Trends in antifungal use in US hospitals, 2006–12

Abstract

Background

Although trends in antibiotic use have been characterized, less is known about antifungal use. Data on antifungal use are important for understanding practice patterns, assessing emergence of antifungal resistance and developing antifungal stewardship programmes. We estimated national trends in inpatient antifungal use in the USA.

Methods

Using billing data for antifungals from the Truven Health MarketScan^® Hospital Drug Database during 2006–12, we estimated the proportion of discharges at which antifungals were given and days of therapy (DOT)/1000 patient days (PDs) by antifungal drug type, year, patient and facility characteristics. We created national estimates using weights generated from Centers for Medicare and Medicaid Services data and assessed trends over time.

Results

Overall, 2.7% of all inpatients and 7.7% of those in ICUs received antifungals. The estimated DOT/1000 PDs for any antifungal was 35.0 for all inpatients and 73.7 for ICU patients. Azoles accounted for 80% of all antifungal use (28.5/1000 PDs), followed by echinocandins (5.0/1000 PDs). By multivariable trend analysis, DOT/1000 PDs for azoles (−21%) and polyenes (−47%) decreased between 2006 and 2012, whereas echinocandins increased 11% during 2006–10 and declined after 2011. Unspecified septicaemia, HIV and antineoplastic therapy were among the top primary diagnosis codes for patients who received antifungals.

Conclusions

Antifungals were most frequently used in ICU settings and fluconazole accounted for a large, but declining, proportion of antifungal use. Antifungal stewardship efforts may have the most impact if focused in ICUs, among certain patient groups (e.g. HIV and malignancy) and on stopping empirical antifungal therapy for unspecified sepsis when not indicated.

Introduction

Measuring antimicrobial use is an essential step in limiting inappropriate use and combating the spread of drug-resistant organisms.^1,2 Substantial strides have been made in quantifying amounts, trends and variability in antibiotic use in the USA through antimicrobial use prevalence surveys in US acute care hospitals³ and studies based on electronic and administrative data.^4–6 In these studies, use of antibacterial medications was common in acute care hospitals, with at least half of patients receiving these drugs while hospitalized. An estimated one-third of this prescribing could be eliminated through targeted use of diagnostic tests and reconsideration of dose and duration. Reducing inappropriate use can help slow the development of resistance, reduce wasted resources and improve patient outcomes by reducing adverse events such as Clostridium difficile infection.⁷

To date, most antimicrobial use surveillance has focused on antibacterial medications, with few studies examining antifungal use. Antibiotics are used with greater frequency than antifungals; however, given growing antifungal resistance, including emergence of resistant fungal species such as Candida auris, a better understanding of antifungal use practice patterns is needed to inform antifungal stewardship efforts. Stewardship efforts are particularly important because there are only a limited number of antifungals available (three major classes), antifungals are associated with a considerable number of side effects and interactions with other drugs and are expensive. An important step in any antimicrobial stewardship effort is to assess usage so that targets can be established for more optimal use. A better understanding of current inpatient antifungal use patterns is a priority in antifungal stewardship.

Several important events may have impacted recent trends in antifungal use, including approval of new antifungal drugs including micafungin, posaconazole and isavuconazole,^8–10 and changes in or release of new guidelines related to treatment and prophylaxis of fungal infections, which, importantly, have been in response to emerging resistance.^11–13 Our primary objective was to develop national estimates of and trends in antifungal use among inpatients in the USA. To this end, we used a proprietary administrative data set from a large and diverse population of US short-term acute care hospitals during 2006–12, the latest years for which comprehensive data exist, and extrapolated findings to all US short-stay acute care hospitals. A smaller data set was examined for 2013–14 to assess later trends.

