together with their families, are forced to navigate a traumatic life event characterized by considerable physical, cognitive, psychological, social, and financial burden. ICU care should therefore be conceived as a continuum, aimed at mitigating the long-term burden borne by patients and families and improving outcomes that truly matter to them. This will require moving beyond short-term outcome measures in future trials, further refining risk stratification and leveraging machine learning and AI to better inform care pathways that optimally support survivorship and patient-centered recovery.