Critical care resource allocation: trying to PREEDICCT outcomes without a crystal ball

Despite pandemic influenza's long reign atop the list of potential medical catastrophes, the first protocol designed to support critical care triage in a pandemic was published only in 2006. InFACT (the International Forum of Acute Care Trialists) was formed in 2009 and provided a platform for international critical care research collaboration during the 2009-2010 Influenza A(H1N1) pandemic. Over the past 2 years, a number of working groups have emerged from InFACT focused upon improving the investigation and care of patients with severe respiratory illness. Arising from these efforts, in June of 2012, an international group of clinicians convened the first meeting of the PREEDICCT (Providing Resources for Effective and Ethical Decisions In Critical Care Triage) study group. The group's aim is to develop decision support tools appropriate for use in triaging critically ill adult patients during epidemics, mass-casualty scenarios or other resource limited settings. This meeting identified a number of knowledge gaps and research priorities in this area, and suggested a revised framework for the requirements of an adequate triage decision support tool.

Despite pandemic infl uenza's long reign atop the list of potential medical catastrophes, the fi rst protocol designed to support critical care triage in a pandemic was published only in 2006 [1]. Additional protocols followed, in attempts to address the goal of developing stan dardized, transparent and equitable tools for allocating critical care resources to those patients most likely to benefi t [2][3][4][5][6][7]. Most of these protocols used the Sequential Organ Failure Assessment score as the quantitative underpinning for triage decision-making due to its ease of use. Th ese protocols have been shown to generally direct resources to those most likely to benefi t [8], in addition to making resources available for surge patients [9]. However, the Sequential Organ Failure Assessment score does not always diff erentiate well between survivors and nonsurvivors of critical illness for some patient populations [10,11].
Th e International Forum of Acute Care Trialists (InFACT) was formed in 2009 and provided a platform for international critical care research collaboration during the 2009/10 infl uenza A(H1N1) pandemic [12]. Over the past 2 years, a number of working groups have emerged from InFACT focused upon improving the investigation and care of patients with severe respiratory illness. Arising from these eff orts, in June 2012 an international group of clinicians convened the fi rst meeting of the Providing Resources for Eff ective and Ethical Decisions In Critical Care Triage (PREEDICCT) Study Group. Th e study group's aim is to develop decision support tools appropriate for triaging critically ill adult patients during epidemics, mass-casualty scenarios or other resource-limited settings. Th is meeting identifi ed a number of knowledge gaps and research priorities in this area, and suggested a revised framework for the requirements of an adequate triage decision support tool.
While purpose-built triage protocols focus on specifi c events (for example, pandemics), resource allocation decisions are part of everyday practice for critical care physicians worldwide. Several PREEDICCT members work in settings where there are chronically insuffi cient critical care resources to meet the demand [13]. Critical care physicians also make resource allocation decisions every day in high-income countries, as they decide who might benefi t from ICU care, when to accept outside transfers and when insuffi cient capacity dictates external transfer of patients. Yet intensivists lack objective tools to support these decision-making processes. Further, practices and specifi c decisions are likely to vary widely by country, by hospital and by individual provider.
Th e fi rst signifi cant shift in direction advocated by our group is to move away from attempting to use a physiologic score alone to predict outcomes. Th e rationale for basing triage tools on a physiologic score is that all critically ill patients compete for a single pool of critical care resources, regardless of whether they are part of the mass-casualty event or not [1]. However, there are at least two ways to compare diff erent types of patients. Th e fi rst method is to use the same tool to measure all patients, such as with a physiological prediction score (for example, the Sequential Organ Failure Assessment, Multiple Organ Dysfunction Score or Acute Physiology and Chronic Health Evaluation scoring systems). Th e second approach is to use diff erent scores tailored to diff erent diseases (for example, a burn score for a burn patient) that all produce a standard measure which can be compared. Th e potential benefi t of using disease-specifi c scores, where available, is improved prognostica tion to overcome the defi ciencies identifi ed with general ized physiological scores.
For diff erent predictive scores to be used when making resource allocation decisions, the scoring tool must allow comparison across diff erent groups of patients based on a common metric. Th e PREEDICCT study group recommends that any such metric receive input from three important outcome dimensions: survival, quality of life, and resource consump tion. Additionally, it is important to recognize that it is not an absolute measure of these factors which is primarily important but rather the incremental diff erence in the measure made by the provision of critical care resources [14,15].
A critical care resource allocation decision support tool based upon a combination of disease-specifi c and general physiological measures with common outcomes should be deployable in a variety of environments, including resource-limited countries. Th is utility probably requires a technological solution that combines often-complex scores into a single tool that facilitates rapid decisionmaking. Th e inter face to facilitate this decision-making would use standardized categories of critical illness phenotypes (such as penetrating trauma, blunt trauma and pneu monia). Given the proliferation of mobile computing devices and the increasing presence of Internet access, we believe it will be feasible to create platform-independent software solutions that can meaningfully augment clinicians' capability to calculate and compare multiple variables in order to optimize utilization of potentially scarce critical care resources. Such tools will not supplant clinical decision-making; instead, they will provide additional data that clinicians can integrate with clinical experience to ensure that critical care admission is based on appropriate, clinically signifi cant factors.
Th e fi rst step in advancing this project will be to utilize the existing InFACT network in a truly global eff ort, involving both resource-rich and resource-poor countries, to better defi ne current triage practices in high-income, middle-income, and low-income countries. Factors of interest include the type and frequency of resources and the method with which resource allocation decisions are currently made. Second, PREEDICT will survey existing disease-specifi c predict ive scores -many of which presently only report a single endpoint, such as survivalto determine whether crosscutting surrogate markers may permit alignment and comparison across disease categories. Once candidate metrics are identifi ed and assembled into the tool, the team will use predictive modeling methodologies to forecast the impact of diff erent thresholds for critical care admission on patient outcomes and also on facility-level and regional capacity and functioning. To be maximally useful, these modeling eff orts will require richly descriptive data regarding health system functioning and patient out comes, the types of which are increasingly being captured in stateof-the-art electronic medical records as well as research databases and registries.
To provide best care to patients during pandemics, environmental disasters or, indeed, day-to-day operations in resource-challenged settings, the global community of acute and critical care clinicians must increasingly see the practice of critical care as caring for critically ill patients, not only caring for patients in an ICU. When there are imbalances in demand and capacity, we should look to each other to help with context-appropriate ways to right this balance. When current and future challenges dictate that patient triage must occur, we must strive to develop decision-making tools to provide the optimal balance of survival and quality of life with the resources available.

Abbreviations
InFACT, International Forum of Acute Care Trialists; PREEDICT, Providing Resources for Eff ective and Ethical Decisions In Critical Care Triage.
Competing interests DF has received grant funds and unrestricted educational funds from Novartis, GlaxoSmithKline and Sanofi -Pasteur vaccine divisions, all of which manufacture infl uenza vaccines. The other authors declare that they have no competing interests.