LAUSR

Abstract Background Classifying fever of unknown origin (FUO) into categorical etiologies (ie, infections, noninfectious inflammatory, oncologic, miscellaneous, and undiagnosed disorders) remains unstandardized and subject to discrepancies. As some disease classifications change, a systematic review of studies would help physicians anticipate the frequency of illness types they may encounter that could influence care. Methods We systematically reviewed prospective FUO studies published across the Medline (PubMed), Embase, Scopus, and Web of Science databases from January 1, 1997, to July 31, 2022. We performed a meta-analysis to estimate associated pooled proportions between the investigator-determined choice of disease category and those determined by the International Classification of Diseases, 10th edition (ICD-10), methodology. Results The proportion of patients with a difference between the investigator and ICD-10-adjusted noninfectious inflammatory disorder category was 1.2% (95% CI, 0.005–0.021; P < .001), and the proportion was similar for the miscellaneous category at 1.5% (95% CI, 0.007–0.025; P < .001). The miscellaneous and noninfectious inflammatory disorders categories demonstrated significant across-study heterogeneity in the proportions of patients changing categories, with 52.7% (P = .007) and 51.0% (P = .010) I2, respectively. Conclusions Adjusting FUO-associated diagnoses by ICD-10 methodology was associated with a statistically significant risk of over- or underestimation of disease category frequency approximation when using a 5 FUO category system. An FUO diagnostic classification system that better reflects mechanistic understanding would assist future research and enhance comparability across heterogenous populations and different geographic regions. We propose an updated FUO classification scheme that streamlines categorizations, aligns with the current understanding of disease mechanisms, and should facilitate empirical decisions, if necessary.