Abstract
We evaluated healthcare facility use of International Classification of Diseases, Tenth Revision (ICD-10) codes for culture-confirmed candidemia cases detected by active public health surveillance during 2019–2020. Most cases (56%) did not receive a candidiasis code, suggesting that studies relying on ICD-10 codes likely underestimate disease burden.
Keywords: candidemia, candidiasis, international classification of diseases, public health surveillance, sepsis
Candidemia, a life-threatening bloodstream infection (BSI) caused by the commensal yeast Candida, is associated with prolonged hospitalizations and substantial healthcare costs [1]. Candidemia is one of the most common BSIs in the United States, with an average incidence of approximately 9 cases per 100 000 population during 2012–2016 [2, 3]. Ongoing, active, population-based surveillance for candidemia is critical for characterizing its epidemiology and consequently informing treatment and prevention strategies. An alternative approach to studying medical conditions (including invasive candidiasis) more broadly involves using International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) diagnosis codes to query large administrative databases [4]. These databases are commonly used to characterize epidemiologic patterns and estimate disease burden because they offer opportunities for relatively quick and inexpensive analyses on large sample sizes [5]. However, analyses relying on ICD-10-CM codes may be inherently biased because these codes primarily exist for billing and payment purposes [5, 6]. The performance of ICD-10-CM codes for ascertaining invasive candidiasis cases has not been extensively evaluated and affects the conclusions that can be drawn from analyses based on ICD-10-CM codes. Therefore, we aimed to describe ICD-10-CM codes among candidemia cases identified by active, laboratory-based public health surveillance.
METHODS
The Emerging Infections Program (EIP) conducts active, population-based surveillance for candidemia in specific counties in 10 US states: California, Colorado, Connecticut, Georgia, Maryland, Minnesota, New Mexico, New York, Oregon, and Tennessee, representing approximately 21.5 million people. A candidemia case is defined as a positive blood culture for Candida species in a surveillance area resident. Any positive blood cultures for Candida species within 30 days of the initial positive culture (incident specimen) are considered part of the same case, whereas a positive blood culture for Candida species after the 30-day period is considered a new case in the same patient.
Detailed surveillance methods have been described elsewhere [3, 7]; in brief, EIP surveillance personnel collect demographic and clinical information from medical and laboratory records using a standardized case report form (CRF). The form also captures whether the case was assigned certain ICD-10-CM codes: B37 (candidiasis), P37.5 (neonatal candidiasis), B48 (other mycoses, not classified elsewhere), B49 (unspecified mycoses), T80.211 (BSI due to central venous catheter), A41.9 (sepsis, unspecified organism), and R65.2 (severe sepsis).
We limited the analysis to cases reported in 2019–2020 among patients hospitalized within 0–7 days of incident specimen collection (95.2% of all cases during 2019–2020). We excluded hospitalized cases without ICD-10-CM data available from the CRF (4.8%). We described and compared demographic and clinical features among cases with versus without candidiasis ICD-10-CM codes (B37 and P37.5) and other related codes using χ2 tests (α < 0.05). We performed multivariable logistic regression to identify independent predictors of receiving a candidiasis ICD-10-CM code. Model selection began with all demographic and clinical features assessed in bivariate analyses, using backwards stepwise selection, and considering 2-way interaction terms. Statistical analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC).
RESULTS
Among 2838 laboratory-confirmed candidemia cases, 1258 (44.3%) had ≥1 ICD-10-CM code for candidiasis; among those, B37.7 (“Candidal sepsis”) was the most common subcode (n = 834, 66.3%). Subcodes for other specified sites of candidiasis were uncommon (Table 1).
Table 1.