Methods

The Truven Health MarketScan^® Hospital Drug Database (HDD) is a proprietary database containing hospital discharge records for patients discharged from participating US short-stay acute care hospitals. Information contained in the HDD includes patient demographic characteristics, information on inpatient stays, inpatient drug utilization data based on billing records and facility descriptors. All adult and paediatric patients discharged from any inpatient unit 1 January 2006–31 December 2012, in participating hospitals were included. For each hospital discharge, we identified antifungal doses charged to the patient during the inpatient stay. We searched the HDD for doses of any antifungal listed in the US FDA’s National Drug Code Directory¹⁴ or the WHO’s Anatomical Therapeutic Chemical (ATC)/DDD classification system.¹⁵ We eliminated any antifungals not administered by the oral or parenteral routes. Each antifungal was categorized into one of three types [azoles (triazoles, including fluconazole, itraconazole, voriconazole and posaconazole and imidazoles, including ketoconazole), echinocandins (anidulafungin, caspofungin and micafungin) and polyenes (any formulation of amphotericin B)] and the remaining were categorized as ‘other’ types. Similar to a past study, we further excluded the 2% of participating hospitals that did not submit any antifungal use data within a given year.⁴ As complete data from all participating hospitals were not available after 2012, we used a limited subset of hospitals that reported information for at least 5 years to conduct an analysis of data from 2009 to 2014.

We estimated measures of antifungal use in two ways, i.e. (i) proportion of hospital discharges in which a patient received at least one dose of an antifungal during their stay, and (ii) days of therapy (DOT) per 1000 patient days (PDs). One DOT represents the administration of a single agent on a given day regardless of the number of doses administered or dosage amount.¹⁶ In addition to calculating overall use measures, we stratified these measures by antifungal type, year, and other patient and facility characteristics.

To create national estimates of inpatient antifungal use, we performed a weighted extrapolation of data from the subset of US hospitals reporting to the HDD. These methods have been previously described and validated.⁴ We used data submitted to the Centers for Medicare and Medicaid Services Healthcare Cost Report Information System (HCRIS) to generate estimated weights to apply to HDD data based on the number of PDs and number of discharges among all US non-federal acute care hospitals for each year of the study, stratified by bed size (<300, ≥300), US census division, urban/rural status and teaching status.¹⁷ For the DOT measure, we used weights based on PDs and stratified them by critical care days versus all other inpatient days; for the proportion of discharges, weights based on discharges were used. Results are reported as the weighted national estimates except where otherwise noted. Previously,⁴ we validated the extrapolated national estimates by comparing with estimates from the Healthcare Cost and Utilization Project (HCUP) Nationwide Inpatient Sample (Agency for Healthcare Research and Quality, Rockville, MD, USA, http://hcupnet.ahrq.gov/) and found them to be similar.

To assess potential trends in antifungal use over time, we developed a multivariable model using linear regression based on the approximately normal distribution of the outcome and using generalized estimating equations (GEE) to account for the inter-hospital covariance. Our outcome was DOT/1000 PDs for each facility. Our main variable of interest was calendar year and we included facility characteristics such as case mix index, average age, bed size category, teaching status, facility urban or rural location, proportion of surgical discharges, average comorbidity score for the facility¹⁸ and facility geographical location as determined by US census division. We used revenue codes to identify days in which patients received care from ICUs; for those days, antifungal use was attributed to an intensive care location. We also included in the model, the proportion of inpatient days in which the primary ICD-9-CM diagnosis code for that admission was related to an infection, which we previously found to be an important predictor of hospital antibiotic use.¹⁹ To examine temporal trends in the proportion of discharges in which the patient received at least one antifungal dose and the proportion of days of oral azole use out of all days of azole use, we used a multivariable logistic model, which also used GEE to account for the inter-hospital covariance, and included the same covariates as above. However, since the model for posaconazole did not converge using GEE, a standard logistic model was used.

All data were analysed using SAS version 9.3 (SAS Institute Inc., Cary, NC, USA). This study did not require review by an investigational review board as the HDD contains no personally identifiable data and the data use agreement precluded any access by the investigators to such identifiers; therefore, the analysis did not involve human subjects.

