Discussion

This multicenter survey of PICU clinician knowledge, attitudes, and practices regarding empiric vancomycin use was part of a formative evaluation to design a stewardship intervention focused on reducing vancomycin overutilization. Our survey revealed several potential determinants of vancomycin overuse that can inform these stewardship efforts. Clinician knowledge of the spectrum of activity of vancomycin was low, and perceptions of whether clinicians perceived vancomycin overuse to be a significant problem varied across institutions. Most clinicians agreed that reducing vancomycin use would reduce antibiotic-associated adverse events. Importantly, although clinicians were confident in their ability to determine when vancomycin should be started in children with sepsis and when it can be safely discontinued, fewer clinicians felt confident about whether vancomycin should be started in a hemodynamically stable child. These findings provide new insights into the factors that shape empiric vancomycin use in the PICU setting and offer several insights for stewardship intervention development.

Knowledge gaps around antibiotic spectrum, as well as limited familiarity with local epidemiology of multidrug-resistant organisms, have been reported in prior studies. ^{Reference Kim, Szymczak and Schnellinger24,Reference Salsgiver, Bernstein and Simon26} In our study, just 13% of clinicians correctly identified organisms uniquely covered by vancomycin, beyond those covered by β-lactam antibiotics typically used in combination with vancomycin for suspected sepsis in the PICU, a common indication in this setting. Clinicians also endorsed an empiric coverage threshold of 80% of potential pathogens, an estimate consistent with other published literature, ^{Reference Cressman, MacFadden, Verma, Razak and Daneman25} despite the prevalence of MRSA infection among patients being evaluated for serious infections or sepsis reported consistently as <10%. ^{Reference Waters and Caraccio12–Reference Chiotos, Newland and Starnes14} Clinicians therefore may be selecting empiric antibiotic regimens inclusive of vancomycin based on both an overestimation of vancomycin’s spectrum of activity as well as an underappreciation of the incremental gain in coverage afforded by routinely including vancomycin in empiric antibiotic therapy.

Despite misperceptions around the utility of empiric vancomycin, half of clinicians acknowledged that vancomycin overuse was a problem in their unit and nearly 80% agreed that reductions in vancomycin use would reduce antibiotic-associated adverse events. Nevertheless, vancomycin remains one of the most commonly utilized antibiotics in hospitalized children, including those in the PICU. ^{Reference Brogan, Thurm and Hersh1,Reference Chiotos, Blumenthal and Boguniewicz3} Although not addressed in the present study, fear of missing a potentially treatable infection has also been demonstrated to influence antibiotic decision making among ICU clinicians in prior work. ^{Reference Pandolfo, Horne and Jani23,Reference Wunderink, Srinivasan and Barie27–Reference Chiotos, Marshall and Kellom29} These fears may be amplified in the time-pressured PICU environment in which early antibiotic administration is emphasized, whereas the clinical consequences of over-treatment are both less immediate and less salient than those of potential undertreatment. ^{Reference Paul, Niedner and Riggs30–Reference Levy, Gesten and Phillips32}

Clinicians most often rated illness severity as a primary driver of their decision making around whether or not to administer empiric vancomycin, a theme reflected in other work focused on ICU-based stewardship interventions. ^{Reference Pandolfo, Horne and Jani23,Reference Wunderink, Srinivasan and Barie27} Although there are undoubtedly severely ill children who do not benefit from vancomycin, focusing stewardship efforts on children who are not severely ill may hold most immediate promise in reducing vancomycin overuse. Prior work suggests that clinicians may broaden or start empiric antibiotics “just in case” in these lower-risk patients. ^{Reference Woods-Hill, Koontz and King28,Reference Chiotos, Marshall and Kellom29} In our study, nearly 40% of clinicians did not feel confident in determining whether vancomycin was necessary in a hemodynamically stable patient. Institutionally endorsed practice guidelines, clinical decision support tools, and feedback to clinicians on local MRSA epidemiology may reduce uncertainty and support more consistent decision making in these scenarios. For example, implementation of a computerized order entry prompts providing guidance about MDRO risk has been associated with reduced use of extended-spectrum antibiotics for common infections in adults hospitalized in non-ICU settings, including pneumonia and skin and soft tissue infections. ^{Reference Gohil, Septimus and Kleinman33,Reference Gohil, Septimus and Kleinman34}