Demographic Characteristics, Clinical Features, Outcomes, and Other ICD-10-CM Codes Among Candidemia Cases With and Without ICD-10-CM Codes for Candidiasis, United States, 2019–2020
| Characteristics | Total Cases (n = 2838) | Cases With a Candidiasis Code (B37 or P37.5) (n = 1258) | Cases Without a Candidiasis Code (n = 1580) | P Value |
|---|---|---|---|---|
| n | n (%) | n (%) | ||
| Median age in years (IQR) | 61 (47–72) | 61 (46–72) | 61 (47–72) | .167 |
| Age Group (Years) | … | … | … | .267 |
| ≤18 | 102 | 54 (4.3) | 48 (3.0) | … |
| 19–44 | 532 | 241 (19.2) | 291 (18.4) | … |
| 45–64 | 1013 | 436 (34.7) | 577 (36.7) | … |
| ≥65 | 1191 | 527 (41.9) | 664 (42.0) | … |
| Sex | … | … | … | .059 |
| Male | 1618 | 742 (59.0) | 876 (55.4) | … |
| Female | 1220 | 516 (41.0) | 704 (44.6) | … |
| Race (n = 2563) | … | … | … | .038 |
| White | 1625 | 763 (66.0) | 862 (61.3) | … |
| Black | 796 | 330 (28.6) | 466 (33.1) | … |
| Other or multiple race | 142 | 63 (5.5) | 79 (5.6) | … |
| Ethnicity (n = 2598) | … | … | … | .481 |
| Hispanic/Latino | 264 | 114 (9.7) | 150 (10.5) | … |
| Non-Hispanic/Latino | 2334 | 1061 (90.3) | 1273 (89.5) | … |
| Surveillance Site | … | … | … | <.0001 |
| California | 197 | 110 (8.7) | 87 (5.5) | … |
| Colorado | 255 | 102 (8.1) | 153 (9.7) | … |
| Connecticuta | 497 | 267 (21.2) | 230 (14.6) | … |
| Georgia | 630 | 272 (21.7) | 350 (22.6) | … |
| Maryland | 339 | 119 (9.5) | 220 (13.9) | … |
| Minnesota | 336 | 118 (9.4) | 218 (13.8) | … |
| New Mexico | 75 | 37 (2.9) | 38 (2.4) | … |
| New York | 133 | 74 (5.9) | 59 (3.7) | … |
| Oregon | 106 | 32 (2.5) | 74 (4.7) | … |
| Tennessee | 270 | 126 (10.0) | 144 (9.1) | … |
| Underlying Conditions | … | … | … | … |
| Diabetes (n = 2837) | 1061 | 481 (38.3) | 580 (36.7) | .395 |
| Malignancy | 692 | 299 (23.8) | 393 (24.9) | .496 |
| Liver disease | 468 | 192 (15.3) | 276 (17.5) | .116 |
| Chronic renal disease | 730 | 317 (25.2) | 413 (26.1) | .569 |
| Transplant | 106 | 47 (3.7) | 59 (3.7) | 1.000 |
| HIV | 53 | 25 (2.0) | 28 (1.8) | .674 |
| Surgery in the 90 days before DISC | 796 | 352 (28.0) | 444 (28.1) | .944 |
| Injection drug use in the past year | 252 | 118 (9.4) | 134 (8.5) | .403 |
| Clinical Features and Outcomes | … | … | … | … |
| Hospitalized in the 90 days before DISC (n = 2795) | 1392 | 645 (51.9) | 747 (48.2) | .053 |
| Had a previous candidemia case (n = 2836) | 161 | 84 (6.7) | 77 (4.9) | .039 |
| Had a CVC in the 2 days before DISC (n = 2821) | 1737 | 771 (61.7) | 966 (61.5) | .918 |
| Neutropenia in the 2 days before DISC (n = 2391) | 152 | 70 (6.5) | 82 (6.2) | .796 |
| Additional bloodstream organisms isolated in the 7 days before DISC (n = 2827) | 815 | 360 (28.7) | 455 (28.9) | .880 |
| Subsequent positive Candida blood culture in the 30 days after DISC (n = 2825) | 757 | 425 (33.9) | 332 (21.2) | <.0001 |
| ICU 14 days before or after DISC | 1775 | 742 (59.0) | 1033 (65.4) | .0005 |
| Antifungal treatment in the 14 days before DISC (n = 2808) | 406 | 205 (16.5) | 201 (12.9) | .007 |
| Antifungal treatment after DISC (n = 2819) | 2413 | 1235 (98.4) | 1178 (75.3) | <.0001 |
| COVID-19 in the 30 days before DISC (n = 1363; 2020 cases only) | 226 | 69 (12.9) | 157 (19.0) | .003 |
| Died during candidemia-related hospitalization (n = 2836) | 950 | 293 (23.3) | 657 (41.6) | <.0001 |
| Died within 48 hours of DISC (n = 2719) | 296 | 16 (1.3) | 280 (18.9) | <.0001 |
| Discharged alive within 48 hours of DISC (n = 1880 discharged alive) | 111 | 16 (1.7) | 95 (10.3) | <.0001 |
| Candida Species From First Positive Blood Culture | … | … | … | … |
| Candida albicans | 1089 | 490 (39.0) | 599 (37.9) | .572 |
| Candida glabrata | 855 | 352 (28.0) | 503 (31.8) | .026 |
| Candida parapsilosis | 369 | 162 (12.9) | 207 (13.1) | .860 |
