Integrating infection and sepsis management through holistic early warning systems and heuristic approaches: a concept proposal

Introduction

Sepsis is a major cause of morbidity and mortality in healthcare systems worldwide. It affects millions of people every year, and represents one of the most common reasons for admission to the emergency department (ED) and the intensive care unit (ICU). As with other medical emergencies, early recognition and appropriate management of patients with sepsis is associated with more favorable patient outcomes [1].

Unlike other medical emergencies, the decision-making and diagnostic process for septic patients is far from straightforward. Sepsis is defined as an aberrant or “dysregulated” host response to infections leading to the presence of organ dysfunction. There is no perfect tool to prompt a clinician to think sepsis: the challenge with sepsis is that it is a heterogeneous and dynamic syndrome caused by immune and metabolic dysfunctions, which involves different aspects temporally and in different components of the immune and metabolic systems [2]. It is common knowledge that, among patients with sepsis, mortality rates increase proportionally with a delay in administration of appropriate and adequate antimicrobial chemotherapy, which mandates early treatment [3].

The new ‘Sepsis-3’ definition extrapolates the diagnosis of sepsis from our previously slightly naïve concept of a stepwise evolving pattern. In doing so, however, we have placed the transition from infection toward sepsis in the domain of uncertainty and time-dependency, with the term ‘infection’ now including a broad spectrum of potential clinical outcomes. The International Classification of Diseases (ICD) coding in many healthcare systems, including in High Income Countries (HIC), is not yet aligned to the implications of the new definition for administrative data. It is likely that most patients with infection (some of whom will be very sick), and some patients with sepsis (therefore with organ failure) who, in some organizations, will continue to be classified according to the old definitions criteria, will be assigned to medical or surgical wards.

This now demands that clinical judgment be used in the risk stratification of patients with infection, and that pragmatic local solutions be used to prompt clinicians to evaluate formally for sepsis [4]. Moreover, to transform the reliability of recognition and management of sepsis in a healthcare system forces a dynamic pull-model collaboration and resource-sharing among sectors normally siloed within discrete areas of the organizational grid.

Despite the endeavors carried out to date, mortality for sepsis both in HIC and in middle and low income countries remains unacceptably high [5]. Among the reasons for this, the failures of healthcare systems to preventing, diagnose, rescue, and treat patients with sepsis, are considered in this article.

We see it as essential to propose the integration of three apparently separate workflows – which we will refer to as the pillars of infection management – dealing with infections and sepsis: infection prevention and control (IPC), infection prevention and control and surveillance (IPCS), antimicrobial stewardship (AMS), and rapid identification and management of sepsis (RIMS). They are ultimately addressing the same issues: the alteration of adaptive resonse leading to infection and, eventually, to sepsis; together with ensuring the continued availability of effective interventions.

Healthcare workers are embedded into systems which separate IPCS, AMS, and RIMS, despite these having a high degree of complementarity as yet only partially explored. Such issues become manifest at the organizational level and they engender dissonance in operative plans, communicative approaches, and ultimately in the mental models of health workers and patients: this creates ambiguity and not sustainable volatility. A patient may develop an infection at home, in a health care facility with poor IPC standards or a facility with good IPC standards; a patient may be colonized in one environment and develop infections at a different time and in a different environment. If the lenses the healthcare people use to anticipate, prevent and treat infections are not coherent it is unlikely that patients and citizens will understand the interrelationship between these pillars and become allies in the fight against infections.

Technological advances in biomarkers [6] and rapid molecular diagnostic tests ([RDTs]): ‘fast microbiology’ (FM)] are currently enhancing our capability to understand – as the care process are unfolding – the host responses and pathogen characteristics [7], [8], [9]. The recent approach in sepsis research are likely to demand healthcare to reconfiguring routinely available clinical data in order to retrieve sepsis phenotypes [10] and, in the next future, treat the endotypes [11].We feel it is necessary to prompt the development of a new generation of concepts and models providing clinicians with heuristics allowing the embracing of uncertainty under time constraints. In the present scenario heuristics should be based on “good-enough” relevant predictors likely to give raise to future AI based decision support systems. Nevertheless the opportunities of the coming technological advances demands today for a meaningful integration supported by a risk-based thinking and human factor/ergonomics (HF/E) approach [12].