A Delphi Consensus study aimed at establishing priority pediatric antimicrobial stewardship interventions in the US flagged the ICU setting as one of two high-priority clinical contexts caring for children. ^{Reference Obeng, Tetteh and Malone35} Implementing stewardship in the ICU is challenging because critically ill children frequently present with nonspecific signs of infection, rapid clinical deterioration, and a high risk of morbidity and mortality, prompting early empiric broad spectrum antibiotic use. Diagnostic uncertainty, the heterogeneity of ICU patients, fear of undertreating sepsis, and the need to make urgent decisions about aspects of treatment such that modifications to antibiotic therapy are forgotten complicate efforts to improve antibiotic decision-making in this context. ^{Reference Pandolfo, Horne and Jani23,Reference Wunderink, Srinivasan and Barie27,Reference Janssen, Oerlemans and Bos36} Our study provides insights into the multilevel drivers of vancomycin overuse in the PICU that suggest different implementation strategies including prospective audit and feedback, development of local clinical practice guidelines, educational outreach, multidisciplinary stewardship rounds, audit and feedback of performance, and use of clinical champions will be needed to improve prescribing. ^{Reference Chorafa, Komatsiouli, Iosifidis, Kourti, Sdougka and Roilides37–Reference Liberati, Brigadoi, Barbieri, Giaquinto and Donà39}

Our study is subject to several limitations. First, our survey relied on stated practices rather than actual practice regarding use of vancomycin. It is likely that some respondents were aware of the goals for the study and therefore may have been biased toward providing desirable answers. We attempted to mitigate this both by making the survey anonymous and by circulating the survey prior to any educational efforts regarding the intervention. Second, because our assessment of vancomycin knowledge was based on a limited number of questions, it is possible that we underestimated clinicians’ knowledge base that would have been captured with a more extensive assessment. Third, we sampled clinicians primarily working at academic medical centers, our results may not be generalizable to all settings. Fourth, we cannot assess response bias because we do not know the characteristics of people who did not respond to the survey. Fifth, we did not use a validated instrument but instead utilized a multistage process informed by the literature and expert consensus to ensure questions were relevant and captured key determinants of vancomycin use in the understudied PICU context. Despite these limitations, our relatively high response rate from multiple geographically diverse PICUs strengthens the generalizability of our findings.

Overall, our study describes PICU clinician beliefs and decision-making for empiric vancomycin to inform future work to optimize its use. Addressing both cognitive and behavioral drivers of vancomycin use is critical for effective and sustainable stewardship efforts in this clinical context, learnings that may be applied to use of other broad-spectrum antibiotics in the PICU setting.

Figure 2.

Vancomycin use practices in common infections in the PICU. (a) Would you prescribe vancomycin to a previously healthy toddler with no central line or other indwelling hardware presenting with community-onset septic shock requiring an epinephrine infusion to maintain blood pressure? (b) Would you prescribe vancomycin to an adolescent with short bowel syndrome and a central venous catheter who has been hospitalized in your PICU for three weeks and develops a fever? He is hemodynamically stable and has no personal history of MRSA. (c) Would you prescribe vancomycin to a child with spina bifida and recurrent urinary tract infections who presents with fever, tachycardia, delayed capillary refill, and a urinalysis with large leukocyte esterase and bacteria? (d) Would you prescribe vancomycin to an infant with RSV and a patchy RML infiltrate versus atelectasis admitted to the PICU on high flow nasal canula and requiring 30% FiO2?

Table 2.