Results

This study included Truven HDD data from 2006 to 2012, with over 34 million discharges from 552 short-stay acute care hospitals, representing 166 million PDs. During those 7 years, the number of hospitals contributing data in each year ranged from 300 to 383. Further details on hospital characteristics included in the study can be found in Table S1 (available as Supplementary data at JAC Online).

Table 1.

Extrapolated estimates of antifungal use in the Truven MarketScan^® HDD by year and various characteristics, 2006–12

	Year							All years
	2006	2007	2008	2009	2010	2011	2012
Proportion of discharges at which patients received at least one dose of antifungal given
antifungal class
all systemic antifungals	2.66%	2.70%	2.73%	2.73%	2.75%	2.74%	2.67%	2.71%
polyenes	0.09%	0.09%	0.08%	0.07%	0.07%	0.06%	0.06%	0.07%
echinocandins	0.22%	0.25%	0.29%	0.32%	0.33%	0.35%	0.33%	0.30%
azoles	2.51%	2.52%	2.54%	2.51%	2.52%	2.51%	2.44%	2.51%
fluconazole	2.40%	2.40%	2.42%	2.39%	2.38%	2.38%	2.30%	2.38%
ketoconazole	0.018%	0.021%	0.020%	0.018%	0.020%	0.015%	0.015%	0.02%
posaconazole	0.000%	0.004%	0.010%	0.012%	0.013%	0.009%	0.010%	0.009%
voriconazole	0.105%	0.108%	0.108%	0.113%	0.115%	0.114%	0.117%	0.11%
other	0.03%	0.03%	0.04%	0.04%	0.04%	0.04%	0.04%	0.04%
Antifungal DOT/1000 PDs
antifungal class
all systemic antifungals	35.5	35.4	35.9	35.8	35.0	34.3	32.9	35.0
polyenes	1.4	1.2	1.1	1.0	1.0	0.9	0.8	1.1
echinocandins	3.9	4.4	5.0	5.4	5.5	5.5	5.1	5.0
azoles	29.9	29.4	29.4	29.0	28.1	27.4	26.5	28.5
fluconazole	27.3	26.7	26.7	26.1	25.2	24.7	23.8	25.8
ketoconazole	0.2	0.2	0.2	0.2	0.2	0.2	0.1	0.2
posaconazole	0.0	0.1	0.2	0.2	0.2	0.2	0.2	0.1
voriconazole	1.9	2.0	2.0	2.1	2.2	2.1	2.1	2.1
other	0.40	0.35	0.39	0.40	0.44	0.45	0.38	0.40
census division
New England (CT, MA, ME, NH, RI, VT)	29.2	27.6	27.6	26.0	24.8	24.1	24.0	26.2
Mid Atlantic (NJ, NY, PA)	30.7	31.4	31.8	31.3	30.7	29.6	29.4	30.7
South Atlantic (DC, DE, FL, GA, MD, NC, SC, VA, WV)	42.0	39.8	40.4	40.7	39.3	39.7	39.3	40.2
Northeast Central (IL, IN, MI, OH, WI)	34.8	32.7	32.1	33.3	31.3	30.1	29.5	32.0
Southeast Central (AL, KY, MS, TN)	39.2	40.9	39.3	38.5	38.0	35.2	33.8	37.9
Northwest Central (IA, KS, MN, MO, ND, NE, SD)	28.0	27.0	28.0	25.1	24.4	23.8	25.0	26.0
Southwest Central (AR, LA, OK, TX)	40.4	44.5	47.2	45.0	43.9	43.1	41.1	43.6
Mountain (AZ, CO, ID, MT, NM, NV, UT, WY)	42.4	39.4	38.8	38.5	40.5	39.4	34.6	39.2
Pacific (AK, CA, HI, OR, WA)	26.0	28.3	29.9	34.0	35.1	34.7	29.1	31.2
age category (years)
0–17	12.0	11.6	10.4	10.4	10.8	10.1	10.1	10.8
18–44	32.7	32.7	33.0	31.9	30.9	30.2	29.0	31.5
45–64	48.0	46.8	46.8	48.6	46.5	45.4	43.4	46.5
65–84	39.1	39.3	40.7	39.6	39.7	38.5	36.6	39.1
≥85	24.4	24.7	25.3	26.0	24.6	23.8	23.1	24.6
sex
male	37.5	37.3	37.5	37.6	36.1	35.5	34.1	36.5
female	33.9	33.8	34.6	34.3	34.2	33.2	31.8	33.7
hospital location
urban	36.5	36.1	36.5	36.6	35.9	35.0	33.3	35.7
rural	27.5	28.8	30.1	28.2	27.1	27.0	28.3	28.2
hospital teaching status
teaching	37.7	36.9	36.8	36.0	34.9	33.7	32.8	35.6
non-teaching	33.1	33.7	34.9	35.5	35.2	34.8	32.9	34.3
hospital bed size
<300	32.5	32.7	34.8	34.5	32.9	32.4	31.8	33.1
≥300	38.2	37.8	36.8	36.8	36.8	35.8	33.8	36.6
care location
ICU	78.3	77.0	76.4	72.7	72.0	71.6	67.4	73.7
non-ICU	31.3	31.2	31.6	31.8	31.2	30.6	29.4	31.0
case mix index
first quartile	24.1	26.7	29.3	29.5	26.5	24.9	26.5	26.7
second quartile	28.4	27.6	28.0	30.1	29.7	29.8	25.8	28.5
third quartile	37.5	38.4	38.7	36.5	37.6	35.0	35.2	37.0
fourth quartile	41.9	41.5	40.2	41.6	39.1	39.6	37.7	40.2

