A Practical Approach to Clinical Antibiotic Stewardship in the ICU Patient With Severe Infection

Jan Fierens, MD; Pieter O. Depuydt, MD, PhD; Jan J. De Waele, MD, PhD


Semin Respir Crit Care Med. 2019;40(4):435-446. 

In This Article

Antibiotic Therapy in Patients With Severe Infections

We will use a pragmatic approach to highlight the importance of ASPs, describing the current evidence and evolving concepts and considerations at the three pivotal moments during an antibiotic therapy: initiation of the antibiotic treatment, the moment of re-evaluation during treatment—also known as the antibiotic time-out—and the discontinuation of an antibiotic treatment (Figure 1).

Figure 1.

A pragmatic approach to apply ASP in daily practice in the ICU patient with severe infection. ASP, antibiotic stewardship program; ICU, intensive care unit.

Initiation of Antibiotic Therapy

Sense and Nonsense of "Watchful Waiting" in Severely Ill Patients. Obviously, the first and foremost important question is whether the patient needs the administration of an antibiotic. There is sufficiently strong evidence suggesting that swift antibiotic administration improves survival in sepsis and septic shock.[14–16] Combined with the potential fear of missing this presumed window of opportunity and potentially also peer pressure among other reasons, this problem will turn many clinicians to the "safe path" of an antibiotic treatment.[17] Nevertheless the diagnosis of hospital-acquired infections poses quite a challenge in the critically ill patient, as up to 50% of febrile episodes are of noninfectious origin, e.g., in burn patients, trauma patients, and patients with severe pancreatitis.[18] A daring quasi-experimental before and after study suggested that a restrictive antibiotic approach in hemodynamically stable patients might be beneficial with a 50% reduction in adjusted mortality, a shorter duration of therapy, and higher rates of appropriate initial therapy.[19] It should be emphasized that in this study only nonseverely ill patients were included. In clinically stable patients, this approach of "watchful waiting" might realize a more appropriate antibiotic use. On the other hand, in every severely ill patient, presenting with septic shock or progressive organ failure, where an underlying infection cannot be ruled out yet, an antibiotic treatment should be started aimed at the likely infectious origin of the patient's deterioration. It goes without saying that before administration of antibiotics, obtaining cultures to identify possible pathogens and appropriate documentation in the patient's file are important to further refine therapy at a later stage.

Empirical Treatment and MDR Pathogens. The selection of appropriate empirical therapy is hampered by the increasing antimicrobial resistance rates, justifying empirical broad-spectrum antibiotic therapy to cover all likely pathogens—including MDR pathogens—for patients with suspected hospital- and ICU-acquired infections in international guidelines.[20–23] It is however pointless to thoughtlessly apply these recommendations and cover pathogens that are unlikely to be involved in the infection that requires therapy. The antibiotic resistance feature of a causative pathogen does not necessarily relate to the severity of illness, as illustrated by the high mortality in, for example, necrotizing fasciitis and severe community-acquired pneumonia, both caused by pathogens generally sensitive to narrow-spectrum β-lactam antibiotics.[24–26] Therefore, guidelines for empirical antibiotic treatment in community-acquired infections, which are usually restrained in their choice of spectrum, may apply even if these patients are critically ill.[27–29] Antibiotic therapy based on local microbiological data, proper use of surveillance cultures, and patient's MDR risk factors have been associated with increased appropriate empirical treatment and a reduced use of broad-spectrum antibiotics, compared with a general guideline-based approach in hospital-acquired pneumonia.[30–32] Currently, no evidence is available confirming this in other forms of severe infections. Nonetheless, it seems logical to cautiously extrapolate these findings to other types of severe infection in the ICU. Identifying MDR risk factors, such as prior health care exposure, prior antibiotic courses, and prior colonization status combined with local epidemiology, is paramount in adjusting empirical treatment in severe infections before pathogen identification and antimicrobial susceptibility testing become available.[33–36] The antibiotic regimen should therefore be based on local guidelines, taking into account the individual patient's current clinical status, MDR risk factors, known drug intolerances, and anticipated side effects.[30,32,37]

