OBJECTIVE: To validate a novel coding method using Current Procedural Terminology, Fourth Edition (CPT-4) codes for identifying infants who underwent a full evaluation for serious bacterial infection (SBI).
METHODS: We performed a multicenter, retrospective examination to determine the accuracy of a combination of CPT-4 codes for blood, cerebrospinal fluid (CSF), and urine cultures to identify previously healthy infants ≤90 days old admitted to a general care floor and fully evaluated for SBI. Full SBI evaluation was defined as blood, CSF, and urine cultures performed during the emergency department encounter or corresponding hospitalization. Cases were defined as infants who had codes for blood, CSF, and urine cultures (87040, 87070, and either 87086 or 87088), and these were compared with all other encounters. We validated these findings by comparing medical record documentation of blood, CSF, and urine cultures to the corresponding CPT-4 codes, with calculation of sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV).
RESULTS: We identified 8548 qualifying encounters, and 347 (4%) had a combination of CPT-4 codes 87040, 87070, and either 87086 or 87088. This combination had a sensitivity of 100% (95% confidence interval, 98.9–100) and specificity of 98.2% (95% confidence interval, 97.3–98.8) for identifying infants who underwent full SBI evaluation for an unknown source.
CONCLUSIONS: CPT-4 codes provide an accurate means to identify infants who underwent complete SBI evaluation.
Infants ≤90 days old presenting with nonspecific symptoms often undergo a full evaluation for serious bacterial infection (SBI), including a complete blood cell count, blood, urine, and cerebrospinal fluid (CSF) cultures, hospital admission, and empirical antibiotics pending test results.1,2 The Agency for Healthcare Research and Quality has identified the diagnosis and management of these infants as an area requiring additional study, largely because the harms of different management strategies are not well studied.2
Prospective analyses are difficult because of their resource-intensive nature and the low prevalence of SBI in otherwise healthy infants (meningitis <0.5%, bacteremia <2%).3 Large patient cohorts can be rapidly and inexpensively derived retrospectively from existing databases; however, we are aware of no validated methods by which infants who underwent SBI evaluation can be identified. Retrospective studies using International Classification of Diseases, Ninth Revision (ICD-9) codes vary in their ability to accurately identify children with specific diagnoses.4–7 Because most young infants who undergo SBI evaluation are ultimately diagnosed with a non-SBI illness, identification of these infants by using ICD-9 codes poses a significant challenge.7 Furthermore, large database studies that use ICD-9 codes to identify this patient population often attempt to define the population via procedural codes.8,9 To our knowledge, there has not been a previous validation of procedural codes used to identify infants who underwent SBI.
We aimed to validate Current Procedural Terminology, Fourth Edition (CPT-4) codes to identify infants who underwent a full SBI evaluation and hypothesize that a combination of codes for blood, CSF, and urine cultures would accurately identify infants undergoing complete SBI evaluation. If successful, these codes could then be used as a validated filter after a preliminary screening by ICD-9 code (or the expected transition to ICD, 10th edition) or be used alone to identify infants who undergo full SBI evaluation.
This was a multicenter, retrospective examination of data from 3 tertiary care children’s hospitals in the United States (Golisano Children’s Hospital at University of Rochester Medical Center, Rochester, NY; Floating Hospital for Children/Tufts Medical Center, Boston, MA; and Children’s Mercy Hospital, Kansas City, MO). All 3 sites use electronic medical record systems. The institutional review board at each participating site approved this study.
We first identified all infants ≤90 days of age admitted to a general ward via the emergency department (ED) at 1 of the 3 participating sites from April 1, 2011 to April 1, 2014. Infants admitted directly to an ICU, and those with past medical history or procedures that place the infant at elevated risk for SBI for reasons unrelated to the primary encounter (eg, central line placement or invasive surgeries), were excluded (Table 1). Specifically, we aimed to identify hospitalized infants ≤90 days of age who had a full evaluation for SBI performed in the ED or general ward. We defined full SBI evaluation as having blood, urine, and CSF cultures performed at any time during the ED encounter or corresponding hospitalization. Because we sought to identify infants who underwent empirical evaluation for SBI rather than infants with known disease processes for whom the evaluation was part of a different workup (eg, an infant with known endocarditis), manual chart review was performed to determine the validity of the abstraction procedure for the additional criterion of SBI evaluation for infection of an unknown source.