Based on the weighting methodology described above, ∼2.7% of all hospital discharges involved patients who received at least one dose of an antifungal during their hospitalization. Among discharges in which the patient spent at least 1 day in an ICU, 7.7% involved patients who received an antifungal. Among discharges where the patient did not spend any time in the ICU, 2.2% involved patients who received an antifungal. The overall antifungal DOT for all study years combined was 35.0 per 1000 PDs (Table 1). Antifungal DOT in ICUs was more than twice as high compared with non-ICUs (73.7 versus 31.0 DOT/1000 PDs; P < 0.0001). Azoles, primarily fluconazole, were the most frequently used antifungal class, accounting for >80% of antifungal DOT (28.5/1000 PDs; 2.5% of all discharges), followed by echinocandins (5.0/1000 PDs; 0.3% of all discharges). Antifungal use was significantly greater in ICUs for all three types of antifungals (azoles, echinocandins and polyenes; P < 0.0001) than in non-intensive care locations.

Multivariable trend analysis showed a small, but statistically significant, decrease in overall antifungal use during 2006–12 as measured by DOT (−18.2% decrease; P < 0.001) (Figure 1). Temporal trends varied by antifungal type. Among the azoles, there was a significant decrease for fluconazole (−22.4%; P < 0.0001), a significant increase for posaconazole (0.0 in 2006 to 0.2 DOT/1000 PDs, in 2012; P = 0.0067) and no temporal trend for voriconazole (P = 0.48). Ketoconazole use declined from 0.2 to 0.1 DOT/1000 PDs in the last 2 years of the study period). There was a 21% increase in the odds of oral azole use during the study period (P = 0.0361). Polyene use also declined significantly over the study period (−46.8%; P < 0.0001). Echinocandin use varied significantly by year (P = 0.0005), with echinocandin use increasing from 2006 to 2010, but then decreasing between 2010 and 2012.

Figure 1.