Holistic early warning systems for infection management

Healthcare Early Warning Systems (HEWS) based on validated clinical scores, usually involving an aggregate of scores weighted against the degree of deviance of a parameter from the physiological norm, are considered the most effective means to effectively highlight patients at risk of deterioration (including from sepsis). HEWS aim to elicit the perception of both “hazards” and “vulnerabilities” in order to prompt assessment of risks and avoidance of harms. EWS trigger an organized non-routine reconfiguration of local or external resources in organizations. Indeed, many quality improvements carried out in healthcare have focused mainly on the timely identification of patients’ deterioration by means of EWS [13].

Pragmatically, we see an opportunity for EWS to gather, throughout the patient journey, information regarding alerts of deterioration and infectious hazards and vulnerabilities and warnings communication. This should foster knowledge construction, sense-making, decision-making, communication, and appropriate responses with reduced lag time (Table 1).

Risk knowledge

Risk is assessed through “the interaction of hazards and vulnerabilities.” [14]. An infectious hazard can be described as the presence of dangerous micro-organisms or those with activated resistance mechanisms (including community-acquired infections [CAIs] and hospital acquired infections [HAIs]). Vulnerability can be defined as a condition accounting for the degree of susceptibility to hazards determined, on the patient side by specific conditions such as co-morbidities and impaired host response; and on the healthcare system side, by organizations not implementing effective IPCS, AMS, and RIMS in a coordinated manner.

We do prevent infections by putting into action all the preventive measures – those which are known to be scientifically validated – which might influence the development of sepsis. Prevention may succeed and may fail for multiple reasons; because of incomplete or ineffective actions, or through comprehensive endeavors which address system change or cultural promotion in isolation from related work streams (the context [15], [16], [17]). Ultimately, prevention may fail because the knowledge used to anticipate the risk of infections is not comprehensive.

In healthcare, the wide variety of EWS worldwide all share common components – they are based on the detection of patient vital signs, parameters that are predictors of patient mortality and morbidity [18].The items considered in sepsis risk stratification tools are oriented to the identification of host vulnerabilities and hazards. They may be further enhanced by encompassing infectious hazards regarding the pathogen, and vulnerabilities and hazards referring to the environment and health care organization. Jeopardized hand hygiene or surgical site infections bundles and antibiotic-resistant bacteria colonization [19] are already included as source of risks into IPC implementation strategies in some healthcare facilities [20]. Alerts around risk zoonotic microorganisms [21]in the geographical area are now being considered at policy level [22]. In order to flesh out into an approachable framework the complexity of infections, we have arranged the four principle theories regarding infection: microbiological, immunological, genetic, and ecological [23] within a framework (Table 2) suggestive of the interaction between the host, pathogen, and environment according to endogenous and exogenous factors, vulnerabilities, and hazards. We propose to complement the existing sepsis risk stratification tools with the following alerts.

1. Alert related to pathogens which might alter the balance between “good” and “bad” bacteria in the patient microbiome (disbiosys), including post-operative pathobiome emergence and pregnancy disbiosys.

2. Alert related to the environment in which the host is living or is treated which elicit the existence of conditions favoring infection such as facilities with high rates of HAI, contaminated areas, and multi drug resistant organisms (MDROs).

3. Alert related to the host pertaining to immunosuppression or immunoparalysis including relative immunosuppression.

4. Alerts related to suspicion of sepsis which refer to the clinical judgment as a fast heuristics for identifying at risk patients [24], [25]

In presence of a suspected infection, are healthcare workers always able to refer to alerts underscoring relevant characteristics of the host response? To consider the ontogenetic characteristics of bacteria, viruses, fungi, and their modifications due to resistance mechanisms? Is a patient’s recent medical history always recorded and considered? And most importantly, do we consider the interactions of all of the above in order to stratify risk in patients with suspected sepsis or to manage the care of patients with sepsis?

To develop an extensive prevention service for infection, health care professionals must be able to retrieve and access information regarding hazards and vulnerabilities relating to infections (Table 2). In order to reduce the impact of infections, a wise approach is to monitor hazards and mitigate vulnerabilities to reduce the occurrence of failure to prevent infections and sepsis. In this, the environmental hazards and vulnerabilities which may facilitate infections should be comprehensively addressed in the prevention strategies in healthcare organizations.