| Candida tropicalis | 165 | 86 (6.8) | 79 (5.0) | .038 |
| Other non-albicans speciesb | 272 | 126 (10.0) | 146 (9.2) | .486 |
| >1 Candida species | 78 | 38 (3.0) | 40 (2.5) | .429 |
| Non-Candidiasis ICD-10-CM Codes | … | … | … | … |
| B48 Other mycoses, not elsewhere classified | 28 | 13 (1.0) | 15 (1.0) | .822 |
| B49 Unspecified mycoses | 640 | 176 (14.0) | 464 (29.4) | <.0001 |
| T80.211 BSI due to CVC | 231 | 160 (12.7) | 71 (4.5) | <.0001 |
| A41.9 Sepsis, unspecified organism | 917 | 302 (24.0) | 615 (38.9) | <.0001 |
| R65.2 Severe sepsis | 1037 | 480 (38.2) | 557 (35.3) | .111 |
| Candidiasis ICD-10-CM Codes | … | … | … | … |
| B37.0 Candidal stomatitis | … | 57 (4.5) | … | … |
| B37.1 Pulmonary candidiasis | … | 24 (1.9) | … | … |
| B37.2 Candidiasis of skin and nail | … | 21 (1.7) | … | … |
| B37.3 Candidiasis of vulva and vagina | … | 10 (0.8) | … | … |
| B37.4 Candidiasis of other urogenital sites | … | 96 (7.6) | … | … |
| B37.5 Candidal meningitis | … | 13 (1.0) | … | … |
| B37.6 Candidal endocarditis | … | 30 (2.4) | … | … |
| B37.7 Candidal sepsis | … | 834 (66.3) | … | … |
| B37.8 Candidiasis of other sites | … | 130 (10.3) | … | … |
| B37.9 Candidiasis, unspecified | … | 210 (16.7) | … | … |
| P37.5 Neonatal candidiasis | … | 11 (0.9) | … | … |
Abbreviations: BSI, bloodstream infection; COVID-19, coronavirus disease 2019; CVC, central venous catheter; DISC, date of incident specimen collection; HIV, human immunodeficiency virus; ICD-10-CM, International Classification of Diseases, Tenth Revision; ICU, intensive care unit; IQR, interquartile range.
Connecticut is the only statewide surveillance site.
Sixty-seven Candida dubliniensis, 12 Candida guilliermondii, 62 Candida krusei, 83 Candida lusitaniae, and 48 other Candida species.
By surveillance site, the percentage of cases with a candidiasis code ranged from 30.2% (Oregon) to 55.8% (California). Compared with those with a candidiasis code, patients without a candidiasis code were more likely to die at any time during hospitalization (41.6% vs 23.3%, P < .001) and die within 48 hours of incident specimen collection (18.9% vs 1.3%, P < .001). Overall, 38.9% of cases and 30.6% of cases among patients who survived to hospital discharge did not have any fungal code (B37, P37.5, B48, or B49).
Cases without a candidiasis code were more likely to have the ICD-10-CM code B49 for “Unspecified mycoses” (29.4% vs 14.0% with a candidiasis code, P < .001) and code A41.9 for “Sepsis, unspecified organism” (38.9% vs 24.0%, P < .001). However, cases with a candidiasis code were more likely to have ICD-10-CM code T80.211 for “BSI due to central venous catheter” (12.7% vs 4.5% without a candidiasis code, P < .001). Having additional non-Candida organisms isolated from the bloodstream was not associated with receiving a candidiasis code but was associated with receiving code R65.2 for “Severe sepsis” (32.7% vs 26.6% without this code, P = .0007) (data not shown in table). Among cases reported in 2020 (n = 1363), those with coronavirus disease 2019 (COVID-19) were less likely to receive a candidiasis code than those without COVID-19 (30.5% vs 41.5%, P = .003).
In the age-adjusted multivariable analysis, Black race (adjusted odds ratio [aOR], 0.79; 95% confidence interval [CI], 0.64–0.98) and death during hospitalization (aOR, 0.58; 95% CI, .48–.70) were associated with lower odds of receiving a candidiasis ICD-10-CM code, whereas antifungal treatment (on or after incident specimen collection date) was associated with higher odds (aOR, 20.21; 95% CI, 12.36–33.06) (Table 2). The odds of receiving a candidiasis ICD-10-CM code varied significantly by surveillance site (P < .001), with cases from certain sites having over 4 times the odds compared with others.