Estimated trends in antifungal use measured by DOT/1000 PDs, Truven MarketScan^® HDD, USA, 2006–12. Estimated trends in antifungal use in GEE models controlling for year, case mix index, average age, bed size category, teaching status, facility urban or rural location, proportion of surgical discharges, average comorbidity score, facility geographical location, critical care setting and proportion of inpatient days in which the primary ICD-9-CM diagnosis code was related to an infection.

Trends in antifungal use measured as a proportion of discharges were similar to trends in DOT/1000 PDs. The proportion of discharges in which a patient received an antifungal declined from 2.9% in 2006 to 2.4% in 2012, yielding an adjusted OR of 0.83 (95% CI: 0.79–0.87). Results for azoles varied by agent. Fluconazole use declined (P < 0.0001), whereas posaconazole use increased (P < 0.0001) and no temporal trend was observed for voriconazole (P = 0.1661). For echinocandins, similar to DOTs, use increased between 2006 and 2010 followed by a decline in 2011 and 2012 (P = 0.0004).

Data were available for a limited number of hospitals (∼100) through 2014 because of a change in the data source. For hospitals with at least 5 years of data (including data from 2013 or 2014), we extended the primary analysis and found that similar overall trends continued for 2013–14 (Table 2). In particular, overall antifungal use continued to decline. Use of azoles, particularly fluconazole, and polyenes continued to decline and echinocandins returned to levels lower than 2009. Data were not extrapolated to the national level for this analysis.

Table 2.

Estimates of antifungal DOT/1000 PDs for participating hospitals with at least 5 years of data, 2009–14

	Year
	2009	2010	2011	2012	2013	2014
Number of hospitals	83	102	101	104	104	92
DOT/1000 PDs
all	36.88	36.42	35.25	33.72	32.52	30.5
polyene	0.75	0.77	0.77	0.63	0.62	0.66
echinocandin	5.54	4.84	5.05	4.41	4.63	4.61
azole	30.22	30.4	28.97	28.22	26.8	24.73
other	0.37	0.4	0.45	0.46	0.47	0.51

Data were not extrapolated to the national level for the estimates in Table 2 because of small sample size and limited geographical coverage for 2013–14.

Antifungal use varied by geographical region of the hospital even after controlling for facility characteristics in the multivariable model (P < 0.0001) (Figure 2). The South Atlantic, Southeast Central, Southwest Central and Mountain divisions had the highest DOT overall, whereas New England, Mid Atlantic and Northwest Central divisions had the lowest. Trends in antifungal use over time did not vary significantly by geographical region (P = 0.1399).

Figure 2.

Mean antifungal DOT/1000 PDs by census division in the USA, Truven MarketScan^® HDD, 2006–12. DOT by census division estimated in GEE models controlling for year, case mix index, average age, bed size category, teaching status, facility urban or rural location, proportion of surgical discharges, average comorbidity score, facility geographical location, critical care setting and proportion of inpatient days in which the primary ICD-9-CM diagnosis code was related to an infection.

Antifungal use did not vary significantly by facility bed size or teaching status (P = 0.5937 and P = 0.0723, respectively) (Table 1). Based on our GEE model, hospitals identified as urban had higher antifungal DOT compared with rural hospitals (P = 0.0045). Antifungal use was significantly associated with higher hospital case mix index (P = 0.0003). Hospitals in the highest quartile for case mix had ∼20% higher antifungal DOT/1000 PDs compared with hospitals in the lowest quartile in our adjusted estimates (P = 0.0004).

Patients who received antifungals were older (mean: 59.4 versus 50.0 years) and had considerably longer lengths of stay (mean: 13.4 days versus 4.7 days) than patients who did not receive them (Table 3). Twenty-nine percent of patients who received an antifungal were in an ICU compared with 10.1% of those who did not receive an antifungal. Unspecified septicaemia was the most common disease category for which antifungals were administered (9.3%). Forty-one percent of patients admitted with HIV and 16% of those admitted for antineoplastic chemotherapy received antifungals.