We propose to identify alerts complementing the existing ones [26] referring to co-morbidities and recent medical history. Whenever noted in patients, alerts may suggest an imbalance in the interactions between host and microorganism. This creates the condition for producing alert signals that need to be mindfully addressed, and combined with severity scores targeting vital parameters, bio-markers, and clinical microbiology.

The alert signs need to be paired with warning scores currently obtained from vital signs monitoring and soon to be complemented with biomarkers and RDTs. Alerts combined with risk stratification should yield warnings triggering the activation of dedicated medical and instrumental resources in order to prevent organic damage (see scenarios below); a risk stratification of sepsis by means of prognostic scores leading to resource allocation and therefore mitigation. The HF/E applied to patient safety [27] introduced into healthcare the concepts of risk prevention and risk anticipation. Risk anticipation is based on experience-the knowledge extracted from past undesirable occurrences such as adverse events. Nevertheless, prediction of future events based on past adverse events is not considered the only strategy for patient and care safety [28].The knowledge collected after system failures obtained by means of incident reporting systems (lagging indicators) [29] “off-line” should be effectively combined with the real-time learning of risks “on-line” during everyday healthcare activity (leading indicators) [30].

Monitoring, alert detection, and warning communication

Currently, the National Early Warning Score (NEWS) [31] is a severity score-generating warning system widely used to elicit deterioration signs related to infections and sepsis according to vital signs. National healthcare systems such as in the UK and Australia, where the use of early warning score and systems is structured, also show a decrease in sepsis mortality registered in administrative data [5].

The diagnosis of infection underlying sepsis is not straightforward. It is dependent on several categories of information relating to the interaction between host, pathogen, and environment. The latter, referring to healthcare environments, should be made available through effective IPCS.

Other information is provided from clinical microbiology services. Several novel diagnostic technologies are now available which consistently reduce the turn-around time (TAT) to pathogen identification and/or antimicrobial susceptibility testing (AST). These novel technologies, which rely upon molecular techniques typically to amplify and detect pathogenic genetic material, are often referred to as FM [32].

Other information on host response is provided by laboratory services. Diagnostic biomarkers can help discriminate sepsis from non-infectious critical illness, may help discriminate the causative organisms, and most importantly assess the severity of disease. Patients with sepsis can have different pathophysiological features and mechanisms despite similar clinical manifestations, The outlook on this will be informed through research on human host-omics [33] and on pathobiome emergence [34] in order to classify the immune responsiveness during critical illness and to help prognosticate on recovery from organ failure.

However, this potential will remain entirely untapped if not meaningfully and sustainably integrated into the clinical workflows [35].

Pragmatically, a reliable Holistic EWS for infection identification and treatment is based on:

1. The framing of alerts on infectious hazards and vulnerabilities referring, into a clinical condition, to the unbalanced interaction of host-pathogen and enviroment by means of:

   1. The detection of severity relevant predictors addressing the relevant parameters, complemented with biomarkers, RDTs, and imaging

   2. The timely and collaborative communication of content according to pre-shared standards and warning scores

   3. The interpretation of and faith in the information, including anticipation around the efficacy of the prescribed treatment including surgical source control

Standardized and collaborative communication is fundamental [36], whilst absent, dysfunctional, delayed, or unidirectional communication between the various clinical settings and support services including the imaging, clinical microbiology, and clinical chemistry can cause loss of, or mistimed communication of, vital information. This is likely to mean constructing cross-sectorial competencies based on pro-active organizational response being activated on the basis of alerts and risks escalation. Moreover, a high degree of situational awareness [37] is a pre-condition for healthcare workers to take into considerations the alert and warning signs related to infection and sepsis. This is favored not only by trained clinicians but also must be embodied into an effective organizational milieu.

Response capability to sepsis in context: a heuristic approach

Infections cannot always be prevented, in which case mitigation of harm depends on effective identification and treatment. Take the scenario of an ED where sepsis is more likely to be spotted [18]. In EDs, healthcare workers are exposed to information with variable levels of intensity, redundancy, availability, and reliability. Tasks are interruptions-laden, the accelerations of work intensity are difficult to anticipate and work patterns are detectable only retrospectively [38]. In such conditions, optimal decision-making is sometimes impossible. The analytical processing of a huge amount of data is simply not doable.

