Using Probability Estimates to Evaluate a Patient With Weakness.

In this case report, we review how probabilistic reasoning can be implemented in retrospect to refine the diagnostic process. A 67-year-old female with a history of polymyalgia rheumatica (PMR) and a recent dental procedure presented with weakness, falls, and chills ongoing for two weeks. She reported pain in her shoulders and lower back. On presentation, she was febrile, and labs were notable for leukocytosis with neutrophilic predominance and an elevated erythrocyte sedimentation rate (ESR). Chest radiograph revealed a left lower lung opacity, which was not seen on a repeat film. She was treated with antibiotics for community-acquired pneumonia and steroids for an exacerbation of PMR. After eight days of hospitalization, she was transferred to a subacute rehabilitation facility. A month later, she was readmitted with worsening lower back pain and right lower extremity weakness. Imaging revealed discitis and osteomyelitis at L1-L2. A spinal epidural abscess was present, leading to severe compression of the cauda equina nerve roots. Aspirate was positive for group B streptococcus. With antibiotic treatment alone, she recovered with resolution of her weakness. In reviewing the literature, it becomes evident where improvements could have been made in the diagnostic process. Fever, leukocytosis, and neurological weakness are not commonly associated with PMR exacerbations. Lack of cough or shortness of breath, a persistently elevated erythrocyte sedimentation rate and C-reactive protein despite antibiotic treatment, and a repeat chest radiograph without an opacity suggest an alternative diagnosis to pneumonia. Persistent back pain with an insidious onset is a feature of untreated spinal epidural abscess. Steroid use and dental procedures are possible risk factors for spinal epidural abscess. By shedding light on how probabilities should be estimated, we hope to encourage probabilistic thinking to improve diagnostic accuracy. As with the best political forecasters, making precise probability estimates and frequently updating them may improve diagnostic accuracy for clinicians.