Faculty Bibliography

The expression of diagnostic certainty in radiology reports is common and used to inform clinicians and patients about the degree of confidence a radiologist has in a finding or differential diagnosis. However, radiologists express diagnostic certainty in myriad ways, which leads to ambiguity and risk of misdirected clinical management. Standardizing how diagnostic certainty is expressed is likely to improve the fidelity of information transfer between radiologists and referring providers and patients. In recognition of the need to improve the clarity of diagnostic tests, the American College of Radiology Commission on Quality and Safety convened a working group of radiologists with expertise and interest in radiology reporting and diagnostic communication. The group's task was to summarize current knowledge and challenges in the expression of diagnostic certainty in radiology reports, and to offer recommendations for establishing greater consensus. Although there is insufficient agreement at this time to formally recommend one diagnostic certainty scale over another, two examples from single-site experiences and one consensus scale favored by this writing group are provided for consideration and pilot testing by radiology practices.

OBJECTIVE:The Neiman Imaging Comorbidity Index (NICI) was developed and validated in a claims dataset encompassing >10 million privately insured beneficiaries, in which it outperformed the commonly used Charlson Comorbidity Index (CCI) in predicting advanced imaging use. This external validation assessed the broader generalizability of NICI for predicting receipt of advanced imaging in nationally representative populations, including patients insured by Medicare, Medicaid, and private payers. METHODS:All 2018 to 2019 patient-level claims from the CMS Medicare 5% Research Identifiable File, CMS Medicaid 100% Research Identifiable File, and private insurance (commercial and Medicare Advantage) claims from Inovalon Insights, LLC, were included. Using 2018 comorbidity data, beneficiaries were assigned CCI and NICI. Area under the receiver operator characteristic curves (AUCs) measured index performance predicting advanced imaging in 2019. AUCs for NICI and CCI were compared overall, across age groups, and after adjusting for age and sex. RESULTS:A total of 108,846,549 beneficiaries were included across Medicare (n = 2,536,403), Medicaid (n = 49,685,052), and private insurance (n = 56,625,094) datasets. NICI outperformed CCI in Medicare (AUC: 0.7709, 95 confidence interval [CI]: 0.7702-0.7716 versus AUC: 0.7503, 95% CI: 0.7496-0.7510; P < .001), Medicaid (AUC: 0.6876, 95% CI: 0.6874-0.6878 versus AUC: 0.6798 95% CI: 0.6796-0.6800]; P < .001), and private insurance data (AUC: 0.6658, 95% CI: 0.6656-0.6660 versus AUC: 0.6479, 95% CI: 0.6477-0.6481; P < .001). NICI outperformed CCI in adjusted models and in nearly all age strata across the three cohorts. DISCUSSION/CONCLUSIONS:The NICI outperformed CCI in predicting advanced imaging in populations insured by numerous different payers. Validation data support NICI as the preferred index to adjust for patient comorbidities when studying advanced imaging as an outcome, but further investigations are warranted.