Nevertheless, the ED delivers safe and high quality care. The reason for this is partly because of heuristics: practical decision strategies ignoring part of the information deemed ‘ideal’ to inform decisions but which is unavailable, and focusing instead on the few relevant ‘at hand’ predictors [7]. Professionals cope with fatigue and cognitive load by applying heuristics which constantly tune with the environment in which they are used. Mostly effective, limited by definitions heuristics may lead to diagnostic errors or liability in incident analysis [39], [40]. Heuristics are “good-enough” strategies engendering tolerance to volatile and uncertain situations.

One of the activities we are proposing is to design novel heuristics allowing clinicians to navigate the uncertainty of infection diagnosis based on the relevant information gathered by means of holistic EWS system. To illustrate this point, here is a detailed example of a specific time-dependent decision-making process in a healthcare setting. We identified the assumptions likely (Table 3) to inform the process of RIMS. The scenario we are using is the inter-professional collaboration of nurse and physician: a kind of “universal couple” usually present in most of the healthcare settings.

The organizational response needs to be activated proportionally to the level of severity assessed. The effort should produce the characterization of the host response by means of biomarkers, information regarding the possible source of infection, and the identification of the pathogen by means of RDTs. The targeted treatment, including antimicrobial therapy and surgical source control, should be tailored as soon as possible to the specific patient’s endotype. The PSIS (portal of entry [P], source control [S], immune status [I], septic shock [S]) stepwise approach is a pragmatic framework [41] for approaching suspicion of sepsis.

To this end, we propose the concept of the Sepsis Risk Bioscore, a contextualized medical heuristic complementing available severity scores with biomarkers suggestive of infections and organ failure informed by RDTs of FM. In order to navigate the uncertainty of sepsis diagnosis it may be helpful to embody within the decision-making process relevant predictors of specific sepsis endotypes. Such a concept might acquire particular relevance if embedded within a healthcare system adopting risk-based thinking around the anticipation of infection and sepsis.

Individual biomarkers are an integral part of the clinical practice of critical care, single measurements are not able to fully understand the biological complexity, the heterogeneity of infections and sepsis to an extent that decreases their clinical utility. For this reason the use of biomarker panels rather than individual biomarkers is more efficient and increases diagnostic accuracy [42]. Combining biomarkers that reflect different aspects of infections and sepsis pathobiology potentially provides more useful information. Biomarkers from distinct pathophysiological pathways should be increasingly used in the critical care setting, to complement clinical judgment and interpretation of other diagnostic and prognostic tests, to improve patient care. The main purposes of such biomarkers are to improve infection diagnosis, to help in the early risk stratification and provide prognostic information regarding the risk for mortality and other adverse outcomes, and to optimize antibiotic tailoring to individual needs of patients. Besides some biomarkers, such as procalcitonin (PCT), C-reactive protein (CRM), lactate which are well established and have shown positive effects in regard to clinical outcomes, there is a growing number of novel markers from different pathophysiological pathways, where the final proof of an added value to clinical judgment and ultimately clinical benefit to patients is still lacking [43], [44], [45], [46], [47], [48], [49].

Among the relevant predictors, biomarkers suggestive of infections and organ failure such as PCT, CRP, MR-proAdrenomedullin (MR-proADM), and presepsin (CD14) might be integrated in the infection EWS [50].

The aim of the Sepsis Risk Bioscore is 1) to stratify the risk of sepsis developing, 2) to identify patients most severely affected and exposed to major risks, and 3) to prioritise them for the use of a FM pathway. Building on the concept of heuristics, we propose the contextualization of different versions of the Sepsis Risk Bioscore for different clinical settings.

In the ED setting, clinicians may pair relevant predictors of patient severity of NEWS and SOFA, or alternative pragmatic locally applied scores such as Red Flag Sepsis, with biomarkers such as PCT (diagnostic/prognostic marker), CRP (diagnostic/prognostic marker), presepsin (diagnostic/prognostic marker), lactate (prognostic marker), MR-proADM (prognostic marker). Eventually, a score for infective risk such as PIRO (predisposition, insult, response, and organ dysfunction) [51] may also be considered as a relevant predictor.

In the medical and surgical settings, the Sepsis Risk Bioscore should be based on the assessment of specific patient risks in order to predict and identify the clinical deterioration.

