Real-world virtual patient simulation to improve diagnostic performance through deliberate practice: a prospective quasi-experimental study

Abstract

Abstract Objectives Diagnostic errors are pervasive in medicine and most often caused by clinical reasoning failures. Clinical presentations characterized by nonspecific symptoms with broad differential diagnoses (e.g., dizziness) are especially prone to such errors. Methods We hypothesized that novice clinicians could achieve proficiency diagnosing dizziness by training with virtual patients (VPs). This was a prospective, quasi-experimental, pretest-posttest study (2019) at a single academic medical center. Internal medicine interns (intervention group) were compared to second/third year residents (control group). A case library of VPs with dizziness was developed from a clinical trial (AVERT-NCT02483429). The approach (VIPER – Virtual Interactive Practice to build Expertise using Real cases) consisted of brief lectures combined with 9 h of supervised deliberate practice. Residents were provided dizziness-related reading and teaching modules. Both groups completed pretests and posttests. Results For interns (n=22) vs. residents (n=18), pretest median diagnostic accuracy did not differ (33% [IQR 18–46] vs. 31% [IQR 13–50], p=0.61) between groups, while posttest accuracy did (50% [IQR 42–67] vs. 20% [IQR 17–33], p=0.001). Pretest median appropriate imaging did not differ (33% [IQR 17–38] vs. 31% [IQR 13–38], p=0.89) between groups, while posttest appropriateness did (65% [IQR 52–74] vs. 25% [IQR 17–36], p<0.001). Conclusions Just 9 h of deliberate practice increased diagnostic skills (both accuracy and testing appropriateness) of medicine interns evaluating real-world dizziness ‘in silico’ more than ∼1.7 years of residency training. Applying condensed educational experiences such as VIPER across a broad range of common presentations could significantly enhance diagnostic education and translate to improved patient care.