"Bridging the Gap" with Combination Therapy in Severely Ill Patients. Starting empirical therapy with a combination of antimicrobials has the advantage of an extended coverage of potentially dangerous MDR pathogens and polymicrobial infections, a possible synergistic effect due to enhanced antimicrobial uptake and prevention of resistance development.[38] This however increases the cost and potential risk of antibiotic toxicity. Broad-spectrum β-lactam antibiotics continue to remain the cornerstone in empirical antibiotic therapy of hospitalized patients worldwide.[39,40] By adding for instance an aminoglycoside to β-lactam therapy, gram-negative antibiotic coverage is improved, but this increases the risk for nephro- and ototoxicity. Adding vancomycin (or linezolid) has the potential to add MRSA coverage to the empirical regimen, but increases the risk for renal damage and possibly fungal superinfection. Several retrospective studies hinted at a potential mortality benefit for combination therapy, but these findings were not always consistent in prospective studies.[27,41–48] A stratification according to severity of illness was not available in most of these studies, making it difficult to extrapolate these findings to the critically ill patient group. A meta-analysis by Kumar et al did however demonstrate a potential benefit in the subgroup of severely ill patients with septic shock, but at the same time the overall effect of combination therapy was detrimental if patient mortality was below 15%, therefore recommending double therapy exclusively to the severely ill patient group.[49] Combination therapy has also been proposed to limit antibacterial resistance. This concept has been adequately proven in mycobacterial infections, where monotherapy with antituberculous agents has to be avoided because of rapid emergence of resistance.[50] Currently there is no evidence supporting such an effect in the ICU in other infections.

In summary, decisions about combination therapy should be made adhering to current guidelines, based on the scarcely available high-quality evidence, wherein empiric combination therapy is reserved for the severely ill patient group, such as septic shock and infections caused by possible MDR pathogens.[3,20] This general conclusion should be complemented by a few clinical entities wherein combination therapy has a proven advantageous effect and should be standard of care, such as β-lactam and macrolide combination therapy in severe community-acquired pneumonia and the association of aminoglycosides in empirical therapy for infectious endocarditis.[44,51,52]

Influence of Immune Status on Empirical Antibiotic Choice. The immune status of a patient is an important factor when choosing empirical treatment. It can be affected by immunosuppressive drugs after solid organ transplantation (SOT) and hematopoietic stem cell transplantation (HSCT), as well as by congenital and acquired immune-deficiency disorders. This creates a new group of severely ill patients due to the short-term complication of organ and bone marrow transplantation and the use of long-term immunosuppressive drugs.[53,54]

SOT recipients may not typically present with the classic features of sepsis as nontransplanted patients, such as fever and leucocytosis. The site of infection may vary, depending on the transplanted organ and the "net state" of immune suppression.[55] In the first month after transplantation, nosocomial infections of lungs, urinary tract, abdomen, and surgical site are the most common.[56] One to 6 months after transplantation, nosocomial infections gradually give way to community-acquired infections. After 6 months, immune suppression is at a steady state and community-acquired infections are most commonly encountered, though the risk for opportunistic infection never really disappears.[55,57] This timeline needs to be taken into account when initiating empirical treatment: initially broad-spectrum antibiotics covering a wide array of Gram-negative and positive microorganisms should be commenced, with a gradual shift to smaller spectrum antibiotics only covering community-acquired pathogens 1 to 6 months after transplantation. One month after SOT, several opportunistic respiratory pathogens, such as Pneumocystis jirovecii, Aspergillus sp, and non-Aspergillus molds should be considered as causative pathogens in any respiratory deterioration, warranting adequate diagnostic workup and extended empirical treatment with antifungals.

Compared with their SOT counterparts, the prolonged period of neutropenia and stepwise recuperation of bone marrow functions exposes HSCT patients to several clinical entities. The post-HSCT period is divided into three distinct periods, indicating recovery of different immunologically active cell populations.[58] These different phases in engraftment and cell recovery give rise to several common infectious complications in HSCT patients (Table 1), and this should be incorporated in the empirical strategy when these patients present with severe infection. A complete overview of infectious complication after HSCT is beyond the scope of this review. Empirical antibiotic treatment in the pretransplant or immediate posttransplant period is comparable to the neutropenia induced by cytoreductive therapy. A prolonged period of neutropenia and chemotherapy-associated mucositis is associated with enteric and oral translocation of both Gram-positive and negative bacteria with the development of neutropenic sepsis, warranting the start of a combination of a broad-spectrum antipseudomonal β-lactam antibiotic with added vancomycin or linezolid—especially when suspecting catheter-related bloodstream infection, skin infections, or pneumonia—in patients presenting with septic shock, preferably guided by patient's MDR status and previous colonization cultures if available.[59,60] When fever persists after 4 to 7 days in patients with high-risk neutropenia, i.e., neutropenia persisting for more than 7 days, an antifungal, covering Candida sp and Aspergillus sp, should be added to the therapy, when reassessment does not yield any other probable source of infection.[60–62]