In ICU/HDU (intensive care unit/high dependency unit) setting, the scores used to assess severity should be complemented with several other biomarkers proposed to improve prognostic work-up of sepsis patients and their different phenotypes. These include markers based on the complex pathophysiology of sepsis, characterized by activating pro- and anti-inflammatory responses (cytokines, chemokines, cell surface markers [presepsin, CD64], soluble receptors [suPAR: urokinase type plasminogen activator receptor, TREM-1: triggering receptor expressed on myeloid cells] combined with reactions and modification in non-immunological pathways: cardiovascular [troponin, BNP: B-type natriuretic peptide], renal [NGAL: neutrophil gelatinose associated lipocalin, PENK: proenkephalin], coagulation [DIC Score, AT: antithrombin, PC: protein C], neuronal and metabolism). The scores should be also complemented with scores accounting for the risk of HAI and colonization by specific MDR pathogens and the risk of immunosuppression (lymphocyte death and deactivation, T-lymphocyte anergy, HLA-DR, IL-10).

On the basis of the alert, the transfer should allow continuity and seamless care in ED where vital parameters, and biomarkers obtained as soon as possible should enable to detect if the host response is adaptive or likely to become abnormal and evolve into sepsis. Patients should be treated according to the warning sign detected, including for example if MDRO or fungal infection is predicted.

This simple risk assessment should be considered as a trigger to immediately implement evidence-based interventions, including as evidence builds the use of FM and the rapid (hours) re-assessment of the empirical antibiotic chemotherapy [52] (Table 4).

Integrate the three workflows of infection management

We see the necessity to explore further seeking more symbiotic relationship between the models dealing with the integration of the three pillars of infection management: IPC, IPCS, AMS complemented with RDTs and RIMS (Figure 1).

Figure 1:

The three workflows: rapid identification and management of sepsis (RIMS – sepsis stewardship); antimicrobial stewardship (AMS); infection, prevention, control and surveillance (IPCS).

Workflows 1 and 2 are characterized as care workflows (Table 5). These are core activities of healthcare systems taking place whenever healthcare workers are aware of a hospitalized patient affected by an infection. The focus is primarily on understanding the host–pathogen interaction in order to mitigate undesirable consequences with appropriate and timely treatment. Workflow 3 is characterized as being “off-line” – peripheral with respect to patient care, but important because it directly influences the organization in which care is delivered.

The workflow of RIMS holds a clinical approach, it is focused on the interaction between host and pathogen. It applies evidenced-based practice to treatment of cases in a time-critical dimension, and generates information while the activity is unfolding that can then be used to re-orient the activity (leading indicators). The potential for failure in this workflow lies in not rescuing a patient likely to benefit from rescue because the system is not aware of the deterioration going on.

Likewise, the workflow of AMS is clinical and targets the host–pathogen interaction by means of appropriate timely and conservative use of antimicrobials informed by clinical and FM. It aims to achieve the best patient outcomes on the basis of the available information. The information is produced by a continuous interaction between clinicians and microbiologist. The focus is the transition from empirical to targeted antimicrobial chemotherapy and further adaptations based on response and new information. The potential for failure in this workflow lies in not effectively treating a patient likely to benefit from an appropriate treatment because the information needed was not acquired, and because of the complicated transition from infection into sepsis and septic shock.

Unlike the others, the workflow of infection prevention control and surveillance holds an epidemiological approach, it is applied to healthcare processes, and seeks to create barriers to contagion. It considers the presence and diffusion of microorganisms. The timescale is retrospective and the approach is analytical. This workflow retrospectively accounts for trends of cases in patient cohorts, and the risk assessment performed on the basis of past critical events (lagging indicators).The failure to prevent is positioned at the logical beginning of the workflow and accounts for the avoidable dysregulated host response because of infections.

The three workflows are all stemming from the current knowledge we have on the interactions of host–pathogen and environment. Operationally, the three workflows are viewed as initiated by an infection either preventable or not preventable.

The capability of avoiding failures is dependent on the level of system awareness induced by sense-making at any stage of the infection management workflows. To this aim, information gathering and sharing is central to all the workflows presented. Integration requires utilization of information regarding the type of pathogen interacting with the host and the specificity of the response. This information will eventually increase the capacity to classify the patient endotype as soon as possible in order to prompt precision medicine for sepsis management.