Table 2.
Multivariable Model of Factors Associated With Receiving an ICD-10-CM Code for Candidiasis, United States, 2019–2020a
| Characteristics | Adjusted Odds Ratio | 95% Confidence Interval | P Value |
|---|---|---|---|
| Age Group (Years) | … | … | .502 |
| ≤18 | Reference | Reference | … |
| 19–44 | 0.74 | 0.45–1.23 | … |
| 45–64 | 0.82 | 0.50–1.34 | … |
| ≥65 | 0.87 | 0.53–1.41 | … |
| Race | … | … | .085 |
| White | Reference | Reference | … |
| Black | 0.79 | 0.64–0.98 | … |
| Other or multiple race | 0.84 | 0.56–1.25 | … |
| Surveillance Site | … | … | <.0001 |
| California | 4.27 | 2.30–7.91 | … |
| Colorado | 1.81 | 1.02–3.19 | … |
| Connecticut | 3.51 | 2.06–5.99 | … |
| Georgia | 2.06 | 1.21–3.50 | … |
| Maryland | 1.33 | 0.76–2.30 | … |
| Minnesota | 1.14 | 0.66–1.96 | … |
| New Mexico | 4.18 | 1.91–9.12 | … |
| New York | 3.01 | 1.63–5.57 | … |
| Oregon | Reference | Reference | … |
| Tennessee | 2.26 | 1.30–3.93 | … |
| Antifungal treatment after DISC | 20.21 | 12.36–33.06 | <.0001 |
| Died during candidemia-related hospitalization | 0.58 | 0.48–0.70 | <.0001 |
Abbreviations: DISC, date of incident specimen collection; ICD-10-CM, International Classification of Diseases, Tenth Revision.
In exploratory analyses, having a subsequent positive Candida culture was also a significant independent predictor and had significant interaction with antifungal treatment. However, the 2 variables were strongly associated (P < .0001), with 98.3% of the cases who had a subsequent positive Candida culture receiving antifungal treatment. Therefore, we removed subsequent positive culture from the final model, given that it may reflect not only illness duration but also facility-dependent testing practices, whereas antifungal treatment is more likely to indicate clinical certainty for candidemia.
DISCUSSION
We evaluated the use of selected ICD-10-CM codes among candidemia cases identified through active laboratory-based surveillance in 10 US locations. Despite having culture-confirmed candidemia, more than half of cases did not receive an ICD-10-CM code for candidiasis and instead received codes for Unspecified mycoses and Sepsis, unspecified organism. These nonspecific codes were particularly common among patients who died in the hospital. Variation in coding by race and surveillance site suggests that analyses of candidemia based on ICD-10-CM codes might undercount cases among certain populations.
Our results suggest that the sole use of ICD-10-CM code B37.7, “Candidal sepsis” for candidemia case ascertainment might miss approximately 70% of cases, similar to a study that showed the sensitivity of code B37.7 was 43% for detecting invasive candidiasis among pediatric cancer patients in Australia [8]. The low proportion of candidemia cases with a candidiasis code might reflect limitations of the ICD-10-CM coding system, because no single ICD-10-CM code for candidemia exists. The code B37.7 Candidal sepsis is closely related to candidemia, supported by our finding that it was the most common B37 subcode; however, candidemia does not always result in sepsis. Presumably, all candidemia patients surviving until hospital discharge should have received at least one B37 subcode, if not specifically B37.7; however, other subcodes were uncommonly listed and could represent additional sites of infection beyond BSI, a possibility that we did not evaluate. Approximately one third of candidemia cases without code B37 received code B49 for Unspecified mycoses, which poses a considerable challenge for interpreting analyses of administrative data and comprises a large portion of the overall burden of fungal disease [9]. More consistent coding for candidemia would undoubtedly help improve analyses of administrative data. Administrative data remain a valuable resource, despite ICD-10-CM codes’ varying sensitivity (eg, ∼60% for invasive mold infection; 98% for COVID-19) [5, 10, 11].