In summary, these special patient groups present to the ICU with severe infections, complicated by an atypical presentation and diagnostic uncertainty. Depending on the underlying immunosuppressive pathology, empirical regimens should therefore be based on combination therapy to include possible co-infections (bacterial, viral, fungal, and parasitical).

Importance of Source Control. Source control is an essential component in the management of severe infections in many ICU patients, particularly in abdominal and skin and soft tissue infection, but its importance extends beyond those types of patients. In an era where implantable devices—such as orthopaedic prostheses, cardiac valves, implantable venous access devices, vascular graft prostheses, cardiovascular implantable electronic devices, and left ventricular assist devices—are becoming essential tools of medical care, source control is an often underestimated but important strategy. It is typically defined as controlling the source of infection by drainage or removal of inflammatory material, abscesses, and pus, as well as adequate debridement of necrotic and/or infected tissue.[63] Although pathophysiologically obvious, the evidence supporting adequate and timely source control is scarce. Nonetheless, several studies have reported that inappropriate source control is independently associated with increased mortality.[64–67] Even more so, there appears to be a linear relationship between mortality and time to source control when it was delayed beyond 6 hours after presentation.[68] While it used to be an almost exclusively surgical issue, percutaneous drainage is becoming the first choice in many patients, also in the ICU. Ultrasound and computed tomography guided drainage proves to be an acceptable alternative to invasive surgery, both as a diagnostic tool to obtain cultures and as a temporizing measure during initial resuscitation and correction of metabolic status of the critically ill patient with deranged physiology.[69] Diagnosis and treatment in these cases of severely ill patients is becoming increasingly more complex, warranting a multidisciplinary approach with intensivists, surgeons, interventional radiologists, and infectious disease (ID) specialists to decide on the preferred method of source control. From an antibiotic stewardship perspective, source control is important and indispensable for adequate antibiotic action. The presence of necrotic material or infected foreign devices with biofilm does not allow antibiotics to penetrate into the infected tissue and inevitably antibiotic resistance will occur. In the case of ongoing contamination from the gastrointestinal tract, the infected tissue (in case the peritoneal cavity) will be continuously exposed to new intestinal content for which antibiotics are not the solution. Lack of source control has been associated with prolonged antibiotic courses and development of antibiotic resistance, particularly in abdominal infections and pancreatic necrosis.[70,71] It is therefore imperative that source control has adequate priority to achieve the goals of antibiotic stewardship.

Antibiotic Therapy Re-evaluation

At several moments during the antibiotic therapy clinicians should review the empirical antibiotic treatment; some refer to this as an antibiotic time-out (ATO).[72] We would like to focus on two separate time-out moments, where new clinical, biochemical, and microbiological information could impact the course of antibiotic treatment.

First Antibiotic Time-Out to Re-evaluate Likelihood of Infection and Re-check Antibiotic Coverage. After 24 hours of empirical therapy, a systematic clinical re-evaluation of the patient is warranted to re-evaluate the probability of infection on the one hand, and to check diagnostic work-up and potential shortcomings in antimicrobial coverage on the other hand. This short time window already allows a better differentiation between infectious and noninfectious causes of patient deterioration and may enable discontinuing antibiotics that were in retrospect unnecessary. Some typical examples are starting empirical antibiotics in severe pancreatitis, which induces a severe inflammatory state with a "septic shock-like" clinical image, and in flash pulmonary edema requiring ventilatory support. The absence of sepsis and septic shock, absence of leucocytosis or leukopenia, fever, or other clinical symptoms (such as hypoxemia in pneumonia) should stimulate the clinician to look for alternative diagnoses and consider stopping antibiotic treatment. These kinds of diagnostic decision trees have been widely applied to ventilator-associated pneumonia (VAP) and could be extrapolated to other ICU-related infections.[73,74]

