Background: Early accurate diagnosis and risk assessment for malaria are crucial for improving patients’ terminal prognosis, and preventing them from progressing to severe or critical state. This study aims to assess the ability of medical institutions at all levels to diagnose malaria and develop a predictive model to guide prevention and reduce diagnostic error.
Methods: A retrospective study was conducted on 494 patients admitted to hospitals with a diagnosis of malaria from January 2014 through December 2016. Descriptive statistics were calculated using Stata 15. Decision tree analysis using the IBM® SPSS® Statistics Version 25 was used to build a model to predict the patients who may be misdiagnosed.
Results: Of the 494 patients included in this study, 136 (27.5%) patients sought care through a county-level hospital at their first visit; and 130 (26.3%), 128 (25.9%), and 41 (8.3%) of the patients went to a prefecture-level hospital, village clinic, and provincial-level hospital, respectively. Nearly 60% of malaria patients were misdiagnosed on their first visit, and 18.8% of patients had complications. The median time from onset to the first visit was 2 days (IQR: 0 - 3 days), and the median time from the first visit to diagnosis was 3 days (IQR: 0 - 4 days). The decision tree classification of malaria patients being misdiagnosed consisted of six categorical variables: healthcare facilities for the initial diagnosis, time interval between onset and initial diagnosis, region, residence type, insurance status, and age. Furthermore, the predictive accuracy of the model was 85.0%.
Conclusion: Insufficient diagnostic capacity of healthcare facilities with lower administrative levels for the first visit was found to be the most important risk factors in misdiagnosing patients. Researchers and clinicians should consider strengthening the primary care facilities, time interval between onset and initial diagnosis, administrated residence, and health insurance status to reduce diagnostic errors.