Factors related to the coding process likely influenced our findings. In-hospital mortality, particularly within 48 hours of initial blood culture, seemed to be a barrier to receiving a candidiasis code, probably because blood cultures for Candida take an average of 2–3 days to turn positive [12]. Medical coders might not revisit culture results and providers might not adjust documentation of medical conditions if a patient died before results became available, resulting in incomplete coding [13]. Therefore, given the high mortality rates, candidemia might be undercounted in data on in-hospital deaths. Controlling for mortality, cases that received antifungal treatment were more likely to receive a candidiasis ICD-10-CM code, likely reflecting clinical suspicion of the diagnosis.
The relationship between candidemia and sepsis is challenging to disentangle with ICD-10-CM codes. That approximately 40% of candidemia cases without a candidiasis code received code A41.9 for Sepsis, unspecified organism perhaps indicates uncertainty about the etiology. In contrast, cases with code R65.2 for Severe sepsis were more likely to have polymicrobial infections, illustrating the medical complexity of candidemia patients. Sepsis is challenging to diagnose and difficult to measure with coding data, particularly because initiatives to improve sepsis recognition have resulted in increased sepsis coding over time [14]. It is unfortunate that we were not able to assess temporal trends in candidemia coding practices with only 2 years of data. Further research could help clarify the clinical, process-related, and policy-related factors, including hospital reimbursement, that influence candidemia coding in the context of sepsis.
Racial differences in candidemia are well documented, with rates in Black patients over twice as high as non-Black patients, possibly because of disparities in socioeconomic status, access to healthcare, and underlying conditions that increase risk for candidemia [3]. Factors related to quality of care or facility-dependent practices might have influenced the lower candidiasis coding we observed for cases in Black patients, particularly given the significant variation by surveillance site. These findings suggest that ICD-10-CM code-based analyses might underestimate candidemia cases among Black patients and in certain geographic areas. Similarly, undercounting of COVID-19-associated candidiasis may occur, based on our bivariate results.
CONCLUSIONS
In conclusion, this study provides context for ICD-10-CM code-based analyses of candidemia and highlights the importance of continued public health surveillance and attention to improved coding practices for candidemia.
Acknowledgments
We thank Malavika Rajeev from the Mycotic Diseases Branch, Centers for Disease Control and Prevention; Joelle Nadle from the California Emerging Infections Program; Hazal Kayalioglu from the Connecticut Emerging Infections Program; Stephanie Thomas, Lewis Perry, Torrey Knight, and Annie Coffin from the Georgia Emerging Infections Program; Lindsay Bonner and Vijitha Lahanda Wadu from the Maryland Emerging Infections Program; Anita Gellert from the New York Emerging Infections Program; Adel Mansour from the Oregon Emerging Infections Program; Tiffanie Markus, Caroline Graber, and Sandra Hardin from the Tennessee Emerging Infections Program; and the participating Emerging Infections Program Laboratories.
Disclaimer. The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention (CDC). This activity was reviewed by the CDC and was conducted consistent with applicable federal law and CDC policy (eg, 45 C.F.R. part 46.102(l)(2), 21 C.F.R. part 56; 42 U.S.C. §241(d); 5 U.S.C. §552a; 44 U.S.C. §3501 et seq.).
Financial support. This work was funded by the Centers for Disease Control and Prevention (Grant Number CDC-RFA-CK17-1701).
Contributor Information
Kaitlin Benedict, Division of Foodborne, Waterborne, and Environmental Diseases, National Center for Emerging and Zoonotic Infectious Diseases, Centers for Disease Control and Prevention, Atlanta, Georgia, USA.
Jeremy A W Gold, Division of Foodborne, Waterborne, and Environmental Diseases, National Center for Emerging and Zoonotic Infectious Diseases, Centers for Disease Control and Prevention, Atlanta, Georgia, USA.
Emily N Jenkins, Division of Foodborne, Waterborne, and Environmental Diseases, National Center for Emerging and Zoonotic Infectious Diseases, Centers for Disease Control and Prevention, Atlanta, Georgia, USA; ASRT, Inc., Atlanta, Georgia, USA.
Jeremy Roland, California Emerging Infections Program, Oakland, California, USA.
Devra Barter, Colorado Department of Public Health and Environment, Denver, Colorado, USA.
Christopher A Czaja, Colorado Department of Public Health and Environment, Denver, Colorado, USA.
Helen Johnston, Colorado Department of Public Health and Environment, Denver, Colorado, USA.
Paula Clogher, Connecticut Emerging Infections Program, Yale School of Public Health, New Haven, Connecticut, USA.