Table 3.

Extrapolated estimates of demographic and clinical characteristics of patients receiving antifungals in the Truven MarketScan^® HDD, 2012

Characteristic	Percentage (unless otherwise stated)		Percentage of patients with a particular characteristic/ condition who received an antifungal
	did not receive antifungal	received antifungal
Age (years), mean (median)	50.0 (55.0)	59.4 (62.0)
Length of stay (days), mean (median)	4.7 (3.0)	13.4 (8.0)
Sex
male	43.2	41.6	2.6
female	56.8	58.4	2.7
Age category (years)
0–17	13.8	2.6	0.5
18–44	24.5	18.4	2.0
45–64	25.9	34.9	3.6
65–84	27.3	35.6	3.5
≥85	8.5	8.6	2.7
Intensive care	10.1	28.6	7.2
Diagnostic-related group
medical	73.9	71.3	2.6
surgical	26.1	28.7	2.9
Top six primary diagnosis codes
038.9: unspecified septicaemia	2.0	9.3	11.6
486: pneumonia, organism unspecified	2.2	3.7	4.4
V57.89: care involving other specified rehabilitation procedure	1.3	2.1	4.5
042: HIV disease	0.1	2.0	40.7
584.9: acute kidney failure, unspecified	1.3	1.9	3.9
V58.11: encounter for antineoplastic chemotherapy	0.3	1.8	15.8

Discussion

This study is the first, to our knowledge, to provide national estimates of antifungal use in US acute care hospitals. We estimate that overall, 2.7% of discharges from US short-stay acute care hospitals during 2006–12 involved patients who received systemic antifungals. Overall antifungal use amounted to 35 DOT/1000 PDs. Antifungal use was higher in ICU settings, where 7.7% of patients received an antifungal, and antifungal DOT in this setting was twice as high at 73.7/1000 PDs as in other inpatient settings. Fluconazole was the most commonly administered antifungal, though its use declined by 20% during the study period. We observed a trend of increasing use of newer azoles such as posaconazole, which was first approved for use in 2006. Ketoconazole use declined in the last 2 years of the study, possibly reflecting the increasing recognition of associated hepatotoxicity and adrenal insufficiency, which resulted in the FDA restricting the use of oral ketoconazole in 2013. Echinocandins made up ∼15% of overall antifungals used in acute care hospitals and their use increased initially during 2006–10, but declined during 2011–14. Septicaemia was the most common diagnosis for which antifungals were administered and a substantial proportion of patients admitted with HIV and patients admitted for chemotherapy received antifungals.

The finding that 2.7% of US inpatients received antifungals is similar to the 4% reported in a recent European point prevalence survey of healthcare-associated infections and antimicrobial use in acute care settings.²⁰ The proportion of patients who received antifungals is much lower than the estimated 55% of US inpatients who receive antibacterial drugs during a hospitalization.⁴ Overall antifungal DOT was also an order of magnitude lower than antibacterial DOT (35 versus 775 DOT/1000 PDs).⁴ Lower overall use of antifungals is not surprising as there are more classes of antibacterial drugs and a wider range of indications for antibacterial drugs. Antifungals, however, were much more commonly used in intensive care settings than general hospital settings. Antifungals are also often more expensive than antibacterials and can contribute disproportionately to the cost of drug expenditure for hospitals.