Vancomycin attitudes by institution

Table 2. Long description

The table presents data on vancomycin attitudes by institution across different sites. It has six rows and six columns. The columns are labeled Agree, Neither agree nor disagree, Disagree, and P value. The rows are labeled with different statements and sites. Row 1: Overuse of vancomycin is a significant problem in my unit. All (N 133), 72 (54); 36 (27); 25 (19); P.01. Site 1 (N 10), 6 (60); 3 (30); 1 (10). Site 2 (N 24), 9 (37); 10 (42); 5 (21). Site 3 (N 68), 45 (65); 16 (24); 8 (12). Site 4 (N 25), 13 (52); 5 (20); 7 (28). Site 5 (N 6), 0 (0); 2 (33); 4 (67). Row 2: Reducing vancomycin overuse would reduce antibiotic associated adverse events in patients admitted to our PICU. All (N 132), 104 (79); 19 (14); 9 (7); P.06. Site 1 (N 10), 7 (70); 3 (30); 0 (0). Site 2 (N 24), 15 (63); 4 (17); 5 (20). Site 3 (N 67), 45 (82); 9 (13); 3 (4). Site 4 (N 25), 23 (92); 2 (8); 0 (0). Site 5 (N 6), 4 (67); 1 (17); 1 (17). Row 3: I am confident in my ability to know when a PICU patient with fever has sepsis and requires antibiotics. All (N 130), 88 (68); 34 (26); 8 (6); P.12. Site 1 (N 10), 8 (80); 2 (20); 0 (0). Site 2 (N 24), 17 (66); 7 (29); 1 (4). Site 3 (N 65), 38 (59); 23 (35); 4 (6). Site 4 (N 25), 21 (84); 2 (8); 2 (8). Site 5 (N 6), 5 (83); 0 (0); 1 (17). Row 4: I am confident in my ability to determine when vancomycin specifically should be started in a PICU patient with fever who is hemodynamically stable. All (N 130), 79 (61); 39 (30); 12 (9); P.95. Site 1 (N 10), 8 (80); 2 (20); 0 (0). Site 2 (N 23), 14 (61); 7 (30); 2 (9). Site 3 (N 66), 38 (58); 20 (30); 8 (12). Site 4 (N 25), 14 (56); 9 (36); 2 (8). Site 5 (N 6), 5 (83); 1 (17); 0 (0). Row 5: I am confident in my ability to determine when vancomycin specifically should be started in a PICU patient presenting with sepsis. All (N 133), 106 (80); 21 (16); 6 (5); P.98. Site 1 (N 10), 9 (90); 1 (10); 0 (0). Site 2 (N 24), 20 (83); 3 (13); 1 (4). Site 3 (N 68), 52 (76); 12 (18); 4 (6). Site 4 (N 25), 19 (76); 5 (20); 1 (4). Site 5 (N 6), 6 (100); 0 (0); 0 (0). Row 6: I use institutional antimicrobial prescribing guidelines to help guide my decision-making around antibiotics for sepsis. All (N 133), 105 (79); 16 (12); 12 (9); P.59. Site 1 (N 10), 9 (90); 1 (10); 0 (0). Site 2 (N 24), 17 (71); 2 (8); 5 (21). Site 3 (N 68), 54 (79); 10 (15); 4 (6). Site 4 (N 25), 19 (74); 3 (12); 3 (12). Site 5 (N 6), 6 (100); 0 (0); 0 (0). Row 7: I am confident in my ability to determine when vancomycin can be stopped based on culture and other diagnostic data. All (N 133), 112 (84); 17 (13); 4 (3); P.08. Site 1 (N 10), 10 (100); 0 (0); 0 (0). Site 2 (N 24), 22 (92); 1 (4); 1 (4). Site 3 (N 68), 56 (82); 12 (18); 0 (0). Site 4 (N 25), 18 (72); 4 (16); 3 (12). Site 5 (N 6), 6 (100); 0 (0); 0 (0).

PICU, pediatric intensive care unit.

Table 3.

Vancomycin attitudes by occupational role

Table 3. Long description

The table presents data on attitudes towards vancomycin use among different occupational roles in a Pediatric Intensive Care Unit (PICU). It has six rows and five columns. The columns are labeled ‘Agree’, ‘Neither agree nor disagree’, ‘Disagree’, and ‘P-value’. The rows are labeled with different statements about vancomycin use and the occupational roles: All, Attendings, Fellows, APPs, and Hospitalists. Each cell contains the number and percentage of respondents who agree, neither agree nor disagree, or disagree with the statement. The table also includes a P-value for each statement, indicating the statistical significance of the results. The statements cover topics such as reducing vancomycin overuse, confidence in determining when to start vancomycin, use of institutional guidelines, and confidence in determining when to stop vancomycin.

APP, advance practice provider.

^a APPs include nurse practitioners (N = 29) and physician assistants (N = 1).