Identification of Infants
Filters corresponding with our study criteria were applied to billing databases, and the presence or absence of 6 CPT-4 codes was recorded for each encounter (Table 2). Because no specific CPT-4 code exists for CSF culture, the search included 2 codes for which CSF cultures may have been recorded (87070, 87071).
Two sites reviewed the medical records of all infants for which ≥1 of the CPT-4 codes was documented. The third site reviewed records for all 145 infants with all 3 codes (blood, urine, and CSF cultures) and a random selection of 314 charts from the 1289 that had ≥1, but not all 3, codes (because of a large sample size and feasibility of data collection). All sites reviewed a random sample of 10% of all infant encounters for which none of the 6 corresponding CPT-4 codes were documented to allow calculations of sensitivity, specificity, and negative predictive value (NPV) that are based on the population of infants who may be evaluated empirically. Medical record review included determination of whether the infant had blood, urine, or CSF cultures performed during the ED encounter or corresponding hospital admission. Additionally, charts were examined to determine whether the infant had a full SBI evaluation performed for infection of unknown source. We did so to ensure that cultures were not performed as follow-up to known infections or as part of larger workups (eg, an oncologic workup).
The primary outcome was accuracy of the combination of codes 87040, 87070, and either 87086 or 87088 for identifying full SBI evaluation. We determined accuracy by calculating sensitivity, specificity, positive predictive value (PPV), and NPV. True-positives were defined as infants with the full combination of codes 87040, 87070, and either 87086 or 87088 who had blood, CSF, and urine cultures performed during the encounter. False-positives were defined as having the aforementioned combination of codes but without the combination of blood, CSF, and urine cultures obtained. True-negatives had any other combination (or none) of the 6 CPT-4 codes and did not have blood, CSF, and urine cultures obtained. False-negatives had any other combination (or none) of the 6 CPT-4 codes and had blood, CSF, and urine cultures obtained. Confidence intervals (CIs) were estimated via the Wilson score interval method.
Our search identified 8548 infants ≤90 days of age who met study criteria before evaluation for CPT codes (3032 from site 1, 591 from site 2, 4925 from site 3). Of these, 347 (4%) encounters had a combination of codes 87040, 87070, and either 87086 or 87088 (Fig 1). All 87040 (n = 687) codes corresponded to a blood culture, and all 87086/8 (n = 630) codes corresponded to a urine culture during the encounter of record. Of the 87070 (n = 418) codes, 357 (85%) corresponded to a CSF culture and 61 (15%) coded for culture of a specimen other than blood, urine, or CSF (46 wound cultures, 12 eye cultures, 2 joint cultures, and 1 sputum culture). The 87081 (n = 66) codes all coded for a culture other than blood, urine, or CSF. No 87071 codes were identified.
Of the 347 subjects with the combination of codes, 333 (96%) underwent a full SBI evaluation (blood, urine, and CSF cultures). The remaining 14 did not undergo a full evaluation because all had 87070 codes for a specimen source other than CSF. Of the 333 cases who underwent a full SBI evaluation, 323 (97%) had evaluations performed for infection of unknown source, and 10 evaluations were performed in the setting of an obvious source of symptoms, including omphalitis (3), joint swelling (1), intussusception (1), pyloric stenosis (1), frequent seizures (2), and breast abscess (n = 2). Additionally, 8201 encounters did not have the combination of codes, 87040, 87070, and either 87086 or 87088. Of these, 1600 encounters included ≥1 of the codes, and 625 were evaluated by chart review (chart review included all charts from sites 1 and 2 and 314 of 1289 charts from site 3 because of feasibility of data collection). None of the patients corresponding to the 625 encounters underwent a full SBI evaluation for unknown source in the ED or corresponding hospitalization. The 6601 remaining encounters had none of the codes. Of these, 665 (10%) underwent chart review, and none were found to have had a full SBI evaluation performed. This resulted in a total of 1290 true-negatives and no false-negatives.
