Sepsis after Scotland: enough with the averages, show us the effect modifiers

Cuthbertson and colleagues demonstrate that survivors of severe sepsis face, in general continued ongoing high mortality and quite poor quality of life. This cohort caps the initial, problem-definition stage on long-term outcomes after critical illness. Having compellingly demonstrated the opportunities for improvements in outcomes, epidemiologic and behavioral research must now to turn to understanding the mechanisms by which these outcomes can be improved. Such fundamental research will provide the evidence base to drive informed and successful interventional trials.

Such fi ndings can seem bleak and lead to an easy nihilism about the futility of intensive care unit care. Our patients seem to think otherwise. One hundred percent of the Scottish survivors -the same folks with all those problems -say they would go through intensive care again [1]. Th is suggests that we are at a critical intellectual turning point in the fi eld. We have clearly defi ned that there is a litany of adverse outcomes among survivors of critical illness. We have not yet discovered how to make this better.
Th e need here is not merely for additional clinical trials. New clinical trials are unlikely to succeed until we have additional rigorous epidemiologic and behavioral research to understand the mechanisms by which some patients do better and other patients do worse.
Specifi cally, we desperately need to understand the eff ect modifi ers of the disablement process as it leads to key long-term outcomes after critical illness. Figure 1 presents the disablement process model [9], modifi ed to explicitly consider an acute disruption [10][11][12]. Baseline status -composed of a patient's physiology, activity limita tions (diffi culty performing standardized tasks in standardized environments such as 6-minute walks), and social disability (diffi culty participating in social roles in one's usual environment) -is infl uenced by an acute insult, such as severe sepsis, sequentially leading to tissue pathology, activity limitations, and participation restrictions. Th e Scottish group goes even further, extending the process to quality of life.
Clinical practice has long taught us that there is no linear correlation between diff erent steps of the disablement process. Neither ejection fraction nor FEV1 (forced expiratory volume in 1 second) is suffi cient to explain which patients with heart failure or chronic obstructive pulmonary disease will be debilitated; although both play a role, clearly there are other factors that infl uence the development and magnitude of disability.
Current research to date has defi ned this new problem space and demonstrated this problem's population impact along the main tissue pathology/activity limitations/

Abstract
Cuthbertson and colleagues demonstrate that survivors of severe sepsis face, in general continued ongoing high mortality and quite poor quality of life. This cohort caps the initial, problem-defi nition stage on long-term outcomes after critical illness. Having compellingly demonstrated the opportunities for improvements in outcomes, epidemiologic and behavioral research must now to turn to understanding the mechanisms by which these outcomes can be improved. Such fundamental research will provide the evidence base to drive informed and successful interventional trials. disability/quality of life axis of Figure 1. Increasingly, this work has focused on cognitive, psychiatric, and muscular problems as the physiologic basis underlying the longterm problems. Research into fi rst-generation long-term outcomes has defi ned the frequencies of diff erent sorts of problems among survivors. Th e conceptual model of Figure 1 suggests that there are at least two ripe domains for a next generation of research. We should explore variation in patient characteristics and hospitalization characteristics that lead to diff erences in subsequent outcomes [8,13]. But we also need to understand those eff ect modifi ers -particularly those external to the patient -which change the conditional probabilities of an outcome: that is, what are those characteristics which explain why patients with the same physiology have diff erent disability? Such data help us explore not only prevention of adverse long-term outcomes but harm mitigation of whatever irreversible physiologic damage may be done.
Th ere is an almost overwhelming variety of potential eff ect modifi ers one might consider in next-generation long-term outcome studies. A few seem particularly promising. Th e fi rst is to look at the impact of social support on recovery [14]. We know that families play an essential but challenging role in decision-making in the intensive care unit [15,16]. But we have much less evidence to test the plausible and actionable hypothesis that patients discharged into high-social support environments make a more rapid recovery.
Also, we have not taken patients and families seriously enough as actors and partners in their own care. Th e social science jargon for this would be that we have neglected the agency of patients and families. Rather than waiting passively for assistance from medical professionals, survivors are experimenting with new ways to heal and adapt. How can we learn from those survivors and facilitate their collective wisdom and resilience? How can we crowd-source recovery?
Cuthbertson and colleagues have confi rmed the substantial burdens under which survivors labor and have emphasized the resilience and capacity of those survivors. Th e authors have set the stage for a next generation of research that moves beyond these average eff ectsessential though they are in defi ning the problem -to leverage the variation in patients' recovery to discover new ways to improve it.