Reverse transcriptase polymerase chain reaction (RT-PCR) is a key tool to diagnose Covid-19. Yet it may not be the most efficient test in all patients. In this paper, we develop a clinical strategy for prescribing RT-PCR to patients based on data from COVIDOM, a French cohort of 54,000 patients with clinically suspected Covid-19, including 12,810 patients tested by RT-PCR. We use a machine-learning algorithm (decision tree) in order to predict RT-PCR results based on the clinical presentation. We show that symptoms alone are sufficient to predict RT-PCR outcome with a mean average precision of 86%. We identify combinations of symptoms that are predictive of RT-PCR positivity (90% for anosmia/ageusia) or negativity (only 30% of RT-PCR+ for a subgroup with cardiopulmonary symptoms): in both cases, RT-PCR provides little added diagnostic value. We propose a prescribing strategy based on clinical presentation that can improve the global efficiency of RT-PCR testing.
【저자키워드】 Infectious diseases, Diseases, DNA, 【초록키워드】 Cardiopulmonary, Symptom, outcome, RT-PCR, Anosmia, polymerase chain reaction, PCR testing, Cohort, Reverse transcriptase polymerase chain reaction, Ageusia, Algorithm, Patient, Decision tree, patients, predict, machine-learning, diagnose, Combination, Efficiency, reverse transcriptase, Predictive, Chain Reaction, RT-PCR testing, diagnostic value, combinations, Anosmia/ageusia, French, negativity, mean average precision, polymerase chain, IMPROVE, tested, identify, develop, provide, clinically suspected, added, 【제목키워드】 Cohort, Outpatient, Analysis, French,