The decline observed in inpatient antifungal use during 2006–12 is perhaps surprising given the growing population of immunocompromised patients susceptible to opportunistic fungal infections. However, these findings are consistent with a previous report in which spending on antifungals in US hospital settings decreased during 2005–15.²¹ Although trends in expenditure may not reflect actual use of antifungals, the concordant findings of these two studies suggest that the decline in use is real. The decline in overall antifungal use may partly be explained by the decline in specific fungal infections, including candidaemia²² and HIV-related opportunistic infections.²³ Consistent with an increasing proportion of fungal infections attributed to invasive mould infections²⁴ and drug-resistant Candida,²⁵ we observed an increase in the proportion of patients receiving voriconazole, posaconazole and echinocandins, a finding that might also be explained by the availability of these newer agents.^8–10 Another reason for the decrease in inpatient antifungal use may be that care is increasingly shifted to outpatient settings. Whereas costs of antifungals in inpatient settings declined during 2005–15 by nearly 20%, costs in outpatient settings increased 200%.²¹ With the caveat that antifungal expenditure may not reflect the actual number of doses used, much of the growth in outpatient expenditure was for echinocandins and polyenes, indicating that parenteral outpatient antifungal therapy may be becoming more widespread. Further investigation into reasons for the decline in antifungal use in inpatient settings is needed.

Decreases in use of antifungals as seen in this study does not necessarily mean that antifungals are being used appropriately. We were not able to assess the appropriateness of antifungal use in our study. However, a Spanish study of antifungal use in a tertiary care hospital demonstrated that antifungals were used unnecessarily in 16% of cases, the wrong antifungal was used in about one-third of cases and the suboptimal duration of treatment occurred in almost half of patients.²⁶ In another study of antifungal use in surgical ICU patients, 88% of antifungals prescribed were given empirically or pre-emptively, whereas only 12% had candidiasis.²⁷ The most common condition for which antifungals were administered in this study was unspecified septicaemia. Antifungal use in the setting of sepsis is usually empirical and prompted by a lack of improvement with broad-spectrum antibacterials. Although fungi cause a small proportion of sepsis cases,²⁸ the high mortality from sepsis in the absence of early treatment^29,30 drives empirical antifungal use. The problem is compounded by the lack of fast and sensitive diagnostic tests for ruling out fungal infections. Improvement in fungal diagnostics could lead to lower use of antifungal agents and help define appropriate use. New diagnostic tests, including those that are culture-independent (e.g. T2 Candida) or that can identify an organism early during culture (e.g. MALDI-TOF MS), can reduce time to effective and optimal antifungal therapy and allow de-escalation of empirical therapy.^31,32

Decline in antifungal use also does not indicate that antifungal stewardship efforts in the context of larger antimicrobial stewardship programmes are not needed. There are no studies assessing the proportion of US hospitals that have included antifungal stewardship as part of antimicrobial stewardship programmes, but the experience in other countries such as the UK³³ suggests it is likely to be low. Antifungal stewardship efforts have the potential to improve patient outcomes by decreasing inappropriate antifungal use, preventing resistance and preserving antifungal treatment options. Antifungal stewardship efforts focused on empirical use in ICUs and other units where most antifungal use takes place (e.g. oncology units), may have the greatest impact.

Intensive antifungal use, as we observed in ICU settings, has been correlated with higher minimum inhibitory concentrations to particular antifungal drugs among certain Candida species and a rise in drug-resistant Candida species.²⁸ Because of rising resistance to fluconazole, initial therapy for suspected fungal infections now mostly consists of echinocandins. The 2016 IDSA’s candidiasis treatment guidelines recommend the use of echinocandins as first-line therapy for invasive candidiasis in all adult patient populations.¹² This is a shift from the 2009 guidelines, which reserved the use of echinocandins for certain populations, such as patients with neutropenia or previous candidaemia treatment.¹¹ Because of the limited number of antifungal drug classes available, echinocandins often serve as both first-line treatment for candidaemia and last-line treatment for drug-resistant fungi. In addition to ensuring that echinocandins are administered for the correct indication at the appropriate dose for the appropriate duration, it is important to preserve their efficacy by de-escalating therapy to fluconazole when an organism is found to be susceptible. Although our study found a decline in inpatient echinocandin use during 2010–14, echinocandin use may increase again following these new guidelines. Ongoing monitoring of antifungal use trends will be essential to document shifts in use based on these new practice guidelines.

