Procalcitonin Use: Variation Across Hospitals and Trends Over Time Free

This multicenter, cross-sectional cohort study used the Public Health Information System, an administrative database from 52 tertiary-care pediatric hospitals across the United States. We included children ≤18 years of age hospitalized between 2014–2018 with pneumonia, sepsis, or fever in infants <2 months of age on the basis of previously published case-identification strategies for administrative databases using admission and/or discharge diagnosis codes (Supplemental Table 2).^14–16  Of note, these diagnoses were not mutually exclusive. We excluded hospitals without PCT testing available, indicated by an aggregate PCT testing rate of 0% for all diagnoses from 2017–2018 (n = 10). We also excluded hospitals with incomplete data (n = 1) and patients transferred from outside hospitals.

In the primary analysis, PCT and CRP use, defined as the proportion of encounters from 2017–2018 with at least one PCT and CRP test, respectively, ordered at any time during the hospitalization, were summarized by diagnosis. We included encounters with biomarker testing at any time, rather than exclusively on the first day of admission, to describe use more broadly. With this method, we captured PCT use for patients who were diagnosed with fever, sepsis, or pneumonia at the time of admission or anytime during the hospitalization. We generated a heat map depicting variation in PCT testing use across hospitals for each diagnosis from 2017–2018, and hospitals were ranked on the basis of the median aggregate testing rate across all diagnoses. We also evaluated the percent of encounters with PCT testing that had testing on the first day of admission (day 0 or 1 in PHIS), repeat PCT testing (>1 PCT test performed for a given encounter), and CRP testing performed on the same day to examine timing and frequency of testing for each diagnosis. As a secondary analysis, we modeled PCT and CRP testing rates over time from 2014–2018 for each diagnosis using hierarchical binomial logit mixed effects regression with an interaction term between testing and year as well as random intercept and slope to account for clustering patients within hospitals. Annual predicted proportions within hospitals were generated from model estimates and then averaged across hospitals and plotted by diagnosis cohort. The annual trend in odds of testing was estimated by linear combination of parameter estimates, and differences between PCT and CRP trends were evaluated with the interaction term of model. This study was deemed exempt from review by the university’s institutional review board.

Over the entire study period there were 91 138 encounters for one of the three diagnoses, and 5317 (6%) of them had PCT testing (Table 1). In our primary analysis, we evaluated use from 2017–2018 and found that PCT testing was performed in 3988 of 34 231 (12%) hospitalizations. Young infants with a fever had the highest proportion of PCT testing (18%; 990 of 5406) followed by children with sepsis (15%; 865 of 5839) and pneumonia (9%; 2133 of 22 986). CRP was used at a significantly higher rate than PCT (26% to 33% across diagnoses, P < .001). Of encounters with PCT testing, 58% had CRP testing on the same day as PCT testing. Among encounters with PCT testing, most had testing performed on the first day of admission (83% for febrile infants, 64% for pneumonia, and 64% for sepsis). Repeat testing among febrile infants who had an initial PCT test was uncommon (3%) with more repeat testing for those with pneumonia who had an initial PCT test (21%) and sepsis (44%). There was across-hospital variation in PCT use for pneumonia (1% to 44%) and sepsis (1% to 46%) and, particularly for febrile young infants (1% to 88%; Fig 1). Hospitals with higher PCT use for one diagnosis tended to have higher PCT use for others.

From 2014–2018, both PCT and CRP testing increased annually for all diagnoses (Fig 2). For febrile young infants, the odds of PCT testing increased at a significantly greater rate per year compared with that of CRP (odds ratio [OR] for PCT 3.38 [2.99–3.83] vs OR for CRP: 1.32 [1.20–1.45]). Similar patterns were found in pneumonia (OR: 1.81 [1.73–1.90] vs 1.11 [1.07–1.15]) and sepsis (OR: 1.41 [1.34–1.49] vs 1.07 [1.03–1.11]).

In this multicenter study of children admitted with three common diagnoses, there was across-hospital variation in PCT use, even when limited only to hospitals with PCT available. We also found increased PCT testing rates over time out of proportion to rises in CRP testing.

Across-hospital variability in PCT use may be due to variability in turnaround times between institutions, integration of PCT into clinical pathways at individual hospitals, and a lack of consensus on PCT testing recommendations nationally. Increased PCT use over time for all diagnoses may be explained by improved availability (ie, faster result time) or growing interest in and perceived utility of PCT. Alternatively, it may suggest exploratory use of PCT because of its novelty and availability, which is supported by the fact that PCT is commonly ordered in conjunction with CRP, a more well-known biomarker. PCT use increased for patients with pneumonia and sepsis, potentially because of previous literature revealing its value.^5,7,12,17  However, previous studies have conflicting evidence; thus, national guidelines for these diagnoses do not recommend PCT use.^18,19  With our findings, we highlight the need to evaluate the impact of PCT testing on medical decision-making and patient outcomes in pneumonia and sepsis, including length of stay and antibiotic use, to guide future PCT recommendations and best practices.

For febrile young infants, the rise in PCT testing may also be due to the dissemination of robust clinical prediction rules that incorporate PCT.^8,9  Recent guidelines by the American Academy of Pediatrics published after this study period now recommend PCT for the workup of febrile young infants.¹³  We hypothesize that PCT use for febrile infants will continue to increase, although, consistent with other guideline uptake, adoption may take time.²⁰  We also hypothesize that comfort with and use of PCT in febrile young infants may increase PCT use for other diagnoses,²¹  despite the lack of guidelines recommending its use. Understanding PCT use before formal guidelines will help assess the impact of guidelines on overall use and practice variation over time.

This study has several limitations. First, we are unable to assess reasons for ordering PCT and diagnoses were identified retrospectively using diagnosis codes, which may vary on the basis of provider discretion. Second, we used previously published case-identification strategy for febrile young infants that includes admission and discharge diagnoses of fever and has a sensitivity of 77% and specificity of 98%.¹⁴  Thus, we may have missed some patients, particularly those diagnosed with a serious bacterial infection because many of them did not have a concurrent diagnosis of fever in the case-identification study and thus would not have been included in our study cohort.¹⁴  Finally, the data used are from children’s hospitals that are associated with PHIS; therefore, our results may not be generalizable to other settings.²² 

In summary, we found variability in PCT use across hospitals and rising rates of PCT use over time. Although some rising use of PCT is on the basis of evidence (particularly for febrile young infants), less is known about the impact of PCT on patient outcomes for children with pneumonia or sepsis. With an emphasis on high-value care, further robust research is needed to evaluate the impact of PCT on patient outcomes for children with pneumonia and sepsis to identify best practices for this relatively novel biomarker.