In Workflows 1 and 2, information is produced at two stages by the Infection Risk Bioscore/Sepsis Risk Bioscore and the clinical chemistry, microbiology lab and imaging services. This in turn requires a process-making room to retain some sort of information and central devices in information processing. This will enable hardwired integration among the workflows informing the response capability: from the organizational point of view the monitoring and detection of alert and communication of warnings; from the clinical point of view the initiation of an early diagnosis and from the patient safety point of view the reduction of failure to rescue and to treat.

In Workflow 3, the information produced is extracted from the fundamental activity of infection prevention control and surveillance and enhanced by the utilization of the Microbiology and Clinical Chemistry lab services for detecting alerts and vulnerabilities of healthcare processes with respect to dangerous interactions of pathogen, host and environment. Moreover, the clinical information is used retrospectively to produce lag indicators regarding hazards and vulnerabilities at a system level.

In the phase “monitoring, detection of alert, and warning communications” is positioned the backbone connecting the three workflows. The information generated from the interaction between the Sepsis Risk Bioscore and the imaging and laboratory services is crucial for the flexibility of the processes and for the sedimentation of knowledge.

The identification of a peculiar kind of interaction between the host and the pathogen through clinical microbiology needs to sediment in the conceptual base used to propagate prevention by corroborating the understanding of alerts. The above-mentioned linkages should incrementally generate novel information as described in Figure 2. This will consequently indirectly lead to increased survival opportunity through an extraordinary feedback loop on the future knowledge on infections.

Figure 2:

Risks knowledge and prevention of infections and sepsis.

Without mindful integration of clinical microbiological and epidemiological data, the construction of a sound risk knowledge is not attainable and hence prevention, diagnosis and treatment may produce jeopardized improvements.

The holistic EWS underpins the heuristic approach enabling clinicians to extract information from alerts, vulnerabilities and uncertainty in a situated approach named Infection Risk Bioscore/Sepsis Risk Bioscore. The clinicians should be allowed to act with scalable degrees of freedom in the situation they are facing. This is achievable not through centrally pre-determined tasks but by means of decentralized risk assessment based on tools such as EWS enabling the dynamic management of resources. The aim is providing flexibility in structures and processes so that healthcare units can learn and implement versatile solutions to the problems posed by a variable diagnosis such as sepsis.

Conclusions

We explore the notion that a risk-based thinking paradigm is likely to create integration among the three workflows of infection management. All these workflows, in order to contribute to improved infections management, need to have in common the ability to create novel information. Drawing an image of HEWS is a way to foresee the future. This was illustrated in the way healthcare units cope with changes by means of heuristics. We aimed to propose heuristics informing infection management allowing clinicians to anticipating what infections management should be like, measuring differences with expectations and adjusting to reality. We imagined a system able to extract novel information for everyday activity in order to sediment the knowledge base of infections. We suggest a novel point of view, through the introduction of a comprehensive risk-based approach into healthcare programs for tackling the epochal challenges posed to human existence by the alteration of the homeostasis between microorganism and human life. The arena in which such an approach was developed is sepsis identification and treatment. The model includes the assessment of alerts and hazards emerging within the healthcare organizations and processes. These are fundamental categories for enhancing current procedures of infections managements. This poses the novel challenge of mitigating alerts and assessing the healthcare settings in which patients are assisted. The model draws on the existing research on the importance of monitoring the host response and introduces novel alerts to be accounted for: that is, relating the host response and vulnerabilities of healthcare organizations. We refer in proposing this model to the current technological outlook on the technological advances in the physical chemistry domain, omics science and the immune-monitoring that are already available and need to be mindfully integrated into everyday workflow in order to unleash the potential. The model is also based on a truly interdisciplinary approach considering the point of views of other disciplines such as HF/E, communication science, engineering, and safety science.

The model proposes also the idea that the support of laboratory services for informing treatment is deemed necessary and need to considered also in low and middle income countries. The emphasis placed on risks should here be considered as the need to orient the many active forces to tackle the consequences of a compromised homeostasis between microorganism and human life due to anti-microbial resistance (AMR). In general, the approach to the understanding of the interactions of microorganism, hosts and environment should be prompted by the will of favoring life and human evolution.