In addition to preventing resistance, antifungal stewardship can also be cost saving because antifungal drugs are expensive. Each year, >$300 million are spent on antifungals at non-federal acute care hospitals in the USA.²¹ In a cost study at a major medical centre, implementation of an antimicrobial stewardship programme was associated with ∼$3 million dollars in cost savings over 3 years, driven largely by a reduction in antifungal use.³⁴ Other groups have also reported cost savings from antifungal stewardship programmes.^26,35

Finally, similar to the highly variable rates of antibacterial use by US region,³⁶ antifungal use differed substantially by region, even when adjusting for hospital-level factors. Reasons for these differences could include differences in rates of fungal infections (e.g. if mould infections are more common in the southern and Mountain census divisions than in the northeastern ones).²⁴ Likewise, some of the geographical variability may be explained by geographically restricted fungal infections, particularly coccidioidomycosis and histoplasmosis, which are more common in some divisions where antifungal use rates were high.³⁷ Other reasons for such differences in antifungal use could include different practice patterns and underlying patient populations. It is important to note that even though regional differences in levels of antifungal use existed, the overall trends in antifungal use over time were similar across all regions.

This study had a number of strengths. Over 300 hospitals were included for each year, representing a diversity of sizes, teaching status and urban/rural locations in each US geographical division. In addition, including data over several years allowed for trend analysis. Detailed data contained information on a range of hospital- and patient-specific demographic and clinical characteristics, allowing inclusion of this information in multivariate models. Further, data from HCRIS allowed us to weight our estimates and extrapolate findings to the national level.

Administrative data are widely used in healthcare research,^38–41 but have limitations, in part because they are collected primarily for billing purposes and require adaptation for research. There is likely to be misclassification in pharmacy data as well as the clinical and facility information; however, this bias is most likely non-differential. Further, use of pharmacy charge data to estimate medication use in hospitals was previously validated in a small sample with excellent agreement⁴² and more recently in a small group of paediatric hospitals.⁴³ Likewise, some geographical regions are over- or under-represented by the distribution of hospitals in this sample compared with the distribution of all US hospitals. Regional adjustment using nationally representative HCRIS data helped mitigate this problem; however, the small number of hospitals in some regions may still limit the reliability of certain geographical estimates and rankings. In addition, extrapolated estimates were limited to acute care, non-federal hospitals and do not adequately represent children’s hospitals, which constitute <1% of included hospitals. Finally, owing to changes in the data source, we were unable to generate national estimates after 2012, as the limited 2013–14 data sets did not include hospitals from each geographical region. However, the limited analysis of 2013–14 data yielded trends similar to those observed through 2012. The time frame of this study did not extend to when the newest antifungal drug isavuconazole was approved and put into use. Therefore, we could not assess the trends in isavuconazole use.

In conclusion, to the best of our knowledge, this study provides the first national estimate of antifungal use in inpatients in the USA. Although overall antifungal use among inpatients is low compared with antibacterial use, higher use was found in specific settings such as large urban hospitals that treat patients with high-acuity conditions, particularly in ICUs where ∼1 in 14 patients receive antifungals, and among specific patient groups, such as those being treated for HIV and receiving chemotherapy. Despite low overall use, the cost of antifungals is high and overuse can lead to adverse patient outcomes and development of resistance. A better understanding of antifungal use allows for more targeted antifungal stewardship efforts within existing antimicrobial stewardship programmes. Further studies are needed to define where antifungal stewardship efforts could have the most impact. Continued monitoring of antifungal use trends, particularly in light of candidiasis guideline changes, will shed light on the best ways to encourage appropriate antifungal use in the USA.

Acknowledgements

We would like to thank Dr Lauri Hicks for her input on this paper.

Funding

This study was carried out as part of our routine work.

Transparency declarations

None to declare.

Disclaimer

The findings and conclusions in this manuscript are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention.

Supplementary data

Table S1 is available as Supplementary data at JAC Online.