It should again be stressed that severely ill patients should remain on antibiotics, as long as there is clinical doubt about the presence of infection. Also the initial tentative infection diagnosis may have proved incorrect, and the implications for empirical antimicrobial therapy of the new infection diagnosis need to be considered. The first microbiological data may also become available at this moment. These may include standard techniques such as direct examination with Gram stain and a direct antibiogram, but also newer and more sophisticated rapid molecular diagnostic techniques, such as polymerase chain reaction (PCR)-based techniques, matrix-assisted laser desorption ionization time-of-flight (MALDI-TOF), and next-generation sequencing (NGS). These new techniques offer the clinician more diagnostic information to eliminate diagnostic uncertainty and alter the empirical treatment more quickly.

The usefulness of a direct examination of a clinical sample may appear limited, but it can offer the clinician valuable information to adapt empirical antibiotic treatment. This technique can be applied to abdominal, respiratory, cerebrospinal fluid, and other samples. Both the presence and absence of microorganisms may impact the empirical antibiotic therapy. These samples also allow direct susceptibility testing. This technique has some disadvantages and technical limitations, but may provide early susceptibility details, shortening turnaround time by 24 hours.[75–78]

A wide array of commercial PCR-based techniques are available to identify pathogens and certain antibiotic-resistant markers. They mostly use a syndrome-based approach targeting the most frequent causative pathogens in several community- and hospital-acquired infections, such as pneumonia, meningitis, and bloodstream infections.[12] While these expensive tests offer a rapid pathogen identification, they offer limited information about susceptibility, as only a limited number of resistance genes are tested, necessitating the need to perform conventional cultures anyhow. The evidence supporting PCR-based testing in the critically ill is promising but currently limited, warranting more studies to adequately position these techniques in diagnosing infection and altering empirical antibiotic regimens.[79,80]

MALDI-TOF is a variant of mass spectrometry, which allows rapid pathogen identification after culture. Another advantage of this technique is the recognition of proteins associated with antibacterial resistance to, e.g., methicillin in S. aureus and to carbapenemase in Gram-negative isolates.[81–84] This technique allows the clinician to expand empirical antibiotic coverage to include unexpected pathogens, such as Stenotrophomonas, Pseudomonas, and Acinetobacter spp., and potentially antibiotic-resistant pathogens, such as CPE and MRSA. While it is easy to use and low cost, the inability to identify pathogens directly from a clinical specimen without culture remains a drawback to the technique.[85]

NGS in medical microbiology offers the possibility to use nucleic acids extracted from a clinical sample to identify all probable pathogens and their resistance genes. Improvements in nucleic acid extraction, shortening turnaround time, and adequate data extrapolation to clinical practice are significant hurdles to overcome before NGS can find its place in routine clinical practice.[86] Currently, no real evidence exists examining the use of these new techniques in the ICU environment.

A combination of clinical re-evaluation in combination with microbiological information, using older and new techniques, can motivate the clinician to discontinue antibiotics or adapt the empirical therapy to the newly available information after 24 hours of treatment.

Second Antibiotic Time-Out to Discontinue or De-escalate the Antibiotic Treatment. After 48 to 72 hours of empirical antibiotic treatment, the clinical picture is complete and usually definite diagnostic information is available, including identification and susceptibility of causative microorganism. This time-out prompts the clinician to do a final review of the appropriateness of treatment, including the need for de-escalation of antibiotics.[72] As previously stated, the nonspecific signs of infection lead to more severely ill patients receiving more antibiotics than necessary. During an antibiotic time-out, a complete and adequate clinical and microbiological assessment can identify low probability infections, where antibiotic therapy can be discontinued. Although once started, there is often reluctance to stop treatment, especially in the most critically ill, several studies have confirmed this to be safe, also in severe infections.[87]