Monica M Farley, Emory University School of Medicine, Atlanta, Georgia, USA; Atlanta VA Medical Center, Atlanta, Georgia, USA.
Andrew Revis, Atlanta VA Medical Center, Atlanta, Georgia, USA; Georgia Emerging Infections Program, Atlanta, Georgia, USA; Foundation for Atlanta Veterans Education and Research, Atlanta, Georgia, USA.
Lee H Harrison, Department of International Health, Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland, USA.
Laura Tourdot, Minnesota Department of Health, St. Paul, Minnesota, USA.
Sarah Shrum Davis, New Mexico Emerging Infections Program, Albuquerque, New Mexico, USA.
Erin C Phipps, New Mexico Emerging Infections Program, Albuquerque, New Mexico, USA; University of New Mexico, Albuquerque, New Mexico, USA.
Christina B Felsen, University of Rochester School of Medicine, Rochester, New York, USA.
Brenda L Tesini, University of Rochester School of Medicine, Rochester, New York, USA.
Gabriela Escutia, Public Health Division, Oregon Health Authority, Portland, Oregon, USA.
Rebecca Pierce, Public Health Division, Oregon Health Authority, Portland, Oregon, USA.
Alexia Zhang, Public Health Division, Oregon Health Authority, Portland, Oregon, USA.
William Schaffner, Vanderbilt University Medical Center, Nashville, Tennessee, USA.
Meghan Lyman, Division of Foodborne, Waterborne, and Environmental Diseases, National Center for Emerging and Zoonotic Infectious Diseases, Centers for Disease Control and Prevention, Atlanta, Georgia, USA.
References
- 1. Pfaller MA, Diekema DJ. Epidemiology of invasive candidiasis: a persistent public health problem. Clin Microbiol Rev 2007; 20:133–63. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Magill SS, O’Leary E, Janelle SJ, et al. Changes in prevalence of health care–associated infections in U. S. hospitals. N Engl J Med 2018; 379:1732–44. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Toda M, Williams SR, Berkow EL, et al. Population-based active surveillance for culture-confirmed candidemia—four sites, United States, 2012–2016. MMWR Surveill Summ 2019; 68:1–15. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Strollo S, Lionakis MS, Adjemian J, Steiner CA, Prevots DR. Epidemiology of hospitalizations associated with invasive candidiasis, United States, 2002–2012. Emerg Infect Dis 2016; 23:7–13. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Sarrazin MSV, Rosenthal GE. Finding pure and simple truths with administrative data. JAMA 2012; 307:1433–5. [DOI] [PubMed] [Google Scholar]
- 6. Tsay SV, Mu Y, Williams S, et al. Burden of candidemia in the United States 2017. Clin Infect Dis 2020; 71:e449–53. [DOI] [PubMed] [Google Scholar]
- 7. Centers for Disease Control and Prevention . Candida Bloodstream Infections. Available at: https://www.cdc.gov/hai/eip/candida.html. Accessed 30 March 2022.
- 8. Valentine JC, Worth LJ, Verspoor KM, et al. Classification performance of administrative coding data for detection of invasive fungal infection in paediatric cancer patients. PLoS One 2020; 15:e0238889. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Benedict K, Whitham HK, Jackson BR. Economic burden of fungal diseases in the United States. Open Forum Infect Dis 2022; 9:ofac097. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Gold JAW, Revis A, Thomas S, et al. Clinical characteristics, health care utilization, and outcomes among patients in a pilot surveillance system for invasive mold disease—Georgia, United States 2017–2019. Open Forum Infect Dis 2022; 9:ofac215. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Kadri SS, Gundrum J, Warner S, et al. Uptake and accuracy of the diagnosis code for COVID-19 among US hospitalizations. JAMA 2020; 324:2553–4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Clancy CJ, Nguyen MH. Finding the “missing 50%” of invasive candidiasis: how nonculture diagnostics will improve understanding of disease spectrum and transform patient care. Clin Infect Dis 2013; 56:1284–92. [DOI] [PubMed] [Google Scholar]
- 13. Alexander S, Conner T, Slaughter T. Overview of inpatient coding. Am J Health-Syst Pharmacy 2003; 60:S11–S4. [DOI] [PubMed] [Google Scholar]
- 14. Rhee C, Gohil S, Klompas M. Regulatory mandates for sepsis care—reasons for caution. N Engl J Med 2014; 370:1673–6. [DOI] [PMC free article] [PubMed] [Google Scholar]