In total, for infants evaluated fully for SBI, there were 333 true-positives, 1290 true-negatives, 14 false-positives, and no false-negatives. For infants who underwent full SBI evaluation for infection of unknown source, there were 323 true-positives, 1290 true-negatives, 24 false-positives, and no false-negatives. The sensitivity and specificity of the combination of codes 87040, 87070, and either 87086 or 87088 for full SBI evaluation for infection of unknown source was 100% (95% CI, 98.8–100) and 98.2% (95% CI, 97.3–98.8), respectively (Table 3).
This multicenter study evaluated the performance of a combination of CPT-4 codes in identifying previously healthy infants ≤90 days old admitted to a general care floor who underwent full evaluation for SBI. Additionally, we examined the performance of the same codes in identifying infants who specifically underwent this evaluation because of concern for infection of unknown or uncertain source. A combination of the codes 87040 (representing blood culture), 87070 (representing CSF culture), and either 87086 or 87088 (representing urine culture) reliably identified our target population.
Fever is a common reason for performing a full SBI evaluation in infants ≤90 days old.1,10 Our codes do not specifically look for infants with fever; however, the full SBI evaluation is performed based on perceived risk, and infants with SBI can also present with hypothermia or normothermia. Requiring the presence of fever as part of the presenting symptoms necessary for SBI evaluation can significantly decrease the sensitivity of identifying at-risk infants.11 Our coding structure does not take into account the reason for the evaluation; thus, users can narrow the search parameters based on any variable available in their database.
Because CPT-4 codes can only identify services rendered, our proposed coding structure targets infants who underwent full SBI evaluation in the ED or corresponding hospitalization. We think this approach allows the user to target a very relevant population; however, there are instances in which infants may undergo only partial SBI evaluation, and our coding structure will not identify any of the aforementioned infants. Given the variability in current practice,12,13 it would be difficult to arrive at 1 algorithm that has sufficient sensitivity to identify all possible evaluation strategies while also maintaining an appropriate level of specificity. In the future, procedure codes for urinalysis and lumbar puncture (without obtainment of adequate CSF) may be investigated to attempt to identify infants who have some workup performed but who may not necessarily have a resulting culture and therefore would not have one of the 6 CPT-4 codes in our study.
Our study has other limitations. We performed this review at 3 sites, and although we found consistency between these sites, our specific CPT-4 coding may not be generalizable to all institutions. However, we examined a range of CPT-4 codes for which blood, CSF, and urine cultures were likely to be coded, and our data suggest that CPT-4 codes are consistently used for specific procedures. Our determination of whether an SBI evaluation was performed for infection without a source was subjective. Although we chose to define full SBI evaluation as occurring only in infants without a confirmed source of infection, this distinction could arguably be removed, because infants ≤90 days of age are at elevated risk of disseminated infection from a variety of pathogens that can manifest initially as focal infection.14,15
Our proposed patient identification strategy suggests that CPT-4 codes can be used to accurately identify infants ≤90 days of age who undergo full SBI evaluations in the ED or during the corresponding hospitalization.
We thank members of the Academic Pediatric Association’s Research Scholar’s Program who aided in study design.
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
FUNDING: Partially funded by a Young Investigator Award given to Dr Biondi from the Academic Pediatric Association.
POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.
- ↵Cincinnati Children’s Hospital Medical Center. Evidence-Based Clinical Care Guideline for Fever of Uncertain Source in Infants 60 Days of Age or Less. October 2010. Available at: www.cincinnatichildrens.org/svc/alpha/h/health-policy/ev-based/default.htm
- Hui C,
- Neto G,
- Tsertsvadze A,
- et al
- Byington CL,
- Reynolds CC,
- Korgenski K,
- et al
- Tieder JS,
- Hall M,
- Auger KA,
- et al
- Jain S,
- Cheng J,
- Alpern ER,
- et al
- Aronson PL,
- Thurm C,
- Alpern ER,
- et al
- Nigrovic LE,
- Fine AM,
- Monuteaux MC,
- Shah SS,
- Neuman MI
- Bachur RG,
- Harper MB
- Copyright © 2016 by the American Academy of Pediatrics