Another consideration to be made during this second time-out is to de-escalate current empirical antibiotic treatment. De-escalation aims to maintain adequate antibiotic therapy, while minimizing needless exposure to broad-spectrum antibiotics and thereby limiting selection of MDR pathogens. There is no consensus on the exact definition, but it can be characterized as the switch from the initial empirical broad-spectrum antibiotic to a narrower spectrum and/or stopping combination therapy in favor of one single product.[88] Antibiotic therapy can be de-escalated in most ICU patients with infections when no MDR pathogens are involved or when the causative pathogen is susceptible to a more narrow spectrum antibiotic than empirically prescribed.[3,20] For example, vancomycin or linezolid should be stopped if MRSA is not found and broad-spectrum agents, like meropenem, should be reserved for pathogens with exclusive susceptibility to one of these agents. Side notes need to be made for several MDR infections, such as CPE, VISA, and pan-resistant Gram-negative organisms, where the synergy provided by combination therapy is the only solution to treat these pathogens, and for bloodstream infections with ESBL-producing bacteria, where a carbapenem-sparing strategy with piperacillin–tazobactam can no longer be advised.[89,90] In these patients it is often difficult to de-escalate. These logical arguments notwithstanding, antibiotic de-escalation is not consistently applied across ICU departments.[91–94] This illustrates the reluctance of clinicians to change an effective antibiotic regimen (especially in culture-negative infections and severe infections) after careful interpretation of the available data.[95] Observational data have demonstrated the safety of de-escalation, even in severe infections such as bloodstream infections, VAP, or in patients with neutropenia or presenting with septic shock.[93,96–101] A landmark prospective noninferiority trial could not confirm a worse outcome in patients with a de-escalated antibiotic therapy.[102] While no definitive conclusion about mortality could be made, a 2016 systematic review did associate de-escalation to a better patient outcome. The heterogeneity, the lack of clear definition of de-escalation, and the risk of bias in several of the included studies however do impede generalization of these findings, especially focusing on the critically ill patient group with severe infections where no good prospective data are available.[88] More high-quality randomized controlled trials are therefore necessary to demonstrate the effect of de-escalation on mortality, recurrent infections, and antimicrobial resistance. Meanwhile, de-escalation does bring about the opportunity to reduce the overuse of broad-spectrum antibiotics and should be considered as a valid option in daily clinical practice, even in the severely ill patients.

Discontinuation of Antibiotic Treatment

Protracted antibiotic courses have been associated with increased antimicrobial resistance and antibiotic toxicity.[103] Therefore, when deciding to extend the duration of treatment, a balancing exercise needs to be made, deciding between the possible benefit of continued exposure to antibiotics and the harm of increased resistance. Physicians are encouraged by many international guidelines to shorten antibiotic courses to 7 to 10 days for most community- and hospital-acquired infections, unless predictors of poor prognosis are present (e.g., undrainable infectious foci, treatment failure).[3,20,22] Several studies however suggested that prolonged courses of antibiotics are still being prescribed.[104–107] The trend toward shorter antibiotic courses gained firm support by a landmark trial in patients with VAP, where a shorter course of 8 days of antibiotic treatment did not result in worse outcome.[108] Patients admitted to the ICU with severe abdominal infections could be managed successfully with a 7-day course of antibiotics.[109] Even a 4-day treatment could suffice, if adequate source control was obtained, but in this study most patients were not in the ICU.[70] Subsequently numerous other trials found that shorter courses of antibiotics were considered safe and effective in many types of infections, including community-acquired pneumonia, urinary tract infections, vertebral osteomyelitis, and even bloodstream infections.[104,109–112] Most of these trials however do not separately describe the subset of septic shock and/or critically ill patients with multiple organ failure, so extrapolating these findings to the severely ill could be difficult. Apart from several conditions where extended courses of antibiotics are currently considered standard-of-care (such as infective endocarditis, prosthetic joint infections, and osteomyelitis), longer courses may be required in severe infections without acceptable microbial eradication, such as persistent gastrointestinal leakage, necrotizing pneumonia, or inaccessible infectious foci. It needs to be noted that these prolonged courses of antibiotics are expert opinion based, without any evidence of improved outcome, and that efforts at obtaining adequate source control may prove to be more effective than continuing antibiotic therapy. Reducing the duration of antibiotic treatment may prove to be an ideal target for ASP interventions. A "plan ahead approach" with the antibiotic strategy documented in the medical file and automated stop orders, counting total days of adequate treatment including the empirical antibiotics, are tools to achieve this goal.